Version: dev

NoahSnowCompaction

Snow-pack compaction in WRF's unified Noah land-surface model, SUBROUTINE SNOWPACK of phys/module_sf_noahlsm.F (WRF 4.8, sf_surface_physics = 2): the densification of an existing pack under its own weight, and the melt water it retains once its surface reaches freezing. THE PHYSICS. Anderson's compaction equation (NOAA Technical Report NWS 19, eq. 3.29) makes the density at a point in the pack grow exponentially in the water equivalent lying above it, at a rate that rises with temperature and falls with the density already reached: d(rho)/dt = rho c1 exp(ctemp tavgc - c2 rho) w. Koren's 1995 approximate solution integrates that over the DEPTH of the pack rather than over time, so the pack-mean density after a step of dtsec is sndens (exp(x) - 1)/x with x = bfac esdcx, and bfac carries the whole of the step length. WRF evaluates (exp(x) - 1)/x through its Taylor series to four terms, because the closed form is 0/0 at x = 0 -- its own comment says so at length. Once the snow surface reaches freezing a second process runs: 13 % of the pack's pore space per day fills with melt water, which relaxes the density towards that of water. The depth is not integrated at all; it is the quotient esd/sndens at the end. HOW IT JOINS AN ODE DOCUMENT. SNOWPACK is an in-place sub-step, not a rate -- Koren's solution is a closed form in the step length and there is no dt-free rate inside it -- so this component takes dtsec as a PARAMETER OF THE SCHEME and exposes `dsndens_dt` = increment/dtsec as the observed a consumer integrating the pack adds, which is the same shape WSM6's operator-split stages use. A mounting document must prescribe dtsec and couple it in; it is not the integrator's step, and the result differs from the dtsec -> 0 limit by xarg/3 + O(xarg^2), measured at 2.2e-6 relative at WRF's own 60 s step. WHY THE INCREMENT IS NEVER FORMED AS A DIFFERENCE. The compaction moves the density by a factor 1 + O(dtsec): 9.4e-5 of it at WRF's 60 s step on the wettest reference column, 4.5e-9 on the thinnest pack, and 7.5e-17 at a 1e-6 s step. Writing the increment as sndens_out - sndens_in therefore returns round-off at any step small enough to be a dt -> 0 reference, and writing it as sndens (pexp - 1) is no better because pexp is stored as 1 + O(1e-13). Both the component and its reference form it instead as sndens_in pexp_m1 + dw (1 - dsx), an exact rearrangement in which nothing cancels, and the reference is taken from a publisher that emits the series sum BEFORE WRF adds one to it (FORTRAN_BUGS N99, the same move N97 and N98 forced for the frozen-soil sink and the melt rate one routine over). REFERENCE STEP. Because nothing is differenced there is no cancellation floor pushing the reference step down, and the usual 'shrink dt until the increment is the derivative' recipe is actively WRONG here: the compaction increment is proportional to dtsec, so a SMALLER step puts LESS of an assertion's budget on the physics. The primary regime is therefore WRF's own 60 s step, where a rel 1e-9 assertion on sndens_out tests the compaction to between 1e-5 and 2e-1 of itself, and a second regime at 1e-6 s pins the tendency in the dt -> 0 limit, where the naive form would be 100 % noise. NOT TRANSCRIBED: the whole UA_PHYS branch (the snomlt-limited retention fraction), which is off in every reference run; SNOW_NEW, which sets the density of freshly fallen snow and is a separate routine; and the caller's own esd > 0 guard, which belongs to SNOPAC and is reported here as the indicator `pack_present`. MEASURED AGREEMENT: every one of the 1659 assertions passes at rel 3e-15, a few ulps of binary64, over all 134 tests of both regimes; the published bound is the 1e-9 the project sets for a double-precision kernel reference, and the six orders between them are headroom, not accuracy that was needed.

Domain
land_surface
Subdomain
noah
Type
model
.esm schema
1.1.0
Source
components/land_surface/noah/snow_compaction.esm
Tags
#land_surface #noah #model #wrf #land-surface #snow #snow-density #compaction

Description

Snow-pack compaction in WRF’s unified Noah land-surface model, SUBROUTINE SNOWPACK of phys/module_sf_noahlsm.F (WRF 4.8, sf_surface_physics = 2): the densification of an existing pack under its own weight, and the melt water it retains once its surface reaches freezing. THE PHYSICS. Anderson’s compaction equation (NOAA Technical Report NWS 19, eq. 3.29) makes the density at a point in the pack grow exponentially in the water equivalent lying above it, at a rate that rises with temperature and falls with the density already reached: d(rho)/dt = rho c1 exp(ctemp tavgc - c2 rho) w. Koren’s 1995 approximate solution integrates that over the DEPTH of the pack rather than over time, so the pack-mean density after a step of dtsec is sndens (exp(x) - 1)/x with x = bfac esdcx, and bfac carries the whole of the step length. WRF evaluates (exp(x) - 1)/x through its Taylor series to four terms, because the closed form is 0/0 at x = 0 – its own comment says so at length. Once the snow surface reaches freezing a second process runs: 13 % of the pack’s pore space per day fills with melt water, which relaxes the density towards that of water. The depth is not integrated at all; it is the quotient esd/sndens at the end. HOW IT JOINS AN ODE DOCUMENT. SNOWPACK is an in-place sub-step, not a rate – Koren’s solution is a closed form in the step length and there is no dt-free rate inside it – so this component takes dtsec as a PARAMETER OF THE SCHEME and exposes dsndens_dt = increment/dtsec as the observed a consumer integrating the pack adds, which is the same shape WSM6’s operator-split stages use. A mounting document must prescribe dtsec and couple it in; it is not the integrator’s step, and the result differs from the dtsec -> 0 limit by xarg/3 + O(xarg^2), measured at 2.2e-6 relative at WRF’s own 60 s step. WHY THE INCREMENT IS NEVER FORMED AS A DIFFERENCE. The compaction moves the density by a factor 1 + O(dtsec): 9.4e-5 of it at WRF’s 60 s step on the wettest reference column, 4.5e-9 on the thinnest pack, and 7.5e-17 at a 1e-6 s step. Writing the increment as sndens_out - sndens_in therefore returns round-off at any step small enough to be a dt -> 0 reference, and writing it as sndens (pexp - 1) is no better because pexp is stored as 1 + O(1e-13). Both the component and its reference form it instead as sndens_in pexp_m1 + dw (1 - dsx), an exact rearrangement in which nothing cancels, and the reference is taken from a publisher that emits the series sum BEFORE WRF adds one to it (FORTRAN_BUGS N99, the same move N97 and N98 forced for the frozen-soil sink and the melt rate one routine over). REFERENCE STEP. Because nothing is differenced there is no cancellation floor pushing the reference step down, and the usual ‘shrink dt until the increment is the derivative’ recipe is actively WRONG here: the compaction increment is proportional to dtsec, so a SMALLER step puts LESS of an assertion’s budget on the physics. The primary regime is therefore WRF’s own 60 s step, where a rel 1e-9 assertion on sndens_out tests the compaction to between 1e-5 and 2e-1 of itself, and a second regime at 1e-6 s pins the tendency in the dt -> 0 limit, where the naive form would be 100 % noise. NOT TRANSCRIBED: the whole UA_PHYS branch (the snomlt-limited retention fraction), which is off in every reference run; SNOW_NEW, which sets the density of freshly fallen snow and is a separate routine; and the caller’s own esd > 0 guard, which belongs to SNOPAC and is reported here as the indicator pack_present. MEASURED AGREEMENT: every one of the 1659 assertions passes at rel 3e-15, a few ulps of binary64, over all 134 tests of both regimes; the published bound is the 1e-9 the project sets for a double-precision kernel reference, and the six orders between them are headroom, not accuracy that was needed.

Reference

  • Anderson, E. A. (1976). A point energy and mass balance model of a snow cover. NOAA Technical Report NWS 19, equation (3.29).
  • Koren, V., Schaake, J., Mitchell, K., Duan, Q.-Y., Chen, F., & Baker, J. M. (1999). A parameterization of snowpack and frozen ground intended for NCEP weather and climate models. J. Geophys. Res., 104(D16), 19569-19585.
  • Kojima, K. (1967). Densification of seasonal snow cover. Physics of Snow and Ice, 1(2), 929-952.
  • wrf-model/WRF v4.8.0 phys/module_sf_noahlsm.F SUBROUTINE SNOWPACK lines 3417-3549, called from SNOPAC at line 3403 (cold-season reference run instrumented in ctessum-claude/WRF branch earthsciml-instrumented-noahcold).

Parameters

NameUnitsDefaultDescription
esdm0.06Water equivalent of the pack that SNOWPACK is handed, WRF’s ESD – the SWE AFTER SNOPAC has taken this step’s sublimation and melt out of it, not the pack the step started with. Consumer-supplied, because it is the snow-energy component’s output that sets it.
sndens_in10.2Snow density entering the compaction, as a fraction of the density of water (WRF’s SNDENS; its own comment reads ‘G/CM3=DIMENSIONLESS FRACTION OF H2O DENSITY’, and the two are numerically equal). This is NOT the density implied by the dumped snow state: SFLX resets it to SNEQV/SNOWH and then, on a snowing step, SNOW_NEW moves it again before SNOPAC is reached, so it is supplied by the consumer and is published from inside SNOWPACK for the tests.
tsnowK265.0Snow surface temperature, WRF’s TSNOW dummy, which SNOPAC fills with the skin temperature T1 its own energy balance has just produced. Its position relative to the freezing point selects the melt-water retention branch.
tsoilK268.0Soil temperature at the top of the soil column, WRF’s TSOIL dummy, which SNOPAC fills with YY = stc[1] - 0.5 ssoil zsoil[1]/df1 – the linearised surface value the soil-heat closure forms, not stc[1] itself.
dtsecs60.0The sub-step the compaction is applied over, WRF’s DTSEC. A PARAMETER OF THE SCHEME, not the integrator’s step: Koren’s solution is a closed form in the step length, so the increment it returns is O(dtsec) and there is no dt-free rate hiding inside it. Prescribe it in the mounting document and couple it in, exactly as WSM6’s dtcld is prescribed; a parent test cannot set a mounted component’s parameter.

Observed

NameUnitsDefaultDescription
tsnowcKSnow surface temperature above freezing, WRF’s TSNOWC = TSNOW - 273.15.
tsoilcKSoil surface temperature above freezing, WRF’s TSOILC.
tavgcKMean temperature of the pack above freezing, the arithmetic mean of the snow surface and the soil below it. It is the only temperature the compaction sees: warm snow compacts fast.
esdccmWater equivalent in SNOWPACK’s own simulation units, centimetres.
esdcxcmWater equivalent floored at par.snow_esdc_min, WRF’s ESDCX. It bounds the overburden a vanishing pack is allowed to feel. It is used ONLY in the compaction argument; the depth diagnostic at the end uses the unfloored esdc, and that distinction matters for the nine reference steps whose pack is below the floor.
dthrhThe sub-step in hours, the unit SNOWPACK’s rate constants are expressed in.
comp_exponent1Anderson’s compaction exponent, snow_ctemp tavgc - snow_c2 sndens_in: compaction accelerates with temperature and is suppressed by the density already reached. Declared dimensionless in its own right so that the exponential is handed a quantity of scale one.
bfac1/cmThe compaction rate per centimetre of overlying water equivalent, over the whole sub-step: dthr snow_c1 exp(comp_exponent). It is proportional to dtsec, which is where all of the step dependence enters.
xarg1bfac esdcx, the dimensionless compaction of the pack’s BASE over the sub-step. It is proportional to dtsec: 6.7e-6 at WRF’s 60 s on a 6 cm pack at -12 C, and 1.1e-13 at 1e-6 s.
pexp_m11The mean compaction over the DEPTH of the pack, minus one. Anderson’s equation makes the density at depth w grow as exp(bfac w), so the pack-mean factor is the depth average of that over 0 <= w <= esdcx, i.e. (exp(xarg) - 1)/xarg; WRF replaces it by its Taylor series to IPOL = 4 terms because the closed form is 0/0 at xarg = 0, and this is that series MINUS ONE, xarg/2 + xarg^2/6 + xarg^3/24 + xarg^4/120, written in WRF’s own Horner nesting. It is the whole of the compaction signal and it is the quantity the tests are pinned to: (exp(x)-1)/x itself differs from 1 by O(x), so at any small step the stored pexp is 1 + O(1e-13) and carries no information at all.
pexp1WRF’s PEXP, 1 + pexp_m1.
dsx_raw1Compacted density before the density limits, sndens_in pexp.
dsx_clamp1Which density limit fired on dsx_raw: 0 neither, 1 capped at snow_dens_max, 2 floored at snow_dens_min. WRF applies the two tests in sequence, cap first, and because the cap is above the floor the second test can only fire on a value that the first left alone, so the two are mutually exclusive. UNTESTED, not merely unused: neither limit fires anywhere in the cold-season reference set, all 67 of whose compaction calls report 0.
dsx1Compacted density after the limits, WRF’s DSX.
ddsx1The compaction increment, dsx - sndens_in, written CANCELLATION-FREE as sndens_in pexp_m1 where no limit fired. That is an exact algebraic rearrangement, not an approximation, and it is the reason this component can be referenced at any step size: the increment is 7.5e-17 of the density at a 1e-6 s step, so forming it as a difference would return pure round-off. Where a limit fires the increment is a difference against a constant and is formed directly.
wet1Indicator, 1 where the snow surface has reached freezing and SNOWPACK runs its melt-water retention branch. 29 of the 67 compaction calls in the cold-season reference set; the other 38 are dry.
dw1Fraction of the pack’s pore space filled by retained melt water over the sub-step, snow_dwrate dthr/24: 13 % per day. Zero on the dry branch. Like bfac it is proportional to dtsec. The UA_PHYS alternative, DW = min(DW, 0.13 snomltc/(esdcx + 0.13 snomltc)), is NOT transcribed – UA_PHYS is off in every reference run.
sndens_wet1Density after retention and before its own cap, WRF’s dsx (1 - dw) + dw: a relaxation of the density towards 1 at rate snow_dwrate, i.e. towards pure water, which is what saturating the pore space with melt water would give.
wet_capped1Indicator for WRF’s second density cap, applied only on the retention branch. UNTESTED: no reference step reaches it, because 0.40 is far above any density these four columns produce.
dsndens1The total density increment over the sub-step, ddsx + dw (1 - dsx), formed without ever differencing two densities. On the dry branch this is the compaction alone.
sndens_out1Density SNOWPACK returns, sndens_in + dsndens.
snowh_outmDepth SNOWPACK returns, esd/sndens_out. It is a PURE DIAGNOSTIC of the water equivalent and the density, with no step dependence of its own, and it uses the UNFLOORED esd. SNOWPACK’s SNOWH argument is declared INTENT(INOUT) but its incoming value is never read: SNOWHC = SNOWH*100 at the top of the routine is overwritten by SNOWHC = ESDC/SNDENS at the bottom before anything uses it (FORTRAN_BUGS N100), so the depth entering the routine is not an input of this component.
dsndens_dt1/sdsndens/dtsec – the density tendency a consumer integrating the pack must add. This is the operator-split form: SNOWPACK is an in-place sub-step, not a rate, so the way it joins an ODE document is as increment over sub-step, exactly as WSM6’s stages do. Its dtsec -> 0 limit is analytic and clean, sndens_in snow_c1 exp(comp_exponent_0) esdcx/(2*3600) + wet snow_dwrate (1 - sndens_in)/86400 with comp_exponent_0 the exponent at dtsec-free bfac; the finite-step value exceeds it by xarg/3 + O(xarg^2), which is 2.2e-6 relative at WRF’s 60 s and 3.7e-14 at 1e-6 s.
pack_present1Indicator for the caller’s guard: SNOPAC calls SNOWPACK only where esd > 0, and where it does not it sets the depth, the density and the cover to zero and the snow conductivity to one instead. Reported here so a consumer knows when this component applies; it is asserted against the reference’s own record of whether SNOWPACK ran.

Observed expressions

$$ tsnowc = tsnow - par.t0 $$

Snow surface temperature above freezing, WRF’s TSNOWC = TSNOW - 273.15.

$$ tsoilc = tsoil - par.t0 $$

Soil surface temperature above freezing, WRF’s TSOILC.

$$ tavgc = 0.5 \cdot \left(tsnowc + tsoilc\right) $$

Mean temperature of the pack above freezing, the arithmetic mean of the snow surface and the soil below it. It is the only temperature the compaction sees: warm snow compacts fast.

$$ esdc = 100.0 \cdot esd $$

Water equivalent in SNOWPACK’s own simulation units, centimetres.

$$ esdcx = \mathrm{max}\left(esdc, par.snow_{esdc\,min}\right) $$

Water equivalent floored at par.snow_esdc_min, WRF’s ESDCX. It bounds the overburden a vanishing pack is allowed to feel. It is used ONLY in the compaction argument; the depth diagnostic at the end uses the unfloored esdc, and that distinction matters for the nine reference steps whose pack is below the floor.

$$ dthr = \frac{dtsec}{3600.0} $$

The sub-step in hours, the unit SNOWPACK’s rate constants are expressed in.

$$ comp_{exponent} = par.snow_{ctemp} \cdot tavgc - par.snow_{c2} \cdot sndens_{in} $$

Anderson’s compaction exponent, snow_ctemp tavgc - snow_c2 sndens_in: compaction accelerates with temperature and is suppressed by the density already reached. Declared dimensionless in its own right so that the exponential is handed a quantity of scale one.

$$ bfac = dthr \cdot par.snow_{c1} \cdot e^{comp_{exponent}} $$

The compaction rate per centimetre of overlying water equivalent, over the whole sub-step: dthr snow_c1 exp(comp_exponent). It is proportional to dtsec, which is where all of the step dependence enters.

$$ xarg = bfac \cdot esdcx $$

bfac esdcx, the dimensionless compaction of the pack’s BASE over the sub-step. It is proportional to dtsec: 6.7e-6 at WRF’s 60 s on a 6 cm pack at -12 C, and 1.1e-13 at 1e-6 s.

$$ pexp_{m1} = \frac{\left(1.0 + \frac{\left(1.0 + \frac{\left(1.0 + \frac{\left(1.0 + 0.0\right) \cdot xarg}{5.0}\right) \cdot xarg}{4.0}\right) \cdot xarg}{3.0}\right) \cdot xarg}{2.0} $$

The mean compaction over the DEPTH of the pack, minus one. Anderson’s equation makes the density at depth w grow as exp(bfac w), so the pack-mean factor is the depth average of that over 0 <= w <= esdcx, i.e. (exp(xarg) - 1)/xarg; WRF replaces it by its Taylor series to IPOL = 4 terms because the closed form is 0/0 at xarg = 0, and this is that series MINUS ONE, xarg/2 + xarg^2/6 + xarg^3/24 + xarg^4/120, written in WRF’s own Horner nesting. It is the whole of the compaction signal and it is the quantity the tests are pinned to: (exp(x)-1)/x itself differs from 1 by O(x), so at any small step the stored pexp is 1 + O(1e-13) and carries no information at all.

$$ pexp = 1.0 + pexp_{m1} $$

WRF’s PEXP, 1 + pexp_m1.

$$ dsx_{raw} = sndens_{in} \cdot pexp $$

Compacted density before the density limits, sndens_in pexp.

$$ dsx_{clamp} = \mathrm{ifelse}\left(\mathrm{>}\left(dsx_{raw}, par.snow_{dens\,max}\right), 1.0, \mathrm{ifelse}\left(\mathrm{<}\left(dsx_{raw}, par.snow_{dens\,min}\right), 2.0, 0.0\right)\right) $$

Which density limit fired on dsx_raw: 0 neither, 1 capped at snow_dens_max, 2 floored at snow_dens_min. WRF applies the two tests in sequence, cap first, and because the cap is above the floor the second test can only fire on a value that the first left alone, so the two are mutually exclusive. UNTESTED, not merely unused: neither limit fires anywhere in the cold-season reference set, all 67 of whose compaction calls report 0.

$$ dsx = \mathrm{ifelse}\left(\mathrm{>}\left(dsx_{raw}, par.snow_{dens\,max}\right), par.snow_{dens\,max}, \mathrm{ifelse}\left(\mathrm{<}\left(dsx_{raw}, par.snow_{dens\,min}\right), par.snow_{dens\,min}, dsx_{raw}\right)\right) $$

Compacted density after the limits, WRF’s DSX.

$$ ddsx = \mathrm{ifelse}\left(dsx_{clamp} = 0.0, sndens_{in} \cdot pexp_{m1}, dsx - sndens_{in}\right) $$

The compaction increment, dsx - sndens_in, written CANCELLATION-FREE as sndens_in pexp_m1 where no limit fired. That is an exact algebraic rearrangement, not an approximation, and it is the reason this component can be referenced at any step size: the increment is 7.5e-17 of the density at a 1e-6 s step, so forming it as a difference would return pure round-off. Where a limit fires the increment is a difference against a constant and is formed directly.

$$ wet = \mathrm{ifelse}\left(\mathrm{>=}\left(tsnowc, 0.0\right), 1.0, 0.0\right) $$

Indicator, 1 where the snow surface has reached freezing and SNOWPACK runs its melt-water retention branch. 29 of the 67 compaction calls in the cold-season reference set; the other 38 are dry.

$$ dw = \mathrm{ifelse}\left(wet = 1.0, \frac{par.snow_{dwrate} \cdot dthr}{24.0}, 0.0\right) $$

Fraction of the pack’s pore space filled by retained melt water over the sub-step, snow_dwrate dthr/24: 13 % per day. Zero on the dry branch. Like bfac it is proportional to dtsec. The UA_PHYS alternative, DW = min(DW, 0.13 snomltc/(esdcx + 0.13 snomltc)), is NOT transcribed – UA_PHYS is off in every reference run.

$$ sndens_{wet} = \mathrm{ifelse}\left(wet = 1.0, dsx \cdot \left(1.0 - dw\right) + dw, dsx\right) $$

Density after retention and before its own cap, WRF’s dsx (1 - dw) + dw: a relaxation of the density towards 1 at rate snow_dwrate, i.e. towards pure water, which is what saturating the pore space with melt water would give.

$$ wet_{capped} = \mathrm{ifelse}\left(\mathrm{and}\left(wet = 1.0, \mathrm{>=}\left(sndens_{wet}, par.snow_{dens\,max}\right)\right), 1.0, 0.0\right) $$

Indicator for WRF’s second density cap, applied only on the retention branch. UNTESTED: no reference step reaches it, because 0.40 is far above any density these four columns produce.

$$ dsndens = \mathrm{ifelse}\left(wet_{capped} = 1.0, par.snow_{dens\,max} - sndens_{in}, ddsx + dw \cdot \left(1.0 - dsx\right)\right) $$

The total density increment over the sub-step, ddsx + dw (1 - dsx), formed without ever differencing two densities. On the dry branch this is the compaction alone.

$$ sndens_{out} = sndens_{in} + dsndens $$

Density SNOWPACK returns, sndens_in + dsndens.

$$ snowh_{out} = \frac{esd}{sndens_{out}} $$

Depth SNOWPACK returns, esd/sndens_out. It is a PURE DIAGNOSTIC of the water equivalent and the density, with no step dependence of its own, and it uses the UNFLOORED esd. SNOWPACK’s SNOWH argument is declared INTENT(INOUT) but its incoming value is never read: SNOWHC = SNOWH*100 at the top of the routine is overwritten by SNOWHC = ESDC/SNDENS at the bottom before anything uses it (FORTRAN_BUGS N100), so the depth entering the routine is not an input of this component.

$$ dsndens_{dt} = \frac{dsndens}{dtsec} $$

dsndens/dtsec – the density tendency a consumer integrating the pack must add. This is the operator-split form: SNOWPACK is an in-place sub-step, not a rate, so the way it joins an ODE document is as increment over sub-step, exactly as WSM6’s stages do. Its dtsec -> 0 limit is analytic and clean, sndens_in snow_c1 exp(comp_exponent_0) esdcx/(2*3600) + wet snow_dwrate (1 - sndens_in)/86400 with comp_exponent_0 the exponent at dtsec-free bfac; the finite-step value exceeds it by xarg/3 + O(xarg^2), which is 2.2e-6 relative at WRF’s 60 s and 3.7e-14 at 1e-6 s.

$$ pack_{present} = \mathrm{ifelse}\left(\mathrm{>}\left(esd, 0.0\right), 1.0, 0.0\right) $$

Indicator for the caller’s guard: SNOPAC calls SNOWPACK only where esd > 0, and where it does not it sets the depth, the density and the cover to zero and the snow conductivity to one instead. Reported here so a consumer knows when this component applies; it is asserted against the reference’s own record of whether SNOWPACK ran.

Raw .esm

Raw .esm JSON (this component)
{
  "subsystems": {
    "par": {
      "ref": "./noah_parameters.esm"
    }
  },
  "variables": {
    "esd": {
      "type": "parameter",
      "units": "m",
      "default": 0.06,
      "description": "Water equivalent of the pack that SNOWPACK is handed, WRF's ESD -- the SWE AFTER SNOPAC has taken this step's sublimation and melt out of it, not the pack the step started with. Consumer-supplied, because it is the snow-energy component's output that sets it."
    },
    "sndens_in": {
      "type": "parameter",
      "units": "1",
      "default": 0.2,
      "description": "Snow density entering the compaction, as a fraction of the density of water (WRF's SNDENS; its own comment reads 'G/CM3=DIMENSIONLESS FRACTION OF H2O DENSITY', and the two are numerically equal). This is NOT the density implied by the dumped snow state: SFLX resets it to SNEQV/SNOWH and then, on a snowing step, SNOW_NEW moves it again before SNOPAC is reached, so it is supplied by the consumer and is published from inside SNOWPACK for the tests."
    },
    "tsnow": {
      "type": "parameter",
      "units": "K",
      "default": 265.0,
      "description": "Snow surface temperature, WRF's TSNOW dummy, which SNOPAC fills with the skin temperature T1 its own energy balance has just produced. Its position relative to the freezing point selects the melt-water retention branch."
    },
    "tsoil": {
      "type": "parameter",
      "units": "K",
      "default": 268.0,
      "description": "Soil temperature at the top of the soil column, WRF's TSOIL dummy, which SNOPAC fills with YY = stc[1] - 0.5 ssoil zsoil[1]/df1 -- the linearised surface value the soil-heat closure forms, not stc[1] itself."
    },
    "dtsec": {
      "type": "parameter",
      "units": "s",
      "default": 60.0,
      "description": "The sub-step the compaction is applied over, WRF's DTSEC. A PARAMETER OF THE SCHEME, not the integrator's step: Koren's solution is a closed form in the step length, so the increment it returns is O(dtsec) and there is no dt-free rate hiding inside it. Prescribe it in the mounting document and couple it in, exactly as WSM6's dtcld is prescribed; a parent test cannot set a mounted component's parameter."
    },
    "tsnowc": {
      "type": "unknown",
      "units": "K",
      "description": "Snow surface temperature above freezing, WRF's TSNOWC = TSNOW - 273.15."
    },
    "tsoilc": {
      "type": "unknown",
      "units": "K",
      "description": "Soil surface temperature above freezing, WRF's TSOILC."
    },
    "tavgc": {
      "type": "unknown",
      "units": "K",
      "description": "Mean temperature of the pack above freezing, the arithmetic mean of the snow surface and the soil below it. It is the only temperature the compaction sees: warm snow compacts fast."
    },
    "esdc": {
      "type": "unknown",
      "units": "cm",
      "description": "Water equivalent in SNOWPACK's own simulation units, centimetres."
    },
    "esdcx": {
      "type": "unknown",
      "units": "cm",
      "description": "Water equivalent floored at par.snow_esdc_min, WRF's ESDCX. It bounds the overburden a vanishing pack is allowed to feel. It is used ONLY in the compaction argument; the depth diagnostic at the end uses the unfloored esdc, and that distinction matters for the nine reference steps whose pack is below the floor."
    },
    "dthr": {
      "type": "unknown",
      "units": "h",
      "description": "The sub-step in hours, the unit SNOWPACK's rate constants are expressed in."
    },
    "comp_exponent": {
      "type": "unknown",
      "units": "1",
      "description": "Anderson's compaction exponent, snow_ctemp tavgc - snow_c2 sndens_in: compaction accelerates with temperature and is suppressed by the density already reached. Declared dimensionless in its own right so that the exponential is handed a quantity of scale one."
    },
    "bfac": {
      "type": "unknown",
      "units": "1/cm",
      "description": "The compaction rate per centimetre of overlying water equivalent, over the whole sub-step: dthr snow_c1 exp(comp_exponent). It is proportional to dtsec, which is where all of the step dependence enters."
    },
    "xarg": {
      "type": "unknown",
      "units": "1",
      "description": "bfac esdcx, the dimensionless compaction of the pack's BASE over the sub-step. It is proportional to dtsec: 6.7e-6 at WRF's 60 s on a 6 cm pack at -12 C, and 1.1e-13 at 1e-6 s."
    },
    "pexp_m1": {
      "type": "unknown",
      "units": "1",
      "description": "The mean compaction over the DEPTH of the pack, minus one. Anderson's equation makes the density at depth w grow as exp(bfac w), so the pack-mean factor is the depth average of that over 0 <= w <= esdcx, i.e. (exp(xarg) - 1)/xarg; WRF replaces it by its Taylor series to IPOL = 4 terms because the closed form is 0/0 at xarg = 0, and this is that series MINUS ONE, xarg/2 + xarg^2/6 + xarg^3/24 + xarg^4/120, written in WRF's own Horner nesting. It is the whole of the compaction signal and it is the quantity the tests are pinned to: (exp(x)-1)/x itself differs from 1 by O(x), so at any small step the stored pexp is 1 + O(1e-13) and carries no information at all."
    },
    "pexp": {
      "type": "unknown",
      "units": "1",
      "description": "WRF's PEXP, 1 + pexp_m1."
    },
    "dsx_raw": {
      "type": "unknown",
      "units": "1",
      "description": "Compacted density before the density limits, sndens_in pexp."
    },
    "dsx_clamp": {
      "type": "unknown",
      "units": "1",
      "description": "Which density limit fired on dsx_raw: 0 neither, 1 capped at snow_dens_max, 2 floored at snow_dens_min. WRF applies the two tests in sequence, cap first, and because the cap is above the floor the second test can only fire on a value that the first left alone, so the two are mutually exclusive. UNTESTED, not merely unused: neither limit fires anywhere in the cold-season reference set, all 67 of whose compaction calls report 0."
    },
    "dsx": {
      "type": "unknown",
      "units": "1",
      "description": "Compacted density after the limits, WRF's DSX."
    },
    "ddsx": {
      "type": "unknown",
      "units": "1",
      "description": "The compaction increment, dsx - sndens_in, written CANCELLATION-FREE as sndens_in pexp_m1 where no limit fired. That is an exact algebraic rearrangement, not an approximation, and it is the reason this component can be referenced at any step size: the increment is 7.5e-17 of the density at a 1e-6 s step, so forming it as a difference would return pure round-off. Where a limit fires the increment is a difference against a constant and is formed directly."
    },
    "wet": {
      "type": "unknown",
      "units": "1",
      "description": "Indicator, 1 where the snow surface has reached freezing and SNOWPACK runs its melt-water retention branch. 29 of the 67 compaction calls in the cold-season reference set; the other 38 are dry."
    },
    "dw": {
      "type": "unknown",
      "units": "1",
      "description": "Fraction of the pack's pore space filled by retained melt water over the sub-step, snow_dwrate dthr/24: 13 % per day. Zero on the dry branch. Like bfac it is proportional to dtsec. The UA_PHYS alternative, DW = min(DW, 0.13 snomltc/(esdcx + 0.13 snomltc)), is NOT transcribed -- UA_PHYS is off in every reference run."
    },
    "sndens_wet": {
      "type": "unknown",
      "units": "1",
      "description": "Density after retention and before its own cap, WRF's dsx (1 - dw) + dw: a relaxation of the density towards 1 at rate snow_dwrate, i.e. towards pure water, which is what saturating the pore space with melt water would give."
    },
    "wet_capped": {
      "type": "unknown",
      "units": "1",
      "description": "Indicator for WRF's second density cap, applied only on the retention branch. UNTESTED: no reference step reaches it, because 0.40 is far above any density these four columns produce."
    },
    "dsndens": {
      "type": "unknown",
      "units": "1",
      "description": "The total density increment over the sub-step, ddsx + dw (1 - dsx), formed without ever differencing two densities. On the dry branch this is the compaction alone."
    },
    "sndens_out": {
      "type": "unknown",
      "units": "1",
      "description": "Density SNOWPACK returns, sndens_in + dsndens."
    },
    "snowh_out": {
      "type": "unknown",
      "units": "m",
      "description": "Depth SNOWPACK returns, esd/sndens_out. It is a PURE DIAGNOSTIC of the water equivalent and the density, with no step dependence of its own, and it uses the UNFLOORED esd. SNOWPACK's SNOWH argument is declared INTENT(INOUT) but its incoming value is never read: SNOWHC = SNOWH*100 at the top of the routine is overwritten by SNOWHC = ESDC/SNDENS at the bottom before anything uses it (FORTRAN_BUGS N100), so the depth entering the routine is not an input of this component."
    },
    "dsndens_dt": {
      "type": "unknown",
      "units": "1/s",
      "description": "dsndens/dtsec -- the density tendency a consumer integrating the pack must add. This is the operator-split form: SNOWPACK is an in-place sub-step, not a rate, so the way it joins an ODE document is as increment over sub-step, exactly as WSM6's stages do. Its dtsec -> 0 limit is analytic and clean, sndens_in snow_c1 exp(comp_exponent_0) esdcx/(2*3600) + wet snow_dwrate (1 - sndens_in)/86400 with comp_exponent_0 the exponent at dtsec-free bfac; the finite-step value exceeds it by xarg/3 + O(xarg^2), which is 2.2e-6 relative at WRF's 60 s and 3.7e-14 at 1e-6 s."
    },
    "pack_present": {
      "type": "unknown",
      "units": "1",
      "description": "Indicator for the caller's guard: SNOPAC calls SNOWPACK only where esd > 0, and where it does not it sets the depth, the density and the cover to zero and the snow conductivity to one instead. Reported here so a consumer knows when this component applies; it is asserted against the reference's own record of whether SNOWPACK ran."
    }
  },
  "equations": [
    {
      "lhs": "tsnowc",
      "rhs": {
        "op": "-",
        "args": [
          "tsnow",
          "par.t0"
        ]
      }
    },
    {
      "lhs": "tsoilc",
      "rhs": {
        "op": "-",
        "args": [
          "tsoil",
          "par.t0"
        ]
      }
    },
    {
      "lhs": "tavgc",
      "rhs": {
        "op": "*",
        "args": [
          0.5,
          {
            "op": "+",
            "args": [
              "tsnowc",
              "tsoilc"
            ]
          }
        ]
      }
    },
    {
      "lhs": "esdc",
      "rhs": {
        "op": "*",
        "args": [
          100.0,
          "esd"
        ]
      }
    },
    {
      "lhs": "esdcx",
      "rhs": {
        "op": "max",
        "args": [
          "esdc",
          "par.snow_esdc_min"
        ]
      }
    },
    {
      "lhs": "dthr",
      "rhs": {
        "op": "/",
        "args": [
          "dtsec",
          3600.0
        ]
      }
    },
    {
      "lhs": "comp_exponent",
      "rhs": {
        "op": "-",
        "args": [
          {
            "op": "*",
            "args": [
              "par.snow_ctemp",
              "tavgc"
            ]
          },
          {
            "op": "*",
            "args": [
              "par.snow_c2",
              "sndens_in"
            ]
          }
        ]
      }
    },
    {
      "lhs": "bfac",
      "rhs": {
        "op": "*",
        "args": [
          {
            "op": "*",
            "args": [
              "dthr",
              "par.snow_c1"
            ]
          },
          {
            "op": "exp",
            "args": [
              "comp_exponent"
            ]
          }
        ]
      }
    },
    {
      "lhs": "xarg",
      "rhs": {
        "op": "*",
        "args": [
          "bfac",
          "esdcx"
        ]
      }
    },
    {
      "lhs": "pexp_m1",
      "rhs": {
        "op": "/",
        "args": [
          {
            "op": "*",
            "args": [
              {
                "op": "+",
                "args": [
                  1.0,
                  {
                    "op": "/",
                    "args": [
                      {
                        "op": "*",
                        "args": [
                          {
                            "op": "+",
                            "args": [
                              1.0,
                              {
                                "op": "/",
                                "args": [
                                  {
                                    "op": "*",
                                    "args": [
                                      {
                                        "op": "+",
                                        "args": [
                                          1.0,
                                          {
                                            "op": "/",
                                            "args": [
                                              {
                                                "op": "*",
                                                "args": [
                                                  {
                                                    "op": "+",
                                                    "args": [
                                                      1.0,
                                                      0.0
                                                    ]
                                                  },
                                                  "xarg"
                                                ]
                                              },
                                              5.0
                                            ]
                                          }
                                        ]
                                      },
                                      "xarg"
                                    ]
                                  },
                                  4.0
                                ]
                              }
                            ]
                          },
                          "xarg"
                        ]
                      },
                      3.0
                    ]
                  }
                ]
              },
              "xarg"
            ]
          },
          2.0
        ]
      }
    },
    {
      "lhs": "pexp",
      "rhs": {
        "op": "+",
        "args": [
          1.0,
          "pexp_m1"
        ]
      }
    },
    {
      "lhs": "dsx_raw",
      "rhs": {
        "op": "*",
        "args": [
          "sndens_in",
          "pexp"
        ]
      }
    },
    {
      "lhs": "dsx_clamp",
      "rhs": {
        "op": "ifelse",
        "args": [
          {
            "op": ">",
            "args": [
              "dsx_raw",
              "par.snow_dens_max"
            ]
          },
          1.0,
          {
            "op": "ifelse",
            "args": [
              {
                "op": "<",
                "args": [
                  "dsx_raw",
                  "par.snow_dens_min"
                ]
              },
              2.0,
              0.0
            ]
          }
        ]
      }
    },
    {
      "lhs": "dsx",
      "rhs": {
        "op": "ifelse",
        "args": [
          {
            "op": ">",
            "args": [
              "dsx_raw",
              "par.snow_dens_max"
            ]
          },
          "par.snow_dens_max",
          {
            "op": "ifelse",
            "args": [
              {
                "op": "<",
                "args": [
                  "dsx_raw",
                  "par.snow_dens_min"
                ]
              },
              "par.snow_dens_min",
              "dsx_raw"
            ]
          }
        ]
      }
    },
    {
      "lhs": "ddsx",
      "rhs": {
        "op": "ifelse",
        "args": [
          {
            "op": "==",
            "args": [
              "dsx_clamp",
              0.0
            ]
          },
          {
            "op": "*",
            "args": [
              "sndens_in",
              "pexp_m1"
            ]
          },
          {
            "op": "-",
            "args": [
              "dsx",
              "sndens_in"
            ]
          }
        ]
      }
    },
    {
      "lhs": "wet",
      "rhs": {
        "op": "ifelse",
        "args": [
          {
            "op": ">=",
            "args": [
              "tsnowc",
              0.0
            ]
          },
          1.0,
          0.0
        ]
      }
    },
    {
      "lhs": "dw",
      "rhs": {
        "op": "ifelse",
        "args": [
          {
            "op": "==",
            "args": [
              "wet",
              1.0
            ]
          },
          {
            "op": "/",
            "args": [
              {
                "op": "*",
                "args": [
                  "par.snow_dwrate",
                  "dthr"
                ]
              },
              24.0
            ]
          },
          0.0
        ]
      }
    },
    {
      "lhs": "sndens_wet",
      "rhs": {
        "op": "ifelse",
        "args": [
          {
            "op": "==",
            "args": [
              "wet",
              1.0
            ]
          },
          {
            "op": "+",
            "args": [
              {
                "op": "*",
                "args": [
                  "dsx",
                  {
                    "op": "-",
                    "args": [
                      1.0,
                      "dw"
                    ]
                  }
                ]
              },
              "dw"
            ]
          },
          "dsx"
        ]
      }
    },
    {
      "lhs": "wet_capped",
      "rhs": {
        "op": "ifelse",
        "args": [
          {
            "op": "and",
            "args": [
              {
                "op": "==",
                "args": [
                  "wet",
                  1.0
                ]
              },
              {
                "op": ">=",
                "args": [
                  "sndens_wet",
                  "par.snow_dens_max"
                ]
              }
            ]
          },
          1.0,
          0.0
        ]
      }
    },
    {
      "lhs": "dsndens",
      "rhs": {
        "op": "ifelse",
        "args": [
          {
            "op": "==",
            "args": [
              "wet_capped",
              1.0
            ]
          },
          {
            "op": "-",
            "args": [
              "par.snow_dens_max",
              "sndens_in"
            ]
          },
          {
            "op": "+",
            "args": [
              "ddsx",
              {
                "op": "*",
                "args": [
                  "dw",
                  {
                    "op": "-",
                    "args": [
                      1.0,
                      "dsx"
                    ]
                  }
                ]
              }
            ]
          }
        ]
      }
    },
    {
      "lhs": "sndens_out",
      "rhs": {
        "op": "+",
        "args": [
          "sndens_in",
          "dsndens"
        ]
      }
    },
    {
      "lhs": "snowh_out",
      "rhs": {
        "op": "/",
        "args": [
          "esd",
          "sndens_out"
        ]
      }
    },
    {
      "lhs": "dsndens_dt",
      "rhs": {
        "op": "/",
        "args": [
          "dsndens",
          "dtsec"
        ]
      }
    },
    {
      "lhs": "pack_present",
      "rhs": {
        "op": "ifelse",
        "args": [
          {
            "op": ">",
            "args": [
              "esd",
              0.0
            ]
          },
          1.0,
          0.0
        ]
      }
    }
  ],
  "tests": [
    {
      "id": "cold_wrfdt_caseA_step26",
      "description": "Reference: WRF 4.8 em_scm_xy COLD-SEASON single-column run, data/eqwefic/noahcold_scm, dumped by ctessum-claude/WRF branch earthsciml-instrumented-noahcold and replayed through the real64 (-fdefault-real-8) kernels/noah_driver at ESM_DT = 60 s, WRF's OWN step; tolerances are the 1e-9 that a double-precision kernel build earns. WHY THE LARGEST STEP AND NOT THE SMALLEST. SNOWPACK is not a rate: Koren's solution is a closed form in the step length, so the density increment is proportional to dtsec and there is no dt -> 0 limit to approach by shrinking the step. Nothing here is formed as a difference either -- the increment is written as sndens_in pexp_m1 + dw (1 - dsx), an exact rearrangement in which nothing cancels, and its reference comes from a publisher that emits the Taylor series BEFORE WRF adds one to it (FORTRAN_BUGS N99) -- so there is no cancellation floor pushing the step DOWN either. What is left is that a smaller step puts LESS of the assertion's budget on the physics: at 60 s the compaction and retention move the density by 4.5e-9 to 9.4e-5 of itself, so a rel 1e-9 assertion on sndens_out tests them to between 1e-5 and 2e-1 of themselves, while at 1e-6 s the same assertion would test them to nothing at all -- the increment is 7.5e-17 of the density there. This is the fifth distinct direction the Noah reference step has had to move (surface closure 1e-6 s, tendencies 1e-2 s, runoff 1e-4 s, canopy drip 1e-2 s, compaction 60 s), and it is why the choice is argued per quantity rather than assumed. The companion regime at 1e-6 s pins the tendency in the dt -> 0 limit. Step 26 of case A, snow falling onto initially bare frozen ground. dry branch -- the snow surface is below freezing and only compaction acts; the pack is below the esdcx floor, so the compaction argument uses the floored 0.01 cm while the depth diagnostic keeps the true water equivalent; neither density limit fires. The density moves by 7.794e-08 of itself over the step.",
      "parameter_overrides": {
        "esd": 5.6982396419694386e-06,
        "sndens_in": 0.0787306049123627,
        "tsnow": 264.19925417549115,
        "tsoil": 264.1975341527298,
        "dtsec": 60.0
      },
      "time_span": {
        "start": 0.0,
        "end": 0.0
      },
      "assertions": [
        {
          "variable": "tavgc",
          "time": 0.0,
          "expected": -8.951605835889495,
          "tolerance": {
            "rel": 1e-09,
            "abs": 1e-12
          }
        },
        {
          "variable": "esdcx",
          "time": 0.0,
          "expected": 0.01,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "comp_exponent",
          "time": 0.0,
          "expected": -2.369471170030776,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "bfac",
          "time": 0.0,
          "expected": 1.5588362460237785e-05,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "xarg",
          "time": 0.0,
          "expected": 1.5588362460237786e-07,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "pexp_m1",
          "time": 0.0,
          "expected": 7.794181635113982e-08,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "pexp",
          "time": 0.0,
          "expected": 1.0000000779418163,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsx_raw",
          "time": 0.0,
          "expected": 0.07873061104876904,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsx",
          "time": 0.0,
          "expected": 0.07873061104876904,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsx_clamp",
          "time": 0.0,
          "expected": 0.0,
          "tolerance": {
            "abs": 0.0,
            "rel": 0.0
          }
        },
        {
          "variable": "wet",
          "time": 0.0,
          "expected": 0.0,
          "tolerance": {
            "abs": 0.0,
            "rel": 0.0
          }
        },
        {
          "variable": "dw",
          "time": 0.0,
          "expected": 0.0,
          "tolerance": {
            "rel": 1e-09,
            "abs": 1e-30
          }
        },
        {
          "variable": "sndens_wet",
          "time": 0.0,
          "expected": 0.07873061104876904,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "sndens_out",
          "time": 0.0,
          "expected": 0.07873061104876904,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "snowh_out",
          "time": 0.0,
          "expected": 7.237641834685762e-05,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsndens_dt",
          "time": 0.0,
          "expected": 1.0227343915489199e-10,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "pack_present",
          "time": 0.0,
          "expected": 1.0,
          "tolerance": {
            "abs": 0.0,
            "rel": 0.0
          }
        }
      ]
    },
    {
      "id": "cold_wrfdt_caseA_step51",
      "description": "Reference: WRF 4.8 em_scm_xy COLD-SEASON single-column run, data/eqwefic/noahcold_scm, dumped by ctessum-claude/WRF branch earthsciml-instrumented-noahcold and replayed through the real64 (-fdefault-real-8) kernels/noah_driver at ESM_DT = 60 s, WRF's OWN step; tolerances are the 1e-9 that a double-precision kernel build earns. WHY THE LARGEST STEP AND NOT THE SMALLEST. SNOWPACK is not a rate: Koren's solution is a closed form in the step length, so the density increment is proportional to dtsec and there is no dt -> 0 limit to approach by shrinking the step. Nothing here is formed as a difference either -- the increment is written as sndens_in pexp_m1 + dw (1 - dsx), an exact rearrangement in which nothing cancels, and its reference comes from a publisher that emits the Taylor series BEFORE WRF adds one to it (FORTRAN_BUGS N99) -- so there is no cancellation floor pushing the step DOWN either. What is left is that a smaller step puts LESS of the assertion's budget on the physics: at 60 s the compaction and retention move the density by 4.5e-9 to 9.4e-5 of itself, so a rel 1e-9 assertion on sndens_out tests them to between 1e-5 and 2e-1 of themselves, while at 1e-6 s the same assertion would test them to nothing at all -- the increment is 7.5e-17 of the density there. This is the fifth distinct direction the Noah reference step has had to move (surface closure 1e-6 s, tendencies 1e-2 s, runoff 1e-4 s, canopy drip 1e-2 s, compaction 60 s), and it is why the choice is argued per quantity rather than assumed. The companion regime at 1e-6 s pins the tendency in the dt -> 0 limit. Step 51 of case A, snow falling onto initially bare frozen ground. dry branch -- the snow surface is below freezing and only compaction acts; the pack is below the esdcx floor, so the compaction argument uses the floored 0.01 cm while the depth diagnostic keeps the true water equivalent; neither density limit fires. The density moves by 7.849e-08 of itself over the step.",
      "parameter_overrides": {
        "esd": 8.994054993710539e-05,
        "sndens_in": 0.07815370609839009,
        "tsnow": 264.1468693334496,
        "tsoil": 264.122158310369,
        "dtsec": 60.0
      },
      "time_span": {
        "start": 0.0,
        "end": 0.0
      },
      "assertions": [
        {
          "variable": "tavgc",
          "time": 0.0,
          "expected": -9.015486178090669,
          "tolerance": {
            "rel": 1e-09,
            "abs": 1e-12
          }
        },
        {
          "variable": "esdcx",
          "time": 0.0,
          "expected": 0.01,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "comp_exponent",
          "time": 0.0,
          "expected": -2.3624667223134455,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "bfac",
          "time": 0.0,
          "expected": 1.5697933624852283e-05,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "xarg",
          "time": 0.0,
          "expected": 1.5697933624852282e-07,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "pexp_m1",
          "time": 0.0,
          "expected": 7.84896722313469e-08,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "pexp",
          "time": 0.0,
          "expected": 1.0000000784896723,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsx_raw",
          "time": 0.0,
          "expected": 0.07815371223264887,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsx",
          "time": 0.0,
          "expected": 0.07815371223264887,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsx_clamp",
          "time": 0.0,
          "expected": 0.0,
          "tolerance": {
            "abs": 0.0,
            "rel": 0.0
          }
        },
        {
          "variable": "wet",
          "time": 0.0,
          "expected": 0.0,
          "tolerance": {
            "abs": 0.0,
            "rel": 0.0
          }
        },
        {
          "variable": "dw",
          "time": 0.0,
          "expected": 0.0,
          "tolerance": {
            "rel": 1e-09,
            "abs": 1e-30
          }
        },
        {
          "variable": "sndens_wet",
          "time": 0.0,
          "expected": 0.07815371223264887,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "sndens_out",
          "time": 0.0,
          "expected": 0.07815371223264887,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "snowh_out",
          "time": 0.0,
          "expected": 0.0011508160952018417,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsndens_dt",
          "time": 0.0,
          "expected": 1.0223764625546094e-10,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "pack_present",
          "time": 0.0,
          "expected": 1.0,
          "tolerance": {
            "abs": 0.0,
            "rel": 0.0
          }
        }
      ]
    },
    {
      "id": "cold_wrfdt_caseA_step76",
      "description": "Reference: WRF 4.8 em_scm_xy COLD-SEASON single-column run, data/eqwefic/noahcold_scm, dumped by ctessum-claude/WRF branch earthsciml-instrumented-noahcold and replayed through the real64 (-fdefault-real-8) kernels/noah_driver at ESM_DT = 60 s, WRF's OWN step; tolerances are the 1e-9 that a double-precision kernel build earns. WHY THE LARGEST STEP AND NOT THE SMALLEST. SNOWPACK is not a rate: Koren's solution is a closed form in the step length, so the density increment is proportional to dtsec and there is no dt -> 0 limit to approach by shrinking the step. Nothing here is formed as a difference either -- the increment is written as sndens_in pexp_m1 + dw (1 - dsx), an exact rearrangement in which nothing cancels, and its reference comes from a publisher that emits the Taylor series BEFORE WRF adds one to it (FORTRAN_BUGS N99) -- so there is no cancellation floor pushing the step DOWN either. What is left is that a smaller step puts LESS of the assertion's budget on the physics: at 60 s the compaction and retention move the density by 4.5e-9 to 9.4e-5 of itself, so a rel 1e-9 assertion on sndens_out tests them to between 1e-5 and 2e-1 of themselves, while at 1e-6 s the same assertion would test them to nothing at all -- the increment is 7.5e-17 of the density there. This is the fifth distinct direction the Noah reference step has had to move (surface closure 1e-6 s, tendencies 1e-2 s, runoff 1e-4 s, canopy drip 1e-2 s, compaction 60 s), and it is why the choice is argued per quantity rather than assumed. The companion regime at 1e-6 s pins the tendency in the dt -> 0 limit. Step 76 of case A, snow falling onto initially bare frozen ground. dry branch -- the snow surface is below freezing and only compaction acts; neither density limit fires. The density moves by 1.667e-07 of itself over the step.",
      "parameter_overrides": {
        "esd": 0.00021237832458379983,
        "sndens_in": 0.0778412719019753,
        "tsnow": 264.07449984275536,
        "tsoil": 264.0267531534461,
        "dtsec": 60.0
      },
      "time_span": {
        "start": 0.0,
        "end": 0.0
      },
      "assertions": [
        {
          "variable": "tavgc",
          "time": 0.0,
          "expected": -9.099373501899237,
          "tolerance": {
            "rel": 1e-09,
            "abs": 1e-12
          }
        },
        {
          "variable": "esdcx",
          "time": 0.0,
          "expected": 0.02123783245837998,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "comp_exponent",
          "time": 0.0,
          "expected": -2.3626165900934204,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "bfac",
          "time": 0.0,
          "expected": 1.569558118667148e-05,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "xarg",
          "time": 0.0,
          "expected": 3.3334012357942974e-07,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "pexp_m1",
          "time": 0.0,
          "expected": 1.666700803089894e-07,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "pexp",
          "time": 0.0,
          "expected": 1.0000001666700804,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsx_raw",
          "time": 0.0,
          "expected": 0.07784128487578636,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsx",
          "time": 0.0,
          "expected": 0.07784128487578636,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsx_clamp",
          "time": 0.0,
          "expected": 0.0,
          "tolerance": {
            "abs": 0.0,
            "rel": 0.0
          }
        },
        {
          "variable": "wet",
          "time": 0.0,
          "expected": 0.0,
          "tolerance": {
            "abs": 0.0,
            "rel": 0.0
          }
        },
        {
          "variable": "dw",
          "time": 0.0,
          "expected": 0.0,
          "tolerance": {
            "rel": 1e-09,
            "abs": 1e-30
          }
        },
        {
          "variable": "sndens_wet",
          "time": 0.0,
          "expected": 0.07784128487578636,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "sndens_out",
          "time": 0.0,
          "expected": 0.07784128487578636,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "snowh_out",
          "time": 0.0,
          "expected": 0.0027283507064753388,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsndens_dt",
          "time": 0.0,
          "expected": 2.1623018398760174e-10,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "pack_present",
          "time": 0.0,
          "expected": 1.0,
          "tolerance": {
            "abs": 0.0,
            "rel": 0.0
          }
        }
      ]
    },
    {
      "id": "cold_wrfdt_caseA_step101",
      "description": "Reference: WRF 4.8 em_scm_xy COLD-SEASON single-column run, data/eqwefic/noahcold_scm, dumped by ctessum-claude/WRF branch earthsciml-instrumented-noahcold and replayed through the real64 (-fdefault-real-8) kernels/noah_driver at ESM_DT = 60 s, WRF's OWN step; tolerances are the 1e-9 that a double-precision kernel build earns. WHY THE LARGEST STEP AND NOT THE SMALLEST. SNOWPACK is not a rate: Koren's solution is a closed form in the step length, so the density increment is proportional to dtsec and there is no dt -> 0 limit to approach by shrinking the step. Nothing here is formed as a difference either -- the increment is written as sndens_in pexp_m1 + dw (1 - dsx), an exact rearrangement in which nothing cancels, and its reference comes from a publisher that emits the Taylor series BEFORE WRF adds one to it (FORTRAN_BUGS N99) -- so there is no cancellation floor pushing the step DOWN either. What is left is that a smaller step puts LESS of the assertion's budget on the physics: at 60 s the compaction and retention move the density by 4.5e-9 to 9.4e-5 of itself, so a rel 1e-9 assertion on sndens_out tests them to between 1e-5 and 2e-1 of themselves, while at 1e-6 s the same assertion would test them to nothing at all -- the increment is 7.5e-17 of the density there. This is the fifth distinct direction the Noah reference step has had to move (surface closure 1e-6 s, tendencies 1e-2 s, runoff 1e-4 s, canopy drip 1e-2 s, compaction 60 s), and it is why the choice is argued per quantity rather than assumed. The companion regime at 1e-6 s pins the tendency in the dt -> 0 limit. Step 101 of case A, snow falling onto initially bare frozen ground. dry branch -- the snow surface is below freezing and only compaction acts; neither density limit fires. The density moves by 2.421e-07 of itself over the step.",
      "parameter_overrides": {
        "esd": 0.00030909263193183023,
        "sndens_in": 0.07762557168115519,
        "tsnow": 263.9974345916095,
        "tsoil": 263.9417684365665,
        "dtsec": 60.0
      },
      "time_span": {
        "start": 0.0,
        "end": 0.0
      },
      "assertions": [
        {
          "variable": "tavgc",
          "time": 0.0,
          "expected": -9.18039848591198,
          "tolerance": {
            "rel": 1e-09,
            "abs": 1e-12
          }
        },
        {
          "variable": "esdcx",
          "time": 0.0,
          "expected": 0.030909263193183024,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "comp_exponent",
          "time": 0.0,
          "expected": -2.3645688841772174,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "bfac",
          "time": 0.0,
          "expected": 1.5664968688401788e-05,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "xarg",
          "time": 0.0,
          "expected": 4.84192640102782e-07,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "pexp_m1",
          "time": 0.0,
          "expected": 2.420963591251478e-07,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "pexp",
          "time": 0.0,
          "expected": 1.0000002420963592,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsx_raw",
          "time": 0.0,
          "expected": 0.07762559047402348,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsx",
          "time": 0.0,
          "expected": 0.07762559047402348,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsx_clamp",
          "time": 0.0,
          "expected": 0.0,
          "tolerance": {
            "abs": 0.0,
            "rel": 0.0
          }
        },
        {
          "variable": "wet",
          "time": 0.0,
          "expected": 0.0,
          "tolerance": {
            "abs": 0.0,
            "rel": 0.0
          }
        },
        {
          "variable": "dw",
          "time": 0.0,
          "expected": 0.0,
          "tolerance": {
            "rel": 1e-09,
            "abs": 1e-30
          }
        },
        {
          "variable": "sndens_wet",
          "time": 0.0,
          "expected": 0.07762559047402348,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "sndens_out",
          "time": 0.0,
          "expected": 0.07762559047402348,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "snowh_out",
          "time": 0.0,
          "expected": 0.003981839365656929,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsndens_dt",
          "time": 0.0,
          "expected": 3.1321447131693084e-10,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "pack_present",
          "time": 0.0,
          "expected": 1.0,
          "tolerance": {
            "abs": 0.0,
            "rel": 0.0
          }
        }
      ]
    },
    {
      "id": "cold_wrfdt_caseA_step328",
      "description": "Reference: WRF 4.8 em_scm_xy COLD-SEASON single-column run, data/eqwefic/noahcold_scm, dumped by ctessum-claude/WRF branch earthsciml-instrumented-noahcold and replayed through the real64 (-fdefault-real-8) kernels/noah_driver at ESM_DT = 60 s, WRF's OWN step; tolerances are the 1e-9 that a double-precision kernel build earns. WHY THE LARGEST STEP AND NOT THE SMALLEST. SNOWPACK is not a rate: Koren's solution is a closed form in the step length, so the density increment is proportional to dtsec and there is no dt -> 0 limit to approach by shrinking the step. Nothing here is formed as a difference either -- the increment is written as sndens_in pexp_m1 + dw (1 - dsx), an exact rearrangement in which nothing cancels, and its reference comes from a publisher that emits the Taylor series BEFORE WRF adds one to it (FORTRAN_BUGS N99) -- so there is no cancellation floor pushing the step DOWN either. What is left is that a smaller step puts LESS of the assertion's budget on the physics: at 60 s the compaction and retention move the density by 4.5e-9 to 9.4e-5 of itself, so a rel 1e-9 assertion on sndens_out tests them to between 1e-5 and 2e-1 of themselves, while at 1e-6 s the same assertion would test them to nothing at all -- the increment is 7.5e-17 of the density there. This is the fifth distinct direction the Noah reference step has had to move (surface closure 1e-6 s, tendencies 1e-2 s, runoff 1e-4 s, canopy drip 1e-2 s, compaction 60 s), and it is why the choice is argued per quantity rather than assumed. The companion regime at 1e-6 s pins the tendency in the dt -> 0 limit. Step 328 of case A, snow falling onto initially bare frozen ground. dry branch -- the snow surface is below freezing and only compaction acts; neither density limit fires. The density moves by 7.805e-07 of itself over the step.",
      "parameter_overrides": {
        "esd": 0.0005983630133302496,
        "sndens_in": 0.08080965052100912,
        "tsnow": 271.63441528585224,
        "tsoil": 270.72671825768185,
        "dtsec": 60.0
      },
      "time_span": {
        "start": 0.0,
        "end": 0.0
      },
      "assertions": [
        {
          "variable": "tavgc",
          "time": 0.0,
          "expected": -1.9694332282329299,
          "tolerance": {
            "rel": 1e-09,
            "abs": 1e-12
          }
        },
        {
          "variable": "esdcx",
          "time": 0.0,
          "expected": 0.05983630133302496,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "comp_exponent",
          "time": 0.0,
          "expected": -1.854557319199826,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "bfac",
          "time": 0.0,
          "expected": 2.6087036120136903e-05,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "xarg",
          "time": 0.0,
          "expected": 1.5609517541700182e-06,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "pexp_m1",
          "time": 0.0,
          "expected": 7.804762831802306e-07,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "pexp",
          "time": 0.0,
          "expected": 1.0000007804762832,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsx_raw",
          "time": 0.0,
          "expected": 0.0808097135910248,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsx",
          "time": 0.0,
          "expected": 0.0808097135910248,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsx_clamp",
          "time": 0.0,
          "expected": 0.0,
          "tolerance": {
            "abs": 0.0,
            "rel": 0.0
          }
        },
        {
          "variable": "wet",
          "time": 0.0,
          "expected": 0.0,
          "tolerance": {
            "abs": 0.0,
            "rel": 0.0
          }
        },
        {
          "variable": "dw",
          "time": 0.0,
          "expected": 0.0,
          "tolerance": {
            "rel": 1e-09,
            "abs": 1e-30
          }
        },
        {
          "variable": "sndens_wet",
          "time": 0.0,
          "expected": 0.0808097135910248,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "sndens_out",
          "time": 0.0,
          "expected": 0.0808097135910248,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "snowh_out",
          "time": 0.0,
          "expected": 0.007404592675065579,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsndens_dt",
          "time": 0.0,
          "expected": 1.0511669280621766e-09,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "pack_present",
          "time": 0.0,
          "expected": 1.0,
          "tolerance": {
            "abs": 0.0,
            "rel": 0.0
          }
        }
      ]
    },
    {
      "id": "cold_wrfdt_caseA_step364",
      "description": "Reference: WRF 4.8 em_scm_xy COLD-SEASON single-column run, data/eqwefic/noahcold_scm, dumped by ctessum-claude/WRF branch earthsciml-instrumented-noahcold and replayed through the real64 (-fdefault-real-8) kernels/noah_driver at ESM_DT = 60 s, WRF's OWN step; tolerances are the 1e-9 that a double-precision kernel build earns. WHY THE LARGEST STEP AND NOT THE SMALLEST. SNOWPACK is not a rate: Koren's solution is a closed form in the step length, so the density increment is proportional to dtsec and there is no dt -> 0 limit to approach by shrinking the step. Nothing here is formed as a difference either -- the increment is written as sndens_in pexp_m1 + dw (1 - dsx), an exact rearrangement in which nothing cancels, and its reference comes from a publisher that emits the Taylor series BEFORE WRF adds one to it (FORTRAN_BUGS N99) -- so there is no cancellation floor pushing the step DOWN either. What is left is that a smaller step puts LESS of the assertion's budget on the physics: at 60 s the compaction and retention move the density by 4.5e-9 to 9.4e-5 of itself, so a rel 1e-9 assertion on sndens_out tests them to between 1e-5 and 2e-1 of themselves, while at 1e-6 s the same assertion would test them to nothing at all -- the increment is 7.5e-17 of the density there. This is the fifth distinct direction the Noah reference step has had to move (surface closure 1e-6 s, tendencies 1e-2 s, runoff 1e-4 s, canopy drip 1e-2 s, compaction 60 s), and it is why the choice is argued per quantity rather than assumed. The companion regime at 1e-6 s pins the tendency in the dt -> 0 limit. Step 364 of case A, snow falling onto initially bare frozen ground. dry branch -- the snow surface is below freezing and only compaction acts; neither density limit fires. The density moves by 9.067e-07 of itself over the step.",
      "parameter_overrides": {
        "esd": 0.0006731587530801318,
        "sndens_in": 0.08223270564547115,
        "tsnow": 272.4829484106161,
        "tsoil": 271.4279273700033,
        "dtsec": 60.0
      },
      "time_span": {
        "start": 0.0,
        "end": 0.0
      },
      "assertions": [
        {
          "variable": "tavgc",
          "time": 0.0,
          "expected": -1.1945621096902812,
          "tolerance": {
            "rel": 1e-09,
            "abs": 1e-12
          }
        },
        {
          "variable": "esdcx",
          "time": 0.0,
          "expected": 0.06731587530801318,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "comp_exponent",
          "time": 0.0,
          "expected": -1.8224517873301167,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "bfac",
          "time": 0.0,
          "expected": 2.6938164140396975e-05,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "xarg",
          "time": 0.0,
          "expected": 1.813366098301755e-06,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "pexp_m1",
          "time": 0.0,
          "expected": 9.066835972005604e-07,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "pexp",
          "time": 0.0,
          "expected": 1.000000906683597,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsx_raw",
          "time": 0.0,
          "expected": 0.0822327802045165,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsx",
          "time": 0.0,
          "expected": 0.0822327802045165,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsx_clamp",
          "time": 0.0,
          "expected": 0.0,
          "tolerance": {
            "abs": 0.0,
            "rel": 0.0
          }
        },
        {
          "variable": "wet",
          "time": 0.0,
          "expected": 0.0,
          "tolerance": {
            "abs": 0.0,
            "rel": 0.0
          }
        },
        {
          "variable": "dw",
          "time": 0.0,
          "expected": 0.0,
          "tolerance": {
            "rel": 1e-09,
            "abs": 1e-30
          }
        },
        {
          "variable": "sndens_wet",
          "time": 0.0,
          "expected": 0.0822327802045165,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "sndens_out",
          "time": 0.0,
          "expected": 0.0822327802045165,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "snowh_out",
          "time": 0.0,
          "expected": 0.008186014766933048,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsndens_dt",
          "time": 0.0,
          "expected": 1.2426507560361769e-09,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "pack_present",
          "time": 0.0,
          "expected": 1.0,
          "tolerance": {
            "abs": 0.0,
            "rel": 0.0
          }
        }
      ]
    },
    {
      "id": "cold_wrfdt_caseA_step397",
      "description": "Reference: WRF 4.8 em_scm_xy COLD-SEASON single-column run, data/eqwefic/noahcold_scm, dumped by ctessum-claude/WRF branch earthsciml-instrumented-noahcold and replayed through the real64 (-fdefault-real-8) kernels/noah_driver at ESM_DT = 60 s, WRF's OWN step; tolerances are the 1e-9 that a double-precision kernel build earns. WHY THE LARGEST STEP AND NOT THE SMALLEST. SNOWPACK is not a rate: Koren's solution is a closed form in the step length, so the density increment is proportional to dtsec and there is no dt -> 0 limit to approach by shrinking the step. Nothing here is formed as a difference either -- the increment is written as sndens_in pexp_m1 + dw (1 - dsx), an exact rearrangement in which nothing cancels, and its reference comes from a publisher that emits the Taylor series BEFORE WRF adds one to it (FORTRAN_BUGS N99) -- so there is no cancellation floor pushing the step DOWN either. What is left is that a smaller step puts LESS of the assertion's budget on the physics: at 60 s the compaction and retention move the density by 4.5e-9 to 9.4e-5 of itself, so a rel 1e-9 assertion on sndens_out tests them to between 1e-5 and 2e-1 of themselves, while at 1e-6 s the same assertion would test them to nothing at all -- the increment is 7.5e-17 of the density there. This is the fifth distinct direction the Noah reference step has had to move (surface closure 1e-6 s, tendencies 1e-2 s, runoff 1e-4 s, canopy drip 1e-2 s, compaction 60 s), and it is why the choice is argued per quantity rather than assumed. The companion regime at 1e-6 s pins the tendency in the dt -> 0 limit. Step 397 of case A, snow falling onto initially bare frozen ground. dry branch -- the snow surface is below freezing and only compaction acts; neither density limit fires. The density moves by 1.003e-06 of itself over the step.",
      "parameter_overrides": {
        "esd": 0.0007555211961233569,
        "sndens_in": 0.0837528448684635,
        "tsnow": 272.7314702148211,
        "tsoil": 271.6048331357169,
        "dtsec": 60.0
      },
      "time_span": {
        "start": 0.0,
        "end": 0.0
      },
      "assertions": [
        {
          "variable": "tavgc",
          "time": 0.0,
          "expected": -0.9818483247309757,
          "tolerance": {
            "rel": 1e-09,
            "abs": 1e-12
          }
        },
        {
          "variable": "esdcx",
          "time": 0.0,
          "expected": 0.0755521196123357,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "comp_exponent",
          "time": 0.0,
          "expected": -1.8373576082162115,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "bfac",
          "time": 0.0,
          "expected": 2.65396064846187e-05,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "xarg",
          "time": 0.0,
          "expected": 2.005123523590232e-06,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "pexp_m1",
          "time": 0.0,
          "expected": 1.0025624318821758e-06,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "pexp",
          "time": 0.0,
          "expected": 1.000001002562432,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsx_raw",
          "time": 0.0,
          "expected": 0.08375292883591934,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsx",
          "time": 0.0,
          "expected": 0.08375292883591934,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsx_clamp",
          "time": 0.0,
          "expected": 0.0,
          "tolerance": {
            "abs": 0.0,
            "rel": 0.0
          }
        },
        {
          "variable": "wet",
          "time": 0.0,
          "expected": 0.0,
          "tolerance": {
            "abs": 0.0,
            "rel": 0.0
          }
        },
        {
          "variable": "dw",
          "time": 0.0,
          "expected": 0.0,
          "tolerance": {
            "rel": 1e-09,
            "abs": 1e-30
          }
        },
        {
          "variable": "sndens_wet",
          "time": 0.0,
          "expected": 0.08375292883591934,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "sndens_out",
          "time": 0.0,
          "expected": 0.08375292883591934,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "snowh_out",
          "time": 0.0,
          "expected": 0.009020833141292303,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsndens_dt",
          "time": 0.0,
          "expected": 1.3994575971396232e-09,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "pack_present",
          "time": 0.0,
          "expected": 1.0,
          "tolerance": {
            "abs": 0.0,
            "rel": 0.0
          }
        }
      ]
    },
    {
      "id": "cold_wrfdt_caseA_step519",
      "description": "Reference: WRF 4.8 em_scm_xy COLD-SEASON single-column run, data/eqwefic/noahcold_scm, dumped by ctessum-claude/WRF branch earthsciml-instrumented-noahcold and replayed through the real64 (-fdefault-real-8) kernels/noah_driver at ESM_DT = 60 s, WRF's OWN step; tolerances are the 1e-9 that a double-precision kernel build earns. WHY THE LARGEST STEP AND NOT THE SMALLEST. SNOWPACK is not a rate: Koren's solution is a closed form in the step length, so the density increment is proportional to dtsec and there is no dt -> 0 limit to approach by shrinking the step. Nothing here is formed as a difference either -- the increment is written as sndens_in pexp_m1 + dw (1 - dsx), an exact rearrangement in which nothing cancels, and its reference comes from a publisher that emits the Taylor series BEFORE WRF adds one to it (FORTRAN_BUGS N99) -- so there is no cancellation floor pushing the step DOWN either. What is left is that a smaller step puts LESS of the assertion's budget on the physics: at 60 s the compaction and retention move the density by 4.5e-9 to 9.4e-5 of itself, so a rel 1e-9 assertion on sndens_out tests them to between 1e-5 and 2e-1 of themselves, while at 1e-6 s the same assertion would test them to nothing at all -- the increment is 7.5e-17 of the density there. This is the fifth distinct direction the Noah reference step has had to move (surface closure 1e-6 s, tendencies 1e-2 s, runoff 1e-4 s, canopy drip 1e-2 s, compaction 60 s), and it is why the choice is argued per quantity rather than assumed. The companion regime at 1e-6 s pins the tendency in the dt -> 0 limit. Step 519 of case A, snow falling onto initially bare frozen ground. dry branch -- the snow surface is below freezing and only compaction acts; neither density limit fires. The density moves by 1.159e-06 of itself over the step.",
      "parameter_overrides": {
        "esd": 0.0010818704115051888,
        "sndens_in": 0.08826029243299223,
        "tsnow": 271.11636057979246,
        "tsoil": 270.2355795930687,
        "dtsec": 60.0
      },
      "time_span": {
        "start": 0.0,
        "end": 0.0
      },
      "assertions": [
        {
          "variable": "tavgc",
          "time": 0.0,
          "expected": -2.47402991356941,
          "tolerance": {
            "rel": 1e-09,
            "abs": 1e-12
          }
        },
        {
          "variable": "esdcx",
          "time": 0.0,
          "expected": 0.10818704115051889,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "comp_exponent",
          "time": 0.0,
          "expected": -2.0513885341783897,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "bfac",
          "time": 0.0,
          "expected": 2.142604580316605e-05,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "xarg",
          "time": 0.0,
          "expected": 2.318020499000028e-06,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "pexp_m1",
          "time": 0.0,
          "expected": 1.1590111450370388e-06,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "pexp",
          "time": 0.0,
          "expected": 1.000001159011145,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsx_raw",
          "time": 0.0,
          "expected": 0.08826039472765483,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsx",
          "time": 0.0,
          "expected": 0.08826039472765483,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsx_clamp",
          "time": 0.0,
          "expected": 0.0,
          "tolerance": {
            "abs": 0.0,
            "rel": 0.0
          }
        },
        {
          "variable": "wet",
          "time": 0.0,
          "expected": 0.0,
          "tolerance": {
            "abs": 0.0,
            "rel": 0.0
          }
        },
        {
          "variable": "dw",
          "time": 0.0,
          "expected": 0.0,
          "tolerance": {
            "rel": 1e-09,
            "abs": 1e-30
          }
        },
        {
          "variable": "sndens_wet",
          "time": 0.0,
          "expected": 0.08826039472765483,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "sndens_out",
          "time": 0.0,
          "expected": 0.08826039472765483,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "snowh_out",
          "time": 0.0,
          "expected": 0.012257711002126349,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsndens_dt",
          "time": 0.0,
          "expected": 1.7049110432344368e-09,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "pack_present",
          "time": 0.0,
          "expected": 1.0,
          "tolerance": {
            "abs": 0.0,
            "rel": 0.0
          }
        }
      ]
    },
    {
      "id": "cold_wrfdt_caseA_step679",
      "description": "Reference: WRF 4.8 em_scm_xy COLD-SEASON single-column run, data/eqwefic/noahcold_scm, dumped by ctessum-claude/WRF branch earthsciml-instrumented-noahcold and replayed through the real64 (-fdefault-real-8) kernels/noah_driver at ESM_DT = 60 s, WRF's OWN step; tolerances are the 1e-9 that a double-precision kernel build earns. WHY THE LARGEST STEP AND NOT THE SMALLEST. SNOWPACK is not a rate: Koren's solution is a closed form in the step length, so the density increment is proportional to dtsec and there is no dt -> 0 limit to approach by shrinking the step. Nothing here is formed as a difference either -- the increment is written as sndens_in pexp_m1 + dw (1 - dsx), an exact rearrangement in which nothing cancels, and its reference comes from a publisher that emits the Taylor series BEFORE WRF adds one to it (FORTRAN_BUGS N99) -- so there is no cancellation floor pushing the step DOWN either. What is left is that a smaller step puts LESS of the assertion's budget on the physics: at 60 s the compaction and retention move the density by 4.5e-9 to 9.4e-5 of itself, so a rel 1e-9 assertion on sndens_out tests them to between 1e-5 and 2e-1 of themselves, while at 1e-6 s the same assertion would test them to nothing at all -- the increment is 7.5e-17 of the density there. This is the fifth distinct direction the Noah reference step has had to move (surface closure 1e-6 s, tendencies 1e-2 s, runoff 1e-4 s, canopy drip 1e-2 s, compaction 60 s), and it is why the choice is argued per quantity rather than assumed. The companion regime at 1e-6 s pins the tendency in the dt -> 0 limit. Step 679 of case A, snow falling onto initially bare frozen ground. dry branch -- the snow surface is below freezing and only compaction acts; neither density limit fires. The density moves by 9.532e-07 of itself over the step.",
      "parameter_overrides": {
        "esd": 0.0012447004352495621,
        "sndens_in": 0.08951868511619737,
        "tsnow": 266.7578763233268,
        "tsoil": 266.8632270087088,
        "dtsec": 60.0
      },
      "time_span": {
        "start": 0.0,
        "end": 0.0
      },
      "assertions": [
        {
          "variable": "tavgc",
          "time": 0.0,
          "expected": -6.339448333982176,
          "tolerance": {
            "rel": 1e-09,
            "abs": 1e-12
          }
        },
        {
          "variable": "esdcx",
          "time": 0.0,
          "expected": 0.12447004352495622,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "comp_exponent",
          "time": 0.0,
          "expected": -2.387048254158719,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "bfac",
          "time": 0.0,
          "expected": 1.5316758497431497e-05,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "xarg",
          "time": 0.0,
          "expected": 1.9064775968365413e-06,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "pexp_m1",
          "time": 0.0,
          "expected": 9.532394041946973e-07,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "pexp",
          "time": 0.0,
          "expected": 1.0000009532394043,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsx_raw",
          "time": 0.0,
          "expected": 0.08951877044893544,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsx",
          "time": 0.0,
          "expected": 0.08951877044893544,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsx_clamp",
          "time": 0.0,
          "expected": 0.0,
          "tolerance": {
            "abs": 0.0,
            "rel": 0.0
          }
        },
        {
          "variable": "wet",
          "time": 0.0,
          "expected": 0.0,
          "tolerance": {
            "abs": 0.0,
            "rel": 0.0
          }
        },
        {
          "variable": "dw",
          "time": 0.0,
          "expected": 0.0,
          "tolerance": {
            "rel": 1e-09,
            "abs": 1e-30
          }
        },
        {
          "variable": "sndens_wet",
          "time": 0.0,
          "expected": 0.08951877044893544,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "sndens_out",
          "time": 0.0,
          "expected": 0.08951877044893544,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "snowh_out",
          "time": 0.0,
          "expected": 0.013904351333328263,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsndens_dt",
          "time": 0.0,
          "expected": 1.4222123010742784e-09,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "pack_present",
          "time": 0.0,
          "expected": 1.0,
          "tolerance": {
            "abs": 0.0,
            "rel": 0.0
          }
        }
      ]
    },
    {
      "id": "cold_wrfdt_caseA_step1761",
      "description": "Reference: WRF 4.8 em_scm_xy COLD-SEASON single-column run, data/eqwefic/noahcold_scm, dumped by ctessum-claude/WRF branch earthsciml-instrumented-noahcold and replayed through the real64 (-fdefault-real-8) kernels/noah_driver at ESM_DT = 60 s, WRF's OWN step; tolerances are the 1e-9 that a double-precision kernel build earns. WHY THE LARGEST STEP AND NOT THE SMALLEST. SNOWPACK is not a rate: Koren's solution is a closed form in the step length, so the density increment is proportional to dtsec and there is no dt -> 0 limit to approach by shrinking the step. Nothing here is formed as a difference either -- the increment is written as sndens_in pexp_m1 + dw (1 - dsx), an exact rearrangement in which nothing cancels, and its reference comes from a publisher that emits the Taylor series BEFORE WRF adds one to it (FORTRAN_BUGS N99) -- so there is no cancellation floor pushing the step DOWN either. What is left is that a smaller step puts LESS of the assertion's budget on the physics: at 60 s the compaction and retention move the density by 4.5e-9 to 9.4e-5 of itself, so a rel 1e-9 assertion on sndens_out tests them to between 1e-5 and 2e-1 of themselves, while at 1e-6 s the same assertion would test them to nothing at all -- the increment is 7.5e-17 of the density there. This is the fifth distinct direction the Noah reference step has had to move (surface closure 1e-6 s, tendencies 1e-2 s, runoff 1e-4 s, canopy drip 1e-2 s, compaction 60 s), and it is why the choice is argued per quantity rather than assumed. The companion regime at 1e-6 s pins the tendency in the dt -> 0 limit. Step 1761 of case A, snow falling onto initially bare frozen ground. SNOWPACK's melt-water retention branch RUNS -- the snow surface is at or above freezing; neither density limit fires. The density moves by 9.186e-04 of itself over the step.",
      "parameter_overrides": {
        "esd": 0.0034267619023603288,
        "sndens_in": 0.0897482270718234,
        "tsnow": 273.16918818880464,
        "tsoil": 270.0358225227164,
        "dtsec": 60.0
      },
      "time_span": {
        "start": 0.0,
        "end": 0.0
      },
      "assertions": [
        {
          "variable": "tavgc",
          "time": 0.0,
          "expected": -1.547494644239464,
          "tolerance": {
            "rel": 1e-09,
            "abs": 1e-12
          }
        },
        {
          "variable": "esdcx",
          "time": 0.0,
          "expected": 0.34267619023603285,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "comp_exponent",
          "time": 0.0,
          "expected": -2.0085123400474485,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "bfac",
          "time": 0.0,
          "expected": 2.2364692099179516e-05,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "xarg",
          "time": 0.0,
          "expected": 7.663847484348741e-06,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "pexp_m1",
          "time": 0.0,
          "expected": 3.83193353128617e-06,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "pexp",
          "time": 0.0,
          "expected": 1.0000038319335314,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsx_raw",
          "time": 0.0,
          "expected": 0.08974857098106409,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsx",
          "time": 0.0,
          "expected": 0.08974857098106409,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsx_clamp",
          "time": 0.0,
          "expected": 0.0,
          "tolerance": {
            "abs": 0.0,
            "rel": 0.0
          }
        },
        {
          "variable": "wet",
          "time": 0.0,
          "expected": 1.0,
          "tolerance": {
            "abs": 0.0,
            "rel": 0.0
          }
        },
        {
          "variable": "dw",
          "time": 0.0,
          "expected": 9.027777777777777e-05,
          "tolerance": {
            "rel": 1e-09,
            "abs": 1e-30
          }
        },
        {
          "variable": "sndens_wet",
          "time": 0.0,
          "expected": 0.08983074645729495,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "sndens_out",
          "time": 0.0,
          "expected": 0.08983074645729495,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "snowh_out",
          "time": 0.0,
          "expected": 0.03814687106033782,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsndens_dt",
          "time": 0.0,
          "expected": 1.3753230911927692e-06,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "pack_present",
          "time": 0.0,
          "expected": 1.0,
          "tolerance": {
            "abs": 0.0,
            "rel": 0.0
          }
        }
      ]
    },
    {
      "id": "cold_wrfdt_caseA_step1824",
      "description": "Reference: WRF 4.8 em_scm_xy COLD-SEASON single-column run, data/eqwefic/noahcold_scm, dumped by ctessum-claude/WRF branch earthsciml-instrumented-noahcold and replayed through the real64 (-fdefault-real-8) kernels/noah_driver at ESM_DT = 60 s, WRF's OWN step; tolerances are the 1e-9 that a double-precision kernel build earns. WHY THE LARGEST STEP AND NOT THE SMALLEST. SNOWPACK is not a rate: Koren's solution is a closed form in the step length, so the density increment is proportional to dtsec and there is no dt -> 0 limit to approach by shrinking the step. Nothing here is formed as a difference either -- the increment is written as sndens_in pexp_m1 + dw (1 - dsx), an exact rearrangement in which nothing cancels, and its reference comes from a publisher that emits the Taylor series BEFORE WRF adds one to it (FORTRAN_BUGS N99) -- so there is no cancellation floor pushing the step DOWN either. What is left is that a smaller step puts LESS of the assertion's budget on the physics: at 60 s the compaction and retention move the density by 4.5e-9 to 9.4e-5 of itself, so a rel 1e-9 assertion on sndens_out tests them to between 1e-5 and 2e-1 of themselves, while at 1e-6 s the same assertion would test them to nothing at all -- the increment is 7.5e-17 of the density there. This is the fifth distinct direction the Noah reference step has had to move (surface closure 1e-6 s, tendencies 1e-2 s, runoff 1e-4 s, canopy drip 1e-2 s, compaction 60 s), and it is why the choice is argued per quantity rather than assumed. The companion regime at 1e-6 s pins the tendency in the dt -> 0 limit. Step 1824 of case A, snow falling onto initially bare frozen ground. SNOWPACK's melt-water retention branch RUNS -- the snow surface is at or above freezing; neither density limit fires. The density moves by 8.632e-04 of itself over the step.",
      "parameter_overrides": {
        "esd": 0.00343367226096951,
        "sndens_in": 0.0950186423774112,
        "tsnow": 275.92893839426443,
        "tsoil": 271.8726372208587,
        "dtsec": 60.0
      },
      "time_span": {
        "start": 0.0,
        "end": 0.0
      },
      "assertions": [
        {
          "variable": "tavgc",
          "time": 0.0,
          "expected": 0.7507878075615793,
          "tolerance": {
            "rel": 1e-09,
            "abs": 1e-12
          }
        },
        {
          "variable": "esdcx",
          "time": 0.0,
          "expected": 0.343367226096951,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "comp_exponent",
          "time": 0.0,
          "expected": -1.935328465320709,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "bfac",
          "time": 0.0,
          "expected": 2.4062806374973404e-05,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "xarg",
          "time": 0.0,
          "expected": 8.262379077082646e-06,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "pexp_m1",
          "time": 0.0,
          "expected": 4.131200916382827e-06,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "pexp",
          "time": 0.0,
          "expected": 1.0000041312009165,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsx_raw",
          "time": 0.0,
          "expected": 0.09501903491851367,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsx",
          "time": 0.0,
          "expected": 0.09501903491851367,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsx_clamp",
          "time": 0.0,
          "expected": 0.0,
          "tolerance": {
            "abs": 0.0,
            "rel": 0.0
          }
        },
        {
          "variable": "wet",
          "time": 0.0,
          "expected": 1.0,
          "tolerance": {
            "abs": 0.0,
            "rel": 0.0
          }
        },
        {
          "variable": "dw",
          "time": 0.0,
          "expected": 9.027777777777777e-05,
          "tolerance": {
            "rel": 1e-09,
            "abs": 1e-30
          }
        },
        {
          "variable": "sndens_wet",
          "time": 0.0,
          "expected": 0.0951007345889724,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "sndens_out",
          "time": 0.0,
          "expected": 0.0951007345889724,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "snowh_out",
          "time": 0.0,
          "expected": 0.03610563341923616,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsndens_dt",
          "time": 0.0,
          "expected": 1.3682035260201384e-06,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "pack_present",
          "time": 0.0,
          "expected": 1.0,
          "tolerance": {
            "abs": 0.0,
            "rel": 0.0
          }
        }
      ]
    },
    {
      "id": "cold_wrfdt_caseA_step1849",
      "description": "Reference: WRF 4.8 em_scm_xy COLD-SEASON single-column run, data/eqwefic/noahcold_scm, dumped by ctessum-claude/WRF branch earthsciml-instrumented-noahcold and replayed through the real64 (-fdefault-real-8) kernels/noah_driver at ESM_DT = 60 s, WRF's OWN step; tolerances are the 1e-9 that a double-precision kernel build earns. WHY THE LARGEST STEP AND NOT THE SMALLEST. SNOWPACK is not a rate: Koren's solution is a closed form in the step length, so the density increment is proportional to dtsec and there is no dt -> 0 limit to approach by shrinking the step. Nothing here is formed as a difference either -- the increment is written as sndens_in pexp_m1 + dw (1 - dsx), an exact rearrangement in which nothing cancels, and its reference comes from a publisher that emits the Taylor series BEFORE WRF adds one to it (FORTRAN_BUGS N99) -- so there is no cancellation floor pushing the step DOWN either. What is left is that a smaller step puts LESS of the assertion's budget on the physics: at 60 s the compaction and retention move the density by 4.5e-9 to 9.4e-5 of itself, so a rel 1e-9 assertion on sndens_out tests them to between 1e-5 and 2e-1 of themselves, while at 1e-6 s the same assertion would test them to nothing at all -- the increment is 7.5e-17 of the density there. This is the fifth distinct direction the Noah reference step has had to move (surface closure 1e-6 s, tendencies 1e-2 s, runoff 1e-4 s, canopy drip 1e-2 s, compaction 60 s), and it is why the choice is argued per quantity rather than assumed. The companion regime at 1e-6 s pins the tendency in the dt -> 0 limit. Step 1849 of case A, snow falling onto initially bare frozen ground. SNOWPACK's melt-water retention branch RUNS -- the snow surface is at or above freezing; neither density limit fires. The density moves by 8.431e-04 of itself over the step.",
      "parameter_overrides": {
        "esd": 0.003412985561945523,
        "sndens_in": 0.09707147189276946,
        "tsnow": 276.5390788067697,
        "tsoil": 272.3708377870439,
        "dtsec": 60.0
      },
      "time_span": {
        "start": 0.0,
        "end": 0.0
      },
      "assertions": [
        {
          "variable": "tavgc",
          "time": 0.0,
          "expected": 1.3049582969068183,
          "tolerance": {
            "rel": 1e-09,
            "abs": 1e-12
          }
        },
        {
          "variable": "esdcx",
          "time": 0.0,
          "expected": 0.3412985561945523,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "comp_exponent",
          "time": 0.0,
          "expected": -1.934104245995613,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "bfac",
          "time": 0.0,
          "expected": 2.4092282566533996e-05,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "xarg",
          "time": 0.0,
          "expected": 8.222661255389236e-06,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "pexp_m1",
          "time": 0.0,
          "expected": 4.111341896410803e-06,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "pexp",
          "time": 0.0,
          "expected": 1.0000041113418965,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsx_raw",
          "time": 0.0,
          "expected": 0.0970718709867788,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsx",
          "time": 0.0,
          "expected": 0.0970718709867788,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsx_clamp",
          "time": 0.0,
          "expected": 0.0,
          "tolerance": {
            "abs": 0.0,
            "rel": 0.0
          }
        },
        {
          "variable": "wet",
          "time": 0.0,
          "expected": 1.0,
          "tolerance": {
            "abs": 0.0,
            "rel": 0.0
          }
        },
        {
          "variable": "dw",
          "time": 0.0,
          "expected": 9.027777777777777e-05,
          "tolerance": {
            "rel": 1e-09,
            "abs": 1e-30
          }
        },
        {
          "variable": "sndens_wet",
          "time": 0.0,
          "expected": 0.09715338533175916,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "sndens_out",
          "time": 0.0,
          "expected": 0.09715338533175916,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "snowh_out",
          "time": 0.0,
          "expected": 0.03512986758300668,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsndens_dt",
          "time": 0.0,
          "expected": 1.365223983161654e-06,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "pack_present",
          "time": 0.0,
          "expected": 1.0,
          "tolerance": {
            "abs": 0.0,
            "rel": 0.0
          }
        }
      ]
    },
    {
      "id": "cold_wrfdt_caseA_step1874",
      "description": "Reference: WRF 4.8 em_scm_xy COLD-SEASON single-column run, data/eqwefic/noahcold_scm, dumped by ctessum-claude/WRF branch earthsciml-instrumented-noahcold and replayed through the real64 (-fdefault-real-8) kernels/noah_driver at ESM_DT = 60 s, WRF's OWN step; tolerances are the 1e-9 that a double-precision kernel build earns. WHY THE LARGEST STEP AND NOT THE SMALLEST. SNOWPACK is not a rate: Koren's solution is a closed form in the step length, so the density increment is proportional to dtsec and there is no dt -> 0 limit to approach by shrinking the step. Nothing here is formed as a difference either -- the increment is written as sndens_in pexp_m1 + dw (1 - dsx), an exact rearrangement in which nothing cancels, and its reference comes from a publisher that emits the Taylor series BEFORE WRF adds one to it (FORTRAN_BUGS N99) -- so there is no cancellation floor pushing the step DOWN either. What is left is that a smaller step puts LESS of the assertion's budget on the physics: at 60 s the compaction and retention move the density by 4.5e-9 to 9.4e-5 of itself, so a rel 1e-9 assertion on sndens_out tests them to between 1e-5 and 2e-1 of themselves, while at 1e-6 s the same assertion would test them to nothing at all -- the increment is 7.5e-17 of the density there. This is the fifth distinct direction the Noah reference step has had to move (surface closure 1e-6 s, tendencies 1e-2 s, runoff 1e-4 s, canopy drip 1e-2 s, compaction 60 s), and it is why the choice is argued per quantity rather than assumed. The companion regime at 1e-6 s pins the tendency in the dt -> 0 limit. Step 1874 of case A, snow falling onto initially bare frozen ground. SNOWPACK's melt-water retention branch RUNS -- the snow surface is at or above freezing; neither density limit fires. The density moves by 8.238e-04 of itself over the step.",
      "parameter_overrides": {
        "esd": 0.0033898014747246047,
        "sndens_in": 0.09912014745183785,
        "tsnow": 276.74539795300836,
        "tsoil": 272.64714897931117,
        "dtsec": 60.0
      },
      "time_span": {
        "start": 0.0,
        "end": 0.0
      },
      "assertions": [
        {
          "variable": "tavgc",
          "time": 0.0,
          "expected": 1.5462734661597892,
          "tolerance": {
            "rel": 1e-09,
            "abs": 1e-12
          }
        },
        {
          "variable": "esdcx",
          "time": 0.0,
          "expected": 0.33898014747246047,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "comp_exponent",
          "time": 0.0,
          "expected": -1.9578212191958115,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "bfac",
          "time": 0.0,
          "expected": 2.3527609186850982e-05,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "xarg",
          "time": 0.0,
          "expected": 7.975392431833161e-06,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "pexp_m1",
          "time": 0.0,
          "expected": 3.987706817085125e-06,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "pexp",
          "time": 0.0,
          "expected": 1.000003987706817,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsx_raw",
          "time": 0.0,
          "expected": 0.09912054271392555,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsx",
          "time": 0.0,
          "expected": 0.09912054271392555,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsx_clamp",
          "time": 0.0,
          "expected": 0.0,
          "tolerance": {
            "abs": 0.0,
            "rel": 0.0
          }
        },
        {
          "variable": "wet",
          "time": 0.0,
          "expected": 1.0,
          "tolerance": {
            "abs": 0.0,
            "rel": 0.0
          }
        },
        {
          "variable": "dw",
          "time": 0.0,
          "expected": 9.027777777777777e-05,
          "tolerance": {
            "rel": 1e-09,
            "abs": 1e-30
          }
        },
        {
          "variable": "sndens_wet",
          "time": 0.0,
          "expected": 0.09920187210937498,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "sndens_out",
          "time": 0.0,
          "expected": 0.09920187210937498,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "snowh_out",
          "time": 0.0,
          "expected": 0.03417074096129134,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsndens_dt",
          "time": 0.0,
          "expected": 1.362077625619024e-06,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "pack_present",
          "time": 0.0,
          "expected": 1.0,
          "tolerance": {
            "abs": 0.0,
            "rel": 0.0
          }
        }
      ]
    },
    {
      "id": "cold_wrfdt_caseA_step1899",
      "description": "Reference: WRF 4.8 em_scm_xy COLD-SEASON single-column run, data/eqwefic/noahcold_scm, dumped by ctessum-claude/WRF branch earthsciml-instrumented-noahcold and replayed through the real64 (-fdefault-real-8) kernels/noah_driver at ESM_DT = 60 s, WRF's OWN step; tolerances are the 1e-9 that a double-precision kernel build earns. WHY THE LARGEST STEP AND NOT THE SMALLEST. SNOWPACK is not a rate: Koren's solution is a closed form in the step length, so the density increment is proportional to dtsec and there is no dt -> 0 limit to approach by shrinking the step. Nothing here is formed as a difference either -- the increment is written as sndens_in pexp_m1 + dw (1 - dsx), an exact rearrangement in which nothing cancels, and its reference comes from a publisher that emits the Taylor series BEFORE WRF adds one to it (FORTRAN_BUGS N99) -- so there is no cancellation floor pushing the step DOWN either. What is left is that a smaller step puts LESS of the assertion's budget on the physics: at 60 s the compaction and retention move the density by 4.5e-9 to 9.4e-5 of itself, so a rel 1e-9 assertion on sndens_out tests them to between 1e-5 and 2e-1 of themselves, while at 1e-6 s the same assertion would test them to nothing at all -- the increment is 7.5e-17 of the density there. This is the fifth distinct direction the Noah reference step has had to move (surface closure 1e-6 s, tendencies 1e-2 s, runoff 1e-4 s, canopy drip 1e-2 s, compaction 60 s), and it is why the choice is argued per quantity rather than assumed. The companion regime at 1e-6 s pins the tendency in the dt -> 0 limit. Step 1899 of case A, snow falling onto initially bare frozen ground. SNOWPACK's melt-water retention branch RUNS -- the snow surface is at or above freezing; neither density limit fires. The density moves by 8.053e-04 of itself over the step.",
      "parameter_overrides": {
        "esd": 0.0033652757905897543,
        "sndens_in": 0.10116387049394708,
        "tsnow": 276.91277668971736,
        "tsoil": 272.89875185363144,
        "dtsec": 60.0
      },
      "time_span": {
        "start": 0.0,
        "end": 0.0
      },
      "assertions": [
        {
          "variable": "tavgc",
          "time": 0.0,
          "expected": 1.7557642716744226,
          "tolerance": {
            "rel": 1e-09,
            "abs": 1e-12
          }
        },
        {
          "variable": "esdcx",
          "time": 0.0,
          "expected": 0.33652757905897546,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "comp_exponent",
          "time": 0.0,
          "expected": -1.9839801386389346,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "bfac",
          "time": 0.0,
          "expected": 2.292013246121927e-05,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "xarg",
          "time": 0.0,
          "expected": 7.713256688885158e-06,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "pexp_m1",
          "time": 0.0,
          "expected": 3.856638260183158e-06,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "pexp",
          "time": 0.0,
          "expected": 1.00000385663826,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsx_raw",
          "time": 0.0,
          "expected": 0.10116426064640056,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsx",
          "time": 0.0,
          "expected": 0.10116426064640056,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsx_clamp",
          "time": 0.0,
          "expected": 0.0,
          "tolerance": {
            "abs": 0.0,
            "rel": 0.0
          }
        },
        {
          "variable": "wet",
          "time": 0.0,
          "expected": 1.0,
          "tolerance": {
            "abs": 0.0,
            "rel": 0.0
          }
        },
        {
          "variable": "dw",
          "time": 0.0,
          "expected": 9.027777777777777e-05,
          "tolerance": {
            "rel": 1e-09,
            "abs": 1e-30
          }
        },
        {
          "variable": "sndens_wet",
          "time": 0.0,
          "expected": 0.10124540553953665,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "sndens_out",
          "time": 0.0,
          "expected": 0.10124540553953665,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "snowh_out",
          "time": 0.0,
          "expected": 0.033238800048814104,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsndens_dt",
          "time": 0.0,
          "expected": 1.3589174264930667e-06,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "pack_present",
          "time": 0.0,
          "expected": 1.0,
          "tolerance": {
            "abs": 0.0,
            "rel": 0.0
          }
        }
      ]
    },
    {
      "id": "cold_wrfdt_caseA_step1924",
      "description": "Reference: WRF 4.8 em_scm_xy COLD-SEASON single-column run, data/eqwefic/noahcold_scm, dumped by ctessum-claude/WRF branch earthsciml-instrumented-noahcold and replayed through the real64 (-fdefault-real-8) kernels/noah_driver at ESM_DT = 60 s, WRF's OWN step; tolerances are the 1e-9 that a double-precision kernel build earns. WHY THE LARGEST STEP AND NOT THE SMALLEST. SNOWPACK is not a rate: Koren's solution is a closed form in the step length, so the density increment is proportional to dtsec and there is no dt -> 0 limit to approach by shrinking the step. Nothing here is formed as a difference either -- the increment is written as sndens_in pexp_m1 + dw (1 - dsx), an exact rearrangement in which nothing cancels, and its reference comes from a publisher that emits the Taylor series BEFORE WRF adds one to it (FORTRAN_BUGS N99) -- so there is no cancellation floor pushing the step DOWN either. What is left is that a smaller step puts LESS of the assertion's budget on the physics: at 60 s the compaction and retention move the density by 4.5e-9 to 9.4e-5 of itself, so a rel 1e-9 assertion on sndens_out tests them to between 1e-5 and 2e-1 of themselves, while at 1e-6 s the same assertion would test them to nothing at all -- the increment is 7.5e-17 of the density there. This is the fifth distinct direction the Noah reference step has had to move (surface closure 1e-6 s, tendencies 1e-2 s, runoff 1e-4 s, canopy drip 1e-2 s, compaction 60 s), and it is why the choice is argued per quantity rather than assumed. The companion regime at 1e-6 s pins the tendency in the dt -> 0 limit. Step 1924 of case A, snow falling onto initially bare frozen ground. SNOWPACK's melt-water retention branch RUNS -- the snow surface is at or above freezing; neither density limit fires. The density moves by 7.874e-04 of itself over the step.",
      "parameter_overrides": {
        "esd": 0.0033411190534057425,
        "sndens_in": 0.10320228740684458,
        "tsnow": 276.2779230332568,
        "tsoil": 272.6813468589015,
        "dtsec": 60.0
      },
      "time_span": {
        "start": 0.0,
        "end": 0.0
      },
      "assertions": [
        {
          "variable": "tavgc",
          "time": 0.0,
          "expected": 1.3296349460791816,
          "tolerance": {
            "rel": 1e-09,
            "abs": 1e-12
          }
        },
        {
          "variable": "esdcx",
          "time": 0.0,
          "expected": 0.3341119053405743,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "comp_exponent",
          "time": 0.0,
          "expected": -2.0608772398574016,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "bfac",
          "time": 0.0,
          "expected": 2.1223701869865714e-05,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "xarg",
          "time": 0.0,
          "expected": 7.091091470121143e-06,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "pexp_m1",
          "time": 0.0,
          "expected": 3.545554115671802e-06,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "pexp",
          "time": 0.0,
          "expected": 1.0000035455541156,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsx_raw",
          "time": 0.0,
          "expected": 0.10320265331613944,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsx",
          "time": 0.0,
          "expected": 0.10320265331613944,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsx_clamp",
          "time": 0.0,
          "expected": 0.0,
          "tolerance": {
            "abs": 0.0,
            "rel": 0.0
          }
        },
        {
          "variable": "wet",
          "time": 0.0,
          "expected": 1.0,
          "tolerance": {
            "abs": 0.0,
            "rel": 0.0
          }
        },
        {
          "variable": "dw",
          "time": 0.0,
          "expected": 9.027777777777777e-05,
          "tolerance": {
            "rel": 1e-09,
            "abs": 1e-30
          }
        },
        {
          "variable": "sndens_wet",
          "time": 0.0,
          "expected": 0.10328361418771506,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "sndens_out",
          "time": 0.0,
          "expected": 0.10328361418771506,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "snowh_out",
          "time": 0.0,
          "expected": 0.032348975001333255,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsndens_dt",
          "time": 0.0,
          "expected": 1.3554463478414729e-06,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "pack_present",
          "time": 0.0,
          "expected": 1.0,
          "tolerance": {
            "abs": 0.0,
            "rel": 0.0
          }
        }
      ]
    },
    {
      "id": "cold_wrfdt_caseA_step2089",
      "description": "Reference: WRF 4.8 em_scm_xy COLD-SEASON single-column run, data/eqwefic/noahcold_scm, dumped by ctessum-claude/WRF branch earthsciml-instrumented-noahcold and replayed through the real64 (-fdefault-real-8) kernels/noah_driver at ESM_DT = 60 s, WRF's OWN step; tolerances are the 1e-9 that a double-precision kernel build earns. WHY THE LARGEST STEP AND NOT THE SMALLEST. SNOWPACK is not a rate: Koren's solution is a closed form in the step length, so the density increment is proportional to dtsec and there is no dt -> 0 limit to approach by shrinking the step. Nothing here is formed as a difference either -- the increment is written as sndens_in pexp_m1 + dw (1 - dsx), an exact rearrangement in which nothing cancels, and its reference comes from a publisher that emits the Taylor series BEFORE WRF adds one to it (FORTRAN_BUGS N99) -- so there is no cancellation floor pushing the step DOWN either. What is left is that a smaller step puts LESS of the assertion's budget on the physics: at 60 s the compaction and retention move the density by 4.5e-9 to 9.4e-5 of itself, so a rel 1e-9 assertion on sndens_out tests them to between 1e-5 and 2e-1 of themselves, while at 1e-6 s the same assertion would test them to nothing at all -- the increment is 7.5e-17 of the density there. This is the fifth distinct direction the Noah reference step has had to move (surface closure 1e-6 s, tendencies 1e-2 s, runoff 1e-4 s, canopy drip 1e-2 s, compaction 60 s), and it is why the choice is argued per quantity rather than assumed. The companion regime at 1e-6 s pins the tendency in the dt -> 0 limit. Step 2089 of case A, snow falling onto initially bare frozen ground. SNOWPACK's melt-water retention branch RUNS -- the snow surface is at or above freezing; neither density limit fires. The density moves by 7.300e-04 of itself over the step.",
      "parameter_overrides": {
        "esd": 0.0034099248678107247,
        "sndens_in": 0.11037592919214192,
        "tsnow": 274.7837293937297,
        "tsoil": 272.2815060392348,
        "dtsec": 60.0
      },
      "time_span": {
        "start": 0.0,
        "end": 0.0
      },
      "assertions": [
        {
          "variable": "tavgc",
          "time": 0.0,
          "expected": 0.3826177164822582,
          "tolerance": {
            "rel": 1e-09,
            "abs": 1e-12
          }
        },
        {
          "variable": "esdcx",
          "time": 0.0,
          "expected": 0.34099248678107247,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "comp_exponent",
          "time": 0.0,
          "expected": -2.2872850957164,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "bfac",
          "time": 0.0,
          "expected": 1.6923627357590557e-05,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "xarg",
          "time": 0.0,
          "expected": 5.770829778020994e-06,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "pexp_m1",
          "time": 0.0,
          "expected": 2.8854204394312254e-06,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "pexp",
          "time": 0.0,
          "expected": 1.0000028854204395,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsx_raw",
          "time": 0.0,
          "expected": 0.11037624767310404,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsx",
          "time": 0.0,
          "expected": 0.11037624767310404,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsx_clamp",
          "time": 0.0,
          "expected": 0.0,
          "tolerance": {
            "abs": 0.0,
            "rel": 0.0
          }
        },
        {
          "variable": "wet",
          "time": 0.0,
          "expected": 1.0,
          "tolerance": {
            "abs": 0.0,
            "rel": 0.0
          }
        },
        {
          "variable": "dw",
          "time": 0.0,
          "expected": 9.027777777777777e-05,
          "tolerance": {
            "rel": 1e-09,
            "abs": 1e-30
          }
        },
        {
          "variable": "sndens_wet",
          "time": 0.0,
          "expected": 0.11045656092852243,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "sndens_out",
          "time": 0.0,
          "expected": 0.11045656092852243,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "snowh_out",
          "time": 0.0,
          "expected": 0.03087118446515207,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsndens_dt",
          "time": 0.0,
          "expected": 1.3438622730085425e-06,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "pack_present",
          "time": 0.0,
          "expected": 1.0,
          "tolerance": {
            "abs": 0.0,
            "rel": 0.0
          }
        }
      ]
    },
    {
      "id": "cold_wrfdt_caseA_step2114",
      "description": "Reference: WRF 4.8 em_scm_xy COLD-SEASON single-column run, data/eqwefic/noahcold_scm, dumped by ctessum-claude/WRF branch earthsciml-instrumented-noahcold and replayed through the real64 (-fdefault-real-8) kernels/noah_driver at ESM_DT = 60 s, WRF's OWN step; tolerances are the 1e-9 that a double-precision kernel build earns. WHY THE LARGEST STEP AND NOT THE SMALLEST. SNOWPACK is not a rate: Koren's solution is a closed form in the step length, so the density increment is proportional to dtsec and there is no dt -> 0 limit to approach by shrinking the step. Nothing here is formed as a difference either -- the increment is written as sndens_in pexp_m1 + dw (1 - dsx), an exact rearrangement in which nothing cancels, and its reference comes from a publisher that emits the Taylor series BEFORE WRF adds one to it (FORTRAN_BUGS N99) -- so there is no cancellation floor pushing the step DOWN either. What is left is that a smaller step puts LESS of the assertion's budget on the physics: at 60 s the compaction and retention move the density by 4.5e-9 to 9.4e-5 of itself, so a rel 1e-9 assertion on sndens_out tests them to between 1e-5 and 2e-1 of themselves, while at 1e-6 s the same assertion would test them to nothing at all -- the increment is 7.5e-17 of the density there. This is the fifth distinct direction the Noah reference step has had to move (surface closure 1e-6 s, tendencies 1e-2 s, runoff 1e-4 s, canopy drip 1e-2 s, compaction 60 s), and it is why the choice is argued per quantity rather than assumed. The companion regime at 1e-6 s pins the tendency in the dt -> 0 limit. Step 2114 of case A, snow falling onto initially bare frozen ground. dry branch -- the snow surface is below freezing and only compaction acts; neither density limit fires. The density moves by 1.699e-06 of itself over the step.",
      "parameter_overrides": {
        "esd": 0.0035964985146328165,
        "sndens_in": 0.11054433964872809,
        "tsnow": 265.81404355859604,
        "tsoil": 266.7713316826251,
        "dtsec": 60.0
      },
      "time_span": {
        "start": 0.0,
        "end": 0.0
      },
      "assertions": [
        {
          "variable": "tavgc",
          "time": 0.0,
          "expected": -6.857312379389413,
          "tolerance": {
            "rel": 1e-09,
            "abs": 1e-12
          }
        },
        {
          "variable": "esdcx",
          "time": 0.0,
          "expected": 0.35964985146328166,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "comp_exponent",
          "time": 0.0,
          "expected": -2.8700161229744428,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "bfac",
          "time": 0.0,
          "expected": 9.449668738645357e-06,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "xarg",
          "time": 0.0,
          "expected": 3.398571958231019e-06,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "pexp_m1",
          "time": 0.0,
          "expected": 1.6992879041657041e-06,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "pexp",
          "time": 0.0,
          "expected": 1.0000016992879042,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsx_raw",
          "time": 0.0,
          "expected": 0.11054452749538733,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsx",
          "time": 0.0,
          "expected": 0.11054452749538733,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsx_clamp",
          "time": 0.0,
          "expected": 0.0,
          "tolerance": {
            "abs": 0.0,
            "rel": 0.0
          }
        },
        {
          "variable": "wet",
          "time": 0.0,
          "expected": 0.0,
          "tolerance": {
            "abs": 0.0,
            "rel": 0.0
          }
        },
        {
          "variable": "dw",
          "time": 0.0,
          "expected": 0.0,
          "tolerance": {
            "rel": 1e-09,
            "abs": 1e-30
          }
        },
        {
          "variable": "sndens_wet",
          "time": 0.0,
          "expected": 0.11054452749538733,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "sndens_out",
          "time": 0.0,
          "expected": 0.11054452749538733,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "snowh_out",
          "time": 0.0,
          "expected": 0.032534387690814334,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsndens_dt",
          "time": 0.0,
          "expected": 3.1307776539844818e-09,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "pack_present",
          "time": 0.0,
          "expected": 1.0,
          "tolerance": {
            "abs": 0.0,
            "rel": 0.0
          }
        }
      ]
    },
    {
      "id": "cold_wrfdt_caseA_step2847",
      "description": "Reference: WRF 4.8 em_scm_xy COLD-SEASON single-column run, data/eqwefic/noahcold_scm, dumped by ctessum-claude/WRF branch earthsciml-instrumented-noahcold and replayed through the real64 (-fdefault-real-8) kernels/noah_driver at ESM_DT = 60 s, WRF's OWN step; tolerances are the 1e-9 that a double-precision kernel build earns. WHY THE LARGEST STEP AND NOT THE SMALLEST. SNOWPACK is not a rate: Koren's solution is a closed form in the step length, so the density increment is proportional to dtsec and there is no dt -> 0 limit to approach by shrinking the step. Nothing here is formed as a difference either -- the increment is written as sndens_in pexp_m1 + dw (1 - dsx), an exact rearrangement in which nothing cancels, and its reference comes from a publisher that emits the Taylor series BEFORE WRF adds one to it (FORTRAN_BUGS N99) -- so there is no cancellation floor pushing the step DOWN either. What is left is that a smaller step puts LESS of the assertion's budget on the physics: at 60 s the compaction and retention move the density by 4.5e-9 to 9.4e-5 of itself, so a rel 1e-9 assertion on sndens_out tests them to between 1e-5 and 2e-1 of themselves, while at 1e-6 s the same assertion would test them to nothing at all -- the increment is 7.5e-17 of the density there. This is the fifth distinct direction the Noah reference step has had to move (surface closure 1e-6 s, tendencies 1e-2 s, runoff 1e-4 s, canopy drip 1e-2 s, compaction 60 s), and it is why the choice is argued per quantity rather than assumed. The companion regime at 1e-6 s pins the tendency in the dt -> 0 limit. Step 2847 of case A, snow falling onto initially bare frozen ground. dry branch -- the snow surface is below freezing and only compaction acts; neither density limit fires. The density moves by 1.417e-06 of itself over the step.",
      "parameter_overrides": {
        "esd": 0.003665838407261948,
        "sndens_in": 0.11048201772600909,
        "tsnow": 263.26987657273827,
        "tsoil": 264.26561775220716,
        "dtsec": 60.0
      },
      "time_span": {
        "start": 0.0,
        "end": 0.0
      },
      "assertions": [
        {
          "variable": "tavgc",
          "time": 0.0,
          "expected": -9.382252837527261,
          "tolerance": {
            "rel": 1e-09,
            "abs": 1e-12
          }
        },
        {
          "variable": "esdcx",
          "time": 0.0,
          "expected": 0.3665838407261948,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "comp_exponent",
          "time": 0.0,
          "expected": -3.0707025992483716,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "bfac",
          "time": 0.0,
          "expected": 7.73142514067313e-06,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "xarg",
          "time": 0.0,
          "expected": 2.834215522355017e-06,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "pexp_m1",
          "time": 0.0,
          "expected": 1.4171090999747283e-06,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "pexp",
          "time": 0.0,
          "expected": 1.0000014171091,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsx_raw",
          "time": 0.0,
          "expected": 0.1104821742910818,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsx",
          "time": 0.0,
          "expected": 0.1104821742910818,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsx_clamp",
          "time": 0.0,
          "expected": 0.0,
          "tolerance": {
            "abs": 0.0,
            "rel": 0.0
          }
        },
        {
          "variable": "wet",
          "time": 0.0,
          "expected": 0.0,
          "tolerance": {
            "abs": 0.0,
            "rel": 0.0
          }
        },
        {
          "variable": "dw",
          "time": 0.0,
          "expected": 0.0,
          "tolerance": {
            "rel": 1e-09,
            "abs": 1e-30
          }
        },
        {
          "variable": "sndens_wet",
          "time": 0.0,
          "expected": 0.1104821742910818,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "sndens_out",
          "time": 0.0,
          "expected": 0.1104821742910818,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "snowh_out",
          "time": 0.0,
          "expected": 0.033180360820956954,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsndens_dt",
          "time": 0.0,
          "expected": 2.609417878384945e-09,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "pack_present",
          "time": 0.0,
          "expected": 1.0,
          "tolerance": {
            "abs": 0.0,
            "rel": 0.0
          }
        }
      ]
    },
    {
      "id": "cold_wrfdt_caseA_step2880",
      "description": "Reference: WRF 4.8 em_scm_xy COLD-SEASON single-column run, data/eqwefic/noahcold_scm, dumped by ctessum-claude/WRF branch earthsciml-instrumented-noahcold and replayed through the real64 (-fdefault-real-8) kernels/noah_driver at ESM_DT = 60 s, WRF's OWN step; tolerances are the 1e-9 that a double-precision kernel build earns. WHY THE LARGEST STEP AND NOT THE SMALLEST. SNOWPACK is not a rate: Koren's solution is a closed form in the step length, so the density increment is proportional to dtsec and there is no dt -> 0 limit to approach by shrinking the step. Nothing here is formed as a difference either -- the increment is written as sndens_in pexp_m1 + dw (1 - dsx), an exact rearrangement in which nothing cancels, and its reference comes from a publisher that emits the Taylor series BEFORE WRF adds one to it (FORTRAN_BUGS N99) -- so there is no cancellation floor pushing the step DOWN either. What is left is that a smaller step puts LESS of the assertion's budget on the physics: at 60 s the compaction and retention move the density by 4.5e-9 to 9.4e-5 of itself, so a rel 1e-9 assertion on sndens_out tests them to between 1e-5 and 2e-1 of themselves, while at 1e-6 s the same assertion would test them to nothing at all -- the increment is 7.5e-17 of the density there. This is the fifth distinct direction the Noah reference step has had to move (surface closure 1e-6 s, tendencies 1e-2 s, runoff 1e-4 s, canopy drip 1e-2 s, compaction 60 s), and it is why the choice is argued per quantity rather than assumed. The companion regime at 1e-6 s pins the tendency in the dt -> 0 limit. Step 2880 of case A, snow falling onto initially bare frozen ground. dry branch -- the snow surface is below freezing and only compaction acts; neither density limit fires. The density moves by 1.417e-06 of itself over the step.",
      "parameter_overrides": {
        "esd": 0.0036658262363054666,
        "sndens_in": 0.11048748311453574,
        "tsnow": 263.2923388137898,
        "tsoil": 264.2530140451615,
        "dtsec": 60.0
      },
      "time_span": {
        "start": 0.0,
        "end": 0.0
      },
      "assertions": [
        {
          "variable": "tavgc",
          "time": 0.0,
          "expected": -9.37732357052434,
          "tolerance": {
            "rel": 1e-09,
            "abs": 1e-12
          }
        },
        {
          "variable": "esdcx",
          "time": 0.0,
          "expected": 0.36658262363054667,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "comp_exponent",
          "time": 0.0,
          "expected": -3.0704230310471976,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "bfac",
          "time": 0.0,
          "expected": 7.733586903458206e-06,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "xarg",
          "time": 0.0,
          "expected": 2.8349985771445447e-06,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "pexp_m1",
          "time": 0.0,
          "expected": 1.4175006281093769e-06,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "pexp",
          "time": 0.0,
          "expected": 1.0000014175006282,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsx_raw",
          "time": 0.0,
          "expected": 0.11048763973061246,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsx",
          "time": 0.0,
          "expected": 0.11048763973061246,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsx_clamp",
          "time": 0.0,
          "expected": 0.0,
          "tolerance": {
            "abs": 0.0,
            "rel": 0.0
          }
        },
        {
          "variable": "wet",
          "time": 0.0,
          "expected": 0.0,
          "tolerance": {
            "abs": 0.0,
            "rel": 0.0
          }
        },
        {
          "variable": "dw",
          "time": 0.0,
          "expected": 0.0,
          "tolerance": {
            "rel": 1e-09,
            "abs": 1e-30
          }
        },
        {
          "variable": "sndens_wet",
          "time": 0.0,
          "expected": 0.11048763973061246,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "sndens_out",
          "time": 0.0,
          "expected": 0.11048763973061246,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "snowh_out",
          "time": 0.0,
          "expected": 0.03317860934710318,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsndens_dt",
          "time": 0.0,
          "expected": 2.6102679452179767e-09,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "pack_present",
          "time": 0.0,
          "expected": 1.0,
          "tolerance": {
            "abs": 0.0,
            "rel": 0.0
          }
        }
      ]
    },
    {
      "id": "cold_wrfdt_caseB_step1",
      "description": "Reference: WRF 4.8 em_scm_xy COLD-SEASON single-column run, data/eqwefic/noahcold_scm, dumped by ctessum-claude/WRF branch earthsciml-instrumented-noahcold and replayed through the real64 (-fdefault-real-8) kernels/noah_driver at ESM_DT = 60 s, WRF's OWN step; tolerances are the 1e-9 that a double-precision kernel build earns. WHY THE LARGEST STEP AND NOT THE SMALLEST. SNOWPACK is not a rate: Koren's solution is a closed form in the step length, so the density increment is proportional to dtsec and there is no dt -> 0 limit to approach by shrinking the step. Nothing here is formed as a difference either -- the increment is written as sndens_in pexp_m1 + dw (1 - dsx), an exact rearrangement in which nothing cancels, and its reference comes from a publisher that emits the Taylor series BEFORE WRF adds one to it (FORTRAN_BUGS N99) -- so there is no cancellation floor pushing the step DOWN either. What is left is that a smaller step puts LESS of the assertion's budget on the physics: at 60 s the compaction and retention move the density by 4.5e-9 to 9.4e-5 of itself, so a rel 1e-9 assertion on sndens_out tests them to between 1e-5 and 2e-1 of themselves, while at 1e-6 s the same assertion would test them to nothing at all -- the increment is 7.5e-17 of the density there. This is the fifth distinct direction the Noah reference step has had to move (surface closure 1e-6 s, tendencies 1e-2 s, runoff 1e-4 s, canopy drip 1e-2 s, compaction 60 s), and it is why the choice is argued per quantity rather than assumed. The companion regime at 1e-6 s pins the tendency in the dt -> 0 limit. Step 1 of case B, a 30 cm pack melting under warm advection. dry branch -- the snow surface is below freezing and only compaction acts; neither density limit fires. The density moves by 2.972e-06 of itself over the step.",
      "parameter_overrides": {
        "esd": 0.060000482612748304,
        "sndens_in": 0.2000000198682161,
        "tsnow": 260.52261495701777,
        "tsoil": 262.64608782840094,
        "dtsec": 60.0
      },
      "time_span": {
        "start": 0.0,
        "end": 0.0
      },
      "assertions": [
        {
          "variable": "tavgc",
          "time": 0.0,
          "expected": -11.565648607290626,
          "tolerance": {
            "rel": 1e-09,
            "abs": 1e-12
          }
        },
        {
          "variable": "esdcx",
          "time": 0.0,
          "expected": 6.000048261274831,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "comp_exponent",
          "time": 0.0,
          "expected": -5.125252305815788,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "bfac",
          "time": 0.0,
          "expected": 9.907862134163364e-07,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "xarg",
          "time": 0.0,
          "expected": 5.9447650971037625e-06,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "pexp_m1",
          "time": 0.0,
          "expected": 2.9723884385993117e-06,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "pexp",
          "time": 0.0,
          "expected": 1.0000029723884385,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsx_raw",
          "time": 0.0,
          "expected": 0.20000061434596286,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsx",
          "time": 0.0,
          "expected": 0.20000061434596286,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsx_clamp",
          "time": 0.0,
          "expected": 0.0,
          "tolerance": {
            "abs": 0.0,
            "rel": 0.0
          }
        },
        {
          "variable": "wet",
          "time": 0.0,
          "expected": 0.0,
          "tolerance": {
            "abs": 0.0,
            "rel": 0.0
          }
        },
        {
          "variable": "dw",
          "time": 0.0,
          "expected": 0.0,
          "tolerance": {
            "rel": 1e-09,
            "abs": 1e-30
          }
        },
        {
          "variable": "sndens_wet",
          "time": 0.0,
          "expected": 0.20000061434596286,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "sndens_out",
          "time": 0.0,
          "expected": 0.20000061434596286,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "snowh_out",
          "time": 0.0,
          "expected": 0.30000149154021566,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsndens_dt",
          "time": 0.0,
          "expected": 9.907962446265303e-09,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "pack_present",
          "time": 0.0,
          "expected": 1.0,
          "tolerance": {
            "abs": 0.0,
            "rel": 0.0
          }
        }
      ]
    },
    {
      "id": "cold_wrfdt_caseB_step26",
      "description": "Reference: WRF 4.8 em_scm_xy COLD-SEASON single-column run, data/eqwefic/noahcold_scm, dumped by ctessum-claude/WRF branch earthsciml-instrumented-noahcold and replayed through the real64 (-fdefault-real-8) kernels/noah_driver at ESM_DT = 60 s, WRF's OWN step; tolerances are the 1e-9 that a double-precision kernel build earns. WHY THE LARGEST STEP AND NOT THE SMALLEST. SNOWPACK is not a rate: Koren's solution is a closed form in the step length, so the density increment is proportional to dtsec and there is no dt -> 0 limit to approach by shrinking the step. Nothing here is formed as a difference either -- the increment is written as sndens_in pexp_m1 + dw (1 - dsx), an exact rearrangement in which nothing cancels, and its reference comes from a publisher that emits the Taylor series BEFORE WRF adds one to it (FORTRAN_BUGS N99) -- so there is no cancellation floor pushing the step DOWN either. What is left is that a smaller step puts LESS of the assertion's budget on the physics: at 60 s the compaction and retention move the density by 4.5e-9 to 9.4e-5 of itself, so a rel 1e-9 assertion on sndens_out tests them to between 1e-5 and 2e-1 of themselves, while at 1e-6 s the same assertion would test them to nothing at all -- the increment is 7.5e-17 of the density there. This is the fifth distinct direction the Noah reference step has had to move (surface closure 1e-6 s, tendencies 1e-2 s, runoff 1e-4 s, canopy drip 1e-2 s, compaction 60 s), and it is why the choice is argued per quantity rather than assumed. The companion regime at 1e-6 s pins the tendency in the dt -> 0 limit. Step 26 of case B, a 30 cm pack melting under warm advection. dry branch -- the snow surface is below freezing and only compaction acts; neither density limit fires. The density moves by 3.609e-06 of itself over the step.",
      "parameter_overrides": {
        "esd": 0.06000476640269228,
        "sndens_in": 0.19999847116775002,
        "tsnow": 264.7685973715716,
        "tsoil": 263.2526378054602,
        "dtsec": 60.0
      },
      "time_span": {
        "start": 0.0,
        "end": 0.0
      },
      "assertions": [
        {
          "variable": "tavgc",
          "time": 0.0,
          "expected": -9.139382411484064,
          "tolerance": {
            "rel": 1e-09,
            "abs": 1e-12
          }
        },
        {
          "variable": "esdcx",
          "time": 0.0,
          "expected": 6.000476640269229,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "comp_exponent",
          "time": 0.0,
          "expected": -4.931118487441475,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "bfac",
          "time": 0.0,
          "expected": 1.2030708411257687e-06,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "xarg",
          "time": 0.0,
          "expected": 7.2189984787642275e-06,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "pexp_m1",
          "time": 0.0,
          "expected": 3.609507925054295e-06,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "pexp",
          "time": 0.0,
          "expected": 1.000003609507925,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsx_raw",
          "time": 0.0,
          "expected": 0.1999991930638167,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsx",
          "time": 0.0,
          "expected": 0.1999991930638167,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsx_clamp",
          "time": 0.0,
          "expected": 0.0,
          "tolerance": {
            "abs": 0.0,
            "rel": 0.0
          }
        },
        {
          "variable": "wet",
          "time": 0.0,
          "expected": 0.0,
          "tolerance": {
            "abs": 0.0,
            "rel": 0.0
          }
        },
        {
          "variable": "dw",
          "time": 0.0,
          "expected": 0.0,
          "tolerance": {
            "rel": 1e-09,
            "abs": 1e-30
          }
        },
        {
          "variable": "sndens_wet",
          "time": 0.0,
          "expected": 0.1999991930638167,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "sndens_out",
          "time": 0.0,
          "expected": 0.1999991930638167,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "snowh_out",
          "time": 0.0,
          "expected": 0.300025042518775,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsndens_dt",
          "time": 0.0,
          "expected": 1.2031601111312277e-08,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "pack_present",
          "time": 0.0,
          "expected": 1.0,
          "tolerance": {
            "abs": 0.0,
            "rel": 0.0
          }
        }
      ]
    },
    {
      "id": "cold_wrfdt_caseB_step51",
      "description": "Reference: WRF 4.8 em_scm_xy COLD-SEASON single-column run, data/eqwefic/noahcold_scm, dumped by ctessum-claude/WRF branch earthsciml-instrumented-noahcold and replayed through the real64 (-fdefault-real-8) kernels/noah_driver at ESM_DT = 60 s, WRF's OWN step; tolerances are the 1e-9 that a double-precision kernel build earns. WHY THE LARGEST STEP AND NOT THE SMALLEST. SNOWPACK is not a rate: Koren's solution is a closed form in the step length, so the density increment is proportional to dtsec and there is no dt -> 0 limit to approach by shrinking the step. Nothing here is formed as a difference either -- the increment is written as sndens_in pexp_m1 + dw (1 - dsx), an exact rearrangement in which nothing cancels, and its reference comes from a publisher that emits the Taylor series BEFORE WRF adds one to it (FORTRAN_BUGS N99) -- so there is no cancellation floor pushing the step DOWN either. What is left is that a smaller step puts LESS of the assertion's budget on the physics: at 60 s the compaction and retention move the density by 4.5e-9 to 9.4e-5 of itself, so a rel 1e-9 assertion on sndens_out tests them to between 1e-5 and 2e-1 of themselves, while at 1e-6 s the same assertion would test them to nothing at all -- the increment is 7.5e-17 of the density there. This is the fifth distinct direction the Noah reference step has had to move (surface closure 1e-6 s, tendencies 1e-2 s, runoff 1e-4 s, canopy drip 1e-2 s, compaction 60 s), and it is why the choice is argued per quantity rather than assumed. The companion regime at 1e-6 s pins the tendency in the dt -> 0 limit. Step 51 of case B, a 30 cm pack melting under warm advection. dry branch -- the snow surface is below freezing and only compaction acts; neither density limit fires. The density moves by 3.659e-06 of itself over the step.",
      "parameter_overrides": {
        "esd": 0.06008747665524631,
        "sndens_in": 0.1995954649678384,
        "tsnow": 264.84977564392454,
        "tsoil": 263.26346595836634,
        "dtsec": 60.0
      },
      "time_span": {
        "start": 0.0,
        "end": 0.0
      },
      "assertions": [
        {
          "variable": "tavgc",
          "time": 0.0,
          "expected": -9.093379198854535,
          "tolerance": {
            "rel": 1e-09,
            "abs": 1e-12
          }
        },
        {
          "variable": "esdcx",
          "time": 0.0,
          "expected": 6.008747665524631,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "comp_exponent",
          "time": 0.0,
          "expected": -4.918975100232969,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "bfac",
          "time": 0.0,
          "expected": 1.217769259862967e-06,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "xarg",
          "time": 0.0,
          "expected": 7.31726819734926e-06,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "pexp_m1",
          "time": 0.0,
          "expected": 3.658643022426599e-06,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "pexp",
          "time": 0.0,
          "expected": 1.0000036586430223,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsx_raw",
          "time": 0.0,
          "expected": 0.19959619521639357,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsx",
          "time": 0.0,
          "expected": 0.19959619521639357,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsx_clamp",
          "time": 0.0,
          "expected": 0.0,
          "tolerance": {
            "abs": 0.0,
            "rel": 0.0
          }
        },
        {
          "variable": "wet",
          "time": 0.0,
          "expected": 0.0,
          "tolerance": {
            "abs": 0.0,
            "rel": 0.0
          }
        },
        {
          "variable": "dw",
          "time": 0.0,
          "expected": 0.0,
          "tolerance": {
            "rel": 1e-09,
            "abs": 1e-30
          }
        },
        {
          "variable": "sndens_wet",
          "time": 0.0,
          "expected": 0.19959619521639357,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "sndens_out",
          "time": 0.0,
          "expected": 0.19959619521639357,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "snowh_out",
          "time": 0.0,
          "expected": 0.3010452007369282,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsndens_dt",
          "time": 0.0,
          "expected": 1.217080925354291e-08,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "pack_present",
          "time": 0.0,
          "expected": 1.0,
          "tolerance": {
            "abs": 0.0,
            "rel": 0.0
          }
        }
      ]
    },
    {
      "id": "cold_wrfdt_caseB_step100",
      "description": "Reference: WRF 4.8 em_scm_xy COLD-SEASON single-column run, data/eqwefic/noahcold_scm, dumped by ctessum-claude/WRF branch earthsciml-instrumented-noahcold and replayed through the real64 (-fdefault-real-8) kernels/noah_driver at ESM_DT = 60 s, WRF's OWN step; tolerances are the 1e-9 that a double-precision kernel build earns. WHY THE LARGEST STEP AND NOT THE SMALLEST. SNOWPACK is not a rate: Koren's solution is a closed form in the step length, so the density increment is proportional to dtsec and there is no dt -> 0 limit to approach by shrinking the step. Nothing here is formed as a difference either -- the increment is written as sndens_in pexp_m1 + dw (1 - dsx), an exact rearrangement in which nothing cancels, and its reference comes from a publisher that emits the Taylor series BEFORE WRF adds one to it (FORTRAN_BUGS N99) -- so there is no cancellation floor pushing the step DOWN either. What is left is that a smaller step puts LESS of the assertion's budget on the physics: at 60 s the compaction and retention move the density by 4.5e-9 to 9.4e-5 of itself, so a rel 1e-9 assertion on sndens_out tests them to between 1e-5 and 2e-1 of themselves, while at 1e-6 s the same assertion would test them to nothing at all -- the increment is 7.5e-17 of the density there. This is the fifth distinct direction the Noah reference step has had to move (surface closure 1e-6 s, tendencies 1e-2 s, runoff 1e-4 s, canopy drip 1e-2 s, compaction 60 s), and it is why the choice is argued per quantity rather than assumed. The companion regime at 1e-6 s pins the tendency in the dt -> 0 limit. Step 100 of case B, a 30 cm pack melting under warm advection. dry branch -- the snow surface is below freezing and only compaction acts; neither density limit fires. The density moves by 3.746e-06 of itself over the step.",
      "parameter_overrides": {
        "esd": 0.06030984480044616,
        "sndens_in": 0.19851568117713758,
        "tsnow": 264.7820722282615,
        "tsoil": 263.26078040450875,
        "dtsec": 60.0
      },
      "time_span": {
        "start": 0.0,
        "end": 0.0
      },
      "assertions": [
        {
          "variable": "tavgc",
          "time": 0.0,
          "expected": -9.12857368361486,
          "tolerance": {
            "rel": 1e-09,
            "abs": 1e-12
          }
        },
        {
          "variable": "esdcx",
          "time": 0.0,
          "expected": 6.030984480044617,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "comp_exponent",
          "time": 0.0,
          "expected": -4.899115199409078,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "bfac",
          "time": 0.0,
          "expected": 1.2421957879577893e-06,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "xarg",
          "time": 0.0,
          "expected": 7.491663518350221e-06,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "pexp_m1",
          "time": 0.0,
          "expected": 3.7458411133630084e-06,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "pexp",
          "time": 0.0,
          "expected": 1.0000037458411133,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsx_raw",
          "time": 0.0,
          "expected": 0.19851642478533776,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsx",
          "time": 0.0,
          "expected": 0.19851642478533776,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsx_clamp",
          "time": 0.0,
          "expected": 0.0,
          "tolerance": {
            "abs": 0.0,
            "rel": 0.0
          }
        },
        {
          "variable": "wet",
          "time": 0.0,
          "expected": 0.0,
          "tolerance": {
            "abs": 0.0,
            "rel": 0.0
          }
        },
        {
          "variable": "dw",
          "time": 0.0,
          "expected": 0.0,
          "tolerance": {
            "rel": 1e-09,
            "abs": 1e-30
          }
        },
        {
          "variable": "sndens_wet",
          "time": 0.0,
          "expected": 0.19851642478533776,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "sndens_out",
          "time": 0.0,
          "expected": 0.19851642478533776,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "snowh_out",
          "time": 0.0,
          "expected": 0.3038027954899004,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsndens_dt",
          "time": 0.0,
          "expected": 1.2393470003343085e-08,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "pack_present",
          "time": 0.0,
          "expected": 1.0,
          "tolerance": {
            "abs": 0.0,
            "rel": 0.0
          }
        }
      ]
    },
    {
      "id": "cold_wrfdt_caseB_step435",
      "description": "Reference: WRF 4.8 em_scm_xy COLD-SEASON single-column run, data/eqwefic/noahcold_scm, dumped by ctessum-claude/WRF branch earthsciml-instrumented-noahcold and replayed through the real64 (-fdefault-real-8) kernels/noah_driver at ESM_DT = 60 s, WRF's OWN step; tolerances are the 1e-9 that a double-precision kernel build earns. WHY THE LARGEST STEP AND NOT THE SMALLEST. SNOWPACK is not a rate: Koren's solution is a closed form in the step length, so the density increment is proportional to dtsec and there is no dt -> 0 limit to approach by shrinking the step. Nothing here is formed as a difference either -- the increment is written as sndens_in pexp_m1 + dw (1 - dsx), an exact rearrangement in which nothing cancels, and its reference comes from a publisher that emits the Taylor series BEFORE WRF adds one to it (FORTRAN_BUGS N99) -- so there is no cancellation floor pushing the step DOWN either. What is left is that a smaller step puts LESS of the assertion's budget on the physics: at 60 s the compaction and retention move the density by 4.5e-9 to 9.4e-5 of itself, so a rel 1e-9 assertion on sndens_out tests them to between 1e-5 and 2e-1 of themselves, while at 1e-6 s the same assertion would test them to nothing at all -- the increment is 7.5e-17 of the density there. This is the fifth distinct direction the Noah reference step has had to move (surface closure 1e-6 s, tendencies 1e-2 s, runoff 1e-4 s, canopy drip 1e-2 s, compaction 60 s), and it is why the choice is argued per quantity rather than assumed. The companion regime at 1e-6 s pins the tendency in the dt -> 0 limit. Step 435 of case B, a 30 cm pack melting under warm advection. dry branch -- the snow surface is below freezing and only compaction acts; neither density limit fires. The density moves by 5.264e-06 of itself over the step.",
      "parameter_overrides": {
        "esd": 0.060749940300446455,
        "sndens_in": 0.19596263353876026,
        "tsnow": 270.63755408553794,
        "tsoil": 264.3878774540648,
        "dtsec": 60.0
      },
      "time_span": {
        "start": 0.0,
        "end": 0.0
      },
      "assertions": [
        {
          "variable": "tavgc",
          "time": 0.0,
          "expected": -5.637284230198617,
          "tolerance": {
            "rel": 1e-09,
            "abs": 1e-12
          }
        },
        {
          "variable": "esdcx",
          "time": 0.0,
          "expected": 6.074994030044645,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "comp_exponent",
          "time": 0.0,
          "expected": -4.566198042729854,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "bfac",
          "time": 0.0,
          "expected": 1.7329025323928804e-06,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "xarg",
          "time": 0.0,
          "expected": 1.0527372538935995e-05,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "pexp_m1",
          "time": 0.0,
          "expected": 5.263704740445373e-06,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "pexp",
          "time": 0.0,
          "expected": 1.0000052637047405,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsx_raw",
          "time": 0.0,
          "expected": 0.19596366502820337,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsx",
          "time": 0.0,
          "expected": 0.19596366502820337,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsx_clamp",
          "time": 0.0,
          "expected": 0.0,
          "tolerance": {
            "abs": 0.0,
            "rel": 0.0
          }
        },
        {
          "variable": "wet",
          "time": 0.0,
          "expected": 0.0,
          "tolerance": {
            "abs": 0.0,
            "rel": 0.0
          }
        },
        {
          "variable": "dw",
          "time": 0.0,
          "expected": 0.0,
          "tolerance": {
            "rel": 1e-09,
            "abs": 1e-30
          }
        },
        {
          "variable": "sndens_wet",
          "time": 0.0,
          "expected": 0.19596366502820337,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "sndens_out",
          "time": 0.0,
          "expected": 0.19596366502820337,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "snowh_out",
          "time": 0.0,
          "expected": 0.3100061447192429,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsndens_dt",
          "time": 0.0,
          "expected": 1.7191490718468866e-08,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "pack_present",
          "time": 0.0,
          "expected": 1.0,
          "tolerance": {
            "abs": 0.0,
            "rel": 0.0
          }
        }
      ]
    },
    {
      "id": "cold_wrfdt_caseB_step463",
      "description": "Reference: WRF 4.8 em_scm_xy COLD-SEASON single-column run, data/eqwefic/noahcold_scm, dumped by ctessum-claude/WRF branch earthsciml-instrumented-noahcold and replayed through the real64 (-fdefault-real-8) kernels/noah_driver at ESM_DT = 60 s, WRF's OWN step; tolerances are the 1e-9 that a double-precision kernel build earns. WHY THE LARGEST STEP AND NOT THE SMALLEST. SNOWPACK is not a rate: Koren's solution is a closed form in the step length, so the density increment is proportional to dtsec and there is no dt -> 0 limit to approach by shrinking the step. Nothing here is formed as a difference either -- the increment is written as sndens_in pexp_m1 + dw (1 - dsx), an exact rearrangement in which nothing cancels, and its reference comes from a publisher that emits the Taylor series BEFORE WRF adds one to it (FORTRAN_BUGS N99) -- so there is no cancellation floor pushing the step DOWN either. What is left is that a smaller step puts LESS of the assertion's budget on the physics: at 60 s the compaction and retention move the density by 4.5e-9 to 9.4e-5 of itself, so a rel 1e-9 assertion on sndens_out tests them to between 1e-5 and 2e-1 of themselves, while at 1e-6 s the same assertion would test them to nothing at all -- the increment is 7.5e-17 of the density there. This is the fifth distinct direction the Noah reference step has had to move (surface closure 1e-6 s, tendencies 1e-2 s, runoff 1e-4 s, canopy drip 1e-2 s, compaction 60 s), and it is why the choice is argued per quantity rather than assumed. The companion regime at 1e-6 s pins the tendency in the dt -> 0 limit. Step 463 of case B, a 30 cm pack melting under warm advection. dry branch -- the snow surface is below freezing and only compaction acts; neither density limit fires. The density moves by 5.537e-06 of itself over the step.",
      "parameter_overrides": {
        "esd": 0.06077183292069061,
        "sndens_in": 0.19575512985495824,
        "tsnow": 271.6147926187266,
        "tsoil": 264.55902159623395,
        "dtsec": 60.0
      },
      "time_span": {
        "start": 0.0,
        "end": 0.0
      },
      "assertions": [
        {
          "variable": "tavgc",
          "time": 0.0,
          "expected": -5.063092892519705,
          "tolerance": {
            "rel": 1e-09,
            "abs": 1e-12
          }
        },
        {
          "variable": "esdcx",
          "time": 0.0,
          "expected": 6.077183292069061,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "comp_exponent",
          "time": 0.0,
          "expected": -4.515905158355699,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "bfac",
          "time": 0.0,
          "expected": 1.8222839854948012e-06,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "xarg",
          "time": 0.0,
          "expected": 1.1074353790054025e-05,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "pexp_m1",
          "time": 0.0,
          "expected": 5.537197335302247e-06,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "pexp",
          "time": 0.0,
          "expected": 1.0000055371973353,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsx_raw",
          "time": 0.0,
          "expected": 0.19575621378974165,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsx",
          "time": 0.0,
          "expected": 0.19575621378974165,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsx_clamp",
          "time": 0.0,
          "expected": 0.0,
          "tolerance": {
            "abs": 0.0,
            "rel": 0.0
          }
        },
        {
          "variable": "wet",
          "time": 0.0,
          "expected": 0.0,
          "tolerance": {
            "abs": 0.0,
            "rel": 0.0
          }
        },
        {
          "variable": "dw",
          "time": 0.0,
          "expected": 0.0,
          "tolerance": {
            "rel": 1e-09,
            "abs": 1e-30
          }
        },
        {
          "variable": "sndens_wet",
          "time": 0.0,
          "expected": 0.19575621378974165,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "sndens_out",
          "time": 0.0,
          "expected": 0.19575621378974165,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "snowh_out",
          "time": 0.0,
          "expected": 0.3104465076442712,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsndens_dt",
          "time": 0.0,
          "expected": 1.8065579723410337e-08,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "pack_present",
          "time": 0.0,
          "expected": 1.0,
          "tolerance": {
            "abs": 0.0,
            "rel": 0.0
          }
        }
      ]
    },
    {
      "id": "cold_wrfdt_caseB_step492",
      "description": "Reference: WRF 4.8 em_scm_xy COLD-SEASON single-column run, data/eqwefic/noahcold_scm, dumped by ctessum-claude/WRF branch earthsciml-instrumented-noahcold and replayed through the real64 (-fdefault-real-8) kernels/noah_driver at ESM_DT = 60 s, WRF's OWN step; tolerances are the 1e-9 that a double-precision kernel build earns. WHY THE LARGEST STEP AND NOT THE SMALLEST. SNOWPACK is not a rate: Koren's solution is a closed form in the step length, so the density increment is proportional to dtsec and there is no dt -> 0 limit to approach by shrinking the step. Nothing here is formed as a difference either -- the increment is written as sndens_in pexp_m1 + dw (1 - dsx), an exact rearrangement in which nothing cancels, and its reference comes from a publisher that emits the Taylor series BEFORE WRF adds one to it (FORTRAN_BUGS N99) -- so there is no cancellation floor pushing the step DOWN either. What is left is that a smaller step puts LESS of the assertion's budget on the physics: at 60 s the compaction and retention move the density by 4.5e-9 to 9.4e-5 of itself, so a rel 1e-9 assertion on sndens_out tests them to between 1e-5 and 2e-1 of themselves, while at 1e-6 s the same assertion would test them to nothing at all -- the increment is 7.5e-17 of the density there. This is the fifth distinct direction the Noah reference step has had to move (surface closure 1e-6 s, tendencies 1e-2 s, runoff 1e-4 s, canopy drip 1e-2 s, compaction 60 s), and it is why the choice is argued per quantity rather than assumed. The companion regime at 1e-6 s pins the tendency in the dt -> 0 limit. Step 492 of case B, a 30 cm pack melting under warm advection. dry branch -- the snow surface is below freezing and only compaction acts; neither density limit fires. The density moves by 5.436e-06 of itself over the step.",
      "parameter_overrides": {
        "esd": 0.06078525119373602,
        "sndens_in": 0.19557730896860487,
        "tsnow": 271.0819585036157,
        "tsoil": 264.5298731537984,
        "dtsec": 60.0
      },
      "time_span": {
        "start": 0.0,
        "end": 0.0
      },
      "assertions": [
        {
          "variable": "tavgc",
          "time": 0.0,
          "expected": -5.344084171292934,
          "tolerance": {
            "rel": 1e-09,
            "abs": 1e-12
          }
        },
        {
          "variable": "esdcx",
          "time": 0.0,
          "expected": 6.078525119373602,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "comp_exponent",
          "time": 0.0,
          "expected": -4.534650222044137,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "bfac",
          "time": 0.0,
          "expected": 1.788443319743658e-06,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "xarg",
          "time": 0.0,
          "expected": 1.087109764363774e-05,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "pexp_m1",
          "time": 0.0,
          "expected": 5.4355685186663986e-06,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "pexp",
          "time": 0.0,
          "expected": 1.0000054355685186,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsx_raw",
          "time": 0.0,
          "expected": 0.19557837204246845,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsx",
          "time": 0.0,
          "expected": 0.19557837204246845,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsx_clamp",
          "time": 0.0,
          "expected": 0.0,
          "tolerance": {
            "abs": 0.0,
            "rel": 0.0
          }
        },
        {
          "variable": "wet",
          "time": 0.0,
          "expected": 0.0,
          "tolerance": {
            "abs": 0.0,
            "rel": 0.0
          }
        },
        {
          "variable": "dw",
          "time": 0.0,
          "expected": 0.0,
          "tolerance": {
            "rel": 1e-09,
            "abs": 1e-30
          }
        },
        {
          "variable": "sndens_wet",
          "time": 0.0,
          "expected": 0.19557837204246845,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "sndens_out",
          "time": 0.0,
          "expected": 0.19557837204246845,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "snowh_out",
          "time": 0.0,
          "expected": 0.3107974085219246,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsndens_dt",
          "time": 0.0,
          "expected": 1.7717897726587334e-08,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "pack_present",
          "time": 0.0,
          "expected": 1.0,
          "tolerance": {
            "abs": 0.0,
            "rel": 0.0
          }
        }
      ]
    },
    {
      "id": "cold_wrfdt_caseB_step517",
      "description": "Reference: WRF 4.8 em_scm_xy COLD-SEASON single-column run, data/eqwefic/noahcold_scm, dumped by ctessum-claude/WRF branch earthsciml-instrumented-noahcold and replayed through the real64 (-fdefault-real-8) kernels/noah_driver at ESM_DT = 60 s, WRF's OWN step; tolerances are the 1e-9 that a double-precision kernel build earns. WHY THE LARGEST STEP AND NOT THE SMALLEST. SNOWPACK is not a rate: Koren's solution is a closed form in the step length, so the density increment is proportional to dtsec and there is no dt -> 0 limit to approach by shrinking the step. Nothing here is formed as a difference either -- the increment is written as sndens_in pexp_m1 + dw (1 - dsx), an exact rearrangement in which nothing cancels, and its reference comes from a publisher that emits the Taylor series BEFORE WRF adds one to it (FORTRAN_BUGS N99) -- so there is no cancellation floor pushing the step DOWN either. What is left is that a smaller step puts LESS of the assertion's budget on the physics: at 60 s the compaction and retention move the density by 4.5e-9 to 9.4e-5 of itself, so a rel 1e-9 assertion on sndens_out tests them to between 1e-5 and 2e-1 of themselves, while at 1e-6 s the same assertion would test them to nothing at all -- the increment is 7.5e-17 of the density there. This is the fifth distinct direction the Noah reference step has had to move (surface closure 1e-6 s, tendencies 1e-2 s, runoff 1e-4 s, canopy drip 1e-2 s, compaction 60 s), and it is why the choice is argued per quantity rather than assumed. The companion regime at 1e-6 s pins the tendency in the dt -> 0 limit. Step 517 of case B, a 30 cm pack melting under warm advection. dry branch -- the snow surface is below freezing and only compaction acts; neither density limit fires. The density moves by 5.167e-06 of itself over the step.",
      "parameter_overrides": {
        "esd": 0.060809316080667426,
        "sndens_in": 0.19539130833452667,
        "tsnow": 269.8457760240593,
        "tsoil": 264.3919624110543,
        "dtsec": 60.0
      },
      "time_span": {
        "start": 0.0,
        "end": 0.0
      },
      "assertions": [
        {
          "variable": "tavgc",
          "time": 0.0,
          "expected": -6.031130782443171,
          "tolerance": {
            "rel": 1e-09,
            "abs": 1e-12
          }
        },
        {
          "variable": "esdcx",
          "time": 0.0,
          "expected": 6.080931608066742,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "comp_exponent",
          "time": 0.0,
          "expected": -4.585707937620514,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "bfac",
          "time": 0.0,
          "expected": 1.6994214542893463e-06,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "xarg",
          "time": 0.0,
          "expected": 1.0334065636814837e-05,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "pexp_m1",
          "time": 0.0,
          "expected": 5.167050617272166e-06,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "pexp",
          "time": 0.0,
          "expected": 1.0000051670506174,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsx_raw",
          "time": 0.0,
          "expected": 0.19539231793130704,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsx",
          "time": 0.0,
          "expected": 0.19539231793130704,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsx_clamp",
          "time": 0.0,
          "expected": 0.0,
          "tolerance": {
            "abs": 0.0,
            "rel": 0.0
          }
        },
        {
          "variable": "wet",
          "time": 0.0,
          "expected": 0.0,
          "tolerance": {
            "abs": 0.0,
            "rel": 0.0
          }
        },
        {
          "variable": "dw",
          "time": 0.0,
          "expected": 0.0,
          "tolerance": {
            "rel": 1e-09,
            "abs": 1e-30
          }
        },
        {
          "variable": "sndens_wet",
          "time": 0.0,
          "expected": 0.19539231793130704,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "sndens_out",
          "time": 0.0,
          "expected": 0.19539231793130704,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "snowh_out",
          "time": 0.0,
          "expected": 0.3112165141622702,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsndens_dt",
          "time": 0.0,
          "expected": 1.682661300565887e-08,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "pack_present",
          "time": 0.0,
          "expected": 1.0,
          "tolerance": {
            "abs": 0.0,
            "rel": 0.0
          }
        }
      ]
    },
    {
      "id": "cold_wrfdt_caseB_step561",
      "description": "Reference: WRF 4.8 em_scm_xy COLD-SEASON single-column run, data/eqwefic/noahcold_scm, dumped by ctessum-claude/WRF branch earthsciml-instrumented-noahcold and replayed through the real64 (-fdefault-real-8) kernels/noah_driver at ESM_DT = 60 s, WRF's OWN step; tolerances are the 1e-9 that a double-precision kernel build earns. WHY THE LARGEST STEP AND NOT THE SMALLEST. SNOWPACK is not a rate: Koren's solution is a closed form in the step length, so the density increment is proportional to dtsec and there is no dt -> 0 limit to approach by shrinking the step. Nothing here is formed as a difference either -- the increment is written as sndens_in pexp_m1 + dw (1 - dsx), an exact rearrangement in which nothing cancels, and its reference comes from a publisher that emits the Taylor series BEFORE WRF adds one to it (FORTRAN_BUGS N99) -- so there is no cancellation floor pushing the step DOWN either. What is left is that a smaller step puts LESS of the assertion's budget on the physics: at 60 s the compaction and retention move the density by 4.5e-9 to 9.4e-5 of itself, so a rel 1e-9 assertion on sndens_out tests them to between 1e-5 and 2e-1 of themselves, while at 1e-6 s the same assertion would test them to nothing at all -- the increment is 7.5e-17 of the density there. This is the fifth distinct direction the Noah reference step has had to move (surface closure 1e-6 s, tendencies 1e-2 s, runoff 1e-4 s, canopy drip 1e-2 s, compaction 60 s), and it is why the choice is argued per quantity rather than assumed. The companion regime at 1e-6 s pins the tendency in the dt -> 0 limit. Step 561 of case B, a 30 cm pack melting under warm advection. dry branch -- the snow surface is below freezing and only compaction acts; neither density limit fires. The density moves by 5.458e-06 of itself over the step.",
      "parameter_overrides": {
        "esd": 0.06083357807912103,
        "sndens_in": 0.19514833462133443,
        "tsnow": 270.88227254108716,
        "tsoil": 264.5880081127981,
        "dtsec": 60.0
      },
      "time_span": {
        "start": 0.0,
        "end": 0.0
      },
      "assertions": [
        {
          "variable": "tavgc",
          "time": 0.0,
          "expected": -5.41485967305735,
          "tolerance": {
            "rel": 1e-09,
            "abs": 1e-12
          }
        },
        {
          "variable": "esdcx",
          "time": 0.0,
          "expected": 6.083357807912103,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "comp_exponent",
          "time": 0.0,
          "expected": -4.531303800892611,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "bfac",
          "time": 0.0,
          "expected": 1.794438229448957e-06,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "xarg",
          "time": 0.0,
          "expected": 1.0916209813934281e-05,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "pexp_m1",
          "time": 0.0,
          "expected": 5.458124767627458e-06,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "pexp",
          "time": 0.0,
          "expected": 1.0000054581247677,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsx_raw",
          "time": 0.0,
          "expected": 0.195149399765293,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsx",
          "time": 0.0,
          "expected": 0.195149399765293,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsx_clamp",
          "time": 0.0,
          "expected": 0.0,
          "tolerance": {
            "abs": 0.0,
            "rel": 0.0
          }
        },
        {
          "variable": "wet",
          "time": 0.0,
          "expected": 0.0,
          "tolerance": {
            "abs": 0.0,
            "rel": 0.0
          }
        },
        {
          "variable": "dw",
          "time": 0.0,
          "expected": 0.0,
          "tolerance": {
            "rel": 1e-09,
            "abs": 1e-30
          }
        },
        {
          "variable": "sndens_wet",
          "time": 0.0,
          "expected": 0.195149399765293,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "sndens_out",
          "time": 0.0,
          "expected": 0.195149399765293,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "snowh_out",
          "time": 0.0,
          "expected": 0.3117282356609133,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsndens_dt",
          "time": 0.0,
          "expected": 1.7752399309299276e-08,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "pack_present",
          "time": 0.0,
          "expected": 1.0,
          "tolerance": {
            "abs": 0.0,
            "rel": 0.0
          }
        }
      ]
    },
    {
      "id": "cold_wrfdt_caseB_step658",
      "description": "Reference: WRF 4.8 em_scm_xy COLD-SEASON single-column run, data/eqwefic/noahcold_scm, dumped by ctessum-claude/WRF branch earthsciml-instrumented-noahcold and replayed through the real64 (-fdefault-real-8) kernels/noah_driver at ESM_DT = 60 s, WRF's OWN step; tolerances are the 1e-9 that a double-precision kernel build earns. WHY THE LARGEST STEP AND NOT THE SMALLEST. SNOWPACK is not a rate: Koren's solution is a closed form in the step length, so the density increment is proportional to dtsec and there is no dt -> 0 limit to approach by shrinking the step. Nothing here is formed as a difference either -- the increment is written as sndens_in pexp_m1 + dw (1 - dsx), an exact rearrangement in which nothing cancels, and its reference comes from a publisher that emits the Taylor series BEFORE WRF adds one to it (FORTRAN_BUGS N99) -- so there is no cancellation floor pushing the step DOWN either. What is left is that a smaller step puts LESS of the assertion's budget on the physics: at 60 s the compaction and retention move the density by 4.5e-9 to 9.4e-5 of itself, so a rel 1e-9 assertion on sndens_out tests them to between 1e-5 and 2e-1 of themselves, while at 1e-6 s the same assertion would test them to nothing at all -- the increment is 7.5e-17 of the density there. This is the fifth distinct direction the Noah reference step has had to move (surface closure 1e-6 s, tendencies 1e-2 s, runoff 1e-4 s, canopy drip 1e-2 s, compaction 60 s), and it is why the choice is argued per quantity rather than assumed. The companion regime at 1e-6 s pins the tendency in the dt -> 0 limit. Step 658 of case B, a 30 cm pack melting under warm advection. dry branch -- the snow surface is below freezing and only compaction acts; neither density limit fires. The density moves by 3.944e-06 of itself over the step.",
      "parameter_overrides": {
        "esd": 0.060759980484990636,
        "sndens_in": 0.19523935840389106,
        "tsnow": 263.6961309154277,
        "tsoil": 263.72845004329685,
        "dtsec": 60.0
      },
      "time_span": {
        "start": 0.0,
        "end": 0.0
      },
      "assertions": [
        {
          "variable": "tavgc",
          "time": 0.0,
          "expected": -9.437709520637696,
          "tolerance": {
            "rel": 1e-09,
            "abs": 1e-12
          }
        },
        {
          "variable": "esdcx",
          "time": 0.0,
          "expected": 6.075998048499064,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "comp_exponent",
          "time": 0.0,
          "expected": -4.855043288132728,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "bfac",
          "time": 0.0,
          "expected": 1.298166029152229e-06,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "xarg",
          "time": 0.0,
          "expected": 7.887654259756722e-06,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "pexp_m1",
          "time": 0.0,
          "expected": 3.943837499080428e-06,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "pexp",
          "time": 0.0,
          "expected": 1.000003943837499,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsx_raw",
          "time": 0.0,
          "expected": 0.195240128396194,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsx",
          "time": 0.0,
          "expected": 0.195240128396194,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsx_clamp",
          "time": 0.0,
          "expected": 0.0,
          "tolerance": {
            "abs": 0.0,
            "rel": 0.0
          }
        },
        {
          "variable": "wet",
          "time": 0.0,
          "expected": 0.0,
          "tolerance": {
            "abs": 0.0,
            "rel": 0.0
          }
        },
        {
          "variable": "dw",
          "time": 0.0,
          "expected": 0.0,
          "tolerance": {
            "rel": 1e-09,
            "abs": 1e-30
          }
        },
        {
          "variable": "sndens_wet",
          "time": 0.0,
          "expected": 0.195240128396194,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "sndens_out",
          "time": 0.0,
          "expected": 0.195240128396194,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "snowh_out",
          "time": 0.0,
          "expected": 0.3112064153209965,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsndens_dt",
          "time": 0.0,
          "expected": 1.2833205049494485e-08,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "pack_present",
          "time": 0.0,
          "expected": 1.0,
          "tolerance": {
            "abs": 0.0,
            "rel": 0.0
          }
        }
      ]
    },
    {
      "id": "cold_wrfdt_caseB_step1612",
      "description": "Reference: WRF 4.8 em_scm_xy COLD-SEASON single-column run, data/eqwefic/noahcold_scm, dumped by ctessum-claude/WRF branch earthsciml-instrumented-noahcold and replayed through the real64 (-fdefault-real-8) kernels/noah_driver at ESM_DT = 60 s, WRF's OWN step; tolerances are the 1e-9 that a double-precision kernel build earns. WHY THE LARGEST STEP AND NOT THE SMALLEST. SNOWPACK is not a rate: Koren's solution is a closed form in the step length, so the density increment is proportional to dtsec and there is no dt -> 0 limit to approach by shrinking the step. Nothing here is formed as a difference either -- the increment is written as sndens_in pexp_m1 + dw (1 - dsx), an exact rearrangement in which nothing cancels, and its reference comes from a publisher that emits the Taylor series BEFORE WRF adds one to it (FORTRAN_BUGS N99) -- so there is no cancellation floor pushing the step DOWN either. What is left is that a smaller step puts LESS of the assertion's budget on the physics: at 60 s the compaction and retention move the density by 4.5e-9 to 9.4e-5 of itself, so a rel 1e-9 assertion on sndens_out tests them to between 1e-5 and 2e-1 of themselves, while at 1e-6 s the same assertion would test them to nothing at all -- the increment is 7.5e-17 of the density there. This is the fifth distinct direction the Noah reference step has had to move (surface closure 1e-6 s, tendencies 1e-2 s, runoff 1e-4 s, canopy drip 1e-2 s, compaction 60 s), and it is why the choice is argued per quantity rather than assumed. The companion regime at 1e-6 s pins the tendency in the dt -> 0 limit. Step 1612 of case B, a 30 cm pack melting under warm advection. SNOWPACK's melt-water retention branch RUNS -- the snow surface is at or above freezing; neither density limit fires. The density moves by 3.741e-04 of itself over the step.",
      "parameter_overrides": {
        "esd": 0.06071578320345706,
        "sndens_in": 0.19683397677243142,
        "tsnow": 273.335639338635,
        "tsoil": 264.8236325232234,
        "dtsec": 60.0
      },
      "time_span": {
        "start": 0.0,
        "end": 0.0
      },
      "assertions": [
        {
          "variable": "tavgc",
          "time": 0.0,
          "expected": -4.070364069070763,
          "tolerance": {
            "rel": 1e-09,
            "abs": 1e-12
          }
        },
        {
          "variable": "esdcx",
          "time": 0.0,
          "expected": 6.071578320345706,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "comp_exponent",
          "time": 0.0,
          "expected": -4.459142637746721,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "bfac",
          "time": 0.0,
          "expected": 1.928713445349671e-06,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "xarg",
          "time": 0.0,
          "expected": 1.1710334740944336e-05,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "pexp_m1",
          "time": 0.0,
          "expected": 5.855190225862369e-06,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "pexp",
          "time": 0.0,
          "expected": 1.0000058551902258,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsx_raw",
          "time": 0.0,
          "expected": 0.19683512927280833,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsx",
          "time": 0.0,
          "expected": 0.19683512927280833,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsx_clamp",
          "time": 0.0,
          "expected": 0.0,
          "tolerance": {
            "abs": 0.0,
            "rel": 0.0
          }
        },
        {
          "variable": "wet",
          "time": 0.0,
          "expected": 1.0,
          "tolerance": {
            "abs": 0.0,
            "rel": 0.0
          }
        },
        {
          "variable": "dw",
          "time": 0.0,
          "expected": 9.027777777777777e-05,
          "tolerance": {
            "rel": 1e-09,
            "abs": 1e-30
          }
        },
        {
          "variable": "sndens_wet",
          "time": 0.0,
          "expected": 0.19690763721252677,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "sndens_out",
          "time": 0.0,
          "expected": 0.19690763721252677,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "snowh_out",
          "time": 0.0,
          "expected": 0.30834651242056793,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsndens_dt",
          "time": 0.0,
          "expected": 1.2276740015890432e-06,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "pack_present",
          "time": 0.0,
          "expected": 1.0,
          "tolerance": {
            "abs": 0.0,
            "rel": 0.0
          }
        }
      ]
    },
    {
      "id": "cold_wrfdt_caseB_step1637",
      "description": "Reference: WRF 4.8 em_scm_xy COLD-SEASON single-column run, data/eqwefic/noahcold_scm, dumped by ctessum-claude/WRF branch earthsciml-instrumented-noahcold and replayed through the real64 (-fdefault-real-8) kernels/noah_driver at ESM_DT = 60 s, WRF's OWN step; tolerances are the 1e-9 that a double-precision kernel build earns. WHY THE LARGEST STEP AND NOT THE SMALLEST. SNOWPACK is not a rate: Koren's solution is a closed form in the step length, so the density increment is proportional to dtsec and there is no dt -> 0 limit to approach by shrinking the step. Nothing here is formed as a difference either -- the increment is written as sndens_in pexp_m1 + dw (1 - dsx), an exact rearrangement in which nothing cancels, and its reference comes from a publisher that emits the Taylor series BEFORE WRF adds one to it (FORTRAN_BUGS N99) -- so there is no cancellation floor pushing the step DOWN either. What is left is that a smaller step puts LESS of the assertion's budget on the physics: at 60 s the compaction and retention move the density by 4.5e-9 to 9.4e-5 of itself, so a rel 1e-9 assertion on sndens_out tests them to between 1e-5 and 2e-1 of themselves, while at 1e-6 s the same assertion would test them to nothing at all -- the increment is 7.5e-17 of the density there. This is the fifth distinct direction the Noah reference step has had to move (surface closure 1e-6 s, tendencies 1e-2 s, runoff 1e-4 s, canopy drip 1e-2 s, compaction 60 s), and it is why the choice is argued per quantity rather than assumed. The companion regime at 1e-6 s pins the tendency in the dt -> 0 limit. Step 1637 of case B, a 30 cm pack melting under warm advection. SNOWPACK's melt-water retention branch RUNS -- the snow surface is at or above freezing; neither density limit fires. The density moves by 3.697e-04 of itself over the step.",
      "parameter_overrides": {
        "esd": 0.06048071282993041,
        "sndens_in": 0.1986732035637152,
        "tsnow": 273.6358408786153,
        "tsoil": 264.8815372959442,
        "dtsec": 60.0
      },
      "time_span": {
        "start": 0.0,
        "end": 0.0
      },
      "assertions": [
        {
          "variable": "tavgc",
          "time": 0.0,
          "expected": -3.891310912720229,
          "tolerance": {
            "rel": 1e-09,
            "abs": 1e-12
          }
        },
        {
          "variable": "esdcx",
          "time": 0.0,
          "expected": 6.048071282993041,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "comp_exponent",
          "time": 0.0,
          "expected": -4.483442147855637,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "bfac",
          "time": 0.0,
          "expected": 1.8824114892021506e-06,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "xarg",
          "time": 0.0,
          "expected": 1.1384958870619691e-05,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "pexp_m1",
          "time": 0.0,
          "expected": 5.692501038252748e-06,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "pexp",
          "time": 0.0,
          "expected": 1.0000056925010383,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsx_raw",
          "time": 0.0,
          "expected": 0.19867433451113278,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsx",
          "time": 0.0,
          "expected": 0.19867433451113278,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsx_clamp",
          "time": 0.0,
          "expected": 0.0,
          "tolerance": {
            "abs": 0.0,
            "rel": 0.0
          }
        },
        {
          "variable": "wet",
          "time": 0.0,
          "expected": 1.0,
          "tolerance": {
            "abs": 0.0,
            "rel": 0.0
          }
        },
        {
          "variable": "dw",
          "time": 0.0,
          "expected": 9.027777777777777e-05,
          "tolerance": {
            "rel": 1e-09,
            "abs": 1e-30
          }
        },
        {
          "variable": "sndens_wet",
          "time": 0.0,
          "expected": 0.1987466764114894,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "sndens_out",
          "time": 0.0,
          "expected": 0.1987466764114894,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "snowh_out",
          "time": 0.0,
          "expected": 0.30431056217870955,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsndens_dt",
          "time": 0.0,
          "expected": 1.2245474629032214e-06,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "pack_present",
          "time": 0.0,
          "expected": 1.0,
          "tolerance": {
            "abs": 0.0,
            "rel": 0.0
          }
        }
      ]
    },
    {
      "id": "cold_wrfdt_caseB_step1816",
      "description": "Reference: WRF 4.8 em_scm_xy COLD-SEASON single-column run, data/eqwefic/noahcold_scm, dumped by ctessum-claude/WRF branch earthsciml-instrumented-noahcold and replayed through the real64 (-fdefault-real-8) kernels/noah_driver at ESM_DT = 60 s, WRF's OWN step; tolerances are the 1e-9 that a double-precision kernel build earns. WHY THE LARGEST STEP AND NOT THE SMALLEST. SNOWPACK is not a rate: Koren's solution is a closed form in the step length, so the density increment is proportional to dtsec and there is no dt -> 0 limit to approach by shrinking the step. Nothing here is formed as a difference either -- the increment is written as sndens_in pexp_m1 + dw (1 - dsx), an exact rearrangement in which nothing cancels, and its reference comes from a publisher that emits the Taylor series BEFORE WRF adds one to it (FORTRAN_BUGS N99) -- so there is no cancellation floor pushing the step DOWN either. What is left is that a smaller step puts LESS of the assertion's budget on the physics: at 60 s the compaction and retention move the density by 4.5e-9 to 9.4e-5 of itself, so a rel 1e-9 assertion on sndens_out tests them to between 1e-5 and 2e-1 of themselves, while at 1e-6 s the same assertion would test them to nothing at all -- the increment is 7.5e-17 of the density there. This is the fifth distinct direction the Noah reference step has had to move (surface closure 1e-6 s, tendencies 1e-2 s, runoff 1e-4 s, canopy drip 1e-2 s, compaction 60 s), and it is why the choice is argued per quantity rather than assumed. The companion regime at 1e-6 s pins the tendency in the dt -> 0 limit. Step 1816 of case B, a 30 cm pack melting under warm advection. SNOWPACK's melt-water retention branch RUNS -- the snow surface is at or above freezing; neither density limit fires. The density moves by 3.403e-04 of itself over the step.",
      "parameter_overrides": {
        "esd": 0.054306531152123604,
        "sndens_in": 0.21170159531489063,
        "tsnow": 274.907254172599,
        "tsoil": 265.7723157520076,
        "dtsec": 60.0
      },
      "time_span": {
        "start": 0.0,
        "end": 0.0
      },
      "assertions": [
        {
          "variable": "tavgc",
          "time": 0.0,
          "expected": -2.810215037696679,
          "tolerance": {
            "rel": 1e-09,
            "abs": 1e-12
          }
        },
        {
          "variable": "esdcx",
          "time": 0.0,
          "expected": 5.43065311521236,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "comp_exponent",
          "time": 0.0,
          "expected": -4.670550704628438,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "bfac",
          "time": 0.0,
          "expected": 1.561184932621332e-06,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "xarg",
          "time": 0.0,
          "expected": 8.478253817762636e-06,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "pexp_m1",
          "time": 0.0,
          "expected": 4.2391388890380095e-06,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "pexp",
          "time": 0.0,
          "expected": 1.000004239138889,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsx_raw",
          "time": 0.0,
          "expected": 0.2117024927473562,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsx",
          "time": 0.0,
          "expected": 0.2117024927473562,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsx_clamp",
          "time": 0.0,
          "expected": 0.0,
          "tolerance": {
            "abs": 0.0,
            "rel": 0.0
          }
        },
        {
          "variable": "wet",
          "time": 0.0,
          "expected": 1.0,
          "tolerance": {
            "abs": 0.0,
            "rel": 0.0
          }
        },
        {
          "variable": "dw",
          "time": 0.0,
          "expected": 9.027777777777777e-05,
          "tolerance": {
            "rel": 1e-09,
            "abs": 1e-30
          }
        },
        {
          "variable": "sndens_wet",
          "time": 0.0,
          "expected": 0.21177365849453872,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "sndens_out",
          "time": 0.0,
          "expected": 0.21177365849453872,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "snowh_out",
          "time": 0.0,
          "expected": 0.25643666704432966,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsndens_dt",
          "time": 0.0,
          "expected": 1.2010529941350179e-06,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "pack_present",
          "time": 0.0,
          "expected": 1.0,
          "tolerance": {
            "abs": 0.0,
            "rel": 0.0
          }
        }
      ]
    },
    {
      "id": "cold_wrfdt_caseB_step1841",
      "description": "Reference: WRF 4.8 em_scm_xy COLD-SEASON single-column run, data/eqwefic/noahcold_scm, dumped by ctessum-claude/WRF branch earthsciml-instrumented-noahcold and replayed through the real64 (-fdefault-real-8) kernels/noah_driver at ESM_DT = 60 s, WRF's OWN step; tolerances are the 1e-9 that a double-precision kernel build earns. WHY THE LARGEST STEP AND NOT THE SMALLEST. SNOWPACK is not a rate: Koren's solution is a closed form in the step length, so the density increment is proportional to dtsec and there is no dt -> 0 limit to approach by shrinking the step. Nothing here is formed as a difference either -- the increment is written as sndens_in pexp_m1 + dw (1 - dsx), an exact rearrangement in which nothing cancels, and its reference comes from a publisher that emits the Taylor series BEFORE WRF adds one to it (FORTRAN_BUGS N99) -- so there is no cancellation floor pushing the step DOWN either. What is left is that a smaller step puts LESS of the assertion's budget on the physics: at 60 s the compaction and retention move the density by 4.5e-9 to 9.4e-5 of itself, so a rel 1e-9 assertion on sndens_out tests them to between 1e-5 and 2e-1 of themselves, while at 1e-6 s the same assertion would test them to nothing at all -- the increment is 7.5e-17 of the density there. This is the fifth distinct direction the Noah reference step has had to move (surface closure 1e-6 s, tendencies 1e-2 s, runoff 1e-4 s, canopy drip 1e-2 s, compaction 60 s), and it is why the choice is argued per quantity rather than assumed. The companion regime at 1e-6 s pins the tendency in the dt -> 0 limit. Step 1841 of case B, a 30 cm pack melting under warm advection. SNOWPACK's melt-water retention branch RUNS -- the snow surface is at or above freezing; neither density limit fires. The density moves by 3.365e-04 of itself over the step.",
      "parameter_overrides": {
        "esd": 0.05311041437883895,
        "sndens_in": 0.21350073926848778,
        "tsnow": 274.9277857895529,
        "tsoil": 265.888653646625,
        "dtsec": 60.0
      },
      "time_span": {
        "start": 0.0,
        "end": 0.0
      },
      "assertions": [
        {
          "variable": "tavgc",
          "time": 0.0,
          "expected": -2.7417802819110193,
          "tolerance": {
            "rel": 1e-09,
            "abs": 1e-12
          }
        },
        {
          "variable": "esdcx",
          "time": 0.0,
          "expected": 5.311041437883895,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "comp_exponent",
          "time": 0.0,
          "expected": -4.702857947191125,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "bfac",
          "time": 0.0,
          "expected": 1.5115533981958177e-06,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "xarg",
          "time": 0.0,
          "expected": 8.027922733392203e-06,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "pexp_m1",
          "time": 0.0,
          "expected": 4.013972107974894e-06,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "pexp",
          "time": 0.0,
          "expected": 1.000004013972108,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsx_raw",
          "time": 0.0,
          "expected": 0.21350159625450024,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsx",
          "time": 0.0,
          "expected": 0.21350159625450024,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsx_clamp",
          "time": 0.0,
          "expected": 0.0,
          "tolerance": {
            "abs": 0.0,
            "rel": 0.0
          }
        },
        {
          "variable": "wet",
          "time": 0.0,
          "expected": 1.0,
          "tolerance": {
            "abs": 0.0,
            "rel": 0.0
          }
        },
        {
          "variable": "dw",
          "time": 0.0,
          "expected": 9.027777777777777e-05,
          "tolerance": {
            "rel": 1e-09,
            "abs": 1e-30
          }
        },
        {
          "variable": "sndens_wet",
          "time": 0.0,
          "expected": 0.21357259958261615,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "sndens_out",
          "time": 0.0,
          "expected": 0.21357259958261615,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "snowh_out",
          "time": 0.0,
          "expected": 0.24867616202936316,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsndens_dt",
          "time": 0.0,
          "expected": 1.1976719021394817e-06,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "pack_present",
          "time": 0.0,
          "expected": 1.0,
          "tolerance": {
            "abs": 0.0,
            "rel": 0.0
          }
        }
      ]
    },
    {
      "id": "cold_wrfdt_caseB_step1866",
      "description": "Reference: WRF 4.8 em_scm_xy COLD-SEASON single-column run, data/eqwefic/noahcold_scm, dumped by ctessum-claude/WRF branch earthsciml-instrumented-noahcold and replayed through the real64 (-fdefault-real-8) kernels/noah_driver at ESM_DT = 60 s, WRF's OWN step; tolerances are the 1e-9 that a double-precision kernel build earns. WHY THE LARGEST STEP AND NOT THE SMALLEST. SNOWPACK is not a rate: Koren's solution is a closed form in the step length, so the density increment is proportional to dtsec and there is no dt -> 0 limit to approach by shrinking the step. Nothing here is formed as a difference either -- the increment is written as sndens_in pexp_m1 + dw (1 - dsx), an exact rearrangement in which nothing cancels, and its reference comes from a publisher that emits the Taylor series BEFORE WRF adds one to it (FORTRAN_BUGS N99) -- so there is no cancellation floor pushing the step DOWN either. What is left is that a smaller step puts LESS of the assertion's budget on the physics: at 60 s the compaction and retention move the density by 4.5e-9 to 9.4e-5 of itself, so a rel 1e-9 assertion on sndens_out tests them to between 1e-5 and 2e-1 of themselves, while at 1e-6 s the same assertion would test them to nothing at all -- the increment is 7.5e-17 of the density there. This is the fifth distinct direction the Noah reference step has had to move (surface closure 1e-6 s, tendencies 1e-2 s, runoff 1e-4 s, canopy drip 1e-2 s, compaction 60 s), and it is why the choice is argued per quantity rather than assumed. The companion regime at 1e-6 s pins the tendency in the dt -> 0 limit. Step 1866 of case B, a 30 cm pack melting under warm advection. SNOWPACK's melt-water retention branch RUNS -- the snow surface is at or above freezing; neither density limit fires. The density moves by 3.327e-04 of itself over the step.",
      "parameter_overrides": {
        "esd": 0.0519152605105377,
        "sndens_in": 0.21529487057670674,
        "tsnow": 274.9066451558317,
        "tsoil": 265.99952240699713,
        "dtsec": 60.0
      },
      "time_span": {
        "start": 0.0,
        "end": 0.0
      },
      "assertions": [
        {
          "variable": "tavgc",
          "time": 0.0,
          "expected": -2.6969162185855566,
          "tolerance": {
            "rel": 1e-09,
            "abs": 1e-12
          }
        },
        {
          "variable": "esdcx",
          "time": 0.0,
          "expected": 5.19152605105377,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "comp_exponent",
          "time": 0.0,
          "expected": -4.736945579597686,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "bfac",
          "time": 0.0,
          "expected": 1.4608964149558936e-06,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "xarg",
          "time": 0.0,
          "expected": 7.58428179613458e-06,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "pexp_m1",
          "time": 0.0,
          "expected": 3.792150484973861e-06,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "pexp",
          "time": 0.0,
          "expected": 1.000003792150485,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsx_raw",
          "time": 0.0,
          "expected": 0.2152956870072546,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsx",
          "time": 0.0,
          "expected": 0.2152956870072546,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsx_clamp",
          "time": 0.0,
          "expected": 0.0,
          "tolerance": {
            "abs": 0.0,
            "rel": 0.0
          }
        },
        {
          "variable": "wet",
          "time": 0.0,
          "expected": 1.0,
          "tolerance": {
            "abs": 0.0,
            "rel": 0.0
          }
        },
        {
          "variable": "dw",
          "time": 0.0,
          "expected": 9.027777777777777e-05,
          "tolerance": {
            "rel": 1e-09,
            "abs": 1e-30
          }
        },
        {
          "variable": "sndens_wet",
          "time": 0.0,
          "expected": 0.21536652836884423,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "sndens_out",
          "time": 0.0,
          "expected": 0.21536652836884423,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "snowh_out",
          "time": 0.0,
          "expected": 0.24105538081398523,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsndens_dt",
          "time": 0.0,
          "expected": 1.194296535624878e-06,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "pack_present",
          "time": 0.0,
          "expected": 1.0,
          "tolerance": {
            "abs": 0.0,
            "rel": 0.0
          }
        }
      ]
    },
    {
      "id": "cold_wrfdt_caseB_step2122",
      "description": "Reference: WRF 4.8 em_scm_xy COLD-SEASON single-column run, data/eqwefic/noahcold_scm, dumped by ctessum-claude/WRF branch earthsciml-instrumented-noahcold and replayed through the real64 (-fdefault-real-8) kernels/noah_driver at ESM_DT = 60 s, WRF's OWN step; tolerances are the 1e-9 that a double-precision kernel build earns. WHY THE LARGEST STEP AND NOT THE SMALLEST. SNOWPACK is not a rate: Koren's solution is a closed form in the step length, so the density increment is proportional to dtsec and there is no dt -> 0 limit to approach by shrinking the step. Nothing here is formed as a difference either -- the increment is written as sndens_in pexp_m1 + dw (1 - dsx), an exact rearrangement in which nothing cancels, and its reference comes from a publisher that emits the Taylor series BEFORE WRF adds one to it (FORTRAN_BUGS N99) -- so there is no cancellation floor pushing the step DOWN either. What is left is that a smaller step puts LESS of the assertion's budget on the physics: at 60 s the compaction and retention move the density by 4.5e-9 to 9.4e-5 of itself, so a rel 1e-9 assertion on sndens_out tests them to between 1e-5 and 2e-1 of themselves, while at 1e-6 s the same assertion would test them to nothing at all -- the increment is 7.5e-17 of the density there. This is the fifth distinct direction the Noah reference step has had to move (surface closure 1e-6 s, tendencies 1e-2 s, runoff 1e-4 s, canopy drip 1e-2 s, compaction 60 s), and it is why the choice is argued per quantity rather than assumed. The companion regime at 1e-6 s pins the tendency in the dt -> 0 limit. Step 2122 of case B, a 30 cm pack melting under warm advection. dry branch -- the snow surface is below freezing and only compaction acts; neither density limit fires. The density moves by 1.365e-06 of itself over the step.",
      "parameter_overrides": {
        "esd": 0.045533597469329834,
        "sndens_in": 0.23088594021114783,
        "tsnow": 262.4313633423511,
        "tsoil": 264.39600206349655,
        "dtsec": 60.0
      },
      "time_span": {
        "start": 0.0,
        "end": 0.0
      },
      "assertions": [
        {
          "variable": "tavgc",
          "time": 0.0,
          "expected": -9.736317297076141,
          "tolerance": {
            "rel": 1e-09,
            "abs": 1e-12
          }
        },
        {
          "variable": "esdcx",
          "time": 0.0,
          "expected": 4.553359746932983,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "comp_exponent",
          "time": 0.0,
          "expected": -5.627510128200196,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "bfac",
          "time": 0.0,
          "expected": 5.995869254493754e-07,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "xarg",
          "time": 0.0,
          "expected": 2.7301349711284935e-06,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "pexp_m1",
          "time": 0.0,
          "expected": 1.3650687278379214e-06,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "pexp",
          "time": 0.0,
          "expected": 1.0000013650687278,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsx_raw",
          "time": 0.0,
          "expected": 0.2308862553863245,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsx",
          "time": 0.0,
          "expected": 0.2308862553863245,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsx_clamp",
          "time": 0.0,
          "expected": 0.0,
          "tolerance": {
            "abs": 0.0,
            "rel": 0.0
          }
        },
        {
          "variable": "wet",
          "time": 0.0,
          "expected": 0.0,
          "tolerance": {
            "abs": 0.0,
            "rel": 0.0
          }
        },
        {
          "variable": "dw",
          "time": 0.0,
          "expected": 0.0,
          "tolerance": {
            "rel": 1e-09,
            "abs": 1e-30
          }
        },
        {
          "variable": "sndens_wet",
          "time": 0.0,
          "expected": 0.2308862553863245,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "sndens_out",
          "time": 0.0,
          "expected": 0.2308862553863245,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "snowh_out",
          "time": 0.0,
          "expected": 0.19721224805323256,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsndens_dt",
          "time": 0.0,
          "expected": 5.252919611328233e-09,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "pack_present",
          "time": 0.0,
          "expected": 1.0,
          "tolerance": {
            "abs": 0.0,
            "rel": 0.0
          }
        }
      ]
    },
    {
      "id": "cold_wrfdt_caseB_step2880",
      "description": "Reference: WRF 4.8 em_scm_xy COLD-SEASON single-column run, data/eqwefic/noahcold_scm, dumped by ctessum-claude/WRF branch earthsciml-instrumented-noahcold and replayed through the real64 (-fdefault-real-8) kernels/noah_driver at ESM_DT = 60 s, WRF's OWN step; tolerances are the 1e-9 that a double-precision kernel build earns. WHY THE LARGEST STEP AND NOT THE SMALLEST. SNOWPACK is not a rate: Koren's solution is a closed form in the step length, so the density increment is proportional to dtsec and there is no dt -> 0 limit to approach by shrinking the step. Nothing here is formed as a difference either -- the increment is written as sndens_in pexp_m1 + dw (1 - dsx), an exact rearrangement in which nothing cancels, and its reference comes from a publisher that emits the Taylor series BEFORE WRF adds one to it (FORTRAN_BUGS N99) -- so there is no cancellation floor pushing the step DOWN either. What is left is that a smaller step puts LESS of the assertion's budget on the physics: at 60 s the compaction and retention move the density by 4.5e-9 to 9.4e-5 of itself, so a rel 1e-9 assertion on sndens_out tests them to between 1e-5 and 2e-1 of themselves, while at 1e-6 s the same assertion would test them to nothing at all -- the increment is 7.5e-17 of the density there. This is the fifth distinct direction the Noah reference step has had to move (surface closure 1e-6 s, tendencies 1e-2 s, runoff 1e-4 s, canopy drip 1e-2 s, compaction 60 s), and it is why the choice is argued per quantity rather than assumed. The companion regime at 1e-6 s pins the tendency in the dt -> 0 limit. Step 2880 of case B, a 30 cm pack melting under warm advection. dry branch -- the snow surface is below freezing and only compaction acts; neither density limit fires. The density moves by 1.000e-06 of itself over the step.",
      "parameter_overrides": {
        "esd": 0.045533597469329834,
        "sndens_in": 0.23105834304297074,
        "tsnow": 256.04450233726726,
        "tsoil": 263.10316061250694,
        "dtsec": 60.0
      },
      "time_span": {
        "start": 0.0,
        "end": 0.0
      },
      "assertions": [
        {
          "variable": "tavgc",
          "time": 0.0,
          "expected": -13.576168525112877,
          "tolerance": {
            "rel": 1e-09,
            "abs": 1e-12
          }
        },
        {
          "variable": "esdcx",
          "time": 0.0,
          "expected": 4.553359746932983,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "comp_exponent",
          "time": 0.0,
          "expected": -5.938318685911415,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "bfac",
          "time": 0.0,
          "expected": 4.394097736276611e-07,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "xarg",
          "time": 0.0,
          "expected": 2.0007907756451266e-06,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "pexp_m1",
          "time": 0.0,
          "expected": 1.0003960550168515e-06,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "pexp",
          "time": 0.0,
          "expected": 1.000001000396055,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsx_raw",
          "time": 0.0,
          "expected": 0.2310585741928256,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsx",
          "time": 0.0,
          "expected": 0.2310585741928256,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsx_clamp",
          "time": 0.0,
          "expected": 0.0,
          "tolerance": {
            "abs": 0.0,
            "rel": 0.0
          }
        },
        {
          "variable": "wet",
          "time": 0.0,
          "expected": 0.0,
          "tolerance": {
            "abs": 0.0,
            "rel": 0.0
          }
        },
        {
          "variable": "dw",
          "time": 0.0,
          "expected": 0.0,
          "tolerance": {
            "rel": 1e-09,
            "abs": 1e-30
          }
        },
        {
          "variable": "sndens_wet",
          "time": 0.0,
          "expected": 0.2310585741928256,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "sndens_out",
          "time": 0.0,
          "expected": 0.2310585741928256,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "snowh_out",
          "time": 0.0,
          "expected": 0.1970651711514961,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsndens_dt",
          "time": 0.0,
          "expected": 3.852497580981971e-09,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "pack_present",
          "time": 0.0,
          "expected": 1.0,
          "tolerance": {
            "abs": 0.0,
            "rel": 0.0
          }
        }
      ]
    },
    {
      "id": "cold_wrfdt_caseC_step1",
      "description": "Reference: WRF 4.8 em_scm_xy COLD-SEASON single-column run, data/eqwefic/noahcold_scm, dumped by ctessum-claude/WRF branch earthsciml-instrumented-noahcold and replayed through the real64 (-fdefault-real-8) kernels/noah_driver at ESM_DT = 60 s, WRF's OWN step; tolerances are the 1e-9 that a double-precision kernel build earns. WHY THE LARGEST STEP AND NOT THE SMALLEST. SNOWPACK is not a rate: Koren's solution is a closed form in the step length, so the density increment is proportional to dtsec and there is no dt -> 0 limit to approach by shrinking the step. Nothing here is formed as a difference either -- the increment is written as sndens_in pexp_m1 + dw (1 - dsx), an exact rearrangement in which nothing cancels, and its reference comes from a publisher that emits the Taylor series BEFORE WRF adds one to it (FORTRAN_BUGS N99) -- so there is no cancellation floor pushing the step DOWN either. What is left is that a smaller step puts LESS of the assertion's budget on the physics: at 60 s the compaction and retention move the density by 4.5e-9 to 9.4e-5 of itself, so a rel 1e-9 assertion on sndens_out tests them to between 1e-5 and 2e-1 of themselves, while at 1e-6 s the same assertion would test them to nothing at all -- the increment is 7.5e-17 of the density there. This is the fifth distinct direction the Noah reference step has had to move (surface closure 1e-6 s, tendencies 1e-2 s, runoff 1e-4 s, canopy drip 1e-2 s, compaction 60 s), and it is why the choice is argued per quantity rather than assumed. The companion regime at 1e-6 s pins the tendency in the dt -> 0 limit. Step 1 of case C, a thin pack over soil that thaws and refreezes. dry branch -- the snow surface is below freezing and only compaction acts; neither density limit fires. The density moves by 3.271e-07 of itself over the step.",
      "parameter_overrides": {
        "esd": 0.005000544777839628,
        "sndens_in": 0.20000002607703365,
        "tsnow": 263.66899138754906,
        "tsoil": 266.44599412043084,
        "dtsec": 60.0
      },
      "time_span": {
        "start": 0.0,
        "end": 0.0
      },
      "assertions": [
        {
          "variable": "tavgc",
          "time": 0.0,
          "expected": -8.09250724601003,
          "tolerance": {
            "rel": 1e-09,
            "abs": 1e-12
          }
        },
        {
          "variable": "esdcx",
          "time": 0.0,
          "expected": 0.5000544777839627,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "comp_exponent",
          "time": 0.0,
          "expected": -4.847401127298509,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "bfac",
          "time": 0.0,
          "expected": 1.3081248276381199e-06,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "xarg",
          "time": 0.0,
          "expected": 6.541336775608163e-07,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "pexp_m1",
          "time": 0.0,
          "expected": 3.270669100955645e-07,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "pexp",
          "time": 0.0,
          "expected": 1.00000032706691,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsx_raw",
          "time": 0.0,
          "expected": 0.2000000914904242,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsx",
          "time": 0.0,
          "expected": 0.2000000914904242,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsx_clamp",
          "time": 0.0,
          "expected": 0.0,
          "tolerance": {
            "abs": 0.0,
            "rel": 0.0
          }
        },
        {
          "variable": "wet",
          "time": 0.0,
          "expected": 0.0,
          "tolerance": {
            "abs": 0.0,
            "rel": 0.0
          }
        },
        {
          "variable": "dw",
          "time": 0.0,
          "expected": 0.0,
          "tolerance": {
            "rel": 1e-09,
            "abs": 1e-30
          }
        },
        {
          "variable": "sndens_wet",
          "time": 0.0,
          "expected": 0.2000000914904242,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "sndens_out",
          "time": 0.0,
          "expected": 0.2000000914904242,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "snowh_out",
          "time": 0.0,
          "expected": 0.025002712451654294,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsndens_dt",
          "time": 0.0,
          "expected": 1.0902231758007954e-09,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "pack_present",
          "time": 0.0,
          "expected": 1.0,
          "tolerance": {
            "abs": 0.0,
            "rel": 0.0
          }
        }
      ]
    },
    {
      "id": "cold_wrfdt_caseC_step26",
      "description": "Reference: WRF 4.8 em_scm_xy COLD-SEASON single-column run, data/eqwefic/noahcold_scm, dumped by ctessum-claude/WRF branch earthsciml-instrumented-noahcold and replayed through the real64 (-fdefault-real-8) kernels/noah_driver at ESM_DT = 60 s, WRF's OWN step; tolerances are the 1e-9 that a double-precision kernel build earns. WHY THE LARGEST STEP AND NOT THE SMALLEST. SNOWPACK is not a rate: Koren's solution is a closed form in the step length, so the density increment is proportional to dtsec and there is no dt -> 0 limit to approach by shrinking the step. Nothing here is formed as a difference either -- the increment is written as sndens_in pexp_m1 + dw (1 - dsx), an exact rearrangement in which nothing cancels, and its reference comes from a publisher that emits the Taylor series BEFORE WRF adds one to it (FORTRAN_BUGS N99) -- so there is no cancellation floor pushing the step DOWN either. What is left is that a smaller step puts LESS of the assertion's budget on the physics: at 60 s the compaction and retention move the density by 4.5e-9 to 9.4e-5 of itself, so a rel 1e-9 assertion on sndens_out tests them to between 1e-5 and 2e-1 of themselves, while at 1e-6 s the same assertion would test them to nothing at all -- the increment is 7.5e-17 of the density there. This is the fifth distinct direction the Noah reference step has had to move (surface closure 1e-6 s, tendencies 1e-2 s, runoff 1e-4 s, canopy drip 1e-2 s, compaction 60 s), and it is why the choice is argued per quantity rather than assumed. The companion regime at 1e-6 s pins the tendency in the dt -> 0 limit. Step 26 of case C, a thin pack over soil that thaws and refreezes. dry branch -- the snow surface is below freezing and only compaction acts; neither density limit fires. The density moves by 3.866e-07 of itself over the step.",
      "parameter_overrides": {
        "esd": 0.005004871604160893,
        "sndens_in": 0.19966806782197208,
        "tsnow": 266.0577436730374,
        "tsoil": 268.0419983558295,
        "dtsec": 60.0
      },
      "time_span": {
        "start": 0.0,
        "end": 0.0
      },
      "assertions": [
        {
          "variable": "tavgc",
          "time": 0.0,
          "expected": -6.100128985566528,
          "tolerance": {
            "rel": 1e-09,
            "abs": 1e-12
          }
        },
        {
          "variable": "esdcx",
          "time": 0.0,
          "expected": 0.5004871604160893,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "comp_exponent",
          "time": 0.0,
          "expected": -4.681039743106736,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "bfac",
          "time": 0.0,
          "expected": 1.5448951850354069e-06,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "xarg",
          "time": 0.0,
          "expected": 7.732002042988597e-07,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "pexp_m1",
          "time": 0.0,
          "expected": 3.866002017892084e-07,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "pexp",
          "time": 0.0,
          "expected": 1.0000003866002019,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsx_raw",
          "time": 0.0,
          "expected": 0.1996681450136874,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsx",
          "time": 0.0,
          "expected": 0.1996681450136874,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsx_clamp",
          "time": 0.0,
          "expected": 0.0,
          "tolerance": {
            "abs": 0.0,
            "rel": 0.0
          }
        },
        {
          "variable": "wet",
          "time": 0.0,
          "expected": 0.0,
          "tolerance": {
            "abs": 0.0,
            "rel": 0.0
          }
        },
        {
          "variable": "dw",
          "time": 0.0,
          "expected": 0.0,
          "tolerance": {
            "rel": 1e-09,
            "abs": 1e-30
          }
        },
        {
          "variable": "sndens_wet",
          "time": 0.0,
          "expected": 0.1996681450136874,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "sndens_out",
          "time": 0.0,
          "expected": 0.1996681450136874,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "snowh_out",
          "time": 0.0,
          "expected": 0.02506594932215054,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsndens_dt",
          "time": 0.0,
          "expected": 1.2865285885139291e-09,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "pack_present",
          "time": 0.0,
          "expected": 1.0,
          "tolerance": {
            "abs": 0.0,
            "rel": 0.0
          }
        }
      ]
    },
    {
      "id": "cold_wrfdt_caseC_step56",
      "description": "Reference: WRF 4.8 em_scm_xy COLD-SEASON single-column run, data/eqwefic/noahcold_scm, dumped by ctessum-claude/WRF branch earthsciml-instrumented-noahcold and replayed through the real64 (-fdefault-real-8) kernels/noah_driver at ESM_DT = 60 s, WRF's OWN step; tolerances are the 1e-9 that a double-precision kernel build earns. WHY THE LARGEST STEP AND NOT THE SMALLEST. SNOWPACK is not a rate: Koren's solution is a closed form in the step length, so the density increment is proportional to dtsec and there is no dt -> 0 limit to approach by shrinking the step. Nothing here is formed as a difference either -- the increment is written as sndens_in pexp_m1 + dw (1 - dsx), an exact rearrangement in which nothing cancels, and its reference comes from a publisher that emits the Taylor series BEFORE WRF adds one to it (FORTRAN_BUGS N99) -- so there is no cancellation floor pushing the step DOWN either. What is left is that a smaller step puts LESS of the assertion's budget on the physics: at 60 s the compaction and retention move the density by 4.5e-9 to 9.4e-5 of itself, so a rel 1e-9 assertion on sndens_out tests them to between 1e-5 and 2e-1 of themselves, while at 1e-6 s the same assertion would test them to nothing at all -- the increment is 7.5e-17 of the density there. This is the fifth distinct direction the Noah reference step has had to move (surface closure 1e-6 s, tendencies 1e-2 s, runoff 1e-4 s, canopy drip 1e-2 s, compaction 60 s), and it is why the choice is argued per quantity rather than assumed. The companion regime at 1e-6 s pins the tendency in the dt -> 0 limit. Step 56 of case C, a thin pack over soil that thaws and refreezes. dry branch -- the snow surface is below freezing and only compaction acts; neither density limit fires. The density moves by 4.516e-07 of itself over the step.",
      "parameter_overrides": {
        "esd": 0.005107083774533871,
        "sndens_in": 0.1938666593204854,
        "tsnow": 266.22081534208206,
        "tsoil": 268.2150159674135,
        "dtsec": 60.0
      },
      "time_span": {
        "start": 0.0,
        "end": 0.0
      },
      "assertions": [
        {
          "variable": "tavgc",
          "time": 0.0,
          "expected": -5.932084345252207,
          "tolerance": {
            "rel": 1e-09,
            "abs": 1e-12
          }
        },
        {
          "variable": "esdcx",
          "time": 0.0,
          "expected": 0.5107083774533872,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "comp_exponent",
          "time": 0.0,
          "expected": -4.54576659335037,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "bfac",
          "time": 0.0,
          "expected": 1.768672413704133e-06,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "xarg",
          "time": 0.0,
          "expected": 9.032758186494037e-07,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "pexp_m1",
          "time": 0.0,
          "expected": 4.5163804530926665e-07,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "pexp",
          "time": 0.0,
          "expected": 1.0000004516380454,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsx_raw",
          "time": 0.0,
          "expected": 0.1938667468780445,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsx",
          "time": 0.0,
          "expected": 0.1938667468780445,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsx_clamp",
          "time": 0.0,
          "expected": 0.0,
          "tolerance": {
            "abs": 0.0,
            "rel": 0.0
          }
        },
        {
          "variable": "wet",
          "time": 0.0,
          "expected": 0.0,
          "tolerance": {
            "abs": 0.0,
            "rel": 0.0
          }
        },
        {
          "variable": "dw",
          "time": 0.0,
          "expected": 0.0,
          "tolerance": {
            "rel": 1e-09,
            "abs": 1e-30
          }
        },
        {
          "variable": "sndens_wet",
          "time": 0.0,
          "expected": 0.1938667468780445,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "sndens_out",
          "time": 0.0,
          "expected": 0.1938667468780445,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "snowh_out",
          "time": 0.0,
          "expected": 0.02634326854283359,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsndens_dt",
          "time": 0.0,
          "expected": 1.4592926511023593e-09,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "pack_present",
          "time": 0.0,
          "expected": 1.0,
          "tolerance": {
            "abs": 0.0,
            "rel": 0.0
          }
        }
      ]
    },
    {
      "id": "cold_wrfdt_caseC_step169",
      "description": "Reference: WRF 4.8 em_scm_xy COLD-SEASON single-column run, data/eqwefic/noahcold_scm, dumped by ctessum-claude/WRF branch earthsciml-instrumented-noahcold and replayed through the real64 (-fdefault-real-8) kernels/noah_driver at ESM_DT = 60 s, WRF's OWN step; tolerances are the 1e-9 that a double-precision kernel build earns. WHY THE LARGEST STEP AND NOT THE SMALLEST. SNOWPACK is not a rate: Koren's solution is a closed form in the step length, so the density increment is proportional to dtsec and there is no dt -> 0 limit to approach by shrinking the step. Nothing here is formed as a difference either -- the increment is written as sndens_in pexp_m1 + dw (1 - dsx), an exact rearrangement in which nothing cancels, and its reference comes from a publisher that emits the Taylor series BEFORE WRF adds one to it (FORTRAN_BUGS N99) -- so there is no cancellation floor pushing the step DOWN either. What is left is that a smaller step puts LESS of the assertion's budget on the physics: at 60 s the compaction and retention move the density by 4.5e-9 to 9.4e-5 of itself, so a rel 1e-9 assertion on sndens_out tests them to between 1e-5 and 2e-1 of themselves, while at 1e-6 s the same assertion would test them to nothing at all -- the increment is 7.5e-17 of the density there. This is the fifth distinct direction the Noah reference step has had to move (surface closure 1e-6 s, tendencies 1e-2 s, runoff 1e-4 s, canopy drip 1e-2 s, compaction 60 s), and it is why the choice is argued per quantity rather than assumed. The companion regime at 1e-6 s pins the tendency in the dt -> 0 limit. Step 169 of case C, a thin pack over soil that thaws and refreezes. dry branch -- the snow surface is below freezing and only compaction acts; neither density limit fires. The density moves by 7.273e-07 of itself over the step.",
      "parameter_overrides": {
        "esd": 0.005408387471342063,
        "sndens_in": 0.1797167326084254,
        "tsnow": 267.9895496356169,
        "tsoil": 269.4964369991844,
        "dtsec": 60.0
      },
      "time_span": {
        "start": 0.0,
        "end": 0.0
      },
      "assertions": [
        {
          "variable": "tavgc",
          "time": 0.0,
          "expected": -4.407006682599302,
          "tolerance": {
            "rel": 1e-09,
            "abs": 1e-12
          }
        },
        {
          "variable": "esdcx",
          "time": 0.0,
          "expected": 0.5408387471342063,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "comp_exponent",
          "time": 0.0,
          "expected": -4.126611919384878,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "bfac",
          "time": 0.0,
          "expected": 2.68957688755153e-06,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "xarg",
          "time": 0.0,
          "expected": 1.4546273941844877e-06,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "pexp_m1",
          "time": 0.0,
          "expected": 7.273140497491814e-07,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "pexp",
          "time": 0.0,
          "expected": 1.0000007273140497,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsx_raw",
          "time": 0.0,
          "expected": 0.17971686331893,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsx",
          "time": 0.0,
          "expected": 0.17971686331893,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsx_clamp",
          "time": 0.0,
          "expected": 0.0,
          "tolerance": {
            "abs": 0.0,
            "rel": 0.0
          }
        },
        {
          "variable": "wet",
          "time": 0.0,
          "expected": 0.0,
          "tolerance": {
            "abs": 0.0,
            "rel": 0.0
          }
        },
        {
          "variable": "dw",
          "time": 0.0,
          "expected": 0.0,
          "tolerance": {
            "rel": 1e-09,
            "abs": 1e-30
          }
        },
        {
          "variable": "sndens_wet",
          "time": 0.0,
          "expected": 0.17971686331893,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "sndens_out",
          "time": 0.0,
          "expected": 0.17971686331893,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "snowh_out",
          "time": 0.0,
          "expected": 0.030093934266726016,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsndens_dt",
          "time": 0.0,
          "expected": 2.1785084100187442e-09,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "pack_present",
          "time": 0.0,
          "expected": 1.0,
          "tolerance": {
            "abs": 0.0,
            "rel": 0.0
          }
        }
      ]
    },
    {
      "id": "cold_wrfdt_caseC_step268",
      "description": "Reference: WRF 4.8 em_scm_xy COLD-SEASON single-column run, data/eqwefic/noahcold_scm, dumped by ctessum-claude/WRF branch earthsciml-instrumented-noahcold and replayed through the real64 (-fdefault-real-8) kernels/noah_driver at ESM_DT = 60 s, WRF's OWN step; tolerances are the 1e-9 that a double-precision kernel build earns. WHY THE LARGEST STEP AND NOT THE SMALLEST. SNOWPACK is not a rate: Koren's solution is a closed form in the step length, so the density increment is proportional to dtsec and there is no dt -> 0 limit to approach by shrinking the step. Nothing here is formed as a difference either -- the increment is written as sndens_in pexp_m1 + dw (1 - dsx), an exact rearrangement in which nothing cancels, and its reference comes from a publisher that emits the Taylor series BEFORE WRF adds one to it (FORTRAN_BUGS N99) -- so there is no cancellation floor pushing the step DOWN either. What is left is that a smaller step puts LESS of the assertion's budget on the physics: at 60 s the compaction and retention move the density by 4.5e-9 to 9.4e-5 of itself, so a rel 1e-9 assertion on sndens_out tests them to between 1e-5 and 2e-1 of themselves, while at 1e-6 s the same assertion would test them to nothing at all -- the increment is 7.5e-17 of the density there. This is the fifth distinct direction the Noah reference step has had to move (surface closure 1e-6 s, tendencies 1e-2 s, runoff 1e-4 s, canopy drip 1e-2 s, compaction 60 s), and it is why the choice is argued per quantity rather than assumed. The companion regime at 1e-6 s pins the tendency in the dt -> 0 limit. Step 268 of case C, a thin pack over soil that thaws and refreezes. SNOWPACK's melt-water retention branch RUNS -- the snow surface is at or above freezing; neither density limit fires. The density moves by 4.190e-04 of itself over the step.",
      "parameter_overrides": {
        "esd": 0.005462483625836071,
        "sndens_in": 0.17759838855166268,
        "tsnow": 273.51393854812557,
        "tsoil": 272.9141555134422,
        "dtsec": 60.0
      },
      "time_span": {
        "start": 0.0,
        "end": 0.0
      },
      "assertions": [
        {
          "variable": "tavgc",
          "time": 0.0,
          "expected": 0.06404703078391094,
          "tolerance": {
            "rel": 1e-09,
            "abs": 1e-12
          }
        },
        {
          "variable": "esdcx",
          "time": 0.0,
          "expected": 0.5462483625836071,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "comp_exponent",
          "time": 0.0,
          "expected": -3.724442397122204,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "bfac",
          "time": 0.0,
          "expected": 4.0210916183897624e-06,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "xarg",
          "time": 0.0,
          "expected": 2.1965147123440746e-06,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "pexp_m1",
          "time": 0.0,
          "expected": 1.0982581602852923e-06,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "pexp",
          "time": 0.0,
          "expected": 1.0000010982581602,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsx_raw",
          "time": 0.0,
          "expected": 0.17759858360054215,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsx",
          "time": 0.0,
          "expected": 0.17759858360054215,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsx_clamp",
          "time": 0.0,
          "expected": 0.0,
          "tolerance": {
            "abs": 0.0,
            "rel": 0.0
          }
        },
        {
          "variable": "wet",
          "time": 0.0,
          "expected": 1.0,
          "tolerance": {
            "abs": 0.0,
            "rel": 0.0
          }
        },
        {
          "variable": "dw",
          "time": 0.0,
          "expected": 9.027777777777777e-05,
          "tolerance": {
            "rel": 1e-09,
            "abs": 1e-30
          }
        },
        {
          "variable": "sndens_wet",
          "time": 0.0,
          "expected": 0.17767282817285598,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "sndens_out",
          "time": 0.0,
          "expected": 0.17767282817285598,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "snowh_out",
          "time": 0.0,
          "expected": 0.030744620221397498,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsndens_dt",
          "time": 0.0,
          "expected": 1.2406603532220055e-06,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "pack_present",
          "time": 0.0,
          "expected": 1.0,
          "tolerance": {
            "abs": 0.0,
            "rel": 0.0
          }
        }
      ]
    },
    {
      "id": "cold_wrfdt_caseC_step380",
      "description": "Reference: WRF 4.8 em_scm_xy COLD-SEASON single-column run, data/eqwefic/noahcold_scm, dumped by ctessum-claude/WRF branch earthsciml-instrumented-noahcold and replayed through the real64 (-fdefault-real-8) kernels/noah_driver at ESM_DT = 60 s, WRF's OWN step; tolerances are the 1e-9 that a double-precision kernel build earns. WHY THE LARGEST STEP AND NOT THE SMALLEST. SNOWPACK is not a rate: Koren's solution is a closed form in the step length, so the density increment is proportional to dtsec and there is no dt -> 0 limit to approach by shrinking the step. Nothing here is formed as a difference either -- the increment is written as sndens_in pexp_m1 + dw (1 - dsx), an exact rearrangement in which nothing cancels, and its reference comes from a publisher that emits the Taylor series BEFORE WRF adds one to it (FORTRAN_BUGS N99) -- so there is no cancellation floor pushing the step DOWN either. What is left is that a smaller step puts LESS of the assertion's budget on the physics: at 60 s the compaction and retention move the density by 4.5e-9 to 9.4e-5 of itself, so a rel 1e-9 assertion on sndens_out tests them to between 1e-5 and 2e-1 of themselves, while at 1e-6 s the same assertion would test them to nothing at all -- the increment is 7.5e-17 of the density there. This is the fifth distinct direction the Noah reference step has had to move (surface closure 1e-6 s, tendencies 1e-2 s, runoff 1e-4 s, canopy drip 1e-2 s, compaction 60 s), and it is why the choice is argued per quantity rather than assumed. The companion regime at 1e-6 s pins the tendency in the dt -> 0 limit. Step 380 of case C, a thin pack over soil that thaws and refreezes. SNOWPACK's melt-water retention branch RUNS -- the snow surface is at or above freezing; neither density limit fires. The density moves by 4.001e-04 of itself over the step.",
      "parameter_overrides": {
        "esd": 0.005301135991764331,
        "sndens_in": 0.18451886504883896,
        "tsnow": 278.21430080852065,
        "tsoil": 275.92341262597563,
        "dtsec": 60.0
      },
      "time_span": {
        "start": 0.0,
        "end": 0.0
      },
      "assertions": [
        {
          "variable": "tavgc",
          "time": 0.0,
          "expected": 3.918856717248161,
          "tolerance": {
            "rel": 1e-09,
            "abs": 1e-12
          }
        },
        {
          "variable": "esdcx",
          "time": 0.0,
          "expected": 0.5301135991764331,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "comp_exponent",
          "time": 0.0,
          "expected": -3.5613876286457655,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "bfac",
          "time": 0.0,
          "expected": 4.733231592227724e-06,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "xarg",
          "time": 0.0,
          "expected": 2.509150435091438e-06,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "pexp_m1",
          "time": 0.0,
          "expected": 1.2545762668523613e-06,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "pexp",
          "time": 0.0,
          "expected": 1.0000012545762669,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsx_raw",
          "time": 0.0,
          "expected": 0.18451909654182785,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsx",
          "time": 0.0,
          "expected": 0.18451909654182785,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsx_clamp",
          "time": 0.0,
          "expected": 0.0,
          "tolerance": {
            "abs": 0.0,
            "rel": 0.0
          }
        },
        {
          "variable": "wet",
          "time": 0.0,
          "expected": 1.0,
          "tolerance": {
            "abs": 0.0,
            "rel": 0.0
          }
        },
        {
          "variable": "dw",
          "time": 0.0,
          "expected": 9.027777777777777e-05,
          "tolerance": {
            "rel": 1e-09,
            "abs": 1e-30
          }
        },
        {
          "variable": "sndens_wet",
          "time": 0.0,
          "expected": 0.18459271634561228,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "sndens_out",
          "time": 0.0,
          "expected": 0.18459271634561228,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "snowh_out",
          "time": 0.0,
          "expected": 0.02871801280522376,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsndens_dt",
          "time": 0.0,
          "expected": 1.2308549462215853e-06,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "pack_present",
          "time": 0.0,
          "expected": 1.0,
          "tolerance": {
            "abs": 0.0,
            "rel": 0.0
          }
        }
      ]
    },
    {
      "id": "cold_wrfdt_caseC_step405",
      "description": "Reference: WRF 4.8 em_scm_xy COLD-SEASON single-column run, data/eqwefic/noahcold_scm, dumped by ctessum-claude/WRF branch earthsciml-instrumented-noahcold and replayed through the real64 (-fdefault-real-8) kernels/noah_driver at ESM_DT = 60 s, WRF's OWN step; tolerances are the 1e-9 that a double-precision kernel build earns. WHY THE LARGEST STEP AND NOT THE SMALLEST. SNOWPACK is not a rate: Koren's solution is a closed form in the step length, so the density increment is proportional to dtsec and there is no dt -> 0 limit to approach by shrinking the step. Nothing here is formed as a difference either -- the increment is written as sndens_in pexp_m1 + dw (1 - dsx), an exact rearrangement in which nothing cancels, and its reference comes from a publisher that emits the Taylor series BEFORE WRF adds one to it (FORTRAN_BUGS N99) -- so there is no cancellation floor pushing the step DOWN either. What is left is that a smaller step puts LESS of the assertion's budget on the physics: at 60 s the compaction and retention move the density by 4.5e-9 to 9.4e-5 of itself, so a rel 1e-9 assertion on sndens_out tests them to between 1e-5 and 2e-1 of themselves, while at 1e-6 s the same assertion would test them to nothing at all -- the increment is 7.5e-17 of the density there. This is the fifth distinct direction the Noah reference step has had to move (surface closure 1e-6 s, tendencies 1e-2 s, runoff 1e-4 s, canopy drip 1e-2 s, compaction 60 s), and it is why the choice is argued per quantity rather than assumed. The companion regime at 1e-6 s pins the tendency in the dt -> 0 limit. Step 405 of case C, a thin pack over soil that thaws and refreezes. SNOWPACK's melt-water retention branch RUNS -- the snow surface is at or above freezing; neither density limit fires. The density moves by 3.953e-04 of itself over the step.",
      "parameter_overrides": {
        "esd": 0.005197173078100883,
        "sndens_in": 0.18636313407082133,
        "tsnow": 279.94878251017286,
        "tsoil": 277.0792878275947,
        "dtsec": 60.0
      },
      "time_span": {
        "start": 0.0,
        "end": 0.0
      },
      "assertions": [
        {
          "variable": "tavgc",
          "time": 0.0,
          "expected": 5.364035168883817,
          "tolerance": {
            "rel": 1e-09,
            "abs": 1e-12
          }
        },
        {
          "variable": "esdcx",
          "time": 0.0,
          "expected": 0.5197173078100883,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "comp_exponent",
          "time": 0.0,
          "expected": -3.4845030019765426,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "bfac",
          "time": 0.0,
          "expected": 5.111499512171793e-06,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "xarg",
          "time": 0.0,
          "expected": 2.6565347653385037e-06,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "pexp_m1",
          "time": 0.0,
          "expected": 1.3282685588661928e-06,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "pexp",
          "time": 0.0,
          "expected": 1.000001328268559,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsx_raw",
          "time": 0.0,
          "expected": 0.18636338161111285,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsx",
          "time": 0.0,
          "expected": 0.18636338161111285,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsx_clamp",
          "time": 0.0,
          "expected": 0.0,
          "tolerance": {
            "abs": 0.0,
            "rel": 0.0
          }
        },
        {
          "variable": "wet",
          "time": 0.0,
          "expected": 1.0,
          "tolerance": {
            "abs": 0.0,
            "rel": 0.0
          }
        },
        {
          "variable": "dw",
          "time": 0.0,
          "expected": 9.027777777777777e-05,
          "tolerance": {
            "rel": 1e-09,
            "abs": 1e-30
          }
        },
        {
          "variable": "sndens_wet",
          "time": 0.0,
          "expected": 0.18643683491693963,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "sndens_out",
          "time": 0.0,
          "expected": 0.18643683491693963,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "snowh_out",
          "time": 0.0,
          "expected": 0.027876321116566374,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsndens_dt",
          "time": 0.0,
          "expected": 1.2283474353048761e-06,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "pack_present",
          "time": 0.0,
          "expected": 1.0,
          "tolerance": {
            "abs": 0.0,
            "rel": 0.0
          }
        }
      ]
    },
    {
      "id": "cold_wrfdt_caseC_step434",
      "description": "Reference: WRF 4.8 em_scm_xy COLD-SEASON single-column run, data/eqwefic/noahcold_scm, dumped by ctessum-claude/WRF branch earthsciml-instrumented-noahcold and replayed through the real64 (-fdefault-real-8) kernels/noah_driver at ESM_DT = 60 s, WRF's OWN step; tolerances are the 1e-9 that a double-precision kernel build earns. WHY THE LARGEST STEP AND NOT THE SMALLEST. SNOWPACK is not a rate: Koren's solution is a closed form in the step length, so the density increment is proportional to dtsec and there is no dt -> 0 limit to approach by shrinking the step. Nothing here is formed as a difference either -- the increment is written as sndens_in pexp_m1 + dw (1 - dsx), an exact rearrangement in which nothing cancels, and its reference comes from a publisher that emits the Taylor series BEFORE WRF adds one to it (FORTRAN_BUGS N99) -- so there is no cancellation floor pushing the step DOWN either. What is left is that a smaller step puts LESS of the assertion's budget on the physics: at 60 s the compaction and retention move the density by 4.5e-9 to 9.4e-5 of itself, so a rel 1e-9 assertion on sndens_out tests them to between 1e-5 and 2e-1 of themselves, while at 1e-6 s the same assertion would test them to nothing at all -- the increment is 7.5e-17 of the density there. This is the fifth distinct direction the Noah reference step has had to move (surface closure 1e-6 s, tendencies 1e-2 s, runoff 1e-4 s, canopy drip 1e-2 s, compaction 60 s), and it is why the choice is argued per quantity rather than assumed. The companion regime at 1e-6 s pins the tendency in the dt -> 0 limit. Step 434 of case C, a thin pack over soil that thaws and refreezes. SNOWPACK's melt-water retention branch RUNS -- the snow surface is at or above freezing; neither density limit fires. The density moves by 3.898e-04 of itself over the step.",
      "parameter_overrides": {
        "esd": 0.005057964365671849,
        "sndens_in": 0.1884977882695936,
        "tsnow": 280.5072263014543,
        "tsoil": 277.5130995895714,
        "dtsec": 60.0
      },
      "time_span": {
        "start": 0.0,
        "end": 0.0
      },
      "assertions": [
        {
          "variable": "tavgc",
          "time": 0.0,
          "expected": 5.860162945512883,
          "tolerance": {
            "rel": 1e-09,
            "abs": 1e-12
          }
        },
        {
          "variable": "esdcx",
          "time": 0.0,
          "expected": 0.5057964365671849,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "comp_exponent",
          "time": 0.0,
          "expected": -3.489640518020435,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "bfac",
          "time": 0.0,
          "expected": 5.085306442688802e-06,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "xarg",
          "time": 0.0,
          "expected": 2.572129877564143e-06,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "pexp_m1",
          "time": 0.0,
          "expected": 1.2860660414247985e-06,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "pexp",
          "time": 0.0,
          "expected": 1.0000012860660414,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsx_raw",
          "time": 0.0,
          "expected": 0.18849803069019797,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsx",
          "time": 0.0,
          "expected": 0.18849803069019797,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsx_clamp",
          "time": 0.0,
          "expected": 0.0,
          "tolerance": {
            "abs": 0.0,
            "rel": 0.0
          }
        },
        {
          "variable": "wet",
          "time": 0.0,
          "expected": 1.0,
          "tolerance": {
            "abs": 0.0,
            "rel": 0.0
          }
        },
        {
          "variable": "dw",
          "time": 0.0,
          "expected": 9.027777777777777e-05,
          "tolerance": {
            "rel": 1e-09,
            "abs": 1e-30
          }
        },
        {
          "variable": "sndens_wet",
          "time": 0.0,
          "expected": 0.18857129128464956,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "sndens_out",
          "time": 0.0,
          "expected": 0.18857129128464956,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "snowh_out",
          "time": 0.0,
          "expected": 0.026822557830591613,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsndens_dt",
          "time": 0.0,
          "expected": 1.2250502509326092e-06,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "pack_present",
          "time": 0.0,
          "expected": 1.0,
          "tolerance": {
            "abs": 0.0,
            "rel": 0.0
          }
        }
      ]
    },
    {
      "id": "cold_wrfdt_caseC_step459",
      "description": "Reference: WRF 4.8 em_scm_xy COLD-SEASON single-column run, data/eqwefic/noahcold_scm, dumped by ctessum-claude/WRF branch earthsciml-instrumented-noahcold and replayed through the real64 (-fdefault-real-8) kernels/noah_driver at ESM_DT = 60 s, WRF's OWN step; tolerances are the 1e-9 that a double-precision kernel build earns. WHY THE LARGEST STEP AND NOT THE SMALLEST. SNOWPACK is not a rate: Koren's solution is a closed form in the step length, so the density increment is proportional to dtsec and there is no dt -> 0 limit to approach by shrinking the step. Nothing here is formed as a difference either -- the increment is written as sndens_in pexp_m1 + dw (1 - dsx), an exact rearrangement in which nothing cancels, and its reference comes from a publisher that emits the Taylor series BEFORE WRF adds one to it (FORTRAN_BUGS N99) -- so there is no cancellation floor pushing the step DOWN either. What is left is that a smaller step puts LESS of the assertion's budget on the physics: at 60 s the compaction and retention move the density by 4.5e-9 to 9.4e-5 of itself, so a rel 1e-9 assertion on sndens_out tests them to between 1e-5 and 2e-1 of themselves, while at 1e-6 s the same assertion would test them to nothing at all -- the increment is 7.5e-17 of the density there. This is the fifth distinct direction the Noah reference step has had to move (surface closure 1e-6 s, tendencies 1e-2 s, runoff 1e-4 s, canopy drip 1e-2 s, compaction 60 s), and it is why the choice is argued per quantity rather than assumed. The companion regime at 1e-6 s pins the tendency in the dt -> 0 limit. Step 459 of case C, a thin pack over soil that thaws and refreezes. SNOWPACK's melt-water retention branch RUNS -- the snow surface is at or above freezing; neither density limit fires. The density moves by 3.851e-04 of itself over the step.",
      "parameter_overrides": {
        "esd": 0.0049379398814328,
        "sndens_in": 0.19033344802528646,
        "tsnow": 280.73589555040195,
        "tsoil": 277.7289289224177,
        "dtsec": 60.0
      },
      "time_span": {
        "start": 0.0,
        "end": 0.0
      },
      "assertions": [
        {
          "variable": "tavgc",
          "time": 0.0,
          "expected": 6.082412236409851,
          "tolerance": {
            "rel": 1e-09,
            "abs": 1e-12
          }
        },
        {
          "variable": "esdcx",
          "time": 0.0,
          "expected": 0.49379398814327996,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "comp_exponent",
          "time": 0.0,
          "expected": -3.5104094296182273,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "bfac",
          "time": 0.0,
          "expected": 4.980779376694868e-06,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "xarg",
          "time": 0.0,
          "expected": 2.459478912479959e-06,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "pexp_m1",
          "time": 0.0,
          "expected": 1.229740464413353e-06,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "pexp",
          "time": 0.0,
          "expected": 1.0000012297404643,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsx_raw",
          "time": 0.0,
          "expected": 0.19033368208602922,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsx",
          "time": 0.0,
          "expected": 0.19033368208602922,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsx_clamp",
          "time": 0.0,
          "expected": 0.0,
          "tolerance": {
            "abs": 0.0,
            "rel": 0.0
          }
        },
        {
          "variable": "wet",
          "time": 0.0,
          "expected": 1.0,
          "tolerance": {
            "abs": 0.0,
            "rel": 0.0
          }
        },
        {
          "variable": "dw",
          "time": 0.0,
          "expected": 9.027777777777777e-05,
          "tolerance": {
            "rel": 1e-09,
            "abs": 1e-30
          }
        },
        {
          "variable": "sndens_wet",
          "time": 0.0,
          "expected": 0.19040677696195202,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "sndens_out",
          "time": 0.0,
          "expected": 0.19040677696195202,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "snowh_out",
          "time": 0.0,
          "expected": 0.025933635137469515,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsndens_dt",
          "time": 0.0,
          "expected": 1.2221489444259506e-06,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "pack_present",
          "time": 0.0,
          "expected": 1.0,
          "tolerance": {
            "abs": 0.0,
            "rel": 0.0
          }
        }
      ]
    },
    {
      "id": "cold_wrfdt_caseC_step708",
      "description": "Reference: WRF 4.8 em_scm_xy COLD-SEASON single-column run, data/eqwefic/noahcold_scm, dumped by ctessum-claude/WRF branch earthsciml-instrumented-noahcold and replayed through the real64 (-fdefault-real-8) kernels/noah_driver at ESM_DT = 60 s, WRF's OWN step; tolerances are the 1e-9 that a double-precision kernel build earns. WHY THE LARGEST STEP AND NOT THE SMALLEST. SNOWPACK is not a rate: Koren's solution is a closed form in the step length, so the density increment is proportional to dtsec and there is no dt -> 0 limit to approach by shrinking the step. Nothing here is formed as a difference either -- the increment is written as sndens_in pexp_m1 + dw (1 - dsx), an exact rearrangement in which nothing cancels, and its reference comes from a publisher that emits the Taylor series BEFORE WRF adds one to it (FORTRAN_BUGS N99) -- so there is no cancellation floor pushing the step DOWN either. What is left is that a smaller step puts LESS of the assertion's budget on the physics: at 60 s the compaction and retention move the density by 4.5e-9 to 9.4e-5 of itself, so a rel 1e-9 assertion on sndens_out tests them to between 1e-5 and 2e-1 of themselves, while at 1e-6 s the same assertion would test them to nothing at all -- the increment is 7.5e-17 of the density there. This is the fifth distinct direction the Noah reference step has had to move (surface closure 1e-6 s, tendencies 1e-2 s, runoff 1e-4 s, canopy drip 1e-2 s, compaction 60 s), and it is why the choice is argued per quantity rather than assumed. The companion regime at 1e-6 s pins the tendency in the dt -> 0 limit. Step 708 of case C, a thin pack over soil that thaws and refreezes. dry branch -- the snow surface is below freezing and only compaction acts; neither density limit fires. The density moves by 2.826e-07 of itself over the step.",
      "parameter_overrides": {
        "esd": 0.004365151261136432,
        "sndens_in": 0.20500662216316415,
        "tsnow": 265.2143268777221,
        "tsoil": 267.2724060997736,
        "dtsec": 60.0
      },
      "time_span": {
        "start": 0.0,
        "end": 0.0
      },
      "assertions": [
        {
          "variable": "tavgc",
          "time": 0.0,
          "expected": -6.906633511252124,
          "tolerance": {
            "rel": 1e-09,
            "abs": 1e-12
          }
        },
        {
          "variable": "esdcx",
          "time": 0.0,
          "expected": 0.43651512611364324,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "comp_exponent",
          "time": 0.0,
          "expected": -4.857669746326617,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "bfac",
          "time": 0.0,
          "expected": 1.2947609239885606e-06,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "xarg",
          "time": 0.0,
          "expected": 5.651827280218838e-07,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "pexp_m1",
          "time": 0.0,
          "expected": 2.825914172495354e-07,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "pexp",
          "time": 0.0,
          "expected": 1.0000002825914172,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsx_raw",
          "time": 0.0,
          "expected": 0.20500668009627604,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsx",
          "time": 0.0,
          "expected": 0.20500668009627604,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsx_clamp",
          "time": 0.0,
          "expected": 0.0,
          "tolerance": {
            "abs": 0.0,
            "rel": 0.0
          }
        },
        {
          "variable": "wet",
          "time": 0.0,
          "expected": 0.0,
          "tolerance": {
            "abs": 0.0,
            "rel": 0.0
          }
        },
        {
          "variable": "dw",
          "time": 0.0,
          "expected": 0.0,
          "tolerance": {
            "rel": 1e-09,
            "abs": 1e-30
          }
        },
        {
          "variable": "sndens_wet",
          "time": 0.0,
          "expected": 0.20500668009627604,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "sndens_out",
          "time": 0.0,
          "expected": 0.20500668009627604,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "snowh_out",
          "time": 0.0,
          "expected": 0.02129272694473396,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsndens_dt",
          "time": 0.0,
          "expected": 9.655518650438095e-10,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "pack_present",
          "time": 0.0,
          "expected": 1.0,
          "tolerance": {
            "abs": 0.0,
            "rel": 0.0
          }
        }
      ]
    },
    {
      "id": "cold_wrfdt_caseC_step733",
      "description": "Reference: WRF 4.8 em_scm_xy COLD-SEASON single-column run, data/eqwefic/noahcold_scm, dumped by ctessum-claude/WRF branch earthsciml-instrumented-noahcold and replayed through the real64 (-fdefault-real-8) kernels/noah_driver at ESM_DT = 60 s, WRF's OWN step; tolerances are the 1e-9 that a double-precision kernel build earns. WHY THE LARGEST STEP AND NOT THE SMALLEST. SNOWPACK is not a rate: Koren's solution is a closed form in the step length, so the density increment is proportional to dtsec and there is no dt -> 0 limit to approach by shrinking the step. Nothing here is formed as a difference either -- the increment is written as sndens_in pexp_m1 + dw (1 - dsx), an exact rearrangement in which nothing cancels, and its reference comes from a publisher that emits the Taylor series BEFORE WRF adds one to it (FORTRAN_BUGS N99) -- so there is no cancellation floor pushing the step DOWN either. What is left is that a smaller step puts LESS of the assertion's budget on the physics: at 60 s the compaction and retention move the density by 4.5e-9 to 9.4e-5 of itself, so a rel 1e-9 assertion on sndens_out tests them to between 1e-5 and 2e-1 of themselves, while at 1e-6 s the same assertion would test them to nothing at all -- the increment is 7.5e-17 of the density there. This is the fifth distinct direction the Noah reference step has had to move (surface closure 1e-6 s, tendencies 1e-2 s, runoff 1e-4 s, canopy drip 1e-2 s, compaction 60 s), and it is why the choice is argued per quantity rather than assumed. The companion regime at 1e-6 s pins the tendency in the dt -> 0 limit. Step 733 of case C, a thin pack over soil that thaws and refreezes. dry branch -- the snow surface is below freezing and only compaction acts; neither density limit fires. The density moves by 2.831e-07 of itself over the step.",
      "parameter_overrides": {
        "esd": 0.004366473488273769,
        "sndens_in": 0.20500799408956843,
        "tsnow": 265.2389769448648,
        "tsoil": 267.2901203138098,
        "dtsec": 60.0
      },
      "time_span": {
        "start": 0.0,
        "end": 0.0
      },
      "assertions": [
        {
          "variable": "tavgc",
          "time": 0.0,
          "expected": -6.885451370662679,
          "tolerance": {
            "rel": 1e-09,
            "abs": 1e-12
          }
        },
        {
          "variable": "esdcx",
          "time": 0.0,
          "expected": 0.4366473488273769,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "comp_exponent",
          "time": 0.0,
          "expected": -4.856003985533952,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "bfac",
          "time": 0.0,
          "expected": 1.2969194832942225e-06,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "xarg",
          "time": 0.0,
          "expected": 5.662964540229937e-07,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "pexp_m1",
          "time": 0.0,
          "expected": 2.8314828046011677e-07,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "pexp",
          "time": 0.0,
          "expected": 1.0000002831482804,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsx_raw",
          "time": 0.0,
          "expected": 0.20500805213722942,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsx",
          "time": 0.0,
          "expected": 0.20500805213722942,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsx_clamp",
          "time": 0.0,
          "expected": 0.0,
          "tolerance": {
            "abs": 0.0,
            "rel": 0.0
          }
        },
        {
          "variable": "wet",
          "time": 0.0,
          "expected": 0.0,
          "tolerance": {
            "abs": 0.0,
            "rel": 0.0
          }
        },
        {
          "variable": "dw",
          "time": 0.0,
          "expected": 0.0,
          "tolerance": {
            "rel": 1e-09,
            "abs": 1e-30
          }
        },
        {
          "variable": "sndens_wet",
          "time": 0.0,
          "expected": 0.20500805213722942,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "sndens_out",
          "time": 0.0,
          "expected": 0.20500805213722942,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "snowh_out",
          "time": 0.0,
          "expected": 0.02129903407574896,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsndens_dt",
          "time": 0.0,
          "expected": 9.674610167839847e-10,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "pack_present",
          "time": 0.0,
          "expected": 1.0,
          "tolerance": {
            "abs": 0.0,
            "rel": 0.0
          }
        }
      ]
    },
    {
      "id": "cold_wrfdt_caseC_step1564",
      "description": "Reference: WRF 4.8 em_scm_xy COLD-SEASON single-column run, data/eqwefic/noahcold_scm, dumped by ctessum-claude/WRF branch earthsciml-instrumented-noahcold and replayed through the real64 (-fdefault-real-8) kernels/noah_driver at ESM_DT = 60 s, WRF's OWN step; tolerances are the 1e-9 that a double-precision kernel build earns. WHY THE LARGEST STEP AND NOT THE SMALLEST. SNOWPACK is not a rate: Koren's solution is a closed form in the step length, so the density increment is proportional to dtsec and there is no dt -> 0 limit to approach by shrinking the step. Nothing here is formed as a difference either -- the increment is written as sndens_in pexp_m1 + dw (1 - dsx), an exact rearrangement in which nothing cancels, and its reference comes from a publisher that emits the Taylor series BEFORE WRF adds one to it (FORTRAN_BUGS N99) -- so there is no cancellation floor pushing the step DOWN either. What is left is that a smaller step puts LESS of the assertion's budget on the physics: at 60 s the compaction and retention move the density by 4.5e-9 to 9.4e-5 of itself, so a rel 1e-9 assertion on sndens_out tests them to between 1e-5 and 2e-1 of themselves, while at 1e-6 s the same assertion would test them to nothing at all -- the increment is 7.5e-17 of the density there. This is the fifth distinct direction the Noah reference step has had to move (surface closure 1e-6 s, tendencies 1e-2 s, runoff 1e-4 s, canopy drip 1e-2 s, compaction 60 s), and it is why the choice is argued per quantity rather than assumed. The companion regime at 1e-6 s pins the tendency in the dt -> 0 limit. Step 1564 of case C, a thin pack over soil that thaws and refreezes. dry branch -- the snow surface is below freezing and only compaction acts; neither density limit fires. The density moves by 3.908e-07 of itself over the step.",
      "parameter_overrides": {
        "esd": 0.004405550730550145,
        "sndens_in": 0.20505721283671166,
        "tsnow": 270.10773456323943,
        "tsoil": 270.27697364307807,
        "dtsec": 60.0
      },
      "time_span": {
        "start": 0.0,
        "end": 0.0
      },
      "assertions": [
        {
          "variable": "tavgc",
          "time": 0.0,
          "expected": -2.9576458968412283,
          "tolerance": {
            "rel": 1e-09,
            "abs": 1e-12
          }
        },
        {
          "variable": "esdcx",
          "time": 0.0,
          "expected": 0.4405550730550145,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "comp_exponent",
          "time": 0.0,
          "expected": -4.542813141318243,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "bfac",
          "time": 0.0,
          "expected": 1.7739038243960779e-06,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "xarg",
          "time": 0.0,
          "expected": 7.815023289493837e-07,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "pexp_m1",
          "time": 0.0,
          "expected": 3.9075126626569345e-07,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "pexp",
          "time": 0.0,
          "expected": 1.0000003907512662,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsx_raw",
          "time": 0.0,
          "expected": 0.20505729296307723,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsx",
          "time": 0.0,
          "expected": 0.20505729296307723,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsx_clamp",
          "time": 0.0,
          "expected": 0.0,
          "tolerance": {
            "abs": 0.0,
            "rel": 0.0
          }
        },
        {
          "variable": "wet",
          "time": 0.0,
          "expected": 0.0,
          "tolerance": {
            "abs": 0.0,
            "rel": 0.0
          }
        },
        {
          "variable": "dw",
          "time": 0.0,
          "expected": 0.0,
          "tolerance": {
            "rel": 1e-09,
            "abs": 1e-30
          }
        },
        {
          "variable": "sndens_wet",
          "time": 0.0,
          "expected": 0.20505729296307723,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "sndens_out",
          "time": 0.0,
          "expected": 0.20505729296307723,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "snowh_out",
          "time": 0.0,
          "expected": 0.021484486929920662,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsndens_dt",
          "time": 0.0,
          "expected": 1.335439426214315e-09,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "pack_present",
          "time": 0.0,
          "expected": 1.0,
          "tolerance": {
            "abs": 0.0,
            "rel": 0.0
          }
        }
      ]
    },
    {
      "id": "cold_wrfdt_caseC_step1785",
      "description": "Reference: WRF 4.8 em_scm_xy COLD-SEASON single-column run, data/eqwefic/noahcold_scm, dumped by ctessum-claude/WRF branch earthsciml-instrumented-noahcold and replayed through the real64 (-fdefault-real-8) kernels/noah_driver at ESM_DT = 60 s, WRF's OWN step; tolerances are the 1e-9 that a double-precision kernel build earns. WHY THE LARGEST STEP AND NOT THE SMALLEST. SNOWPACK is not a rate: Koren's solution is a closed form in the step length, so the density increment is proportional to dtsec and there is no dt -> 0 limit to approach by shrinking the step. Nothing here is formed as a difference either -- the increment is written as sndens_in pexp_m1 + dw (1 - dsx), an exact rearrangement in which nothing cancels, and its reference comes from a publisher that emits the Taylor series BEFORE WRF adds one to it (FORTRAN_BUGS N99) -- so there is no cancellation floor pushing the step DOWN either. What is left is that a smaller step puts LESS of the assertion's budget on the physics: at 60 s the compaction and retention move the density by 4.5e-9 to 9.4e-5 of itself, so a rel 1e-9 assertion on sndens_out tests them to between 1e-5 and 2e-1 of themselves, while at 1e-6 s the same assertion would test them to nothing at all -- the increment is 7.5e-17 of the density there. This is the fifth distinct direction the Noah reference step has had to move (surface closure 1e-6 s, tendencies 1e-2 s, runoff 1e-4 s, canopy drip 1e-2 s, compaction 60 s), and it is why the choice is argued per quantity rather than assumed. The companion regime at 1e-6 s pins the tendency in the dt -> 0 limit. Step 1785 of case C, a thin pack over soil that thaws and refreezes. SNOWPACK's melt-water retention branch RUNS -- the snow surface is at or above freezing; neither density limit fires. The density moves by 3.212e-04 of itself over the step.",
      "parameter_overrides": {
        "esd": 0.003761259412088598,
        "sndens_in": 0.21981131880599794,
        "tsnow": 288.34277368071645,
        "tsoil": 283.89919366320333,
        "dtsec": 60.0
      },
      "time_span": {
        "start": 0.0,
        "end": 0.0
      },
      "assertions": [
        {
          "variable": "tavgc",
          "time": 0.0,
          "expected": 12.970983671959914,
          "tolerance": {
            "rel": 1e-09,
            "abs": 1e-12
          }
        },
        {
          "variable": "esdcx",
          "time": 0.0,
          "expected": 0.3761259412088598,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "comp_exponent",
          "time": 0.0,
          "expected": -3.578359001169164,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "bfac",
          "time": 0.0,
          "expected": 4.653579966157857e-06,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "xarg",
          "time": 0.0,
          "expected": 1.7503321447618178e-06,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "pexp_m1",
          "time": 0.0,
          "expected": 8.751665829915685e-07,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "pexp",
          "time": 0.0,
          "expected": 1.000000875166583,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsx_raw",
          "time": 0.0,
          "expected": 0.2198115111775187,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsx",
          "time": 0.0,
          "expected": 0.2198115111775187,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsx_clamp",
          "time": 0.0,
          "expected": 0.0,
          "tolerance": {
            "abs": 0.0,
            "rel": 0.0
          }
        },
        {
          "variable": "wet",
          "time": 0.0,
          "expected": 1.0,
          "tolerance": {
            "abs": 0.0,
            "rel": 0.0
          }
        },
        {
          "variable": "dw",
          "time": 0.0,
          "expected": 9.027777777777777e-05,
          "tolerance": {
            "rel": 1e-09,
            "abs": 1e-30
          }
        },
        {
          "variable": "sndens_wet",
          "time": 0.0,
          "expected": 0.21988194486053741,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "sndens_out",
          "time": 0.0,
          "expected": 0.21988194486053741,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "snowh_out",
          "time": 0.0,
          "expected": 0.017105812914626612,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsndens_dt",
          "time": 0.0,
          "expected": 1.177100908991309e-06,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "pack_present",
          "time": 0.0,
          "expected": 1.0,
          "tolerance": {
            "abs": 0.0,
            "rel": 0.0
          }
        }
      ]
    },
    {
      "id": "cold_wrfdt_caseC_step1810",
      "description": "Reference: WRF 4.8 em_scm_xy COLD-SEASON single-column run, data/eqwefic/noahcold_scm, dumped by ctessum-claude/WRF branch earthsciml-instrumented-noahcold and replayed through the real64 (-fdefault-real-8) kernels/noah_driver at ESM_DT = 60 s, WRF's OWN step; tolerances are the 1e-9 that a double-precision kernel build earns. WHY THE LARGEST STEP AND NOT THE SMALLEST. SNOWPACK is not a rate: Koren's solution is a closed form in the step length, so the density increment is proportional to dtsec and there is no dt -> 0 limit to approach by shrinking the step. Nothing here is formed as a difference either -- the increment is written as sndens_in pexp_m1 + dw (1 - dsx), an exact rearrangement in which nothing cancels, and its reference comes from a publisher that emits the Taylor series BEFORE WRF adds one to it (FORTRAN_BUGS N99) -- so there is no cancellation floor pushing the step DOWN either. What is left is that a smaller step puts LESS of the assertion's budget on the physics: at 60 s the compaction and retention move the density by 4.5e-9 to 9.4e-5 of itself, so a rel 1e-9 assertion on sndens_out tests them to between 1e-5 and 2e-1 of themselves, while at 1e-6 s the same assertion would test them to nothing at all -- the increment is 7.5e-17 of the density there. This is the fifth distinct direction the Noah reference step has had to move (surface closure 1e-6 s, tendencies 1e-2 s, runoff 1e-4 s, canopy drip 1e-2 s, compaction 60 s), and it is why the choice is argued per quantity rather than assumed. The companion regime at 1e-6 s pins the tendency in the dt -> 0 limit. Step 1810 of case C, a thin pack over soil that thaws and refreezes. SNOWPACK's melt-water retention branch RUNS -- the snow surface is at or above freezing; neither density limit fires. The density moves by 3.179e-04 of itself over the step.",
      "parameter_overrides": {
        "esd": 0.003643560027632743,
        "sndens_in": 0.22157488677926387,
        "tsnow": 288.8332593952433,
        "tsoil": 284.4626434642655,
        "dtsec": 60.0
      },
      "time_span": {
        "start": 0.0,
        "end": 0.0
      },
      "assertions": [
        {
          "variable": "tavgc",
          "time": 0.0,
          "expected": 13.497951429754437,
          "tolerance": {
            "rel": 1e-09,
            "abs": 1e-12
          }
        },
        {
          "variable": "esdcx",
          "time": 0.0,
          "expected": 0.3643560027632743,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "comp_exponent",
          "time": 0.0,
          "expected": -3.573236507984186,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "bfac",
          "time": 0.0,
          "expected": 4.677479057026106e-06,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "xarg",
          "time": 0.0,
          "expected": 1.7042675722269615e-06,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "pexp_m1",
          "time": 0.0,
          "expected": 8.521342702016799e-07,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "pexp",
          "time": 0.0,
          "expected": 1.0000008521342703,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsx_raw",
          "time": 0.0,
          "expected": 0.22157507559081832,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsx",
          "time": 0.0,
          "expected": 0.22157507559081832,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsx_clamp",
          "time": 0.0,
          "expected": 0.0,
          "tolerance": {
            "abs": 0.0,
            "rel": 0.0
          }
        },
        {
          "variable": "wet",
          "time": 0.0,
          "expected": 1.0,
          "tolerance": {
            "abs": 0.0,
            "rel": 0.0
          }
        },
        {
          "variable": "dw",
          "time": 0.0,
          "expected": 9.027777777777777e-05,
          "tolerance": {
            "rel": 1e-09,
            "abs": 1e-30
          }
        },
        {
          "variable": "sndens_wet",
          "time": 0.0,
          "expected": 0.22164535006316083,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "sndens_out",
          "time": 0.0,
          "expected": 0.22164535006316083,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "snowh_out",
          "time": 0.0,
          "expected": 0.01643869373571095,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsndens_dt",
          "time": 0.0,
          "expected": 1.1743880649489372e-06,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "pack_present",
          "time": 0.0,
          "expected": 1.0,
          "tolerance": {
            "abs": 0.0,
            "rel": 0.0
          }
        }
      ]
    },
    {
      "id": "cold_wrfdt_caseC_step1838",
      "description": "Reference: WRF 4.8 em_scm_xy COLD-SEASON single-column run, data/eqwefic/noahcold_scm, dumped by ctessum-claude/WRF branch earthsciml-instrumented-noahcold and replayed through the real64 (-fdefault-real-8) kernels/noah_driver at ESM_DT = 60 s, WRF's OWN step; tolerances are the 1e-9 that a double-precision kernel build earns. WHY THE LARGEST STEP AND NOT THE SMALLEST. SNOWPACK is not a rate: Koren's solution is a closed form in the step length, so the density increment is proportional to dtsec and there is no dt -> 0 limit to approach by shrinking the step. Nothing here is formed as a difference either -- the increment is written as sndens_in pexp_m1 + dw (1 - dsx), an exact rearrangement in which nothing cancels, and its reference comes from a publisher that emits the Taylor series BEFORE WRF adds one to it (FORTRAN_BUGS N99) -- so there is no cancellation floor pushing the step DOWN either. What is left is that a smaller step puts LESS of the assertion's budget on the physics: at 60 s the compaction and retention move the density by 4.5e-9 to 9.4e-5 of itself, so a rel 1e-9 assertion on sndens_out tests them to between 1e-5 and 2e-1 of themselves, while at 1e-6 s the same assertion would test them to nothing at all -- the increment is 7.5e-17 of the density there. This is the fifth distinct direction the Noah reference step has had to move (surface closure 1e-6 s, tendencies 1e-2 s, runoff 1e-4 s, canopy drip 1e-2 s, compaction 60 s), and it is why the choice is argued per quantity rather than assumed. The companion regime at 1e-6 s pins the tendency in the dt -> 0 limit. Step 1838 of case C, a thin pack over soil that thaws and refreezes. SNOWPACK's melt-water retention branch RUNS -- the snow surface is at or above freezing; neither density limit fires. The density moves by 3.143e-04 of itself over the step.",
      "parameter_overrides": {
        "esd": 0.003514782561516859,
        "sndens_in": 0.2235453724670667,
        "tsnow": 289.1608321106209,
        "tsoil": 284.91598230451956,
        "dtsec": 60.0
      },
      "time_span": {
        "start": 0.0,
        "end": 0.0
      },
      "assertions": [
        {
          "variable": "tavgc",
          "time": 0.0,
          "expected": 13.888407207570253,
          "tolerance": {
            "rel": 1e-09,
            "abs": 1e-12
          }
        },
        {
          "variable": "esdcx",
          "time": 0.0,
          "expected": 0.35147825615168593,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "comp_exponent",
          "time": 0.0,
          "expected": -3.5833802452027808,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "bfac",
          "time": 0.0,
          "expected": 4.630271772554088e-06,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "xarg",
          "time": 0.0,
          "expected": 1.6274398481256868e-06,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "pexp_m1",
          "time": 0.0,
          "expected": 8.137203654897661e-07,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "pexp",
          "time": 0.0,
          "expected": 1.0000008137203655,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsx_raw",
          "time": 0.0,
          "expected": 0.2235455543704889,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsx",
          "time": 0.0,
          "expected": 0.2235455543704889,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsx_clamp",
          "time": 0.0,
          "expected": 0.0,
          "tolerance": {
            "abs": 0.0,
            "rel": 0.0
          }
        },
        {
          "variable": "wet",
          "time": 0.0,
          "expected": 1.0,
          "tolerance": {
            "abs": 0.0,
            "rel": 0.0
          }
        },
        {
          "variable": "dw",
          "time": 0.0,
          "expected": 9.027777777777777e-05,
          "tolerance": {
            "rel": 1e-09,
            "abs": 1e-30
          }
        },
        {
          "variable": "sndens_wet",
          "time": 0.0,
          "expected": 0.223615650952386,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "sndens_out",
          "time": 0.0,
          "expected": 0.223615650952386,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "snowh_out",
          "time": 0.0,
          "expected": 0.015717963150375616,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsndens_dt",
          "time": 0.0,
          "expected": 1.1713080886549348e-06,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "pack_present",
          "time": 0.0,
          "expected": 1.0,
          "tolerance": {
            "abs": 0.0,
            "rel": 0.0
          }
        }
      ]
    },
    {
      "id": "cold_wrfdt_caseC_step1863",
      "description": "Reference: WRF 4.8 em_scm_xy COLD-SEASON single-column run, data/eqwefic/noahcold_scm, dumped by ctessum-claude/WRF branch earthsciml-instrumented-noahcold and replayed through the real64 (-fdefault-real-8) kernels/noah_driver at ESM_DT = 60 s, WRF's OWN step; tolerances are the 1e-9 that a double-precision kernel build earns. WHY THE LARGEST STEP AND NOT THE SMALLEST. SNOWPACK is not a rate: Koren's solution is a closed form in the step length, so the density increment is proportional to dtsec and there is no dt -> 0 limit to approach by shrinking the step. Nothing here is formed as a difference either -- the increment is written as sndens_in pexp_m1 + dw (1 - dsx), an exact rearrangement in which nothing cancels, and its reference comes from a publisher that emits the Taylor series BEFORE WRF adds one to it (FORTRAN_BUGS N99) -- so there is no cancellation floor pushing the step DOWN either. What is left is that a smaller step puts LESS of the assertion's budget on the physics: at 60 s the compaction and retention move the density by 4.5e-9 to 9.4e-5 of itself, so a rel 1e-9 assertion on sndens_out tests them to between 1e-5 and 2e-1 of themselves, while at 1e-6 s the same assertion would test them to nothing at all -- the increment is 7.5e-17 of the density there. This is the fifth distinct direction the Noah reference step has had to move (surface closure 1e-6 s, tendencies 1e-2 s, runoff 1e-4 s, canopy drip 1e-2 s, compaction 60 s), and it is why the choice is argued per quantity rather than assumed. The companion regime at 1e-6 s pins the tendency in the dt -> 0 limit. Step 1863 of case C, a thin pack over soil that thaws and refreezes. SNOWPACK's melt-water retention branch RUNS -- the snow surface is at or above freezing; neither density limit fires. The density moves by 3.111e-04 of itself over the step.",
      "parameter_overrides": {
        "esd": 0.003404672216433211,
        "sndens_in": 0.22530075234017996,
        "tsnow": 289.26076994858494,
        "tsoil": 285.16995609564384,
        "dtsec": 60.0
      },
      "time_span": {
        "start": 0.0,
        "end": 0.0
      },
      "assertions": [
        {
          "variable": "tavgc",
          "time": 0.0,
          "expected": 14.065363022114411,
          "tolerance": {
            "rel": 1e-09,
            "abs": 1e-12
          }
        },
        {
          "variable": "esdcx",
          "time": 0.0,
          "expected": 0.3404672216433211,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "comp_exponent",
          "time": 0.0,
          "expected": -3.6060867573746265,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "bfac",
          "time": 0.0,
          "expected": 4.526319117640455e-06,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "xarg",
          "time": 0.0,
          "expected": 1.5410632942540944e-06,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "pexp_m1",
          "time": 0.0,
          "expected": 7.705320429398791e-07,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "pexp",
          "time": 0.0,
          "expected": 1.0000007705320428,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsx_raw",
          "time": 0.0,
          "expected": 0.22530092594162893,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsx",
          "time": 0.0,
          "expected": 0.22530092594162893,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsx_clamp",
          "time": 0.0,
          "expected": 0.0,
          "tolerance": {
            "abs": 0.0,
            "rel": 0.0
          }
        },
        {
          "variable": "wet",
          "time": 0.0,
          "expected": 1.0,
          "tolerance": {
            "abs": 0.0,
            "rel": 0.0
          }
        },
        {
          "variable": "dw",
          "time": 0.0,
          "expected": 9.027777777777777e-05,
          "tolerance": {
            "rel": 1e-09,
            "abs": 1e-30
          }
        },
        {
          "variable": "sndens_wet",
          "time": 0.0,
          "expected": 0.2253708640524814,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "sndens_out",
          "time": 0.0,
          "expected": 0.2253708640524814,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "snowh_out",
          "time": 0.0,
          "expected": 0.015106975920544795,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsndens_dt",
          "time": 0.0,
          "expected": 1.1685285383578067e-06,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "pack_present",
          "time": 0.0,
          "expected": 1.0,
          "tolerance": {
            "abs": 0.0,
            "rel": 0.0
          }
        }
      ]
    },
    {
      "id": "cold_wrfdt_caseC_step1888",
      "description": "Reference: WRF 4.8 em_scm_xy COLD-SEASON single-column run, data/eqwefic/noahcold_scm, dumped by ctessum-claude/WRF branch earthsciml-instrumented-noahcold and replayed through the real64 (-fdefault-real-8) kernels/noah_driver at ESM_DT = 60 s, WRF's OWN step; tolerances are the 1e-9 that a double-precision kernel build earns. WHY THE LARGEST STEP AND NOT THE SMALLEST. SNOWPACK is not a rate: Koren's solution is a closed form in the step length, so the density increment is proportional to dtsec and there is no dt -> 0 limit to approach by shrinking the step. Nothing here is formed as a difference either -- the increment is written as sndens_in pexp_m1 + dw (1 - dsx), an exact rearrangement in which nothing cancels, and its reference comes from a publisher that emits the Taylor series BEFORE WRF adds one to it (FORTRAN_BUGS N99) -- so there is no cancellation floor pushing the step DOWN either. What is left is that a smaller step puts LESS of the assertion's budget on the physics: at 60 s the compaction and retention move the density by 4.5e-9 to 9.4e-5 of itself, so a rel 1e-9 assertion on sndens_out tests them to between 1e-5 and 2e-1 of themselves, while at 1e-6 s the same assertion would test them to nothing at all -- the increment is 7.5e-17 of the density there. This is the fifth distinct direction the Noah reference step has had to move (surface closure 1e-6 s, tendencies 1e-2 s, runoff 1e-4 s, canopy drip 1e-2 s, compaction 60 s), and it is why the choice is argued per quantity rather than assumed. The companion regime at 1e-6 s pins the tendency in the dt -> 0 limit. Step 1888 of case C, a thin pack over soil that thaws and refreezes. SNOWPACK's melt-water retention branch RUNS -- the snow surface is at or above freezing; neither density limit fires. The density moves by 3.080e-04 of itself over the step.",
      "parameter_overrides": {
        "esd": 0.0033007433445571954,
        "sndens_in": 0.2270514846566476,
        "tsnow": 289.16088334733814,
        "tsoil": 285.25687927201136,
        "dtsec": 60.0
      },
      "time_span": {
        "start": 0.0,
        "end": 0.0
      },
      "assertions": [
        {
          "variable": "tavgc",
          "time": 0.0,
          "expected": 14.058881309674774,
          "tolerance": {
            "rel": 1e-09,
            "abs": 1e-12
          }
        },
        {
          "variable": "esdcx",
          "time": 0.0,
          "expected": 0.33007433445571954,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "comp_exponent",
          "time": 0.0,
          "expected": -3.6433706730156175,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "bfac",
          "time": 0.0,
          "expected": 4.360667477173186e-06,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "xarg",
          "time": 0.0,
          "expected": 1.439344415310641e-06,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "pexp_m1",
          "time": 0.0,
          "expected": 7.196725529408357e-07,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "pexp",
          "time": 0.0,
          "expected": 1.0000007196725529,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsx_raw",
          "time": 0.0,
          "expected": 0.2270516480593692,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsx",
          "time": 0.0,
          "expected": 0.2270516480593692,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsx_clamp",
          "time": 0.0,
          "expected": 0.0,
          "tolerance": {
            "abs": 0.0,
            "rel": 0.0
          }
        },
        {
          "variable": "wet",
          "time": 0.0,
          "expected": 1.0,
          "tolerance": {
            "abs": 0.0,
            "rel": 0.0
          }
        },
        {
          "variable": "dw",
          "time": 0.0,
          "expected": 9.027777777777777e-05,
          "tolerance": {
            "rel": 1e-09,
            "abs": 1e-30
          }
        },
        {
          "variable": "sndens_wet",
          "time": 0.0,
          "expected": 0.2271214281189194,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "sndens_out",
          "time": 0.0,
          "expected": 0.2271214281189194,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "snowh_out",
          "time": 0.0,
          "expected": 0.014532945534443129,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsndens_dt",
          "time": 0.0,
          "expected": 1.1657243711967949e-06,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "pack_present",
          "time": 0.0,
          "expected": 1.0,
          "tolerance": {
            "abs": 0.0,
            "rel": 0.0
          }
        }
      ]
    },
    {
      "id": "cold_wrfdt_caseC_step1978",
      "description": "Reference: WRF 4.8 em_scm_xy COLD-SEASON single-column run, data/eqwefic/noahcold_scm, dumped by ctessum-claude/WRF branch earthsciml-instrumented-noahcold and replayed through the real64 (-fdefault-real-8) kernels/noah_driver at ESM_DT = 60 s, WRF's OWN step; tolerances are the 1e-9 that a double-precision kernel build earns. WHY THE LARGEST STEP AND NOT THE SMALLEST. SNOWPACK is not a rate: Koren's solution is a closed form in the step length, so the density increment is proportional to dtsec and there is no dt -> 0 limit to approach by shrinking the step. Nothing here is formed as a difference either -- the increment is written as sndens_in pexp_m1 + dw (1 - dsx), an exact rearrangement in which nothing cancels, and its reference comes from a publisher that emits the Taylor series BEFORE WRF adds one to it (FORTRAN_BUGS N99) -- so there is no cancellation floor pushing the step DOWN either. What is left is that a smaller step puts LESS of the assertion's budget on the physics: at 60 s the compaction and retention move the density by 4.5e-9 to 9.4e-5 of itself, so a rel 1e-9 assertion on sndens_out tests them to between 1e-5 and 2e-1 of themselves, while at 1e-6 s the same assertion would test them to nothing at all -- the increment is 7.5e-17 of the density there. This is the fifth distinct direction the Noah reference step has had to move (surface closure 1e-6 s, tendencies 1e-2 s, runoff 1e-4 s, canopy drip 1e-2 s, compaction 60 s), and it is why the choice is argued per quantity rather than assumed. The companion regime at 1e-6 s pins the tendency in the dt -> 0 limit. Step 1978 of case C, a thin pack over soil that thaws and refreezes. SNOWPACK's melt-water retention branch RUNS -- the snow surface is at or above freezing; neither density limit fires. The density moves by 2.971e-04 of itself over the step.",
      "parameter_overrides": {
        "esd": 0.0029944604846105327,
        "sndens_in": 0.23331962970999673,
        "tsnow": 287.2600513118683,
        "tsoil": 284.3804945164556,
        "dtsec": 60.0
      },
      "time_span": {
        "start": 0.0,
        "end": 0.0
      },
      "assertions": [
        {
          "variable": "tavgc",
          "time": 0.0,
          "expected": 12.670272914161984,
          "tolerance": {
            "rel": 1e-09,
            "abs": 1e-12
          }
        },
        {
          "variable": "esdcx",
          "time": 0.0,
          "expected": 0.29944604846105327,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "comp_exponent",
          "time": 0.0,
          "expected": -3.886090390776973,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "bfac",
          "time": 0.0,
          "expected": 3.4209059682025287e-06,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "xarg",
          "time": 0.0,
          "expected": 1.0243767743350808e-06,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "pexp_m1",
          "time": 0.0,
          "expected": 5.121885620588811e-07,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "pexp",
          "time": 0.0,
          "expected": 1.000000512188562,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsx_raw",
          "time": 0.0,
          "expected": 0.23331974921364235,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsx",
          "time": 0.0,
          "expected": 0.23331974921364235,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsx_clamp",
          "time": 0.0,
          "expected": 0.0,
          "tolerance": {
            "abs": 0.0,
            "rel": 0.0
          }
        },
        {
          "variable": "wet",
          "time": 0.0,
          "expected": 1.0,
          "tolerance": {
            "abs": 0.0,
            "rel": 0.0
          }
        },
        {
          "variable": "dw",
          "time": 0.0,
          "expected": 9.027777777777777e-05,
          "tolerance": {
            "rel": 1e-09,
            "abs": 1e-30
          }
        },
        {
          "variable": "sndens_wet",
          "time": 0.0,
          "expected": 0.23338896340294946,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "sndens_out",
          "time": 0.0,
          "expected": 0.23338896340294946,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "snowh_out",
          "time": 0.0,
          "expected": 0.012830343135980055,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsndens_dt",
          "time": 0.0,
          "expected": 1.1555615492123835e-06,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "pack_present",
          "time": 0.0,
          "expected": 1.0,
          "tolerance": {
            "abs": 0.0,
            "rel": 0.0
          }
        }
      ]
    },
    {
      "id": "cold_wrfdt_caseC_step2003",
      "description": "Reference: WRF 4.8 em_scm_xy COLD-SEASON single-column run, data/eqwefic/noahcold_scm, dumped by ctessum-claude/WRF branch earthsciml-instrumented-noahcold and replayed through the real64 (-fdefault-real-8) kernels/noah_driver at ESM_DT = 60 s, WRF's OWN step; tolerances are the 1e-9 that a double-precision kernel build earns. WHY THE LARGEST STEP AND NOT THE SMALLEST. SNOWPACK is not a rate: Koren's solution is a closed form in the step length, so the density increment is proportional to dtsec and there is no dt -> 0 limit to approach by shrinking the step. Nothing here is formed as a difference either -- the increment is written as sndens_in pexp_m1 + dw (1 - dsx), an exact rearrangement in which nothing cancels, and its reference comes from a publisher that emits the Taylor series BEFORE WRF adds one to it (FORTRAN_BUGS N99) -- so there is no cancellation floor pushing the step DOWN either. What is left is that a smaller step puts LESS of the assertion's budget on the physics: at 60 s the compaction and retention move the density by 4.5e-9 to 9.4e-5 of itself, so a rel 1e-9 assertion on sndens_out tests them to between 1e-5 and 2e-1 of themselves, while at 1e-6 s the same assertion would test them to nothing at all -- the increment is 7.5e-17 of the density there. This is the fifth distinct direction the Noah reference step has had to move (surface closure 1e-6 s, tendencies 1e-2 s, runoff 1e-4 s, canopy drip 1e-2 s, compaction 60 s), and it is why the choice is argued per quantity rather than assumed. The companion regime at 1e-6 s pins the tendency in the dt -> 0 limit. Step 2003 of case C, a thin pack over soil that thaws and refreezes. SNOWPACK's melt-water retention branch RUNS -- the snow surface is at or above freezing; neither density limit fires. The density moves by 2.942e-04 of itself over the step.",
      "parameter_overrides": {
        "esd": 0.0029313513491442673,
        "sndens_in": 0.23505094109091298,
        "tsnow": 286.2669097330707,
        "tsoil": 283.7584908909183,
        "dtsec": 60.0
      },
      "time_span": {
        "start": 0.0,
        "end": 0.0
      },
      "assertions": [
        {
          "variable": "tavgc",
          "time": 0.0,
          "expected": 11.86270031199453,
          "tolerance": {
            "rel": 1e-09,
            "abs": 1e-12
          }
        },
        {
          "variable": "esdcx",
          "time": 0.0,
          "expected": 0.2931351349144267,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "comp_exponent",
          "time": 0.0,
          "expected": -3.9870537379496107,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "bfac",
          "time": 0.0,
          "expected": 3.0923832495668746e-06,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "xarg",
          "time": 0.0,
          "expected": 9.064861810688991e-07,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "pexp_m1",
          "time": 0.0,
          "expected": 4.5324322748734666e-07,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "pexp",
          "time": 0.0,
          "expected": 1.0000004532432274,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsx_raw",
          "time": 0.0,
          "expected": 0.23505104762616014,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsx",
          "time": 0.0,
          "expected": 0.23505104762616014,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsx_clamp",
          "time": 0.0,
          "expected": 0.0,
          "tolerance": {
            "abs": 0.0,
            "rel": 0.0
          }
        },
        {
          "variable": "wet",
          "time": 0.0,
          "expected": 1.0,
          "tolerance": {
            "abs": 0.0,
            "rel": 0.0
          }
        },
        {
          "variable": "dw",
          "time": 0.0,
          "expected": 9.027777777777777e-05,
          "tolerance": {
            "rel": 1e-09,
            "abs": 1e-30
          }
        },
        {
          "variable": "sndens_wet",
          "time": 0.0,
          "expected": 0.23512010551769388,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "sndens_out",
          "time": 0.0,
          "expected": 0.23512010551769388,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "snowh_out",
          "time": 0.0,
          "expected": 0.01246746356586536,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsndens_dt",
          "time": 0.0,
          "expected": 1.152740446348557e-06,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "pack_present",
          "time": 0.0,
          "expected": 1.0,
          "tolerance": {
            "abs": 0.0,
            "rel": 0.0
          }
        }
      ]
    },
    {
      "id": "cold_wrfdt_caseC_step2104",
      "description": "Reference: WRF 4.8 em_scm_xy COLD-SEASON single-column run, data/eqwefic/noahcold_scm, dumped by ctessum-claude/WRF branch earthsciml-instrumented-noahcold and replayed through the real64 (-fdefault-real-8) kernels/noah_driver at ESM_DT = 60 s, WRF's OWN step; tolerances are the 1e-9 that a double-precision kernel build earns. WHY THE LARGEST STEP AND NOT THE SMALLEST. SNOWPACK is not a rate: Koren's solution is a closed form in the step length, so the density increment is proportional to dtsec and there is no dt -> 0 limit to approach by shrinking the step. Nothing here is formed as a difference either -- the increment is written as sndens_in pexp_m1 + dw (1 - dsx), an exact rearrangement in which nothing cancels, and its reference comes from a publisher that emits the Taylor series BEFORE WRF adds one to it (FORTRAN_BUGS N99) -- so there is no cancellation floor pushing the step DOWN either. What is left is that a smaller step puts LESS of the assertion's budget on the physics: at 60 s the compaction and retention move the density by 4.5e-9 to 9.4e-5 of itself, so a rel 1e-9 assertion on sndens_out tests them to between 1e-5 and 2e-1 of themselves, while at 1e-6 s the same assertion would test them to nothing at all -- the increment is 7.5e-17 of the density there. This is the fifth distinct direction the Noah reference step has had to move (surface closure 1e-6 s, tendencies 1e-2 s, runoff 1e-4 s, canopy drip 1e-2 s, compaction 60 s), and it is why the choice is argued per quantity rather than assumed. The companion regime at 1e-6 s pins the tendency in the dt -> 0 limit. Step 2104 of case C, a thin pack over soil that thaws and refreezes. SNOWPACK's melt-water retention branch RUNS -- the snow surface is at or above freezing; neither density limit fires. The density moves by 2.829e-04 of itself over the step.",
      "parameter_overrides": {
        "esd": 0.0028142975137242336,
        "sndens_in": 0.24200385594087023,
        "tsnow": 275.7970575572843,
        "tsoil": 275.6493858302569,
        "dtsec": 60.0
      },
      "time_span": {
        "start": 0.0,
        "end": 0.0
      },
      "assertions": [
        {
          "variable": "tavgc",
          "time": 0.0,
          "expected": 2.57322169377062,
          "tolerance": {
            "rel": 1e-09,
            "abs": 1e-12
          }
        },
        {
          "variable": "esdcx",
          "time": 0.0,
          "expected": 0.2814297513724234,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "comp_exponent",
          "time": 0.0,
          "expected": -4.876223239256625,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "bfac",
          "time": 0.0,
          "expected": 1.270960063634206e-06,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "xarg",
          "time": 0.0,
          "expected": 3.57685974712854e-07,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "pexp_m1",
          "time": 0.0,
          "expected": 1.7884300867963834e-07,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "pexp",
          "time": 0.0,
          "expected": 1.0000001788430086,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsx_raw",
          "time": 0.0,
          "expected": 0.24200389922156793,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsx",
          "time": 0.0,
          "expected": 0.24200389922156793,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsx_clamp",
          "time": 0.0,
          "expected": 0.0,
          "tolerance": {
            "abs": 0.0,
            "rel": 0.0
          }
        },
        {
          "variable": "wet",
          "time": 0.0,
          "expected": 1.0,
          "tolerance": {
            "abs": 0.0,
            "rel": 0.0
          }
        },
        {
          "variable": "dw",
          "time": 0.0,
          "expected": 9.027777777777777e-05,
          "tolerance": {
            "rel": 1e-09,
            "abs": 1e-30
          }
        },
        {
          "variable": "sndens_wet",
          "time": 0.0,
          "expected": 0.24207232942511042,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "sndens_out",
          "time": 0.0,
          "expected": 0.24207232942511042,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "snowh_out",
          "time": 0.0,
          "expected": 0.01162585381157696,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsndens_dt",
          "time": 0.0,
          "expected": 1.1412247373367644e-06,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "pack_present",
          "time": 0.0,
          "expected": 1.0,
          "tolerance": {
            "abs": 0.0,
            "rel": 0.0
          }
        }
      ]
    },
    {
      "id": "cold_wrfdt_caseC_step2119",
      "description": "Reference: WRF 4.8 em_scm_xy COLD-SEASON single-column run, data/eqwefic/noahcold_scm, dumped by ctessum-claude/WRF branch earthsciml-instrumented-noahcold and replayed through the real64 (-fdefault-real-8) kernels/noah_driver at ESM_DT = 60 s, WRF's OWN step; tolerances are the 1e-9 that a double-precision kernel build earns. WHY THE LARGEST STEP AND NOT THE SMALLEST. SNOWPACK is not a rate: Koren's solution is a closed form in the step length, so the density increment is proportional to dtsec and there is no dt -> 0 limit to approach by shrinking the step. Nothing here is formed as a difference either -- the increment is written as sndens_in pexp_m1 + dw (1 - dsx), an exact rearrangement in which nothing cancels, and its reference comes from a publisher that emits the Taylor series BEFORE WRF adds one to it (FORTRAN_BUGS N99) -- so there is no cancellation floor pushing the step DOWN either. What is left is that a smaller step puts LESS of the assertion's budget on the physics: at 60 s the compaction and retention move the density by 4.5e-9 to 9.4e-5 of itself, so a rel 1e-9 assertion on sndens_out tests them to between 1e-5 and 2e-1 of themselves, while at 1e-6 s the same assertion would test them to nothing at all -- the increment is 7.5e-17 of the density there. This is the fifth distinct direction the Noah reference step has had to move (surface closure 1e-6 s, tendencies 1e-2 s, runoff 1e-4 s, canopy drip 1e-2 s, compaction 60 s), and it is why the choice is argued per quantity rather than assumed. The companion regime at 1e-6 s pins the tendency in the dt -> 0 limit. Step 2119 of case C, a thin pack over soil that thaws and refreezes. dry branch -- the snow surface is below freezing and only compaction acts; neither density limit fires. The density moves by 1.394e-07 of itself over the step.",
      "parameter_overrides": {
        "esd": 0.00281433349512125,
        "sndens_in": 0.24289350120351655,
        "tsnow": 272.62433875257983,
        "tsoil": 273.0668121189063,
        "dtsec": 60.0
      },
      "time_span": {
        "start": 0.0,
        "end": 0.0
      },
      "assertions": [
        {
          "variable": "tavgc",
          "time": 0.0,
          "expected": -0.30442456425691944,
          "tolerance": {
            "rel": 1e-09,
            "abs": 1e-12
          }
        },
        {
          "variable": "esdcx",
          "time": 0.0,
          "expected": 0.281433349512125,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "comp_exponent",
          "time": 0.0,
          "expected": -5.125117490414401,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "bfac",
          "time": 0.0,
          "expected": 9.909197956616569e-07,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "xarg",
          "time": 0.0,
          "expected": 2.788778771909306e-07,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "pexp_m1",
          "time": 0.0,
          "expected": 1.3943895155761126e-07,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "pexp",
          "time": 0.0,
          "expected": 1.0000001394389515,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsx_raw",
          "time": 0.0,
          "expected": 0.2428935350723317,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsx",
          "time": 0.0,
          "expected": 0.2428935350723317,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsx_clamp",
          "time": 0.0,
          "expected": 0.0,
          "tolerance": {
            "abs": 0.0,
            "rel": 0.0
          }
        },
        {
          "variable": "wet",
          "time": 0.0,
          "expected": 0.0,
          "tolerance": {
            "abs": 0.0,
            "rel": 0.0
          }
        },
        {
          "variable": "dw",
          "time": 0.0,
          "expected": 0.0,
          "tolerance": {
            "rel": 1e-09,
            "abs": 1e-30
          }
        },
        {
          "variable": "sndens_wet",
          "time": 0.0,
          "expected": 0.2428935350723317,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "sndens_out",
          "time": 0.0,
          "expected": 0.2428935350723317,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "snowh_out",
          "time": 0.0,
          "expected": 0.011586695768922645,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsndens_dt",
          "time": 0.0,
          "expected": 5.644802524662622e-10,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "pack_present",
          "time": 0.0,
          "expected": 1.0,
          "tolerance": {
            "abs": 0.0,
            "rel": 0.0
          }
        }
      ]
    },
    {
      "id": "cold_wrfdt_caseC_step2141",
      "description": "Reference: WRF 4.8 em_scm_xy COLD-SEASON single-column run, data/eqwefic/noahcold_scm, dumped by ctessum-claude/WRF branch earthsciml-instrumented-noahcold and replayed through the real64 (-fdefault-real-8) kernels/noah_driver at ESM_DT = 60 s, WRF's OWN step; tolerances are the 1e-9 that a double-precision kernel build earns. WHY THE LARGEST STEP AND NOT THE SMALLEST. SNOWPACK is not a rate: Koren's solution is a closed form in the step length, so the density increment is proportional to dtsec and there is no dt -> 0 limit to approach by shrinking the step. Nothing here is formed as a difference either -- the increment is written as sndens_in pexp_m1 + dw (1 - dsx), an exact rearrangement in which nothing cancels, and its reference comes from a publisher that emits the Taylor series BEFORE WRF adds one to it (FORTRAN_BUGS N99) -- so there is no cancellation floor pushing the step DOWN either. What is left is that a smaller step puts LESS of the assertion's budget on the physics: at 60 s the compaction and retention move the density by 4.5e-9 to 9.4e-5 of itself, so a rel 1e-9 assertion on sndens_out tests them to between 1e-5 and 2e-1 of themselves, while at 1e-6 s the same assertion would test them to nothing at all -- the increment is 7.5e-17 of the density there. This is the fifth distinct direction the Noah reference step has had to move (surface closure 1e-6 s, tendencies 1e-2 s, runoff 1e-4 s, canopy drip 1e-2 s, compaction 60 s), and it is why the choice is argued per quantity rather than assumed. The companion regime at 1e-6 s pins the tendency in the dt -> 0 limit. Step 2141 of case C, a thin pack over soil that thaws and refreezes. dry branch -- the snow surface is below freezing and only compaction acts; neither density limit fires. The density moves by 1.058e-07 of itself over the step.",
      "parameter_overrides": {
        "esd": 0.002814965497024615,
        "sndens_in": 0.24289463687606072,
        "tsnow": 268.83101045249964,
        "tsoil": 269.9544853583889,
        "dtsec": 60.0
      },
      "time_span": {
        "start": 0.0,
        "end": 0.0
      },
      "assertions": [
        {
          "variable": "tavgc",
          "time": 0.0,
          "expected": -3.757252094555696,
          "tolerance": {
            "rel": 1e-09,
            "abs": 1e-12
          }
        },
        {
          "variable": "esdcx",
          "time": 0.0,
          "expected": 0.2814965497024615,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "comp_exponent",
          "time": 0.0,
          "expected": -5.401367541961731,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "bfac",
          "time": 0.0,
          "expected": 7.517347583524858e-07,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "xarg",
          "time": 0.0,
          "expected": 2.1161074076763842e-07,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "pexp_m1",
          "time": 0.0,
          "expected": 1.0580537784700387e-07,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "pexp",
          "time": 0.0,
          "expected": 1.0000001058053778,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsx_raw",
          "time": 0.0,
          "expected": 0.24289466257561954,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsx",
          "time": 0.0,
          "expected": 0.24289466257561954,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsx_clamp",
          "time": 0.0,
          "expected": 0.0,
          "tolerance": {
            "abs": 0.0,
            "rel": 0.0
          }
        },
        {
          "variable": "wet",
          "time": 0.0,
          "expected": 0.0,
          "tolerance": {
            "abs": 0.0,
            "rel": 0.0
          }
        },
        {
          "variable": "dw",
          "time": 0.0,
          "expected": 0.0,
          "tolerance": {
            "rel": 1e-09,
            "abs": 1e-30
          }
        },
        {
          "variable": "sndens_wet",
          "time": 0.0,
          "expected": 0.24289466257561954,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "sndens_out",
          "time": 0.0,
          "expected": 0.24289466257561954,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "snowh_out",
          "time": 0.0,
          "expected": 0.011589243942930372,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsndens_dt",
          "time": 0.0,
          "expected": 4.283259805280401e-10,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "pack_present",
          "time": 0.0,
          "expected": 1.0,
          "tolerance": {
            "abs": 0.0,
            "rel": 0.0
          }
        }
      ]
    },
    {
      "id": "cold_wrfdt_caseC_step2166",
      "description": "Reference: WRF 4.8 em_scm_xy COLD-SEASON single-column run, data/eqwefic/noahcold_scm, dumped by ctessum-claude/WRF branch earthsciml-instrumented-noahcold and replayed through the real64 (-fdefault-real-8) kernels/noah_driver at ESM_DT = 60 s, WRF's OWN step; tolerances are the 1e-9 that a double-precision kernel build earns. WHY THE LARGEST STEP AND NOT THE SMALLEST. SNOWPACK is not a rate: Koren's solution is a closed form in the step length, so the density increment is proportional to dtsec and there is no dt -> 0 limit to approach by shrinking the step. Nothing here is formed as a difference either -- the increment is written as sndens_in pexp_m1 + dw (1 - dsx), an exact rearrangement in which nothing cancels, and its reference comes from a publisher that emits the Taylor series BEFORE WRF adds one to it (FORTRAN_BUGS N99) -- so there is no cancellation floor pushing the step DOWN either. What is left is that a smaller step puts LESS of the assertion's budget on the physics: at 60 s the compaction and retention move the density by 4.5e-9 to 9.4e-5 of itself, so a rel 1e-9 assertion on sndens_out tests them to between 1e-5 and 2e-1 of themselves, while at 1e-6 s the same assertion would test them to nothing at all -- the increment is 7.5e-17 of the density there. This is the fifth distinct direction the Noah reference step has had to move (surface closure 1e-6 s, tendencies 1e-2 s, runoff 1e-4 s, canopy drip 1e-2 s, compaction 60 s), and it is why the choice is argued per quantity rather than assumed. The companion regime at 1e-6 s pins the tendency in the dt -> 0 limit. Step 2166 of case C, a thin pack over soil that thaws and refreezes. dry branch -- the snow surface is below freezing and only compaction acts; neither density limit fires. The density moves by 1.012e-07 of itself over the step.",
      "parameter_overrides": {
        "esd": 0.002815887721949831,
        "sndens_in": 0.2428957064342471,
        "tsnow": 268.2079244189253,
        "tsoil": 269.44896955996074,
        "dtsec": 60.0
      },
      "time_span": {
        "start": 0.0,
        "end": 0.0
      },
      "assertions": [
        {
          "variable": "tavgc",
          "time": 0.0,
          "expected": -4.321553010556954,
          "tolerance": {
            "rel": 1e-09,
            "abs": 1e-12
          }
        },
        {
          "variable": "esdcx",
          "time": 0.0,
          "expected": 0.28158877219498313,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "comp_exponent",
          "time": 0.0,
          "expected": -5.446534075963745,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "bfac",
          "time": 0.0,
          "expected": 7.185368652072819e-07,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "xarg",
          "time": 0.0,
          "expected": 2.023319136505506e-07,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "pexp_m1",
          "time": 0.0,
          "expected": 1.0116596364830952e-07,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "pexp",
          "time": 0.0,
          "expected": 1.0000001011659636,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsx_raw",
          "time": 0.0,
          "expected": 0.2428957310070253,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsx",
          "time": 0.0,
          "expected": 0.2428957310070253,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsx_clamp",
          "time": 0.0,
          "expected": 0.0,
          "tolerance": {
            "abs": 0.0,
            "rel": 0.0
          }
        },
        {
          "variable": "wet",
          "time": 0.0,
          "expected": 0.0,
          "tolerance": {
            "abs": 0.0,
            "rel": 0.0
          }
        },
        {
          "variable": "dw",
          "time": 0.0,
          "expected": 0.0,
          "tolerance": {
            "rel": 1e-09,
            "abs": 1e-30
          }
        },
        {
          "variable": "sndens_wet",
          "time": 0.0,
          "expected": 0.2428957310070253,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "sndens_out",
          "time": 0.0,
          "expected": 0.2428957310070253,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "snowh_out",
          "time": 0.0,
          "expected": 0.011592989758508301,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsndens_dt",
          "time": 0.0,
          "expected": 4.0954630345762505e-10,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "pack_present",
          "time": 0.0,
          "expected": 1.0,
          "tolerance": {
            "abs": 0.0,
            "rel": 0.0
          }
        }
      ]
    },
    {
      "id": "cold_wrfdt_caseC_step2880",
      "description": "Reference: WRF 4.8 em_scm_xy COLD-SEASON single-column run, data/eqwefic/noahcold_scm, dumped by ctessum-claude/WRF branch earthsciml-instrumented-noahcold and replayed through the real64 (-fdefault-real-8) kernels/noah_driver at ESM_DT = 60 s, WRF's OWN step; tolerances are the 1e-9 that a double-precision kernel build earns. WHY THE LARGEST STEP AND NOT THE SMALLEST. SNOWPACK is not a rate: Koren's solution is a closed form in the step length, so the density increment is proportional to dtsec and there is no dt -> 0 limit to approach by shrinking the step. Nothing here is formed as a difference either -- the increment is written as sndens_in pexp_m1 + dw (1 - dsx), an exact rearrangement in which nothing cancels, and its reference comes from a publisher that emits the Taylor series BEFORE WRF adds one to it (FORTRAN_BUGS N99) -- so there is no cancellation floor pushing the step DOWN either. What is left is that a smaller step puts LESS of the assertion's budget on the physics: at 60 s the compaction and retention move the density by 4.5e-9 to 9.4e-5 of itself, so a rel 1e-9 assertion on sndens_out tests them to between 1e-5 and 2e-1 of themselves, while at 1e-6 s the same assertion would test them to nothing at all -- the increment is 7.5e-17 of the density there. This is the fifth distinct direction the Noah reference step has had to move (surface closure 1e-6 s, tendencies 1e-2 s, runoff 1e-4 s, canopy drip 1e-2 s, compaction 60 s), and it is why the choice is argued per quantity rather than assumed. The companion regime at 1e-6 s pins the tendency in the dt -> 0 limit. Step 2880 of case C, a thin pack over soil that thaws and refreezes. dry branch -- the snow surface is below freezing and only compaction acts; neither density limit fires. The density moves by 8.847e-08 of itself over the step.",
      "parameter_overrides": {
        "esd": 0.0028401262188247154,
        "sndens_in": 0.24292904682425578,
        "tsnow": 266.5171266345389,
        "tsoil": 267.59058241409787,
        "dtsec": 60.0
      },
      "time_span": {
        "start": 0.0,
        "end": 0.0
      },
      "assertions": [
        {
          "variable": "tavgc",
          "time": 0.0,
          "expected": -6.096145475681595,
          "tolerance": {
            "rel": 1e-09,
            "abs": 1e-12
          }
        },
        {
          "variable": "esdcx",
          "time": 0.0,
          "expected": 0.28401262188247156,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "comp_exponent",
          "time": 0.0,
          "expected": -5.589201621363899,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "bfac",
          "time": 0.0,
          "expected": 6.230018369710803e-07,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "xarg",
          "time": 0.0,
          "expected": 1.7694038515575262e-07,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "pexp_m1",
          "time": 0.0,
          "expected": 8.847019779585985e-08,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "pexp",
          "time": 0.0,
          "expected": 1.0000000884701978,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsx_raw",
          "time": 0.0,
          "expected": 0.24292906831623662,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsx",
          "time": 0.0,
          "expected": 0.24292906831623662,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsx_clamp",
          "time": 0.0,
          "expected": 0.0,
          "tolerance": {
            "abs": 0.0,
            "rel": 0.0
          }
        },
        {
          "variable": "wet",
          "time": 0.0,
          "expected": 0.0,
          "tolerance": {
            "abs": 0.0,
            "rel": 0.0
          }
        },
        {
          "variable": "dw",
          "time": 0.0,
          "expected": 0.0,
          "tolerance": {
            "rel": 1e-09,
            "abs": 1e-30
          }
        },
        {
          "variable": "sndens_wet",
          "time": 0.0,
          "expected": 0.24292906831623662,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "sndens_out",
          "time": 0.0,
          "expected": 0.24292906831623662,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "snowh_out",
          "time": 0.0,
          "expected": 0.0116911748705492,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsndens_dt",
          "time": 0.0,
          "expected": 3.581996803816935e-10,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "pack_present",
          "time": 0.0,
          "expected": 1.0,
          "tolerance": {
            "abs": 0.0,
            "rel": 0.0
          }
        }
      ]
    },
    {
      "id": "cold_wrfdt_caseD_step1",
      "description": "Reference: WRF 4.8 em_scm_xy COLD-SEASON single-column run, data/eqwefic/noahcold_scm, dumped by ctessum-claude/WRF branch earthsciml-instrumented-noahcold and replayed through the real64 (-fdefault-real-8) kernels/noah_driver at ESM_DT = 60 s, WRF's OWN step; tolerances are the 1e-9 that a double-precision kernel build earns. WHY THE LARGEST STEP AND NOT THE SMALLEST. SNOWPACK is not a rate: Koren's solution is a closed form in the step length, so the density increment is proportional to dtsec and there is no dt -> 0 limit to approach by shrinking the step. Nothing here is formed as a difference either -- the increment is written as sndens_in pexp_m1 + dw (1 - dsx), an exact rearrangement in which nothing cancels, and its reference comes from a publisher that emits the Taylor series BEFORE WRF adds one to it (FORTRAN_BUGS N99) -- so there is no cancellation floor pushing the step DOWN either. What is left is that a smaller step puts LESS of the assertion's budget on the physics: at 60 s the compaction and retention move the density by 4.5e-9 to 9.4e-5 of itself, so a rel 1e-9 assertion on sndens_out tests them to between 1e-5 and 2e-1 of themselves, while at 1e-6 s the same assertion would test them to nothing at all -- the increment is 7.5e-17 of the density there. This is the fifth distinct direction the Noah reference step has had to move (surface closure 1e-6 s, tendencies 1e-2 s, runoff 1e-4 s, canopy drip 1e-2 s, compaction 60 s), and it is why the choice is argued per quantity rather than assumed. The companion regime at 1e-6 s pins the tendency in the dt -> 0 limit. Step 1 of case D, a trace pack over frozen ground in a dry column. dry branch -- the snow surface is below freezing and only compaction acts; the pack is below the esdcx floor, so the compaction argument uses the floored 0.01 cm while the depth diagnostic keeps the true water equivalent; neither density limit fires. The density moves by 5.515e-09 of itself over the step.",
      "parameter_overrides": {
        "esd": 4.9997699939662195e-05,
        "sndens_in": 0.2,
        "tsnow": 262.90190014862435,
        "tsoil": 262.9471643284575,
        "dtsec": 60.0
      },
      "time_span": {
        "start": 0.0,
        "end": 0.0
      },
      "assertions": [
        {
          "variable": "tavgc",
          "time": 0.0,
          "expected": -10.225467761459043,
          "tolerance": {
            "rel": 1e-09,
            "abs": 1e-12
          }
        },
        {
          "variable": "esdcx",
          "time": 0.0,
          "expected": 0.01,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "comp_exponent",
          "time": 0.0,
          "expected": -5.018037420916723,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "bfac",
          "time": 0.0,
          "expected": 1.10291689049954e-06,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "xarg",
          "time": 0.0,
          "expected": 1.10291689049954e-08,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "pexp_m1",
          "time": 0.0,
          "expected": 5.514584472771461e-09,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "pexp",
          "time": 0.0,
          "expected": 1.0000000055145846,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsx_raw",
          "time": 0.0,
          "expected": 0.20000000110291694,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsx",
          "time": 0.0,
          "expected": 0.20000000110291694,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsx_clamp",
          "time": 0.0,
          "expected": 0.0,
          "tolerance": {
            "abs": 0.0,
            "rel": 0.0
          }
        },
        {
          "variable": "wet",
          "time": 0.0,
          "expected": 0.0,
          "tolerance": {
            "abs": 0.0,
            "rel": 0.0
          }
        },
        {
          "variable": "dw",
          "time": 0.0,
          "expected": 0.0,
          "tolerance": {
            "rel": 1e-09,
            "abs": 1e-30
          }
        },
        {
          "variable": "sndens_wet",
          "time": 0.0,
          "expected": 0.20000000110291694,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "sndens_out",
          "time": 0.0,
          "expected": 0.20000000110291694,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "snowh_out",
          "time": 0.0,
          "expected": 0.00024998849831972823,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsndens_dt",
          "time": 0.0,
          "expected": 1.8381948242571537e-11,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "pack_present",
          "time": 0.0,
          "expected": 1.0,
          "tolerance": {
            "abs": 0.0,
            "rel": 0.0
          }
        }
      ]
    },
    {
      "id": "cold_wrfdt_caseD_step26",
      "description": "Reference: WRF 4.8 em_scm_xy COLD-SEASON single-column run, data/eqwefic/noahcold_scm, dumped by ctessum-claude/WRF branch earthsciml-instrumented-noahcold and replayed through the real64 (-fdefault-real-8) kernels/noah_driver at ESM_DT = 60 s, WRF's OWN step; tolerances are the 1e-9 that a double-precision kernel build earns. WHY THE LARGEST STEP AND NOT THE SMALLEST. SNOWPACK is not a rate: Koren's solution is a closed form in the step length, so the density increment is proportional to dtsec and there is no dt -> 0 limit to approach by shrinking the step. Nothing here is formed as a difference either -- the increment is written as sndens_in pexp_m1 + dw (1 - dsx), an exact rearrangement in which nothing cancels, and its reference comes from a publisher that emits the Taylor series BEFORE WRF adds one to it (FORTRAN_BUGS N99) -- so there is no cancellation floor pushing the step DOWN either. What is left is that a smaller step puts LESS of the assertion's budget on the physics: at 60 s the compaction and retention move the density by 4.5e-9 to 9.4e-5 of itself, so a rel 1e-9 assertion on sndens_out tests them to between 1e-5 and 2e-1 of themselves, while at 1e-6 s the same assertion would test them to nothing at all -- the increment is 7.5e-17 of the density there. This is the fifth distinct direction the Noah reference step has had to move (surface closure 1e-6 s, tendencies 1e-2 s, runoff 1e-4 s, canopy drip 1e-2 s, compaction 60 s), and it is why the choice is argued per quantity rather than assumed. The companion regime at 1e-6 s pins the tendency in the dt -> 0 limit. Step 26 of case D, a trace pack over frozen ground in a dry column. dry branch -- the snow surface is below freezing and only compaction acts; the pack is below the esdcx floor, so the compaction argument uses the floored 0.01 cm while the depth diagnostic keeps the true water equivalent; neither density limit fires. The density moves by 4.748e-09 of itself over the step.",
      "parameter_overrides": {
        "esd": 4.9962448544795985e-05,
        "sndens_in": 0.20000037571748935,
        "tsnow": 261.02665625536906,
        "tsoil": 261.08120979942294,
        "dtsec": 60.0
      },
      "time_span": {
        "start": 0.0,
        "end": 0.0
      },
      "assertions": [
        {
          "variable": "tavgc",
          "time": 0.0,
          "expected": -12.096066972603978,
          "tolerance": {
            "rel": 1e-09,
            "abs": 1e-12
          }
        },
        {
          "variable": "esdcx",
          "time": 0.0,
          "expected": 0.01,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "comp_exponent",
          "time": 0.0,
          "expected": -5.167693247875595,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "bfac",
          "time": 0.0,
          "expected": 9.496161410250604e-07,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "xarg",
          "time": 0.0,
          "expected": 9.496161410250604e-09,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "pexp_m1",
          "time": 0.0,
          "expected": 4.7480807201548154e-09,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "pexp",
          "time": 0.0,
          "expected": 1.0000000047480808,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsx_raw",
          "time": 0.0,
          "expected": 0.2000003766671073,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsx",
          "time": 0.0,
          "expected": 0.2000003766671073,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsx_clamp",
          "time": 0.0,
          "expected": 0.0,
          "tolerance": {
            "abs": 0.0,
            "rel": 0.0
          }
        },
        {
          "variable": "wet",
          "time": 0.0,
          "expected": 0.0,
          "tolerance": {
            "abs": 0.0,
            "rel": 0.0
          }
        },
        {
          "variable": "dw",
          "time": 0.0,
          "expected": 0.0,
          "tolerance": {
            "rel": 1e-09,
            "abs": 1e-30
          }
        },
        {
          "variable": "sndens_wet",
          "time": 0.0,
          "expected": 0.2000003766671073,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "sndens_out",
          "time": 0.0,
          "expected": 0.2000003766671073,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "snowh_out",
          "time": 0.0,
          "expected": 0.00024981177224459185,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsndens_dt",
          "time": 0.0,
          "expected": 1.5826965466132174e-11,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "pack_present",
          "time": 0.0,
          "expected": 1.0,
          "tolerance": {
            "abs": 0.0,
            "rel": 0.0
          }
        }
      ]
    },
    {
      "id": "cold_wrfdt_caseD_step51",
      "description": "Reference: WRF 4.8 em_scm_xy COLD-SEASON single-column run, data/eqwefic/noahcold_scm, dumped by ctessum-claude/WRF branch earthsciml-instrumented-noahcold and replayed through the real64 (-fdefault-real-8) kernels/noah_driver at ESM_DT = 60 s, WRF's OWN step; tolerances are the 1e-9 that a double-precision kernel build earns. WHY THE LARGEST STEP AND NOT THE SMALLEST. SNOWPACK is not a rate: Koren's solution is a closed form in the step length, so the density increment is proportional to dtsec and there is no dt -> 0 limit to approach by shrinking the step. Nothing here is formed as a difference either -- the increment is written as sndens_in pexp_m1 + dw (1 - dsx), an exact rearrangement in which nothing cancels, and its reference comes from a publisher that emits the Taylor series BEFORE WRF adds one to it (FORTRAN_BUGS N99) -- so there is no cancellation floor pushing the step DOWN either. What is left is that a smaller step puts LESS of the assertion's budget on the physics: at 60 s the compaction and retention move the density by 4.5e-9 to 9.4e-5 of itself, so a rel 1e-9 assertion on sndens_out tests them to between 1e-5 and 2e-1 of themselves, while at 1e-6 s the same assertion would test them to nothing at all -- the increment is 7.5e-17 of the density there. This is the fifth distinct direction the Noah reference step has had to move (surface closure 1e-6 s, tendencies 1e-2 s, runoff 1e-4 s, canopy drip 1e-2 s, compaction 60 s), and it is why the choice is argued per quantity rather than assumed. The companion regime at 1e-6 s pins the tendency in the dt -> 0 limit. Step 51 of case D, a trace pack over frozen ground in a dry column. dry branch -- the snow surface is below freezing and only compaction acts; the pack is below the esdcx floor, so the compaction argument uses the floored 0.01 cm while the depth diagnostic keeps the true water equivalent; neither density limit fires. The density moves by 4.529e-09 of itself over the step.",
      "parameter_overrides": {
        "esd": 4.995112827880049e-05,
        "sndens_in": 0.20000064964772196,
        "tsnow": 260.43398841500215,
        "tsoil": 260.49147503268733,
        "dtsec": 60.0
      },
      "time_span": {
        "start": 0.0,
        "end": 0.0
      },
      "assertions": [
        {
          "variable": "tavgc",
          "time": 0.0,
          "expected": -12.687268276155237,
          "tolerance": {
            "rel": 1e-09,
            "abs": 1e-12
          }
        },
        {
          "variable": "esdcx",
          "time": 0.0,
          "expected": 0.01,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "comp_exponent",
          "time": 0.0,
          "expected": -5.21499510469458,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "bfac",
          "time": 0.0,
          "expected": 9.057433466262762e-07,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "xarg",
          "time": 0.0,
          "expected": 9.057433466262763e-09,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "pexp_m1",
          "time": 0.0,
          "expected": 4.528716746804232e-09,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "pexp",
          "time": 0.0,
          "expected": 1.0000000045287167,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsx_raw",
          "time": 0.0,
          "expected": 0.20000065055346825,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsx",
          "time": 0.0,
          "expected": 0.20000065055346825,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsx_clamp",
          "time": 0.0,
          "expected": 0.0,
          "tolerance": {
            "abs": 0.0,
            "rel": 0.0
          }
        },
        {
          "variable": "wet",
          "time": 0.0,
          "expected": 0.0,
          "tolerance": {
            "abs": 0.0,
            "rel": 0.0
          }
        },
        {
          "variable": "dw",
          "time": 0.0,
          "expected": 0.0,
          "tolerance": {
            "rel": 1e-09,
            "abs": 1e-30
          }
        },
        {
          "variable": "sndens_wet",
          "time": 0.0,
          "expected": 0.20000065055346825,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "sndens_out",
          "time": 0.0,
          "expected": 0.20000065055346825,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "snowh_out",
          "time": 0.0,
          "expected": 0.00024975482899965135,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsndens_dt",
          "time": 0.0,
          "expected": 1.5095771523856073e-11,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "pack_present",
          "time": 0.0,
          "expected": 1.0,
          "tolerance": {
            "abs": 0.0,
            "rel": 0.0
          }
        }
      ]
    },
    {
      "id": "cold_wrfdt_caseD_step392",
      "description": "Reference: WRF 4.8 em_scm_xy COLD-SEASON single-column run, data/eqwefic/noahcold_scm, dumped by ctessum-claude/WRF branch earthsciml-instrumented-noahcold and replayed through the real64 (-fdefault-real-8) kernels/noah_driver at ESM_DT = 60 s, WRF's OWN step; tolerances are the 1e-9 that a double-precision kernel build earns. WHY THE LARGEST STEP AND NOT THE SMALLEST. SNOWPACK is not a rate: Koren's solution is a closed form in the step length, so the density increment is proportional to dtsec and there is no dt -> 0 limit to approach by shrinking the step. Nothing here is formed as a difference either -- the increment is written as sndens_in pexp_m1 + dw (1 - dsx), an exact rearrangement in which nothing cancels, and its reference comes from a publisher that emits the Taylor series BEFORE WRF adds one to it (FORTRAN_BUGS N99) -- so there is no cancellation floor pushing the step DOWN either. What is left is that a smaller step puts LESS of the assertion's budget on the physics: at 60 s the compaction and retention move the density by 4.5e-9 to 9.4e-5 of itself, so a rel 1e-9 assertion on sndens_out tests them to between 1e-5 and 2e-1 of themselves, while at 1e-6 s the same assertion would test them to nothing at all -- the increment is 7.5e-17 of the density there. This is the fifth distinct direction the Noah reference step has had to move (surface closure 1e-6 s, tendencies 1e-2 s, runoff 1e-4 s, canopy drip 1e-2 s, compaction 60 s), and it is why the choice is argued per quantity rather than assumed. The companion regime at 1e-6 s pins the tendency in the dt -> 0 limit. Step 392 of case D, a trace pack over frozen ground in a dry column. SNOWPACK's melt-water retention branch RUNS -- the snow surface is at or above freezing; the pack is below the esdcx floor, so the compaction argument uses the floored 0.01 cm while the depth diagnostic keeps the true water equivalent; neither density limit fires. The density moves by 3.352e-04 of itself over the step.",
      "parameter_overrides": {
        "esd": 4.578182432774142e-05,
        "sndens_in": 0.2121168851409987,
        "tsnow": 282.1061051446816,
        "tsoil": 282.06111154540895,
        "dtsec": 60.0
      },
      "time_span": {
        "start": 0.0,
        "end": 0.0
      },
      "assertions": [
        {
          "variable": "tavgc",
          "time": 0.0,
          "expected": 8.933608345045286,
          "tolerance": {
            "rel": 1e-09,
            "abs": 1e-12
          }
        },
        {
          "variable": "esdcx",
          "time": 0.0,
          "expected": 0.01,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "comp_exponent",
          "time": 0.0,
          "expected": -3.7397659203573497,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "bfac",
          "time": 0.0,
          "expected": 3.959944022351806e-06,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "xarg",
          "time": 0.0,
          "expected": 3.9599440223518056e-08,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "pexp_m1",
          "time": 0.0,
          "expected": 1.9799720373111643e-08,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "pexp",
          "time": 0.0,
          "expected": 1.0000000197997203,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsx_raw",
          "time": 0.0,
          "expected": 0.21211688934085368,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsx",
          "time": 0.0,
          "expected": 0.21211688934085368,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsx_clamp",
          "time": 0.0,
          "expected": 0.0,
          "tolerance": {
            "abs": 0.0,
            "rel": 0.0
          }
        },
        {
          "variable": "wet",
          "time": 0.0,
          "expected": 1.0,
          "tolerance": {
            "abs": 0.0,
            "rel": 0.0
          }
        },
        {
          "variable": "dw",
          "time": 0.0,
          "expected": 9.027777777777777e-05,
          "tolerance": {
            "rel": 1e-09,
            "abs": 1e-30
          }
        },
        {
          "variable": "sndens_wet",
          "time": 0.0,
          "expected": 0.21218801767723264,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "sndens_out",
          "time": 0.0,
          "expected": 0.21218801767723264,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "snowh_out",
          "time": 0.0,
          "expected": 0.0002157606486403107,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsndens_dt",
          "time": 0.0,
          "expected": 1.1855422705660485e-06,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "pack_present",
          "time": 0.0,
          "expected": 1.0,
          "tolerance": {
            "abs": 0.0,
            "rel": 0.0
          }
        }
      ]
    },
    {
      "id": "cold_wrfdt_caseD_step417",
      "description": "Reference: WRF 4.8 em_scm_xy COLD-SEASON single-column run, data/eqwefic/noahcold_scm, dumped by ctessum-claude/WRF branch earthsciml-instrumented-noahcold and replayed through the real64 (-fdefault-real-8) kernels/noah_driver at ESM_DT = 60 s, WRF's OWN step; tolerances are the 1e-9 that a double-precision kernel build earns. WHY THE LARGEST STEP AND NOT THE SMALLEST. SNOWPACK is not a rate: Koren's solution is a closed form in the step length, so the density increment is proportional to dtsec and there is no dt -> 0 limit to approach by shrinking the step. Nothing here is formed as a difference either -- the increment is written as sndens_in pexp_m1 + dw (1 - dsx), an exact rearrangement in which nothing cancels, and its reference comes from a publisher that emits the Taylor series BEFORE WRF adds one to it (FORTRAN_BUGS N99) -- so there is no cancellation floor pushing the step DOWN either. What is left is that a smaller step puts LESS of the assertion's budget on the physics: at 60 s the compaction and retention move the density by 4.5e-9 to 9.4e-5 of itself, so a rel 1e-9 assertion on sndens_out tests them to between 1e-5 and 2e-1 of themselves, while at 1e-6 s the same assertion would test them to nothing at all -- the increment is 7.5e-17 of the density there. This is the fifth distinct direction the Noah reference step has had to move (surface closure 1e-6 s, tendencies 1e-2 s, runoff 1e-4 s, canopy drip 1e-2 s, compaction 60 s), and it is why the choice is argued per quantity rather than assumed. The companion regime at 1e-6 s pins the tendency in the dt -> 0 limit. Step 417 of case D, a trace pack over frozen ground in a dry column. SNOWPACK's melt-water retention branch RUNS -- the snow surface is at or above freezing; the pack is below the esdcx floor, so the compaction argument uses the floored 0.01 cm while the depth diagnostic keeps the true water equivalent; neither density limit fires. The density moves by 3.317e-04 of itself over the step.",
      "parameter_overrides": {
        "esd": 4.2039052275693785e-05,
        "sndens_in": 0.21389331851641966,
        "tsnow": 283.1070103613335,
        "tsoil": 283.06199859771243,
        "dtsec": 60.0
      },
      "time_span": {
        "start": 0.0,
        "end": 0.0
      },
      "assertions": [
        {
          "variable": "tavgc",
          "time": 0.0,
          "expected": 9.934504479522985,
          "tolerance": {
            "rel": 1e-09,
            "abs": 1e-12
          }
        },
        {
          "variable": "esdcx",
          "time": 0.0,
          "expected": 0.01,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "comp_exponent",
          "time": 0.0,
          "expected": -3.6969993304829734,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "bfac",
          "time": 0.0,
          "expected": 4.132970836586567e-06,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "xarg",
          "time": 0.0,
          "expected": 4.132970836586567e-08,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "pexp_m1",
          "time": 0.0,
          "expected": 2.0664854467623636e-08,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "pexp",
          "time": 0.0,
          "expected": 1.0000000206648545,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsx_raw",
          "time": 0.0,
          "expected": 0.21389332293649396,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsx",
          "time": 0.0,
          "expected": 0.21389332293649396,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsx_clamp",
          "time": 0.0,
          "expected": 0.0,
          "tolerance": {
            "abs": 0.0,
            "rel": 0.0
          }
        },
        {
          "variable": "wet",
          "time": 0.0,
          "expected": 1.0,
          "tolerance": {
            "abs": 0.0,
            "rel": 0.0
          }
        },
        {
          "variable": "dw",
          "time": 0.0,
          "expected": 9.027777777777777e-05,
          "tolerance": {
            "rel": 1e-09,
            "abs": 1e-30
          }
        },
        {
          "variable": "sndens_wet",
          "time": 0.0,
          "expected": 0.21396429090039554,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "sndens_out",
          "time": 0.0,
          "expected": 0.21396429090039554,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "snowh_out",
          "time": 0.0,
          "expected": 0.00019647695463007782,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsndens_dt",
          "time": 0.0,
          "expected": 1.1828730662644209e-06,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "pack_present",
          "time": 0.0,
          "expected": 1.0,
          "tolerance": {
            "abs": 0.0,
            "rel": 0.0
          }
        }
      ]
    },
    {
      "id": "cold_wrfdt_caseD_step521",
      "description": "Reference: WRF 4.8 em_scm_xy COLD-SEASON single-column run, data/eqwefic/noahcold_scm, dumped by ctessum-claude/WRF branch earthsciml-instrumented-noahcold and replayed through the real64 (-fdefault-real-8) kernels/noah_driver at ESM_DT = 60 s, WRF's OWN step; tolerances are the 1e-9 that a double-precision kernel build earns. WHY THE LARGEST STEP AND NOT THE SMALLEST. SNOWPACK is not a rate: Koren's solution is a closed form in the step length, so the density increment is proportional to dtsec and there is no dt -> 0 limit to approach by shrinking the step. Nothing here is formed as a difference either -- the increment is written as sndens_in pexp_m1 + dw (1 - dsx), an exact rearrangement in which nothing cancels, and its reference comes from a publisher that emits the Taylor series BEFORE WRF adds one to it (FORTRAN_BUGS N99) -- so there is no cancellation floor pushing the step DOWN either. What is left is that a smaller step puts LESS of the assertion's budget on the physics: at 60 s the compaction and retention move the density by 4.5e-9 to 9.4e-5 of itself, so a rel 1e-9 assertion on sndens_out tests them to between 1e-5 and 2e-1 of themselves, while at 1e-6 s the same assertion would test them to nothing at all -- the increment is 7.5e-17 of the density there. This is the fifth distinct direction the Noah reference step has had to move (surface closure 1e-6 s, tendencies 1e-2 s, runoff 1e-4 s, canopy drip 1e-2 s, compaction 60 s), and it is why the choice is argued per quantity rather than assumed. The companion regime at 1e-6 s pins the tendency in the dt -> 0 limit. Step 521 of case D, a trace pack over frozen ground in a dry column. SNOWPACK's melt-water retention branch RUNS -- the snow surface is at or above freezing; the pack is below the esdcx floor, so the compaction argument uses the floored 0.01 cm while the depth diagnostic keeps the true water equivalent; neither density limit fires. The density moves by 3.177e-04 of itself over the step.",
      "parameter_overrides": {
        "esd": 1.173524884494185e-05,
        "sndens_in": 0.22124005810918224,
        "tsnow": 282.27871277137297,
        "tsoil": 282.26730000822283,
        "dtsec": 60.0
      },
      "time_span": {
        "start": 0.0,
        "end": 0.0
      },
      "assertions": [
        {
          "variable": "tavgc",
          "time": 0.0,
          "expected": 9.123006389797922,
          "tolerance": {
            "rel": 1e-09,
            "abs": 1e-12
          }
        },
        {
          "variable": "esdcx",
          "time": 0.0,
          "expected": 0.01,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "comp_exponent",
          "time": 0.0,
          "expected": -3.916200709108993,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "bfac",
          "time": 0.0,
          "expected": 3.319436702611561e-06,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "xarg",
          "time": 0.0,
          "expected": 3.3194367026115614e-08,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "pexp_m1",
          "time": 0.0,
          "expected": 1.659718369670214e-08,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "pexp",
          "time": 0.0,
          "expected": 1.0000000165971836,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsx_raw",
          "time": 0.0,
          "expected": 0.2212400617811441,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsx",
          "time": 0.0,
          "expected": 0.2212400617811441,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsx_clamp",
          "time": 0.0,
          "expected": 0.0,
          "tolerance": {
            "abs": 0.0,
            "rel": 0.0
          }
        },
        {
          "variable": "wet",
          "time": 0.0,
          "expected": 1.0,
          "tolerance": {
            "abs": 0.0,
            "rel": 0.0
          }
        },
        {
          "variable": "dw",
          "time": 0.0,
          "expected": 9.027777777777777e-05,
          "tolerance": {
            "rel": 1e-09,
            "abs": 1e-30
          }
        },
        {
          "variable": "sndens_wet",
          "time": 0.0,
          "expected": 0.22131036649778887,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "sndens_out",
          "time": 0.0,
          "expected": 0.22131036649778887,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "snowh_out",
          "time": 0.0,
          "expected": 5.30262049204961e-05,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsndens_dt",
          "time": 0.0,
          "expected": 1.1718064767773888e-06,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "pack_present",
          "time": 0.0,
          "expected": 1.0,
          "tolerance": {
            "abs": 0.0,
            "rel": 0.0
          }
        }
      ]
    },
    {
      "id": "cold_wrfdt_caseD_step553_packgone",
      "description": "Reference: the same cold-season run at ESM_DT = 60 s. THE PACK-GONE CASE, and the only test here that does not run SNOWPACK. SNOPAC calls SNOWPACK only where esd > 0; at this step a full 60 s of sublimation takes the 1.25e-06 m pack to exactly zero (the ESD - ESNOW2 <= ESDMIN branch), SFLX zeroes the depth, the density and the cover instead, and the reference's own snp_called reads 0. The only assertion is that `pack_present` agrees with that: the density supplied is the column's own sneqv/snowh, which is what SFLX would have formed had the pack survived, and every other quantity in the component is a hypothetical at this step and is deliberately not asserted. Without this test the zero branch of the caller's guard would be untested rather than merely unused. Step 553 of case D, a trace pack over frozen ground in a dry column.",
      "parameter_overrides": {
        "esd": 0.0,
        "sndens_in": 0.22348673836568872,
        "tsnow": 281.563297708391,
        "tsoil": 281.562242990759,
        "dtsec": 60.0
      },
      "time_span": {
        "start": 0.0,
        "end": 0.0
      },
      "assertions": [
        {
          "variable": "pack_present",
          "time": 0.0,
          "expected": 0.0,
          "tolerance": {
            "abs": 0.0,
            "rel": 0.0
          }
        }
      ]
    },
    {
      "id": "cold_tiny_caseA_step26",
      "description": "Reference: the same cold-season run and the same real64 kernels/noah_driver, replayed at ESM_DT = 1e-6 s, at rel 1e-9. THE POINT OF THIS REGIME is that it is exactly where a naively written compaction component fails: the density increment is 7.5e-17 of the density, so both sndens_out - sndens_in and sndens (pexp - 1) return pure round-off, and only the cancellation-free form -- sndens_in pexp_m1 + dw (1 - dsx), referenced against the series sum published before WRF adds one to it -- still carries the signal. `dsndens_dt` is the dt -> 0 compaction rate here to twelve digits. Step 26 of case A, snow falling onto initially bare frozen ground. dry branch -- the snow surface is below freezing and only compaction acts; the pack is below the esdcx floor, so the compaction argument uses the floored 0.01 cm while the depth diagnostic keeps the true water equivalent; neither density limit fires. The dt -> 0 compaction rate is 1.0226e-10 per second, and the 60 s regime's tendency for the same step exceeds it by xarg/3 = 5.20e-08 relative.",
      "parameter_overrides": {
        "esd": 4.933989215719804e-06,
        "sndens_in": 0.07875117960259606,
        "tsnow": 264.19918167723114,
        "tsoil": 264.19768091377506,
        "dtsec": 1e-06
      },
      "time_span": {
        "start": 0.0,
        "end": 0.0
      },
      "assertions": [
        {
          "variable": "bfac",
          "time": 0.0,
          "expected": 2.5969458267065256e-13,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "xarg",
          "time": 0.0,
          "expected": 2.5969458267065255e-15,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "pexp_m1",
          "time": 0.0,
          "expected": 1.2984729133532641e-15,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsx",
          "time": 0.0,
          "expected": 0.07875117960259617,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dw",
          "time": 0.0,
          "expected": 0.0,
          "tolerance": {
            "rel": 1e-09,
            "abs": 1e-30
          }
        },
        {
          "variable": "sndens_out",
          "time": 0.0,
          "expected": 0.07875117960259617,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsndens_dt",
          "time": 0.0,
          "expected": 1.0225627360858905e-10,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "pack_present",
          "time": 0.0,
          "expected": 1.0,
          "tolerance": {
            "abs": 0.0,
            "rel": 0.0
          }
        }
      ]
    },
    {
      "id": "cold_tiny_caseA_step51",
      "description": "Reference: the same cold-season run and the same real64 kernels/noah_driver, replayed at ESM_DT = 1e-6 s, at rel 1e-9. THE POINT OF THIS REGIME is that it is exactly where a naively written compaction component fails: the density increment is 7.5e-17 of the density, so both sndens_out - sndens_in and sndens (pexp - 1) return pure round-off, and only the cancellation-free form -- sndens_in pexp_m1 + dw (1 - dsx), referenced against the series sum published before WRF adds one to it -- still carries the signal. `dsndens_dt` is the dt -> 0 compaction rate here to twelve digits. Step 51 of case A, snow falling onto initially bare frozen ground. dry branch -- the snow surface is below freezing and only compaction acts; the pack is below the esdcx floor, so the compaction argument uses the floored 0.01 cm while the depth diagnostic keeps the true water equivalent; neither density limit fires. The dt -> 0 compaction rate is 1.0222e-10 per second, and the 60 s regime's tendency for the same step exceeds it by xarg/3 = 5.23e-08 relative.",
      "parameter_overrides": {
        "esd": 8.478196841182004e-05,
        "sndens_in": 0.07817026589847288,
        "tsnow": 264.1463233102526,
        "tsoil": 264.1230061052533,
        "dtsec": 1e-06
      },
      "time_span": {
        "start": 0.0,
        "end": 0.0
      },
      "assertions": [
        {
          "variable": "bfac",
          "time": 0.0,
          "expected": 2.615444158232097e-13,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "xarg",
          "time": 0.0,
          "expected": 2.615444158232097e-15,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "pexp_m1",
          "time": 0.0,
          "expected": 1.3077220791160499e-15,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsx",
          "time": 0.0,
          "expected": 0.078170265898473,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dw",
          "time": 0.0,
          "expected": 0.0,
          "tolerance": {
            "rel": 1e-09,
            "abs": 1e-30
          }
        },
        {
          "variable": "sndens_out",
          "time": 0.0,
          "expected": 0.078170265898473,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsndens_dt",
          "time": 0.0,
          "expected": 1.0222498264580542e-10,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "pack_present",
          "time": 0.0,
          "expected": 1.0,
          "tolerance": {
            "abs": 0.0,
            "rel": 0.0
          }
        }
      ]
    },
    {
      "id": "cold_tiny_caseA_step76",
      "description": "Reference: the same cold-season run and the same real64 kernels/noah_driver, replayed at ESM_DT = 1e-6 s, at rel 1e-9. THE POINT OF THIS REGIME is that it is exactly where a naively written compaction component fails: the density increment is 7.5e-17 of the density, so both sndens_out - sndens_in and sndens (pexp - 1) return pure round-off, and only the cancellation-free form -- sndens_in pexp_m1 + dw (1 - dsx), referenced against the series sum published before WRF adds one to it -- still carries the signal. `dsndens_dt` is the dt -> 0 compaction rate here to twelve digits. Step 76 of case A, snow falling onto initially bare frozen ground. dry branch -- the snow surface is below freezing and only compaction acts; neither density limit fires. The dt -> 0 compaction rate is 2.1165e-10 per second, and the 60 s regime's tendency for the same step exceeds it by xarg/3 = 1.11e-07 relative.",
      "parameter_overrides": {
        "esd": 0.00020789193520814943,
        "sndens_in": 0.0778515009112724,
        "tsnow": 264.0740483850281,
        "tsoil": 264.0272812184948,
        "dtsec": 1e-06
      },
      "time_span": {
        "start": 0.0,
        "end": 0.0
      },
      "assertions": [
        {
          "variable": "bfac",
          "time": 0.0,
          "expected": 2.615376346533872e-13,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "xarg",
          "time": 0.0,
          "expected": 5.437156499785463e-15,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "pexp_m1",
          "time": 0.0,
          "expected": 2.7185782498927362e-15,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsx",
          "time": 0.0,
          "expected": 0.0778515009112726,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dw",
          "time": 0.0,
          "expected": 0.0,
          "tolerance": {
            "rel": 1e-09,
            "abs": 1e-30
          }
        },
        {
          "variable": "sndens_out",
          "time": 0.0,
          "expected": 0.0778515009112726,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsndens_dt",
          "time": 0.0,
          "expected": 2.1164539709888966e-10,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "pack_present",
          "time": 0.0,
          "expected": 1.0,
          "tolerance": {
            "abs": 0.0,
            "rel": 0.0
          }
        }
      ]
    },
    {
      "id": "cold_tiny_caseA_step101",
      "description": "Reference: the same cold-season run and the same real64 kernels/noah_driver, replayed at ESM_DT = 1e-6 s, at rel 1e-9. THE POINT OF THIS REGIME is that it is exactly where a naively written compaction component fails: the density increment is 7.5e-17 of the density, so both sndens_out - sndens_in and sndens (pexp - 1) return pure round-off, and only the cancellation-free form -- sndens_in pexp_m1 + dw (1 - dsx), referenced against the series sum published before WRF adds one to it -- still carries the signal. `dsndens_dt` is the dt -> 0 compaction rate here to twelve digits. Step 101 of case A, snow falling onto initially bare frozen ground. dry branch -- the snow surface is below freezing and only compaction acts; neither density limit fires. The dt -> 0 compaction rate is 3.1019e-10 per second, and the 60 s regime's tendency for the same step exceeds it by xarg/3 = 1.61e-07 relative.",
      "parameter_overrides": {
        "esd": 0.0003061296884464328,
        "sndens_in": 0.07763229291570702,
        "tsnow": 263.99715943438696,
        "tsoil": 263.94200790609096,
        "dtsec": 1e-06
      },
      "time_span": {
        "start": 0.0,
        "end": 0.0
      },
      "assertions": [
        {
          "variable": "bfac",
          "time": 0.0,
          "expected": 2.6104559065391367e-13,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "xarg",
          "time": 0.0,
          "expected": 7.991380533719763e-15,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "pexp_m1",
          "time": 0.0,
          "expected": 3.995690266859893e-15,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsx",
          "time": 0.0,
          "expected": 0.07763229291570732,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dw",
          "time": 0.0,
          "expected": 0.0,
          "tolerance": {
            "rel": 1e-09,
            "abs": 1e-30
          }
        },
        {
          "variable": "sndens_out",
          "time": 0.0,
          "expected": 0.07763229291570732,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsndens_dt",
          "time": 0.0,
          "expected": 3.101945971973067e-10,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "pack_present",
          "time": 0.0,
          "expected": 1.0,
          "tolerance": {
            "abs": 0.0,
            "rel": 0.0
          }
        }
      ]
    },
    {
      "id": "cold_tiny_caseA_step328",
      "description": "Reference: the same cold-season run and the same real64 kernels/noah_driver, replayed at ESM_DT = 1e-6 s, at rel 1e-9. THE POINT OF THIS REGIME is that it is exactly where a naively written compaction component fails: the density increment is 7.5e-17 of the density, so both sndens_out - sndens_in and sndens (pexp - 1) return pure round-off, and only the cancellation-free form -- sndens_in pexp_m1 + dw (1 - dsx), referenced against the series sum published before WRF adds one to it -- still carries the signal. `dsndens_dt` is the dt -> 0 compaction rate here to twelve digits. Step 328 of case A, snow falling onto initially bare frozen ground. dry branch -- the snow surface is below freezing and only compaction acts; neither density limit fires. The dt -> 0 compaction rate is 1.0506e-09 per second, and the 60 s regime's tendency for the same step exceeds it by xarg/3 = 5.20e-07 relative.",
      "parameter_overrides": {
        "esd": 0.0005980000132755349,
        "sndens_in": 0.08080160941911226,
        "tsnow": 271.6343269782292,
        "tsoil": 270.72712939812857,
        "dtsec": 1e-06
      },
      "time_span": {
        "start": 0.0,
        "end": 0.0
      },
      "assertions": [
        {
          "variable": "bfac",
          "time": 0.0,
          "expected": 4.348629760007544e-13,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "xarg",
          "time": 0.0,
          "expected": 2.6004806542148976e-14,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "pexp_m1",
          "time": 0.0,
          "expected": 1.3002403271074602e-14,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsx",
          "time": 0.0,
          "expected": 0.08080160941911331,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dw",
          "time": 0.0,
          "expected": 0.0,
          "tolerance": {
            "rel": 1e-09,
            "abs": 1e-30
          }
        },
        {
          "variable": "sndens_out",
          "time": 0.0,
          "expected": 0.08080160941911331,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsndens_dt",
          "time": 0.0,
          "expected": 1.0506151106191577e-09,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "pack_present",
          "time": 0.0,
          "expected": 1.0,
          "tolerance": {
            "abs": 0.0,
            "rel": 0.0
          }
        }
      ]
    },
    {
      "id": "cold_tiny_caseA_step364",
      "description": "Reference: the same cold-season run and the same real64 kernels/noah_driver, replayed at ESM_DT = 1e-6 s, at rel 1e-9. THE POINT OF THIS REGIME is that it is exactly where a naively written compaction component fails: the density increment is 7.5e-17 of the density, so both sndens_out - sndens_in and sndens (pexp - 1) return pure round-off, and only the cancellation-free form -- sndens_in pexp_m1 + dw (1 - dsx), referenced against the series sum published before WRF adds one to it -- still carries the signal. `dsndens_dt` is the dt -> 0 compaction rate here to twelve digits. Step 364 of case A, snow falling onto initially bare frozen ground. dry branch -- the snow surface is below freezing and only compaction acts; neither density limit fires. The dt -> 0 compaction rate is 1.2420e-09 per second, and the 60 s regime's tendency for the same step exceeds it by xarg/3 = 6.04e-07 relative.",
      "parameter_overrides": {
        "esd": 0.0006727515137822265,
        "sndens_in": 0.0822234289980726,
        "tsnow": 272.48285645272506,
        "tsoil": 271.4284103665575,
        "dtsec": 1e-06
      },
      "time_span": {
        "start": 0.0,
        "end": 0.0
      },
      "assertions": [
        {
          "variable": "bfac",
          "time": 0.0,
          "expected": 4.4906389840640236e-13,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "xarg",
          "time": 0.0,
          "expected": 3.021084174378552e-14,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "pexp_m1",
          "time": 0.0,
          "expected": 1.510542087189291e-14,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsx",
          "time": 0.0,
          "expected": 0.08222342899807383,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dw",
          "time": 0.0,
          "expected": 0.0,
          "tolerance": {
            "rel": 1e-09,
            "abs": 1e-30
          }
        },
        {
          "variable": "sndens_out",
          "time": 0.0,
          "expected": 0.08222342899807383,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsndens_dt",
          "time": 0.0,
          "expected": 1.2420195005460906e-09,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "pack_present",
          "time": 0.0,
          "expected": 1.0,
          "tolerance": {
            "abs": 0.0,
            "rel": 0.0
          }
        }
      ]
    },
    {
      "id": "cold_tiny_caseA_step397",
      "description": "Reference: the same cold-season run and the same real64 kernels/noah_driver, replayed at ESM_DT = 1e-6 s, at rel 1e-9. THE POINT OF THIS REGIME is that it is exactly where a naively written compaction component fails: the density increment is 7.5e-17 of the density, so both sndens_out - sndens_in and sndens (pexp - 1) return pure round-off, and only the cancellation-free form -- sndens_in pexp_m1 + dw (1 - dsx), referenced against the series sum published before WRF adds one to it -- still carries the signal. `dsndens_dt` is the dt -> 0 compaction rate here to twelve digits. Step 397 of case A, snow falling onto initially bare frozen ground. dry branch -- the snow surface is below freezing and only compaction acts; neither density limit fires. The dt -> 0 compaction rate is 1.3986e-09 per second, and the 60 s regime's tendency for the same step exceeds it by xarg/3 = 6.68e-07 relative.",
      "parameter_overrides": {
        "esd": 0.000754952838188198,
        "sndens_in": 0.08374194307700526,
        "tsnow": 272.7313557783245,
        "tsoil": 271.60545121123533,
        "dtsec": 1e-06
      },
      "time_span": {
        "start": 0.0,
        "end": 0.0
      },
      "assertions": [
        {
          "variable": "bfac",
          "time": 0.0,
          "expected": 4.4243696462661243e-13,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "xarg",
          "time": 0.0,
          "expected": 3.3401904216423244e-14,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "pexp_m1",
          "time": 0.0,
          "expected": 1.6700952108211805e-14,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsx",
          "time": 0.0,
          "expected": 0.08374194307700665,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dw",
          "time": 0.0,
          "expected": 0.0,
          "tolerance": {
            "rel": 1e-09,
            "abs": 1e-30
          }
        },
        {
          "variable": "sndens_out",
          "time": 0.0,
          "expected": 0.08374194307700665,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsndens_dt",
          "time": 0.0,
          "expected": 1.398570180777664e-09,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "pack_present",
          "time": 0.0,
          "expected": 1.0,
          "tolerance": {
            "abs": 0.0,
            "rel": 0.0
          }
        }
      ]
    },
    {
      "id": "cold_tiny_caseA_step519",
      "description": "Reference: the same cold-season run and the same real64 kernels/noah_driver, replayed at ESM_DT = 1e-6 s, at rel 1e-9. THE POINT OF THIS REGIME is that it is exactly where a naively written compaction component fails: the density increment is 7.5e-17 of the density, so both sndens_out - sndens_in and sndens (pexp - 1) return pure round-off, and only the cancellation-free form -- sndens_in pexp_m1 + dw (1 - dsx), referenced against the series sum published before WRF adds one to it -- still carries the signal. `dsndens_dt` is the dt -> 0 compaction rate here to twelve digits. Step 519 of case A, snow falling onto initially bare frozen ground. dry branch -- the snow surface is below freezing and only compaction acts; neither density limit fires. The dt -> 0 compaction rate is 1.6804e-09 per second, and the 60 s regime's tendency for the same step exceeds it by xarg/3 = 7.73e-07 relative.",
      "parameter_overrides": {
        "esd": 0.001064269454511822,
        "sndens_in": 0.08809207245299414,
        "tsnow": 271.11474665254,
        "tsoil": 270.24454768936005,
        "dtsec": 1e-06
      },
      "time_span": {
        "start": 0.0,
        "end": 0.0
      },
      "assertions": [
        {
          "variable": "bfac",
          "time": 0.0,
          "expected": 3.584699297927518e-13,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "xarg",
          "time": 0.0,
          "expected": 3.815085966394231e-14,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "pexp_m1",
          "time": 0.0,
          "expected": 1.9075429831971395e-14,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsx",
          "time": 0.0,
          "expected": 0.08809207245299581,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dw",
          "time": 0.0,
          "expected": 0.0,
          "tolerance": {
            "rel": 1e-09,
            "abs": 1e-30
          }
        },
        {
          "variable": "sndens_out",
          "time": 0.0,
          "expected": 0.08809207245299581,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsndens_dt",
          "time": 0.0,
          "expected": 1.68039414683003e-09,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "pack_present",
          "time": 0.0,
          "expected": 1.0,
          "tolerance": {
            "abs": 0.0,
            "rel": 0.0
          }
        }
      ]
    },
    {
      "id": "cold_tiny_caseA_step679",
      "description": "Reference: the same cold-season run and the same real64 kernels/noah_driver, replayed at ESM_DT = 1e-6 s, at rel 1e-9. THE POINT OF THIS REGIME is that it is exactly where a naively written compaction component fails: the density increment is 7.5e-17 of the density, so both sndens_out - sndens_in and sndens (pexp - 1) return pure round-off, and only the cancellation-free form -- sndens_in pexp_m1 + dw (1 - dsx), referenced against the series sum published before WRF adds one to it -- still carries the signal. `dsndens_dt` is the dt -> 0 compaction rate here to twelve digits. Step 679 of case A, snow falling onto initially bare frozen ground. dry branch -- the snow surface is below freezing and only compaction acts; neither density limit fires. The dt -> 0 compaction rate is 1.4198e-09 per second, and the 60 s regime's tendency for the same step exceeds it by xarg/3 = 6.35e-07 relative.",
      "parameter_overrides": {
        "esd": 0.0012426043395302364,
        "sndens_in": 0.08951729135633105,
        "tsnow": 266.75790199007184,
        "tsoil": 266.86310909949907,
        "dtsec": 1e-06
      },
      "time_span": {
        "start": 0.0,
        "end": 0.0
      },
      "assertions": [
        {
          "variable": "bfac",
          "time": 0.0,
          "expected": 2.5528583822948347e-13,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "xarg",
          "time": 0.0,
          "expected": 3.1721929040457005e-14,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "pexp_m1",
          "time": 0.0,
          "expected": 1.5860964520228673e-14,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsx",
          "time": 0.0,
          "expected": 0.08951729135633246,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dw",
          "time": 0.0,
          "expected": 0.0,
          "tolerance": {
            "rel": 1e-09,
            "abs": 1e-30
          }
        },
        {
          "variable": "sndens_out",
          "time": 0.0,
          "expected": 0.08951729135633246,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsndens_dt",
          "time": 0.0,
          "expected": 1.4198305821497396e-09,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "pack_present",
          "time": 0.0,
          "expected": 1.0,
          "tolerance": {
            "abs": 0.0,
            "rel": 0.0
          }
        }
      ]
    },
    {
      "id": "cold_tiny_caseA_step1761",
      "description": "Reference: the same cold-season run and the same real64 kernels/noah_driver, replayed at ESM_DT = 1e-6 s, at rel 1e-9. THE POINT OF THIS REGIME is that it is exactly where a naively written compaction component fails: the density increment is 7.5e-17 of the density, so both sndens_out - sndens_in and sndens (pexp - 1) return pure round-off, and only the cancellation-free form -- sndens_in pexp_m1 + dw (1 - dsx), referenced against the series sum published before WRF adds one to it -- still carries the signal. `dsndens_dt` is the dt -> 0 compaction rate here to twelve digits. Step 1761 of case A, snow falling onto initially bare frozen ground. SNOWPACK's melt-water retention branch RUNS -- the snow surface is at or above freezing; neither density limit fires. The dt -> 0 compaction rate is 5.7111e-09 per second, and the 60 s regime's tendency for the same step exceeds it by xarg/3 = 2.55e-06 relative.",
      "parameter_overrides": {
        "esd": 0.003412499791248944,
        "sndens_in": 0.08971407185480638,
        "tsnow": 273.16806402555886,
        "tsoil": 270.04228421704977,
        "dtsec": 1e-06
      },
      "time_span": {
        "start": 0.0,
        "end": 0.0
      },
      "assertions": [
        {
          "variable": "bfac",
          "time": 0.0,
          "expected": 3.730919661392771e-13,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "xarg",
          "time": 0.0,
          "expected": 1.273176256566941e-13,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "pexp_m1",
          "time": 0.0,
          "expected": 6.365881282834974e-14,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsx",
          "time": 0.0,
          "expected": 0.0897140718548121,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dw",
          "time": 0.0,
          "expected": 1.5046296296296297e-12,
          "tolerance": {
            "rel": 1e-09,
            "abs": 1e-30
          }
        },
        {
          "variable": "sndens_out",
          "time": 0.0,
          "expected": 0.08971407185618176,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsndens_dt",
          "time": 0.0,
          "expected": 1.3753542702304323e-06,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "pack_present",
          "time": 0.0,
          "expected": 1.0,
          "tolerance": {
            "abs": 0.0,
            "rel": 0.0
          }
        }
      ]
    },
    {
      "id": "cold_tiny_caseA_step1824",
      "description": "Reference: the same cold-season run and the same real64 kernels/noah_driver, replayed at ESM_DT = 1e-6 s, at rel 1e-9. THE POINT OF THIS REGIME is that it is exactly where a naively written compaction component fails: the density increment is 7.5e-17 of the density, so both sndens_out - sndens_in and sndens (pexp - 1) return pure round-off, and only the cancellation-free form -- sndens_in pexp_m1 + dw (1 - dsx), referenced against the series sum published before WRF adds one to it -- still carries the signal. `dsndens_dt` is the dt -> 0 compaction rate here to twelve digits. Step 1824 of case A, snow falling onto initially bare frozen ground. SNOWPACK's melt-water retention branch RUNS -- the snow surface is at or above freezing; neither density limit fires. The dt -> 0 compaction rate is 6.5438e-09 per second, and the 60 s regime's tendency for the same step exceeds it by xarg/3 = 2.75e-06 relative.",
      "parameter_overrides": {
        "esd": 0.00343441148287316,
        "sndens_in": 0.09501853516179763,
        "tsnow": 275.9289370950565,
        "tsoil": 271.87266744600214,
        "dtsec": 1e-06
      },
      "time_span": {
        "start": 0.0,
        "end": 0.0
      },
      "assertions": [
        {
          "variable": "bfac",
          "time": 0.0,
          "expected": 4.0104813991266923e-13,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "xarg",
          "time": 0.0,
          "expected": 1.377364336900993e-13,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "pexp_m1",
          "time": 0.0,
          "expected": 6.88682168450528e-14,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsx",
          "time": 0.0,
          "expected": 0.09501853516180417,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dw",
          "time": 0.0,
          "expected": 1.5046296296296297e-12,
          "tolerance": {
            "rel": 1e-09,
            "abs": 1e-30
          }
        },
        {
          "variable": "sndens_out",
          "time": 0.0,
          "expected": 0.09501853516316583,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsndens_dt",
          "time": 0.0,
          "expected": 1.3682056833449964e-06,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "pack_present",
          "time": 0.0,
          "expected": 1.0,
          "tolerance": {
            "abs": 0.0,
            "rel": 0.0
          }
        }
      ]
    },
    {
      "id": "cold_tiny_caseA_step1849",
      "description": "Reference: the same cold-season run and the same real64 kernels/noah_driver, replayed at ESM_DT = 1e-6 s, at rel 1e-9. THE POINT OF THIS REGIME is that it is exactly where a naively written compaction component fails: the density increment is 7.5e-17 of the density, so both sndens_out - sndens_in and sndens (pexp - 1) return pure round-off, and only the cancellation-free form -- sndens_in pexp_m1 + dw (1 - dsx), referenced against the series sum published before WRF adds one to it -- still carries the signal. `dsndens_dt` is the dt -> 0 compaction rate here to twelve digits. Step 1849 of case A, snow falling onto initially bare frozen ground. SNOWPACK's melt-water retention branch RUNS -- the snow surface is at or above freezing; neither density limit fires. The dt -> 0 compaction rate is 6.6533e-09 per second, and the 60 s regime's tendency for the same step exceeds it by xarg/3 = 2.74e-06 relative.",
      "parameter_overrides": {
        "esd": 0.003413871396318216,
        "sndens_in": 0.09707135072312419,
        "tsnow": 276.5390782475416,
        "tsoil": 272.3708732093106,
        "dtsec": 1e-06
      },
      "time_span": {
        "start": 0.0,
        "end": 0.0
      },
      "assertions": [
        {
          "variable": "bfac",
          "time": 0.0,
          "expected": 4.015396244708003e-13,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "xarg",
          "time": 0.0,
          "expected": 1.370804638469223e-13,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "pexp_m1",
          "time": 0.0,
          "expected": 6.85402319234643e-14,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsx",
          "time": 0.0,
          "expected": 0.09707135072313085,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dw",
          "time": 0.0,
          "expected": 1.5046296296296297e-12,
          "tolerance": {
            "rel": 1e-09,
            "abs": 1e-30
          }
        },
        {
          "variable": "sndens_out",
          "time": 0.0,
          "expected": 0.09707135072448943,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsndens_dt",
          "time": 0.0,
          "expected": 1.3652264920351244e-06,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "pack_present",
          "time": 0.0,
          "expected": 1.0,
          "tolerance": {
            "abs": 0.0,
            "rel": 0.0
          }
        }
      ]
    },
    {
      "id": "cold_tiny_caseA_step1874",
      "description": "Reference: the same cold-season run and the same real64 kernels/noah_driver, replayed at ESM_DT = 1e-6 s, at rel 1e-9. THE POINT OF THIS REGIME is that it is exactly where a naively written compaction component fails: the density increment is 7.5e-17 of the density, so both sndens_out - sndens_in and sndens (pexp - 1) return pure round-off, and only the cancellation-free form -- sndens_in pexp_m1 + dw (1 - dsx), referenced against the series sum published before WRF adds one to it -- still carries the signal. `dsndens_dt` is the dt -> 0 compaction rate here to twelve digits. Step 1874 of case A, snow falling onto initially bare frozen ground. SNOWPACK's melt-water retention branch RUNS -- the snow surface is at or above freezing; neither density limit fires. The dt -> 0 compaction rate is 6.5895e-09 per second, and the 60 s regime's tendency for the same step exceeds it by xarg/3 = 2.66e-06 relative.",
      "parameter_overrides": {
        "esd": 0.0033907520118942394,
        "sndens_in": 0.09912003421401928,
        "tsnow": 276.745397916633,
        "tsoil": 272.64718218123585,
        "dtsec": 1e-06
      },
      "time_span": {
        "start": 0.0,
        "end": 0.0
      },
      "assertions": [
        {
          "variable": "bfac",
          "time": 0.0,
          "expected": 3.921282724628948e-13,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "xarg",
          "time": 0.0,
          "expected": 1.329609728774173e-13,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "pexp_m1",
          "time": 0.0,
          "expected": 6.64804864387116e-14,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsx",
          "time": 0.0,
          "expected": 0.09912003421402586,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dw",
          "time": 0.0,
          "expected": 1.5046296296296297e-12,
          "tolerance": {
            "rel": 1e-09,
            "abs": 1e-30
          }
        },
        {
          "variable": "sndens_out",
          "time": 0.0,
          "expected": 0.09912003421538136,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsndens_dt",
          "time": 0.0,
          "expected": 1.3620802373516737e-06,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "pack_present",
          "time": 0.0,
          "expected": 1.0,
          "tolerance": {
            "abs": 0.0,
            "rel": 0.0
          }
        }
      ]
    },
    {
      "id": "cold_tiny_caseA_step1899",
      "description": "Reference: the same cold-season run and the same real64 kernels/noah_driver, replayed at ESM_DT = 1e-6 s, at rel 1e-9. THE POINT OF THIS REGIME is that it is exactly where a naively written compaction component fails: the density increment is 7.5e-17 of the density, so both sndens_out - sndens_in and sndens (pexp - 1) return pure round-off, and only the cancellation-free form -- sndens_in pexp_m1 + dw (1 - dsx), referenced against the series sum published before WRF adds one to it -- still carries the signal. `dsndens_dt` is the dt -> 0 compaction rate here to twelve digits. Step 1899 of case A, snow falling onto initially bare frozen ground. SNOWPACK's melt-water retention branch RUNS -- the snow surface is at or above freezing; neither density limit fires. The dt -> 0 compaction rate is 6.5045e-09 per second, and the 60 s regime's tendency for the same step exceeds it by xarg/3 = 2.57e-06 relative.",
      "parameter_overrides": {
        "esd": 0.003366285702196814,
        "sndens_in": 0.10116377383023133,
        "tsnow": 276.9127770651018,
        "tsoil": 272.8987805846188,
        "dtsec": 1e-06
      },
      "time_span": {
        "start": 0.0,
        "end": 0.0
      },
      "assertions": [
        {
          "variable": "bfac",
          "time": 0.0,
          "expected": 3.820034278776774e-13,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "xarg",
          "time": 0.0,
          "expected": 1.2859326774547972e-13,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "pexp_m1",
          "time": 0.0,
          "expected": 6.429663387274261e-14,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsx",
          "time": 0.0,
          "expected": 0.10116377383023784,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dw",
          "time": 0.0,
          "expected": 1.5046296296296297e-12,
          "tolerance": {
            "rel": 1e-09,
            "abs": 1e-30
          }
        },
        {
          "variable": "sndens_out",
          "time": 0.0,
          "expected": 0.10116377383159025,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsndens_dt",
          "time": 0.0,
          "expected": 1.358920108206651e-06,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "pack_present",
          "time": 0.0,
          "expected": 1.0,
          "tolerance": {
            "abs": 0.0,
            "rel": 0.0
          }
        }
      ]
    },
    {
      "id": "cold_tiny_caseA_step1924",
      "description": "Reference: the same cold-season run and the same real64 kernels/noah_driver, replayed at ESM_DT = 1e-6 s, at rel 1e-9. THE POINT OF THIS REGIME is that it is exactly where a naively written compaction component fails: the density increment is 7.5e-17 of the density, so both sndens_out - sndens_in and sndens (pexp - 1) return pure round-off, and only the cancellation-free form -- sndens_in pexp_m1 + dw (1 - dsx), referenced against the series sum published before WRF adds one to it -- still carries the signal. `dsndens_dt` is the dt -> 0 compaction rate here to twelve digits. Step 1924 of case A, snow falling onto initially bare frozen ground. SNOWPACK's melt-water retention branch RUNS -- the snow surface is at or above freezing; neither density limit fires. The dt -> 0 compaction rate is 6.1001e-09 per second, and the 60 s regime's tendency for the same step exceeds it by xarg/3 = 2.36e-06 relative.",
      "parameter_overrides": {
        "esd": 0.0033420040272028775,
        "sndens_in": 0.10320220070394324,
        "tsnow": 276.27792321355463,
        "tsoil": 272.6813712485515,
        "dtsec": 1e-06
      },
      "time_span": {
        "start": 0.0,
        "end": 0.0
      },
      "assertions": [
        {
          "variable": "bfac",
          "time": 0.0,
          "expected": 3.537293561974355e-13,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "xarg",
          "time": 0.0,
          "expected": 1.1821649329517105e-13,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "pexp_m1",
          "time": 0.0,
          "expected": 5.910824664758784e-14,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsx",
          "time": 0.0,
          "expected": 0.10320220070394934,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dw",
          "time": 0.0,
          "expected": 1.5046296296296297e-12,
          "tolerance": {
            "rel": 1e-09,
            "abs": 1e-30
          }
        },
        {
          "variable": "sndens_out",
          "time": 0.0,
          "expected": 0.10320220070529869,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsndens_dt",
          "time": 0.0,
          "expected": 1.3554486417412664e-06,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "pack_present",
          "time": 0.0,
          "expected": 1.0,
          "tolerance": {
            "abs": 0.0,
            "rel": 0.0
          }
        }
      ]
    },
    {
      "id": "cold_tiny_caseA_step2089",
      "description": "Reference: the same cold-season run and the same real64 kernels/noah_driver, replayed at ESM_DT = 1e-6 s, at rel 1e-9. THE POINT OF THIS REGIME is that it is exactly where a naively written compaction component fails: the density increment is 7.5e-17 of the density, so both sndens_out - sndens_in and sndens (pexp - 1) return pure round-off, and only the cancellation-free form -- sndens_in pexp_m1 + dw (1 - dsx), referenced against the series sum published before WRF adds one to it -- still carries the signal. `dsndens_dt` is the dt -> 0 compaction rate here to twelve digits. Step 2089 of case A, snow falling onto initially bare frozen ground. SNOWPACK's melt-water retention branch RUNS -- the snow surface is at or above freezing; neither density limit fires. The dt -> 0 compaction rate is 5.2737e-09 per second, and the 60 s regime's tendency for the same step exceeds it by xarg/3 = 1.92e-06 relative.",
      "parameter_overrides": {
        "esd": 0.0033869065813832352,
        "sndens_in": 0.11037496454291075,
        "tsnow": 274.7809347429996,
        "tsoil": 272.29122644070054,
        "dtsec": 1e-06
      },
      "time_span": {
        "start": 0.0,
        "end": 0.0
      },
      "assertions": [
        {
          "variable": "bfac",
          "time": 0.0,
          "expected": 2.821443215191794e-13,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "xarg",
          "time": 0.0,
          "expected": 9.555964594532164e-14,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "pexp_m1",
          "time": 0.0,
          "expected": 4.7779822972662334e-14,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsx",
          "time": 0.0,
          "expected": 0.11037496454291602,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dw",
          "time": 0.0,
          "expected": 1.5046296296296297e-12,
          "tolerance": {
            "rel": 1e-09,
            "abs": 1e-30
          }
        },
        {
          "variable": "sndens_out",
          "time": 0.0,
          "expected": 0.11037496454425459,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsndens_dt",
          "time": 0.0,
          "expected": 1.3438298838755126e-06,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "pack_present",
          "time": 0.0,
          "expected": 1.0,
          "tolerance": {
            "abs": 0.0,
            "rel": 0.0
          }
        }
      ]
    },
    {
      "id": "cold_tiny_caseA_step2114",
      "description": "Reference: the same cold-season run and the same real64 kernels/noah_driver, replayed at ESM_DT = 1e-6 s, at rel 1e-9. THE POINT OF THIS REGIME is that it is exactly where a naively written compaction component fails: the density increment is 7.5e-17 of the density, so both sndens_out - sndens_in and sndens (pexp - 1) return pure round-off, and only the cancellation-free form -- sndens_in pexp_m1 + dw (1 - dsx), referenced against the series sum published before WRF adds one to it -- still carries the signal. `dsndens_dt` is the dt -> 0 compaction rate here to twelve digits. Step 2114 of case A, snow falling onto initially bare frozen ground. dry branch -- the snow surface is below freezing and only compaction acts; neither density limit fires. The dt -> 0 compaction rate is 3.1268e-09 per second, and the 60 s regime's tendency for the same step exceeds it by xarg/3 = 1.13e-06 relative.",
      "parameter_overrides": {
        "esd": 0.003592143068162672,
        "sndens_in": 0.11054866993095654,
        "tsnow": 265.81458701581164,
        "tsoil": 266.7709355885183,
        "dtsec": 1e-06
      },
      "time_span": {
        "start": 0.0,
        "end": 0.0
      },
      "assertions": [
        {
          "variable": "bfac",
          "time": 0.0,
          "expected": 1.5748108599553997e-13,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "xarg",
          "time": 0.0,
          "expected": 5.6569459142560856e-14,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "pexp_m1",
          "time": 0.0,
          "expected": 2.828472957128096e-14,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsx",
          "time": 0.0,
          "expected": 0.11054866993095966,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dw",
          "time": 0.0,
          "expected": 0.0,
          "tolerance": {
            "rel": 1e-09,
            "abs": 1e-30
          }
        },
        {
          "variable": "sndens_out",
          "time": 0.0,
          "expected": 0.11054866993095966,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsndens_dt",
          "time": 0.0,
          "expected": 3.126839233461905e-09,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "pack_present",
          "time": 0.0,
          "expected": 1.0,
          "tolerance": {
            "abs": 0.0,
            "rel": 0.0
          }
        }
      ]
    },
    {
      "id": "cold_tiny_caseA_step2847",
      "description": "Reference: the same cold-season run and the same real64 kernels/noah_driver, replayed at ESM_DT = 1e-6 s, at rel 1e-9. THE POINT OF THIS REGIME is that it is exactly where a naively written compaction component fails: the density increment is 7.5e-17 of the density, so both sndens_out - sndens_in and sndens (pexp - 1) return pure round-off, and only the cancellation-free form -- sndens_in pexp_m1 + dw (1 - dsx), referenced against the series sum published before WRF adds one to it -- still carries the signal. `dsndens_dt` is the dt -> 0 compaction rate here to twelve digits. Step 2847 of case A, snow falling onto initially bare frozen ground. dry branch -- the snow surface is below freezing and only compaction acts; neither density limit fires. The dt -> 0 compaction rate is 2.6094e-09 per second, and the 60 s regime's tendency for the same step exceeds it by xarg/3 = 9.45e-07 relative.",
      "parameter_overrides": {
        "esd": 0.0036658388562500403,
        "sndens_in": 0.11048201772600909,
        "tsnow": 263.26987657273827,
        "tsoil": 264.26561775220716,
        "dtsec": 1e-06
      },
      "time_span": {
        "start": 0.0,
        "end": 0.0
      },
      "assertions": [
        {
          "variable": "bfac",
          "time": 0.0,
          "expected": 1.288570856778855e-13,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "xarg",
          "time": 0.0,
          "expected": 4.723693115811332e-14,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "pexp_m1",
          "time": 0.0,
          "expected": 2.3618465579057033e-14,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsx",
          "time": 0.0,
          "expected": 0.11048201772601168,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dw",
          "time": 0.0,
          "expected": 0.0,
          "tolerance": {
            "rel": 1e-09,
            "abs": 1e-30
          }
        },
        {
          "variable": "sndens_out",
          "time": 0.0,
          "expected": 0.11048201772601168,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsndens_dt",
          "time": 0.0,
          "expected": 2.6094157327665147e-09,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "pack_present",
          "time": 0.0,
          "expected": 1.0,
          "tolerance": {
            "abs": 0.0,
            "rel": 0.0
          }
        }
      ]
    },
    {
      "id": "cold_tiny_caseA_step2880",
      "description": "Reference: the same cold-season run and the same real64 kernels/noah_driver, replayed at ESM_DT = 1e-6 s, at rel 1e-9. THE POINT OF THIS REGIME is that it is exactly where a naively written compaction component fails: the density increment is 7.5e-17 of the density, so both sndens_out - sndens_in and sndens (pexp - 1) return pure round-off, and only the cancellation-free form -- sndens_in pexp_m1 + dw (1 - dsx), referenced against the series sum published before WRF adds one to it -- still carries the signal. `dsndens_dt` is the dt -> 0 compaction rate here to twelve digits. Step 2880 of case A, snow falling onto initially bare frozen ground. dry branch -- the snow surface is below freezing and only compaction acts; neither density limit fires. The dt -> 0 compaction rate is 2.6103e-09 per second, and the 60 s regime's tendency for the same step exceeds it by xarg/3 = 9.45e-07 relative.",
      "parameter_overrides": {
        "esd": 0.003665826749056569,
        "sndens_in": 0.11048748311453574,
        "tsnow": 263.2923388137898,
        "tsoil": 264.2530140451615,
        "dtsec": 1e-06
      },
      "time_span": {
        "start": 0.0,
        "end": 0.0
      },
      "assertions": [
        {
          "variable": "bfac",
          "time": 0.0,
          "expected": 1.2889311505763677e-13,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "xarg",
          "time": 0.0,
          "expected": 4.724998289475109e-14,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "pexp_m1",
          "time": 0.0,
          "expected": 2.3624991447375914e-14,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsx",
          "time": 0.0,
          "expected": 0.11048748311453833,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dw",
          "time": 0.0,
          "expected": 0.0,
          "tolerance": {
            "rel": 1e-09,
            "abs": 1e-30
          }
        },
        {
          "variable": "sndens_out",
          "time": 0.0,
          "expected": 0.11048748311453833,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsndens_dt",
          "time": 0.0,
          "expected": 2.6102658436229977e-09,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "pack_present",
          "time": 0.0,
          "expected": 1.0,
          "tolerance": {
            "abs": 0.0,
            "rel": 0.0
          }
        }
      ]
    },
    {
      "id": "cold_tiny_caseB_step1",
      "description": "Reference: the same cold-season run and the same real64 kernels/noah_driver, replayed at ESM_DT = 1e-6 s, at rel 1e-9. THE POINT OF THIS REGIME is that it is exactly where a naively written compaction component fails: the density increment is 7.5e-17 of the density, so both sndens_out - sndens_in and sndens (pexp - 1) return pure round-off, and only the cancellation-free form -- sndens_in pexp_m1 + dw (1 - dsx), referenced against the series sum published before WRF adds one to it -- still carries the signal. `dsndens_dt` is the dt -> 0 compaction rate here to twelve digits. Step 1 of case B, a 30 cm pack melting under warm advection. dry branch -- the snow surface is below freezing and only compaction acts; neither density limit fires. The dt -> 0 compaction rate is 9.9079e-09 per second, and the 60 s regime's tendency for the same step exceeds it by xarg/3 = 1.98e-06 relative.",
      "parameter_overrides": {
        "esd": 0.06000000238419379,
        "sndens_in": 0.2000000198682161,
        "tsnow": 260.52261495701777,
        "tsoil": 262.64608782840094,
        "dtsec": 1e-06
      },
      "time_span": {
        "start": 0.0,
        "end": 0.0
      },
      "assertions": [
        {
          "variable": "bfac",
          "time": 0.0,
          "expected": 1.651310355693894e-14,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "xarg",
          "time": 0.0,
          "expected": 9.907862527867754e-14,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "pexp_m1",
          "time": 0.0,
          "expected": 4.9539312639340405e-14,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsx",
          "time": 0.0,
          "expected": 0.20000001986822602,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dw",
          "time": 0.0,
          "expected": 0.0,
          "tolerance": {
            "rel": 1e-09,
            "abs": 1e-30
          }
        },
        {
          "variable": "sndens_out",
          "time": 0.0,
          "expected": 0.20000001986822602,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsndens_dt",
          "time": 0.0,
          "expected": 9.907863512125852e-09,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "pack_present",
          "time": 0.0,
          "expected": 1.0,
          "tolerance": {
            "abs": 0.0,
            "rel": 0.0
          }
        }
      ]
    },
    {
      "id": "cold_tiny_caseB_step26",
      "description": "Reference: the same cold-season run and the same real64 kernels/noah_driver, replayed at ESM_DT = 1e-6 s, at rel 1e-9. THE POINT OF THIS REGIME is that it is exactly where a naively written compaction component fails: the density increment is 7.5e-17 of the density, so both sndens_out - sndens_in and sndens (pexp - 1) return pure round-off, and only the cancellation-free form -- sndens_in pexp_m1 + dw (1 - dsx), referenced against the series sum published before WRF adds one to it -- still carries the signal. `dsndens_dt` is the dt -> 0 compaction rate here to twelve digits. Step 26 of case B, a 30 cm pack melting under warm advection. dry branch -- the snow surface is below freezing and only compaction acts; neither density limit fires. The dt -> 0 compaction rate is 1.2031e-08 per second, and the 60 s regime's tendency for the same step exceeds it by xarg/3 = 2.41e-06 relative.",
      "parameter_overrides": {
        "esd": 0.06000409275294473,
        "sndens_in": 0.20000203140698358,
        "tsnow": 264.7685956420138,
        "tsoil": 263.2526439516071,
        "dtsec": 1e-06
      },
      "time_span": {
        "start": 0.0,
        "end": 0.0
      },
      "assertions": [
        {
          "variable": "bfac",
          "time": 0.0,
          "expected": 2.0049685156514463e-14,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "xarg",
          "time": 0.0,
          "expected": 1.203063167798833e-13,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "pexp_m1",
          "time": 0.0,
          "expected": 6.015315838994406e-14,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsx",
          "time": 0.0,
          "expected": 0.20000203140699563,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dw",
          "time": 0.0,
          "expected": 0.0,
          "tolerance": {
            "rel": 1e-09,
            "abs": 1e-30
          }
        },
        {
          "variable": "sndens_out",
          "time": 0.0,
          "expected": 0.20000203140699563,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsndens_dt",
          "time": 0.0,
          "expected": 1.203075387353485e-08,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "pack_present",
          "time": 0.0,
          "expected": 1.0,
          "tolerance": {
            "abs": 0.0,
            "rel": 0.0
          }
        }
      ]
    },
    {
      "id": "cold_tiny_caseB_step51",
      "description": "Reference: the same cold-season run and the same real64 kernels/noah_driver, replayed at ESM_DT = 1e-6 s, at rel 1e-9. THE POINT OF THIS REGIME is that it is exactly where a naively written compaction component fails: the density increment is 7.5e-17 of the density, so both sndens_out - sndens_in and sndens (pexp - 1) return pure round-off, and only the cancellation-free form -- sndens_in pexp_m1 + dw (1 - dsx), referenced against the series sum published before WRF adds one to it -- still carries the signal. `dsndens_dt` is the dt -> 0 compaction rate here to twelve digits. Step 51 of case B, a 30 cm pack melting under warm advection. dry branch -- the snow surface is below freezing and only compaction acts; neither density limit fires. The dt -> 0 compaction rate is 1.2165e-08 per second, and the 60 s regime's tendency for the same step exceeds it by xarg/3 = 2.44e-06 relative.",
      "parameter_overrides": {
        "esd": 0.06008230149754566,
        "sndens_in": 0.19962168444099768,
        "tsnow": 264.8497609014917,
        "tsoil": 263.2635130945321,
        "dtsec": 1e-06
      },
      "time_span": {
        "start": 0.0,
        "end": 0.0
      },
      "assertions": [
        {
          "variable": "bfac",
          "time": 0.0,
          "expected": 2.0285008447400137e-14,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "xarg",
          "time": 0.0,
          "expected": 1.2187699934169557e-13,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "pexp_m1",
          "time": 0.0,
          "expected": 6.093849967085026e-14,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsx",
          "time": 0.0,
          "expected": 0.19962168444100983,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dw",
          "time": 0.0,
          "expected": 0.0,
          "tolerance": {
            "rel": 1e-09,
            "abs": 1e-30
          }
        },
        {
          "variable": "sndens_out",
          "time": 0.0,
          "expected": 0.19962168444100983,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsndens_dt",
          "time": 0.0,
          "expected": 1.216464595160231e-08,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "pack_present",
          "time": 0.0,
          "expected": 1.0,
          "tolerance": {
            "abs": 0.0,
            "rel": 0.0
          }
        }
      ]
    },
    {
      "id": "cold_tiny_caseB_step100",
      "description": "Reference: the same cold-season run and the same real64 kernels/noah_driver, replayed at ESM_DT = 1e-6 s, at rel 1e-9. THE POINT OF THIS REGIME is that it is exactly where a naively written compaction component fails: the density increment is 7.5e-17 of the density, so both sndens_out - sndens_in and sndens (pexp - 1) return pure round-off, and only the cancellation-free form -- sndens_in pexp_m1 + dw (1 - dsx), referenced against the series sum published before WRF adds one to it -- still carries the signal. `dsndens_dt` is the dt -> 0 compaction rate here to twelve digits. Step 100 of case B, a 30 cm pack melting under warm advection. dry branch -- the snow surface is below freezing and only compaction acts; neither density limit fires. The dt -> 0 compaction rate is 1.2390e-08 per second, and the 60 s regime's tendency for the same step exceeds it by xarg/3 = 2.50e-06 relative.",
      "parameter_overrides": {
        "esd": 0.06030664220457646,
        "sndens_in": 0.1985316538739585,
        "tsnow": 264.78206158424587,
        "tsoil": 263.2608076540921,
        "dtsec": 1e-06
      },
      "time_span": {
        "start": 0.0,
        "end": 0.0
      },
      "assertions": [
        {
          "variable": "bfac",
          "time": 0.0,
          "expected": 2.0696333618292883e-14,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "xarg",
          "time": 0.0,
          "expected": 1.2481263864649362e-13,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "pexp_m1",
          "time": 0.0,
          "expected": 6.24063193232494e-14,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsx",
          "time": 0.0,
          "expected": 0.19853165387397087,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dw",
          "time": 0.0,
          "expected": 0.0,
          "tolerance": {
            "rel": 1e-09,
            "abs": 1e-30
          }
        },
        {
          "variable": "sndens_out",
          "time": 0.0,
          "expected": 0.19853165387397087,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsndens_dt",
          "time": 0.0,
          "expected": 1.2389629787431077e-08,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "pack_present",
          "time": 0.0,
          "expected": 1.0,
          "tolerance": {
            "abs": 0.0,
            "rel": 0.0
          }
        }
      ]
    },
    {
      "id": "cold_tiny_caseB_step435",
      "description": "Reference: the same cold-season run and the same real64 kernels/noah_driver, replayed at ESM_DT = 1e-6 s, at rel 1e-9. THE POINT OF THIS REGIME is that it is exactly where a naively written compaction component fails: the density increment is 7.5e-17 of the density, so both sndens_out - sndens_in and sndens (pexp - 1) return pure round-off, and only the cancellation-free form -- sndens_in pexp_m1 + dw (1 - dsx), referenced against the series sum published before WRF adds one to it -- still carries the signal. `dsndens_dt` is the dt -> 0 compaction rate here to twelve digits. Step 435 of case B, a 30 cm pack melting under warm advection. dry branch -- the snow surface is below freezing and only compaction acts; neither density limit fires. The dt -> 0 compaction rate is 1.7168e-08 per second, and the 60 s regime's tendency for the same step exceeds it by xarg/3 = 3.51e-06 relative.",
      "parameter_overrides": {
        "esd": 0.06073277816205019,
        "sndens_in": 0.19603108392743757,
        "tsnow": 270.63745141127583,
        "tsoil": 264.38843147109196,
        "dtsec": 1e-06
      },
      "time_span": {
        "start": 0.0,
        "end": 0.0
      },
      "assertions": [
        {
          "variable": "bfac",
          "time": 0.0,
          "expected": 2.884074312742842e-14,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "xarg",
          "time": 0.0,
          "expected": 1.7515784543867834e-13,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "pexp_m1",
          "time": 0.0,
          "expected": 8.75789227193443e-14,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsx",
          "time": 0.0,
          "expected": 0.19603108392745472,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dw",
          "time": 0.0,
          "expected": 0.0,
          "tolerance": {
            "rel": 1e-09,
            "abs": 1e-30
          }
        },
        {
          "variable": "sndens_out",
          "time": 0.0,
          "expected": 0.19603108392745472,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsndens_dt",
          "time": 0.0,
          "expected": 1.716819114987035e-08,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "pack_present",
          "time": 0.0,
          "expected": 1.0,
          "tolerance": {
            "abs": 0.0,
            "rel": 0.0
          }
        }
      ]
    },
    {
      "id": "cold_tiny_caseB_step463",
      "description": "Reference: the same cold-season run and the same real64 kernels/noah_driver, replayed at ESM_DT = 1e-6 s, at rel 1e-9. THE POINT OF THIS REGIME is that it is exactly where a naively written compaction component fails: the density increment is 7.5e-17 of the density, so both sndens_out - sndens_in and sndens (pexp - 1) return pure round-off, and only the cancellation-free form -- sndens_in pexp_m1 + dw (1 - dsx), referenced against the series sum published before WRF adds one to it -- still carries the signal. `dsndens_dt` is the dt -> 0 compaction rate here to twelve digits. Step 463 of case B, a 30 cm pack melting under warm advection. dry branch -- the snow surface is below freezing and only compaction acts; neither density limit fires. The dt -> 0 compaction rate is 1.8064e-08 per second, and the 60 s regime's tendency for the same step exceeds it by xarg/3 = 3.69e-06 relative.",
      "parameter_overrides": {
        "esd": 0.06077174469828753,
        "sndens_in": 0.19576132745810296,
        "tsnow": 271.61478129558895,
        "tsoil": 264.55907945374287,
        "dtsec": 1e-06
      },
      "time_span": {
        "start": 0.0,
        "end": 0.0
      },
      "assertions": [
        {
          "variable": "bfac",
          "time": 0.0,
          "expected": 3.0367503713184553e-14,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "xarg",
          "time": 0.0,
          "expected": 1.8454861827819503e-13,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "pexp_m1",
          "time": 0.0,
          "expected": 9.22743091391032e-14,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsx",
          "time": 0.0,
          "expected": 0.19576132745812103,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dw",
          "time": 0.0,
          "expected": 0.0,
          "tolerance": {
            "rel": 1e-09,
            "abs": 1e-30
          }
        },
        {
          "variable": "sndens_out",
          "time": 0.0,
          "expected": 0.19576132745812103,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsndens_dt",
          "time": 0.0,
          "expected": 1.8063741247350206e-08,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "pack_present",
          "time": 0.0,
          "expected": 1.0,
          "tolerance": {
            "abs": 0.0,
            "rel": 0.0
          }
        }
      ]
    },
    {
      "id": "cold_tiny_caseB_step492",
      "description": "Reference: the same cold-season run and the same real64 kernels/noah_driver, replayed at ESM_DT = 1e-6 s, at rel 1e-9. THE POINT OF THIS REGIME is that it is exactly where a naively written compaction component fails: the density increment is 7.5e-17 of the density, so both sndens_out - sndens_in and sndens (pexp - 1) return pure round-off, and only the cancellation-free form -- sndens_in pexp_m1 + dw (1 - dsx), referenced against the series sum published before WRF adds one to it -- still carries the signal. `dsndens_dt` is the dt -> 0 compaction rate here to twelve digits. Step 492 of case B, a 30 cm pack melting under warm advection. dry branch -- the snow surface is below freezing and only compaction acts; neither density limit fires. The dt -> 0 compaction rate is 1.7718e-08 per second, and the 60 s regime's tendency for the same step exceeds it by xarg/3 = 3.62e-06 relative.",
      "parameter_overrides": {
        "esd": 0.060786481946686254,
        "sndens_in": 0.19557859079108367,
        "tsnow": 271.0819562615801,
        "tsoil": 264.5298843435946,
        "dtsec": 1e-06
      },
      "time_span": {
        "start": 0.0,
        "end": 0.0
      },
      "assertions": [
        {
          "variable": "bfac",
          "time": 0.0,
          "expected": 2.9806596977885914e-14,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "xarg",
          "time": 0.0,
          "expected": 1.811838169088415e-13,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "pexp_m1",
          "time": 0.0,
          "expected": 9.059190845442623e-14,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsx",
          "time": 0.0,
          "expected": 0.19557859079110138,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dw",
          "time": 0.0,
          "expected": 0.0,
          "tolerance": {
            "rel": 1e-09,
            "abs": 1e-30
          }
        },
        {
          "variable": "sndens_out",
          "time": 0.0,
          "expected": 0.19557859079110138,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsndens_dt",
          "time": 0.0,
          "expected": 1.7717837792591544e-08,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "pack_present",
          "time": 0.0,
          "expected": 1.0,
          "tolerance": {
            "abs": 0.0,
            "rel": 0.0
          }
        }
      ]
    },
    {
      "id": "cold_tiny_caseB_step517",
      "description": "Reference: the same cold-season run and the same real64 kernels/noah_driver, replayed at ESM_DT = 1e-6 s, at rel 1e-9. THE POINT OF THIS REGIME is that it is exactly where a naively written compaction component fails: the density increment is 7.5e-17 of the density, so both sndens_out - sndens_in and sndens (pexp - 1) return pure round-off, and only the cancellation-free form -- sndens_in pexp_m1 + dw (1 - dsx), referenced against the series sum published before WRF adds one to it -- still carries the signal. `dsndens_dt` is the dt -> 0 compaction rate here to twelve digits. Step 517 of case B, a 30 cm pack melting under warm advection. dry branch -- the snow surface is below freezing and only compaction acts; neither density limit fires. The dt -> 0 compaction rate is 1.6791e-08 per second, and the 60 s regime's tendency for the same step exceeds it by xarg/3 = 3.44e-06 relative.",
      "parameter_overrides": {
        "esd": 0.060780186206587865,
        "sndens_in": 0.19549596741898956,
        "tsnow": 269.8456360508011,
        "tsoil": 264.3927261317292,
        "dtsec": 1e-06
      },
      "time_span": {
        "start": 0.0,
        "end": 0.0
      },
      "assertions": [
        {
          "variable": "bfac",
          "time": 0.0,
          "expected": 2.8262213431360096e-14,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "xarg",
          "time": 0.0,
          "expected": 1.7177825949683953e-13,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "pexp_m1",
          "time": 0.0,
          "expected": 8.588912974842468e-14,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsx",
          "time": 0.0,
          "expected": 0.19549596741900635,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dw",
          "time": 0.0,
          "expected": 0.0,
          "tolerance": {
            "rel": 1e-09,
            "abs": 1e-30
          }
        },
        {
          "variable": "sndens_out",
          "time": 0.0,
          "expected": 0.19549596741900635,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsndens_dt",
          "time": 0.0,
          "expected": 1.67909785109434e-08,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "pack_present",
          "time": 0.0,
          "expected": 1.0,
          "tolerance": {
            "abs": 0.0,
            "rel": 0.0
          }
        }
      ]
    },
    {
      "id": "cold_tiny_caseB_step561",
      "description": "Reference: the same cold-season run and the same real64 kernels/noah_driver, replayed at ESM_DT = 1e-6 s, at rel 1e-9. THE POINT OF THIS REGIME is that it is exactly where a naively written compaction component fails: the density increment is 7.5e-17 of the density, so both sndens_out - sndens_in and sndens (pexp - 1) return pure round-off, and only the cancellation-free form -- sndens_in pexp_m1 + dw (1 - dsx), referenced against the series sum published before WRF adds one to it -- still carries the signal. `dsndens_dt` is the dt -> 0 compaction rate here to twelve digits. Step 561 of case B, a 30 cm pack melting under warm advection. dry branch -- the snow surface is below freezing and only compaction acts; neither density limit fires. The dt -> 0 compaction rate is 1.7752e-08 per second, and the 60 s regime's tendency for the same step exceeds it by xarg/3 = 3.64e-06 relative.",
      "parameter_overrides": {
        "esd": 0.06083413213490516,
        "sndens_in": 0.19515089514893563,
        "tsnow": 270.8822675984362,
        "tsoil": 264.58803047309226,
        "dtsec": 1e-06
      },
      "time_span": {
        "start": 0.0,
        "end": 0.0
      },
      "assertions": [
        {
          "variable": "bfac",
          "time": 0.0,
          "expected": 2.9905716554845914e-14,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "xarg",
          "time": 0.0,
          "expected": 1.819288312486517e-13,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "pexp_m1",
          "time": 0.0,
          "expected": 9.096441562433136e-14,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsx",
          "time": 0.0,
          "expected": 0.1951508951489534,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dw",
          "time": 0.0,
          "expected": 0.0,
          "tolerance": {
            "rel": 1e-09,
            "abs": 1e-30
          }
        },
        {
          "variable": "sndens_out",
          "time": 0.0,
          "expected": 0.1951508951489534,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsndens_dt",
          "time": 0.0,
          "expected": 1.7751787135788094e-08,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "pack_present",
          "time": 0.0,
          "expected": 1.0,
          "tolerance": {
            "abs": 0.0,
            "rel": 0.0
          }
        }
      ]
    },
    {
      "id": "cold_tiny_caseB_step658",
      "description": "Reference: the same cold-season run and the same real64 kernels/noah_driver, replayed at ESM_DT = 1e-6 s, at rel 1e-9. THE POINT OF THIS REGIME is that it is exactly where a naively written compaction component fails: the density increment is 7.5e-17 of the density, so both sndens_out - sndens_in and sndens (pexp - 1) return pure round-off, and only the cancellation-free form -- sndens_in pexp_m1 + dw (1 - dsx), referenced against the series sum published before WRF adds one to it -- still carries the signal. `dsndens_dt` is the dt -> 0 compaction rate here to twelve digits. Step 658 of case B, a 30 cm pack melting under warm advection. dry branch -- the snow surface is below freezing and only compaction acts; neither density limit fires. The dt -> 0 compaction rate is 1.2833e-08 per second, and the 60 s regime's tendency for the same step exceeds it by xarg/3 = 2.63e-06 relative.",
      "parameter_overrides": {
        "esd": 0.060759976506233285,
        "sndens_in": 0.19523935840389106,
        "tsnow": 263.6961309154277,
        "tsoil": 263.72845004329685,
        "dtsec": 1e-06
      },
      "time_span": {
        "start": 0.0,
        "end": 0.0
      },
      "assertions": [
        {
          "variable": "bfac",
          "time": 0.0,
          "expected": 2.1636100485870487e-14,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "xarg",
          "time": 0.0,
          "expected": 1.3146089572079932e-13,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "pexp_m1",
          "time": 0.0,
          "expected": 6.573044786040254e-14,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsx",
          "time": 0.0,
          "expected": 0.19523935840390388,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dw",
          "time": 0.0,
          "expected": 0.0,
          "tolerance": {
            "rel": 1e-09,
            "abs": 1e-30
          }
        },
        {
          "variable": "sndens_out",
          "time": 0.0,
          "expected": 0.19523935840390388,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsndens_dt",
          "time": 0.0,
          "expected": 1.2833170467865408e-08,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "pack_present",
          "time": 0.0,
          "expected": 1.0,
          "tolerance": {
            "abs": 0.0,
            "rel": 0.0
          }
        }
      ]
    },
    {
      "id": "cold_tiny_caseB_step1612",
      "description": "Reference: the same cold-season run and the same real64 kernels/noah_driver, replayed at ESM_DT = 1e-6 s, at rel 1e-9. THE POINT OF THIS REGIME is that it is exactly where a naively written compaction component fails: the density increment is 7.5e-17 of the density, so both sndens_out - sndens_in and sndens (pexp - 1) return pure round-off, and only the cancellation-free form -- sndens_in pexp_m1 + dw (1 - dsx), referenced against the series sum published before WRF adds one to it -- still carries the signal. `dsndens_dt` is the dt -> 0 compaction rate here to twelve digits. Step 1612 of case B, a 30 cm pack melting under warm advection. SNOWPACK's melt-water retention branch RUNS -- the snow surface is at or above freezing; neither density limit fires. The dt -> 0 compaction rate is 1.9210e-08 per second, and the 60 s regime's tendency for the same step exceeds it by xarg/3 = 3.90e-06 relative.",
      "parameter_overrides": {
        "esd": 0.06072096899143159,
        "sndens_in": 0.19683397677243142,
        "tsnow": 273.335639338635,
        "tsoil": 264.8236325232234,
        "dtsec": 1e-06
      },
      "time_span": {
        "start": 0.0,
        "end": 0.0
      },
      "assertions": [
        {
          "variable": "bfac",
          "time": 0.0,
          "expected": 3.214522408916119e-14,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "xarg",
          "time": 0.0,
          "expected": 1.9518891551405764e-13,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "pexp_m1",
          "time": 0.0,
          "expected": 9.759445775703517e-14,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsx",
          "time": 0.0,
          "expected": 0.19683397677245065,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dw",
          "time": 0.0,
          "expected": 1.5046296296296297e-12,
          "tolerance": {
            "rel": 1e-09,
            "abs": 1e-30
          }
        },
        {
          "variable": "sndens_out",
          "time": 0.0,
          "expected": 0.19683397677365913,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsndens_dt",
          "time": 0.0,
          "expected": 1.2276773012912367e-06,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "pack_present",
          "time": 0.0,
          "expected": 1.0,
          "tolerance": {
            "abs": 0.0,
            "rel": 0.0
          }
        }
      ]
    },
    {
      "id": "cold_tiny_caseB_step1637",
      "description": "Reference: the same cold-season run and the same real64 kernels/noah_driver, replayed at ESM_DT = 1e-6 s, at rel 1e-9. THE POINT OF THIS REGIME is that it is exactly where a naively written compaction component fails: the density increment is 7.5e-17 of the density, so both sndens_out - sndens_in and sndens (pexp - 1) return pure round-off, and only the cancellation-free form -- sndens_in pexp_m1 + dw (1 - dsx), referenced against the series sum published before WRF adds one to it -- still carries the signal. `dsndens_dt` is the dt -> 0 compaction rate here to twelve digits. Step 1637 of case B, a 30 cm pack melting under warm advection. SNOWPACK's melt-water retention branch RUNS -- the snow surface is at or above freezing; neither density limit fires. The dt -> 0 compaction rate is 1.8853e-08 per second, and the 60 s regime's tendency for the same step exceeds it by xarg/3 = 3.79e-06 relative.",
      "parameter_overrides": {
        "esd": 0.06049393862463843,
        "sndens_in": 0.1986732035637152,
        "tsnow": 273.6358408786153,
        "tsoil": 264.8815372959442,
        "dtsec": 1e-06
      },
      "time_span": {
        "start": 0.0,
        "end": 0.0
      },
      "assertions": [
        {
          "variable": "bfac",
          "time": 0.0,
          "expected": 3.137352482003584e-14,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "xarg",
          "time": 0.0,
          "expected": 1.8979080849018186e-13,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "pexp_m1",
          "time": 0.0,
          "expected": 9.489540424509695e-14,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsx",
          "time": 0.0,
          "expected": 0.19867320356373405,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dw",
          "time": 0.0,
          "expected": 1.5046296296296297e-12,
          "tolerance": {
            "rel": 1e-09,
            "abs": 1e-30
          }
        },
        {
          "variable": "sndens_out",
          "time": 0.0,
          "expected": 0.19867320356493975,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsndens_dt",
          "time": 0.0,
          "expected": 1.2245532148990437e-06,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "pack_present",
          "time": 0.0,
          "expected": 1.0,
          "tolerance": {
            "abs": 0.0,
            "rel": 0.0
          }
        }
      ]
    },
    {
      "id": "cold_tiny_caseB_step1816",
      "description": "Reference: the same cold-season run and the same real64 kernels/noah_driver, replayed at ESM_DT = 1e-6 s, at rel 1e-9. THE POINT OF THIS REGIME is that it is exactly where a naively written compaction component fails: the density increment is 7.5e-17 of the density, so both sndens_out - sndens_in and sndens (pexp - 1) return pure round-off, and only the cancellation-free form -- sndens_in pexp_m1 + dw (1 - dsx), referenced against the series sum published before WRF adds one to it -- still carries the signal. `dsndens_dt` is the dt -> 0 compaction rate here to twelve digits. Step 1816 of case B, a 30 cm pack melting under warm advection. SNOWPACK's melt-water retention branch RUNS -- the snow surface is at or above freezing; neither density limit fires. The dt -> 0 compaction rate is 1.4970e-08 per second, and the 60 s regime's tendency for the same step exceeds it by xarg/3 = 2.83e-06 relative.",
      "parameter_overrides": {
        "esd": 0.05435399711052939,
        "sndens_in": 0.21170159531489063,
        "tsnow": 274.907254172599,
        "tsoil": 265.7723157520076,
        "dtsec": 1e-06
      },
      "time_span": {
        "start": 0.0,
        "end": 0.0
      },
      "assertions": [
        {
          "variable": "bfac",
          "time": 0.0,
          "expected": 2.6019748877022202e-14,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "xarg",
          "time": 0.0,
          "expected": 1.4142773552783652e-13,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "pexp_m1",
          "time": 0.0,
          "expected": 7.071386776392159e-14,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsx",
          "time": 0.0,
          "expected": 0.2117015953149056,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dw",
          "time": 0.0,
          "expected": 1.5046296296296297e-12,
          "tolerance": {
            "rel": 1e-09,
            "abs": 1e-30
          }
        },
        {
          "variable": "sndens_out",
          "time": 0.0,
          "expected": 0.2117015953160917,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsndens_dt",
          "time": 0.0,
          "expected": 1.20106737529547e-06,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "pack_present",
          "time": 0.0,
          "expected": 1.0,
          "tolerance": {
            "abs": 0.0,
            "rel": 0.0
          }
        }
      ]
    },
    {
      "id": "cold_tiny_caseB_step1841",
      "description": "Reference: the same cold-season run and the same real64 kernels/noah_driver, replayed at ESM_DT = 1e-6 s, at rel 1e-9. THE POINT OF THIS REGIME is that it is exactly where a naively written compaction component fails: the density increment is 7.5e-17 of the density, so both sndens_out - sndens_in and sndens (pexp - 1) return pure round-off, and only the cancellation-free form -- sndens_in pexp_m1 + dw (1 - dsx), referenced against the series sum published before WRF adds one to it -- still carries the signal. `dsndens_dt` is the dt -> 0 compaction rate here to twelve digits. Step 1841 of case B, a 30 cm pack melting under warm advection. SNOWPACK's melt-water retention branch RUNS -- the snow surface is at or above freezing; neither density limit fires. The dt -> 0 compaction rate is 1.4296e-08 per second, and the 60 s regime's tendency for the same step exceeds it by xarg/3 = 2.68e-06 relative.",
      "parameter_overrides": {
        "esd": 0.05315843224445422,
        "sndens_in": 0.21350073926848778,
        "tsnow": 274.9277857895529,
        "tsoil": 265.888653646625,
        "dtsec": 1e-06
      },
      "time_span": {
        "start": 0.0,
        "end": 0.0
      },
      "assertions": [
        {
          "variable": "bfac",
          "time": 0.0,
          "expected": 2.5192556636596963e-14,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "xarg",
          "time": 0.0,
          "expected": 1.3391968150311153e-13,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "pexp_m1",
          "time": 0.0,
          "expected": 6.695984075155874e-14,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsx",
          "time": 0.0,
          "expected": 0.2135007392685021,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dw",
          "time": 0.0,
          "expected": 1.5046296296296297e-12,
          "tolerance": {
            "rel": 1e-09,
            "abs": 1e-30
          }
        },
        {
          "variable": "sndens_out",
          "time": 0.0,
          "expected": 0.21350073926968552,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsndens_dt",
          "time": 0.0,
          "expected": 1.1976860668801696e-06,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "pack_present",
          "time": 0.0,
          "expected": 1.0,
          "tolerance": {
            "abs": 0.0,
            "rel": 0.0
          }
        }
      ]
    },
    {
      "id": "cold_tiny_caseB_step1866",
      "description": "Reference: the same cold-season run and the same real64 kernels/noah_driver, replayed at ESM_DT = 1e-6 s, at rel 1e-9. THE POINT OF THIS REGIME is that it is exactly where a naively written compaction component fails: the density increment is 7.5e-17 of the density, so both sndens_out - sndens_in and sndens (pexp - 1) return pure round-off, and only the cancellation-free form -- sndens_in pexp_m1 + dw (1 - dsx), referenced against the series sum published before WRF adds one to it -- still carries the signal. `dsndens_dt` is the dt -> 0 compaction rate here to twelve digits. Step 1866 of case B, a 30 cm pack melting under warm advection. SNOWPACK's melt-water retention branch RUNS -- the snow surface is at or above freezing; neither density limit fires. The dt -> 0 compaction rate is 1.3620e-08 per second, and the 60 s regime's tendency for the same step exceeds it by xarg/3 = 2.53e-06 relative.",
      "parameter_overrides": {
        "esd": 0.05196269601504422,
        "sndens_in": 0.21529487057670674,
        "tsnow": 274.9066451558317,
        "tsoil": 265.99952240699713,
        "dtsec": 1e-06
      },
      "time_span": {
        "start": 0.0,
        "end": 0.0
      },
      "assertions": [
        {
          "variable": "bfac",
          "time": 0.0,
          "expected": 2.4348273582598227e-14,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "xarg",
          "time": 0.0,
          "expected": 1.2652019386636835e-13,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "pexp_m1",
          "time": 0.0,
          "expected": 6.326009693318684e-14,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsx",
          "time": 0.0,
          "expected": 0.21529487057672037,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dw",
          "time": 0.0,
          "expected": 1.5046296296296297e-12,
          "tolerance": {
            "rel": 1e-09,
            "abs": 1e-30
          }
        },
        {
          "variable": "sndens_out",
          "time": 0.0,
          "expected": 0.21529487057790106,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsndens_dt",
          "time": 0.0,
          "expected": 1.1943101626345204e-06,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "pack_present",
          "time": 0.0,
          "expected": 1.0,
          "tolerance": {
            "abs": 0.0,
            "rel": 0.0
          }
        }
      ]
    },
    {
      "id": "cold_tiny_caseB_step2122",
      "description": "Reference: the same cold-season run and the same real64 kernels/noah_driver, replayed at ESM_DT = 1e-6 s, at rel 1e-9. THE POINT OF THIS REGIME is that it is exactly where a naively written compaction component fails: the density increment is 7.5e-17 of the density, so both sndens_out - sndens_in and sndens (pexp - 1) return pure round-off, and only the cancellation-free form -- sndens_in pexp_m1 + dw (1 - dsx), referenced against the series sum published before WRF adds one to it -- still carries the signal. `dsndens_dt` is the dt -> 0 compaction rate here to twelve digits. Step 2122 of case B, a 30 cm pack melting under warm advection. dry branch -- the snow surface is below freezing and only compaction acts; neither density limit fires. The dt -> 0 compaction rate is 5.2529e-09 per second, and the 60 s regime's tendency for the same step exceeds it by xarg/3 = 9.10e-07 relative.",
      "parameter_overrides": {
        "esd": 0.045533597469329834,
        "sndens_in": 0.23088594021114783,
        "tsnow": 262.4313633423511,
        "tsoil": 264.39600206349655,
        "dtsec": 1e-06
      },
      "time_span": {
        "start": 0.0,
        "end": 0.0
      },
      "assertions": [
        {
          "variable": "bfac",
          "time": 0.0,
          "expected": 9.993115424156257e-15,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "xarg",
          "time": 0.0,
          "expected": 4.550224951880823e-14,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "pexp_m1",
          "time": 0.0,
          "expected": 2.2751124759404458e-14,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsx",
          "time": 0.0,
          "expected": 0.23088594021115305,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dw",
          "time": 0.0,
          "expected": 0.0,
          "tolerance": {
            "rel": 1e-09,
            "abs": 1e-30
          }
        },
        {
          "variable": "sndens_out",
          "time": 0.0,
          "expected": 0.23088594021115305,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsndens_dt",
          "time": 0.0,
          "expected": 5.252914830936223e-09,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "pack_present",
          "time": 0.0,
          "expected": 1.0,
          "tolerance": {
            "abs": 0.0,
            "rel": 0.0
          }
        }
      ]
    },
    {
      "id": "cold_tiny_caseB_step2880",
      "description": "Reference: the same cold-season run and the same real64 kernels/noah_driver, replayed at ESM_DT = 1e-6 s, at rel 1e-9. THE POINT OF THIS REGIME is that it is exactly where a naively written compaction component fails: the density increment is 7.5e-17 of the density, so both sndens_out - sndens_in and sndens (pexp - 1) return pure round-off, and only the cancellation-free form -- sndens_in pexp_m1 + dw (1 - dsx), referenced against the series sum published before WRF adds one to it -- still carries the signal. `dsndens_dt` is the dt -> 0 compaction rate here to twelve digits. Step 2880 of case B, a 30 cm pack melting under warm advection. dry branch -- the snow surface is below freezing and only compaction acts; neither density limit fires. The dt -> 0 compaction rate is 3.8525e-09 per second, and the 60 s regime's tendency for the same step exceeds it by xarg/3 = 6.67e-07 relative.",
      "parameter_overrides": {
        "esd": 0.045533597469329834,
        "sndens_in": 0.23105834304297074,
        "tsnow": 256.04450233726726,
        "tsoil": 263.10316061250694,
        "dtsec": 1e-06
      },
      "time_span": {
        "start": 0.0,
        "end": 0.0
      },
      "assertions": [
        {
          "variable": "bfac",
          "time": 0.0,
          "expected": 7.323496227127686e-15,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "xarg",
          "time": 0.0,
          "expected": 3.334651292741878e-14,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "pexp_m1",
          "time": 0.0,
          "expected": 1.6673256463709575e-14,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsx",
          "time": 0.0,
          "expected": 0.2310583430429746,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dw",
          "time": 0.0,
          "expected": 0.0,
          "tolerance": {
            "rel": 1e-09,
            "abs": 1e-30
          }
        },
        {
          "variable": "sndens_out",
          "time": 0.0,
          "expected": 0.2310583430429746,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsndens_dt",
          "time": 0.0,
          "expected": 3.852495011635237e-09,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "pack_present",
          "time": 0.0,
          "expected": 1.0,
          "tolerance": {
            "abs": 0.0,
            "rel": 0.0
          }
        }
      ]
    },
    {
      "id": "cold_tiny_caseC_step1",
      "description": "Reference: the same cold-season run and the same real64 kernels/noah_driver, replayed at ESM_DT = 1e-6 s, at rel 1e-9. THE POINT OF THIS REGIME is that it is exactly where a naively written compaction component fails: the density increment is 7.5e-17 of the density, so both sndens_out - sndens_in and sndens (pexp - 1) return pure round-off, and only the cancellation-free form -- sndens_in pexp_m1 + dw (1 - dsx), referenced against the series sum published before WRF adds one to it -- still carries the signal. `dsndens_dt` is the dt -> 0 compaction rate here to twelve digits. Step 1 of case C, a thin pack over soil that thaws and refreezes. dry branch -- the snow surface is below freezing and only compaction acts; neither density limit fires. The dt -> 0 compaction rate is 1.0901e-09 per second, and the 60 s regime's tendency for the same step exceeds it by xarg/3 = 2.18e-07 relative.",
      "parameter_overrides": {
        "esd": 0.005000000353911652,
        "sndens_in": 0.20000002607703365,
        "tsnow": 263.66899138754906,
        "tsoil": 266.44599412043084,
        "dtsec": 1e-06
      },
      "time_span": {
        "start": 0.0,
        "end": 0.0
      },
      "assertions": [
        {
          "variable": "bfac",
          "time": 0.0,
          "expected": 2.1802080460635333e-14,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "xarg",
          "time": 0.0,
          "expected": 1.0901041001918699e-14,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "pexp_m1",
          "time": 0.0,
          "expected": 5.450520500959369e-15,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsx",
          "time": 0.0,
          "expected": 0.20000002607703476,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dw",
          "time": 0.0,
          "expected": 0.0,
          "tolerance": {
            "rel": 1e-09,
            "abs": 1e-30
          }
        },
        {
          "variable": "sndens_out",
          "time": 0.0,
          "expected": 0.20000002607703476,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsndens_dt",
          "time": 0.0,
          "expected": 1.0901042423252804e-09,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "pack_present",
          "time": 0.0,
          "expected": 1.0,
          "tolerance": {
            "abs": 0.0,
            "rel": 0.0
          }
        }
      ]
    },
    {
      "id": "cold_tiny_caseC_step26",
      "description": "Reference: the same cold-season run and the same real64 kernels/noah_driver, replayed at ESM_DT = 1e-6 s, at rel 1e-9. THE POINT OF THIS REGIME is that it is exactly where a naively written compaction component fails: the density increment is 7.5e-17 of the density, so both sndens_out - sndens_in and sndens (pexp - 1) return pure round-off, and only the cancellation-free form -- sndens_in pexp_m1 + dw (1 - dsx), referenced against the series sum published before WRF adds one to it -- still carries the signal. `dsndens_dt` is the dt -> 0 compaction rate here to twelve digits. Step 26 of case C, a thin pack over soil that thaws and refreezes. dry branch -- the snow surface is below freezing and only compaction acts; neither density limit fires. The dt -> 0 compaction rate is 1.2846e-09 per second, and the 60 s regime's tendency for the same step exceeds it by xarg/3 = 2.58e-07 relative.",
      "parameter_overrides": {
        "esd": 0.005003620404770962,
        "sndens_in": 0.1997457910986596,
        "tsnow": 266.05796185080965,
        "tsoil": 268.0412772808923,
        "dtsec": 1e-06
      },
      "time_span": {
        "start": 0.0,
        "end": 0.0
      },
      "assertions": [
        {
          "variable": "bfac",
          "time": 0.0,
          "expected": 2.5705744252854212e-14,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "xarg",
          "time": 0.0,
          "expected": 1.2862178646340522e-14,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "pexp_m1",
          "time": 0.0,
          "expected": 6.431089323170288e-15,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsx",
          "time": 0.0,
          "expected": 0.19974579109866086,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dw",
          "time": 0.0,
          "expected": 0.0,
          "tolerance": {
            "rel": 1e-09,
            "abs": 1e-30
          }
        },
        {
          "variable": "sndens_out",
          "time": 0.0,
          "expected": 0.19974579109866086,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsndens_dt",
          "time": 0.0,
          "expected": 1.2845830244827924e-09,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "pack_present",
          "time": 0.0,
          "expected": 1.0,
          "tolerance": {
            "abs": 0.0,
            "rel": 0.0
          }
        }
      ]
    },
    {
      "id": "cold_tiny_caseC_step56",
      "description": "Reference: the same cold-season run and the same real64 kernels/noah_driver, replayed at ESM_DT = 1e-6 s, at rel 1e-9. THE POINT OF THIS REGIME is that it is exactly where a naively written compaction component fails: the density increment is 7.5e-17 of the density, so both sndens_out - sndens_in and sndens (pexp - 1) return pure round-off, and only the cancellation-free form -- sndens_in pexp_m1 + dw (1 - dsx), referenced against the series sum published before WRF adds one to it -- still carries the signal. `dsndens_dt` is the dt -> 0 compaction rate here to twelve digits. Step 56 of case C, a thin pack over soil that thaws and refreezes. dry branch -- the snow surface is below freezing and only compaction acts; neither density limit fires. The dt -> 0 compaction rate is 1.4515e-09 per second, and the 60 s regime's tendency for the same step exceeds it by xarg/3 = 3.01e-07 relative.",
      "parameter_overrides": {
        "esd": 0.005102005787278408,
        "sndens_in": 0.19413903381883213,
        "tsnow": 266.22168328348897,
        "tsoil": 268.21243820893034,
        "dtsec": 1e-06
      },
      "time_span": {
        "start": 0.0,
        "end": 0.0
      },
      "assertions": [
        {
          "variable": "bfac",
          "time": 0.0,
          "expected": 2.9307740910151564e-14,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "xarg",
          "time": 0.0,
          "expected": 1.495282637356494e-14,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "pexp_m1",
          "time": 0.0,
          "expected": 7.476413186782508e-15,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsx",
          "time": 0.0,
          "expected": 0.1941390338188336,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dw",
          "time": 0.0,
          "expected": 0.0,
          "tolerance": {
            "rel": 1e-09,
            "abs": 1e-30
          }
        },
        {
          "variable": "sndens_out",
          "time": 0.0,
          "expected": 0.1941390338188336,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsndens_dt",
          "time": 0.0,
          "expected": 1.451463632512332e-09,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "pack_present",
          "time": 0.0,
          "expected": 1.0,
          "tolerance": {
            "abs": 0.0,
            "rel": 0.0
          }
        }
      ]
    },
    {
      "id": "cold_tiny_caseC_step169",
      "description": "Reference: the same cold-season run and the same real64 kernels/noah_driver, replayed at ESM_DT = 1e-6 s, at rel 1e-9. THE POINT OF THIS REGIME is that it is exactly where a naively written compaction component fails: the density increment is 7.5e-17 of the density, so both sndens_out - sndens_in and sndens (pexp - 1) return pure round-off, and only the cancellation-free form -- sndens_in pexp_m1 + dw (1 - dsx), referenced against the series sum published before WRF adds one to it -- still carries the signal. `dsndens_dt` is the dt -> 0 compaction rate here to twelve digits. Step 169 of case C, a thin pack over soil that thaws and refreezes. dry branch -- the snow surface is below freezing and only compaction acts; neither density limit fires. The dt -> 0 compaction rate is 2.1697e-09 per second, and the 60 s regime's tendency for the same step exceeds it by xarg/3 = 4.85e-07 relative.",
      "parameter_overrides": {
        "esd": 0.005403401795857795,
        "sndens_in": 0.1799161790916233,
        "tsnow": 267.99019416622883,
        "tsoil": 269.49490107114605,
        "dtsec": 1e-06
      },
      "time_span": {
        "start": 0.0,
        "end": 0.0
      },
      "assertions": [
        {
          "variable": "bfac",
          "time": 0.0,
          "expected": 4.463733315267788e-14,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "xarg",
          "time": 0.0,
          "expected": 2.4119344611948236e-14,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "pexp_m1",
          "time": 0.0,
          "expected": 1.2059672305974216e-14,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsx",
          "time": 0.0,
          "expected": 0.17991617909162547,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dw",
          "time": 0.0,
          "expected": 0.0,
          "tolerance": {
            "rel": 1e-09,
            "abs": 1e-30
          }
        },
        {
          "variable": "sndens_out",
          "time": 0.0,
          "expected": 0.17991617909162547,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsndens_dt",
          "time": 0.0,
          "expected": 2.1697301623879468e-09,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "pack_present",
          "time": 0.0,
          "expected": 1.0,
          "tolerance": {
            "abs": 0.0,
            "rel": 0.0
          }
        }
      ]
    },
    {
      "id": "cold_tiny_caseC_step268",
      "description": "Reference: the same cold-season run and the same real64 kernels/noah_driver, replayed at ESM_DT = 1e-6 s, at rel 1e-9. THE POINT OF THIS REGIME is that it is exactly where a naively written compaction component fails: the density increment is 7.5e-17 of the density, so both sndens_out - sndens_in and sndens (pexp - 1) return pure round-off, and only the cancellation-free form -- sndens_in pexp_m1 + dw (1 - dsx), referenced against the series sum published before WRF adds one to it -- still carries the signal. `dsndens_dt` is the dt -> 0 compaction rate here to twelve digits. Step 268 of case C, a thin pack over soil that thaws and refreezes. SNOWPACK's melt-water retention branch RUNS -- the snow surface is at or above freezing; neither density limit fires. The dt -> 0 compaction rate is 3.2503e-09 per second, and the 60 s regime's tendency for the same step exceeds it by xarg/3 = 7.32e-07 relative.",
      "parameter_overrides": {
        "esd": 0.005462722387160853,
        "sndens_in": 0.17761091261995568,
        "tsnow": 273.5139880248689,
        "tsoil": 272.91424693302145,
        "dtsec": 1e-06
      },
      "time_span": {
        "start": 0.0,
        "end": 0.0
      },
      "assertions": [
        {
          "variable": "bfac",
          "time": 0.0,
          "expected": 6.70009474148183e-14,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "xarg",
          "time": 0.0,
          "expected": 3.660075754039149e-14,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "pexp_m1",
          "time": 0.0,
          "expected": 1.830037877019597e-14,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsx",
          "time": 0.0,
          "expected": 0.17761091261995893,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dw",
          "time": 0.0,
          "expected": 1.5046296296296297e-12,
          "tolerance": {
            "rel": 1e-09,
            "abs": 1e-30
          }
        },
        {
          "variable": "sndens_out",
          "time": 0.0,
          "expected": 0.17761091262119633,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsndens_dt",
          "time": 0.0,
          "expected": 1.2406413349307457e-06,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "pack_present",
          "time": 0.0,
          "expected": 1.0,
          "tolerance": {
            "abs": 0.0,
            "rel": 0.0
          }
        }
      ]
    },
    {
      "id": "cold_tiny_caseC_step380",
      "description": "Reference: the same cold-season run and the same real64 kernels/noah_driver, replayed at ESM_DT = 1e-6 s, at rel 1e-9. THE POINT OF THIS REGIME is that it is exactly where a naively written compaction component fails: the density increment is 7.5e-17 of the density, so both sndens_out - sndens_in and sndens (pexp - 1) return pure round-off, and only the cancellation-free form -- sndens_in pexp_m1 + dw (1 - dsx), referenced against the series sum published before WRF adds one to it -- still carries the signal. `dsndens_dt` is the dt -> 0 compaction rate here to twelve digits. Step 380 of case C, a thin pack over soil that thaws and refreezes. SNOWPACK's melt-water retention branch RUNS -- the snow surface is at or above freezing; neither density limit fires. The dt -> 0 compaction rate is 3.8608e-09 per second, and the 60 s regime's tendency for the same step exceeds it by xarg/3 = 8.36e-07 relative.",
      "parameter_overrides": {
        "esd": 0.005304833874044777,
        "sndens_in": 0.18452061980804027,
        "tsnow": 278.214324627469,
        "tsoil": 275.9234684230028,
        "dtsec": 1e-06
      },
      "time_span": {
        "start": 0.0,
        "end": 0.0
      },
      "assertions": [
        {
          "variable": "bfac",
          "time": 0.0,
          "expected": 7.888453748714131e-14,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "xarg",
          "time": 0.0,
          "expected": 4.1846936660014224e-14,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "pexp_m1",
          "time": 0.0,
          "expected": 2.0923468330007402e-14,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsx",
          "time": 0.0,
          "expected": 0.18452061980804413,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dw",
          "time": 0.0,
          "expected": 1.5046296296296297e-12,
          "tolerance": {
            "rel": 1e-09,
            "abs": 1e-30
          }
        },
        {
          "variable": "sndens_out",
          "time": 0.0,
          "expected": 0.18452061980927115,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsndens_dt",
          "time": 0.0,
          "expected": 1.2308552491336094e-06,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "pack_present",
          "time": 0.0,
          "expected": 1.0,
          "tolerance": {
            "abs": 0.0,
            "rel": 0.0
          }
        }
      ]
    },
    {
      "id": "cold_tiny_caseC_step405",
      "description": "Reference: the same cold-season run and the same real64 kernels/noah_driver, replayed at ESM_DT = 1e-6 s, at rel 1e-9. THE POINT OF THIS REGIME is that it is exactly where a naively written compaction component fails: the density increment is 7.5e-17 of the density, so both sndens_out - sndens_in and sndens (pexp - 1) return pure round-off, and only the cancellation-free form -- sndens_in pexp_m1 + dw (1 - dsx), referenced against the series sum published before WRF adds one to it -- still carries the signal. `dsndens_dt` is the dt -> 0 compaction rate here to twelve digits. Step 405 of case C, a thin pack over soil that thaws and refreezes. SNOWPACK's melt-water retention branch RUNS -- the snow surface is at or above freezing; neither density limit fires. The dt -> 0 compaction rate is 4.1295e-09 per second, and the 60 s regime's tendency for the same step exceeds it by xarg/3 = 8.86e-07 relative.",
      "parameter_overrides": {
        "esd": 0.005202005617239995,
        "sndens_in": 0.18636313407082133,
        "tsnow": 279.94878251017286,
        "tsoil": 277.0792878275947,
        "dtsec": 1e-06
      },
      "time_span": {
        "start": 0.0,
        "end": 0.0
      },
      "assertions": [
        {
          "variable": "bfac",
          "time": 0.0,
          "expected": 8.519165853619654e-14,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "xarg",
          "time": 0.0,
          "expected": 4.43167486247286e-14,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "pexp_m1",
          "time": 0.0,
          "expected": 2.2158374312364628e-14,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsx",
          "time": 0.0,
          "expected": 0.18636313407082547,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dw",
          "time": 0.0,
          "expected": 1.5046296296296297e-12,
          "tolerance": {
            "rel": 1e-09,
            "abs": 1e-30
          }
        },
        {
          "variable": "sndens_out",
          "time": 0.0,
          "expected": 0.1863631340720497,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsndens_dt",
          "time": 0.0,
          "expected": 1.2283516403187932e-06,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "pack_present",
          "time": 0.0,
          "expected": 1.0,
          "tolerance": {
            "abs": 0.0,
            "rel": 0.0
          }
        }
      ]
    },
    {
      "id": "cold_tiny_caseC_step434",
      "description": "Reference: the same cold-season run and the same real64 kernels/noah_driver, replayed at ESM_DT = 1e-6 s, at rel 1e-9. THE POINT OF THIS REGIME is that it is exactly where a naively written compaction component fails: the density increment is 7.5e-17 of the density, so both sndens_out - sndens_in and sndens (pexp - 1) return pure round-off, and only the cancellation-free form -- sndens_in pexp_m1 + dw (1 - dsx), referenced against the series sum published before WRF adds one to it -- still carries the signal. `dsndens_dt` is the dt -> 0 compaction rate here to twelve digits. Step 434 of case C, a thin pack over soil that thaws and refreezes. SNOWPACK's melt-water retention branch RUNS -- the snow surface is at or above freezing; neither density limit fires. The dt -> 0 compaction rate is 4.0442e-09 per second, and the 60 s regime's tendency for the same step exceeds it by xarg/3 = 8.57e-07 relative.",
      "parameter_overrides": {
        "esd": 0.005062823183753572,
        "sndens_in": 0.1884977882695936,
        "tsnow": 280.5072263014543,
        "tsoil": 277.5130995895714,
        "dtsec": 1e-06
      },
      "time_span": {
        "start": 0.0,
        "end": 0.0
      },
      "assertions": [
        {
          "variable": "bfac",
          "time": 0.0,
          "expected": 8.47551073781467e-14,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "xarg",
          "time": 0.0,
          "expected": 4.2910012257560455e-14,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "pexp_m1",
          "time": 0.0,
          "expected": 2.145500612878053e-14,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsx",
          "time": 0.0,
          "expected": 0.18849778826959765,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dw",
          "time": 0.0,
          "expected": 1.5046296296296297e-12,
          "tolerance": {
            "rel": 1e-09,
            "abs": 1e-30
          }
        },
        {
          "variable": "sndens_out",
          "time": 0.0,
          "expected": 0.18849778827081867,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsndens_dt",
          "time": 0.0,
          "expected": 1.2250544934821264e-06,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "pack_present",
          "time": 0.0,
          "expected": 1.0,
          "tolerance": {
            "abs": 0.0,
            "rel": 0.0
          }
        }
      ]
    },
    {
      "id": "cold_tiny_caseC_step459",
      "description": "Reference: the same cold-season run and the same real64 kernels/noah_driver, replayed at ESM_DT = 1e-6 s, at rel 1e-9. THE POINT OF THIS REGIME is that it is exactly where a naively written compaction component fails: the density increment is 7.5e-17 of the density, so both sndens_out - sndens_in and sndens (pexp - 1) return pure round-off, and only the cancellation-free form -- sndens_in pexp_m1 + dw (1 - dsx), referenced against the series sum published before WRF adds one to it -- still carries the signal. `dsndens_dt` is the dt -> 0 compaction rate here to twelve digits. Step 459 of case C, a thin pack over soil that thaws and refreezes. SNOWPACK's melt-water retention branch RUNS -- the snow surface is at or above freezing; neither density limit fires. The dt -> 0 compaction rate is 3.9047e-09 per second, and the 60 s regime's tendency for the same step exceeds it by xarg/3 = 8.20e-07 relative.",
      "parameter_overrides": {
        "esd": 0.004942660219888693,
        "sndens_in": 0.19033344802528646,
        "tsnow": 280.73589555040195,
        "tsoil": 277.7289289224177,
        "dtsec": 1e-06
      },
      "time_span": {
        "start": 0.0,
        "end": 0.0
      },
      "assertions": [
        {
          "variable": "bfac",
          "time": 0.0,
          "expected": 8.301298961158113e-14,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "xarg",
          "time": 0.0,
          "expected": 4.1030500148719536e-14,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "pexp_m1",
          "time": 0.0,
          "expected": 2.0515250074360052e-14,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsx",
          "time": 0.0,
          "expected": 0.19033344802529034,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dw",
          "time": 0.0,
          "expected": 1.5046296296296297e-12,
          "tolerance": {
            "rel": 1e-09,
            "abs": 1e-30
          }
        },
        {
          "variable": "sndens_out",
          "time": 0.0,
          "expected": 0.1903334480265086,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsndens_dt",
          "time": 0.0,
          "expected": 1.222153022504961e-06,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "pack_present",
          "time": 0.0,
          "expected": 1.0,
          "tolerance": {
            "abs": 0.0,
            "rel": 0.0
          }
        }
      ]
    },
    {
      "id": "cold_tiny_caseC_step708",
      "description": "Reference: the same cold-season run and the same real64 kernels/noah_driver, replayed at ESM_DT = 1e-6 s, at rel 1e-9. THE POINT OF THIS REGIME is that it is exactly where a naively written compaction component fails: the density increment is 7.5e-17 of the density, so both sndens_out - sndens_in and sndens (pexp - 1) return pure round-off, and only the cancellation-free form -- sndens_in pexp_m1 + dw (1 - dsx), referenced against the series sum published before WRF adds one to it -- still carries the signal. `dsndens_dt` is the dt -> 0 compaction rate here to twelve digits. Step 708 of case C, a thin pack over soil that thaws and refreezes. dry branch -- the snow surface is below freezing and only compaction acts; neither density limit fires. The dt -> 0 compaction rate is 9.6554e-10 per second, and the 60 s regime's tendency for the same step exceeds it by xarg/3 = 1.88e-07 relative.",
      "parameter_overrides": {
        "esd": 0.0043650981970140244,
        "sndens_in": 0.20500662216316415,
        "tsnow": 265.2143268777221,
        "tsoil": 267.2724060997736,
        "dtsec": 1e-06
      },
      "time_span": {
        "start": 0.0,
        "end": 0.0
      },
      "assertions": [
        {
          "variable": "bfac",
          "time": 0.0,
          "expected": 2.1579348733142677e-14,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "xarg",
          "time": 0.0,
          "expected": 9.419597624777797e-15,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "pexp_m1",
          "time": 0.0,
          "expected": 4.709798812388913e-15,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsx",
          "time": 0.0,
          "expected": 0.2050066221631651,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dw",
          "time": 0.0,
          "expected": 0.0,
          "tolerance": {
            "rel": 1e-09,
            "abs": 1e-30
          }
        },
        {
          "variable": "sndens_out",
          "time": 0.0,
          "expected": 0.2050066221631651,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsndens_dt",
          "time": 0.0,
          "expected": 9.655399455959332e-10,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "pack_present",
          "time": 0.0,
          "expected": 1.0,
          "tolerance": {
            "abs": 0.0,
            "rel": 0.0
          }
        }
      ]
    },
    {
      "id": "cold_tiny_caseC_step733",
      "description": "Reference: the same cold-season run and the same real64 kernels/noah_driver, replayed at ESM_DT = 1e-6 s, at rel 1e-9. THE POINT OF THIS REGIME is that it is exactly where a naively written compaction component fails: the density increment is 7.5e-17 of the density, so both sndens_out - sndens_in and sndens (pexp - 1) return pure round-off, and only the cancellation-free form -- sndens_in pexp_m1 + dw (1 - dsx), referenced against the series sum published before WRF adds one to it -- still carries the signal. `dsndens_dt` is the dt -> 0 compaction rate here to twelve digits. Step 733 of case C, a thin pack over soil that thaws and refreezes. dry branch -- the snow surface is below freezing and only compaction acts; neither density limit fires. The dt -> 0 compaction rate is 9.6745e-10 per second, and the 60 s regime's tendency for the same step exceeds it by xarg/3 = 1.89e-07 relative.",
      "parameter_overrides": {
        "esd": 0.004366421140731254,
        "sndens_in": 0.20500799408956843,
        "tsnow": 265.2389769448648,
        "tsoil": 267.2901203138098,
        "dtsec": 1e-06
      },
      "time_span": {
        "start": 0.0,
        "end": 0.0
      },
      "assertions": [
        {
          "variable": "bfac",
          "time": 0.0,
          "expected": 2.1615324721570375e-14,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "xarg",
          "time": 0.0,
          "expected": 9.438161082803579e-15,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "pexp_m1",
          "time": 0.0,
          "expected": 4.7190805414018035e-15,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsx",
          "time": 0.0,
          "expected": 0.20500799408956938,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dw",
          "time": 0.0,
          "expected": 0.0,
          "tolerance": {
            "rel": 1e-09,
            "abs": 1e-30
          }
        },
        {
          "variable": "sndens_out",
          "time": 0.0,
          "expected": 0.20500799408956938,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsndens_dt",
          "time": 0.0,
          "expected": 9.674492357398984e-10,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "pack_present",
          "time": 0.0,
          "expected": 1.0,
          "tolerance": {
            "abs": 0.0,
            "rel": 0.0
          }
        }
      ]
    },
    {
      "id": "cold_tiny_caseC_step1564",
      "description": "Reference: the same cold-season run and the same real64 kernels/noah_driver, replayed at ESM_DT = 1e-6 s, at rel 1e-9. THE POINT OF THIS REGIME is that it is exactly where a naively written compaction component fails: the density increment is 7.5e-17 of the density, so both sndens_out - sndens_in and sndens (pexp - 1) return pure round-off, and only the cancellation-free form -- sndens_in pexp_m1 + dw (1 - dsx), referenced against the series sum published before WRF adds one to it -- still carries the signal. `dsndens_dt` is the dt -> 0 compaction rate here to twelve digits. Step 1564 of case C, a thin pack over soil that thaws and refreezes. dry branch -- the snow surface is below freezing and only compaction acts; neither density limit fires. The dt -> 0 compaction rate is 1.3354e-09 per second, and the 60 s regime's tendency for the same step exceeds it by xarg/3 = 2.61e-07 relative.",
      "parameter_overrides": {
        "esd": 0.004405545536428777,
        "sndens_in": 0.20505721283671166,
        "tsnow": 270.10773456323943,
        "tsoil": 270.27697364307807,
        "dtsec": 1e-06
      },
      "time_span": {
        "start": 0.0,
        "end": 0.0
      },
      "assertions": [
        {
          "variable": "bfac",
          "time": 0.0,
          "expected": 2.9565063739934634e-14,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "xarg",
          "time": 0.0,
          "expected": 1.302502345937013e-14,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "pexp_m1",
          "time": 0.0,
          "expected": 6.512511729685094e-15,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsx",
          "time": 0.0,
          "expected": 0.205057212836713,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dw",
          "time": 0.0,
          "expected": 0.0,
          "tolerance": {
            "rel": 1e-09,
            "abs": 1e-30
          }
        },
        {
          "variable": "sndens_out",
          "time": 0.0,
          "expected": 0.205057212836713,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsndens_dt",
          "time": 0.0,
          "expected": 1.3354375038556176e-09,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "pack_present",
          "time": 0.0,
          "expected": 1.0,
          "tolerance": {
            "abs": 0.0,
            "rel": 0.0
          }
        }
      ]
    },
    {
      "id": "cold_tiny_caseC_step1785",
      "description": "Reference: the same cold-season run and the same real64 kernels/noah_driver, replayed at ESM_DT = 1e-6 s, at rel 1e-9. THE POINT OF THIS REGIME is that it is exactly where a naively written compaction component fails: the density increment is 7.5e-17 of the density, so both sndens_out - sndens_in and sndens (pexp - 1) return pure round-off, and only the cancellation-free form -- sndens_in pexp_m1 + dw (1 - dsx), referenced against the series sum published before WRF adds one to it -- still carries the signal. `dsndens_dt` is the dt -> 0 compaction rate here to twelve digits. Step 1785 of case C, a thin pack over soil that thaws and refreezes. SNOWPACK's melt-water retention branch RUNS -- the snow surface is at or above freezing; neither density limit fires. The dt -> 0 compaction rate is 3.2102e-09 per second, and the 60 s regime's tendency for the same step exceeds it by xarg/3 = 5.83e-07 relative.",
      "parameter_overrides": {
        "esd": 0.0037659858352648255,
        "sndens_in": 0.21981131880599794,
        "tsnow": 288.34277368071645,
        "tsoil": 283.89919366320333,
        "dtsec": 1e-06
      },
      "time_span": {
        "start": 0.0,
        "end": 0.0
      },
      "assertions": [
        {
          "variable": "bfac",
          "time": 0.0,
          "expected": 7.755966610263095e-14,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "xarg",
          "time": 0.0,
          "expected": 2.920886039303776e-14,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "pexp_m1",
          "time": 0.0,
          "expected": 1.4604430196519025e-14,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsx",
          "time": 0.0,
          "expected": 0.21981131880600116,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dw",
          "time": 0.0,
          "expected": 1.5046296296296297e-12,
          "tolerance": {
            "rel": 1e-09,
            "abs": 1e-30
          }
        },
        {
          "variable": "sndens_out",
          "time": 0.0,
          "expected": 0.21981131880717505,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsndens_dt",
          "time": 0.0,
          "expected": 1.1771052254880628e-06,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "pack_present",
          "time": 0.0,
          "expected": 1.0,
          "tolerance": {
            "abs": 0.0,
            "rel": 0.0
          }
        }
      ]
    },
    {
      "id": "cold_tiny_caseC_step1810",
      "description": "Reference: the same cold-season run and the same real64 kernels/noah_driver, replayed at ESM_DT = 1e-6 s, at rel 1e-9. THE POINT OF THIS REGIME is that it is exactly where a naively written compaction component fails: the density increment is 7.5e-17 of the density, so both sndens_out - sndens_in and sndens (pexp - 1) return pure round-off, and only the cancellation-free form -- sndens_in pexp_m1 + dw (1 - dsx), referenced against the series sum published before WRF adds one to it -- still carries the signal. `dsndens_dt` is the dt -> 0 compaction rate here to twelve digits. Step 1810 of case C, a thin pack over soil that thaws and refreezes. SNOWPACK's melt-water retention branch RUNS -- the snow surface is at or above freezing; neither density limit fires. The dt -> 0 compaction rate is 3.1509e-09 per second, and the 60 s regime's tendency for the same step exceeds it by xarg/3 = 5.68e-07 relative.",
      "parameter_overrides": {
        "esd": 0.0036482352296973897,
        "sndens_in": 0.22157488677926387,
        "tsnow": 288.8332593952433,
        "tsoil": 284.4626434642655,
        "dtsec": 1e-06
      },
      "time_span": {
        "start": 0.0,
        "end": 0.0
      },
      "assertions": [
        {
          "variable": "bfac",
          "time": 0.0,
          "expected": 7.795798428376844e-14,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "xarg",
          "time": 0.0,
          "expected": 2.844090647002394e-14,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "pexp_m1",
          "time": 0.0,
          "expected": 1.4220453235012107e-14,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsx",
          "time": 0.0,
          "expected": 0.221574886779267,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dw",
          "time": 0.0,
          "expected": 1.5046296296296297e-12,
          "tolerance": {
            "rel": 1e-09,
            "abs": 1e-30
          }
        },
        {
          "variable": "sndens_out",
          "time": 0.0,
          "expected": 0.22157488678043827,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsndens_dt",
          "time": 0.0,
          "expected": 1.1743923851152117e-06,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "pack_present",
          "time": 0.0,
          "expected": 1.0,
          "tolerance": {
            "abs": 0.0,
            "rel": 0.0
          }
        }
      ]
    },
    {
      "id": "cold_tiny_caseC_step1838",
      "description": "Reference: the same cold-season run and the same real64 kernels/noah_driver, replayed at ESM_DT = 1e-6 s, at rel 1e-9. THE POINT OF THIS REGIME is that it is exactly where a naively written compaction component fails: the density increment is 7.5e-17 of the density, so both sndens_out - sndens_in and sndens (pexp - 1) return pure round-off, and only the cancellation-free form -- sndens_in pexp_m1 + dw (1 - dsx), referenced against the series sum published before WRF adds one to it -- still carries the signal. `dsndens_dt` is the dt -> 0 compaction rate here to twelve digits. Step 1838 of case C, a thin pack over soil that thaws and refreezes. SNOWPACK's melt-water retention branch RUNS -- the snow surface is at or above freezing; neither density limit fires. The dt -> 0 compaction rate is 3.0356e-09 per second, and the 60 s regime's tendency for the same step exceeds it by xarg/3 = 5.42e-07 relative.",
      "parameter_overrides": {
        "esd": 0.00351929571472037,
        "sndens_in": 0.2235453724670667,
        "tsnow": 289.1608321106209,
        "tsoil": 284.91598230451956,
        "dtsec": 1e-06
      },
      "time_span": {
        "start": 0.0,
        "end": 0.0
      },
      "assertions": [
        {
          "variable": "bfac",
          "time": 0.0,
          "expected": 7.71711962092348e-14,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "xarg",
          "time": 0.0,
          "expected": 2.7158826011900492e-14,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "pexp_m1",
          "time": 0.0,
          "expected": 1.357941300595037e-14,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsx",
          "time": 0.0,
          "expected": 0.22354537246706974,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dw",
          "time": 0.0,
          "expected": 1.5046296296296297e-12,
          "tolerance": {
            "rel": 1e-09,
            "abs": 1e-30
          }
        },
        {
          "variable": "sndens_out",
          "time": 0.0,
          "expected": 0.22354537246823802,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsndens_dt",
          "time": 0.0,
          "expected": 1.1713122535873843e-06,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "pack_present",
          "time": 0.0,
          "expected": 1.0,
          "tolerance": {
            "abs": 0.0,
            "rel": 0.0
          }
        }
      ]
    },
    {
      "id": "cold_tiny_caseC_step1863",
      "description": "Reference: the same cold-season run and the same real64 kernels/noah_driver, replayed at ESM_DT = 1e-6 s, at rel 1e-9. THE POINT OF THIS REGIME is that it is exactly where a naively written compaction component fails: the density increment is 7.5e-17 of the density, so both sndens_out - sndens_in and sndens (pexp - 1) return pure round-off, and only the cancellation-free form -- sndens_in pexp_m1 + dw (1 - dsx), referenced against the series sum published before WRF adds one to it -- still carries the signal. `dsndens_dt` is the dt -> 0 compaction rate here to twelve digits. Step 1863 of case C, a thin pack over soil that thaws and refreezes. SNOWPACK's melt-water retention branch RUNS -- the snow surface is at or above freezing; neither density limit fires. The dt -> 0 compaction rate is 2.8970e-09 per second, and the 60 s regime's tendency for the same step exceeds it by xarg/3 = 5.14e-07 relative.",
      "parameter_overrides": {
        "esd": 0.003408966818752834,
        "sndens_in": 0.22530075234017996,
        "tsnow": 289.26076994858494,
        "tsoil": 285.16995609564384,
        "dtsec": 1e-06
      },
      "time_span": {
        "start": 0.0,
        "end": 0.0
      },
      "assertions": [
        {
          "variable": "bfac",
          "time": 0.0,
          "expected": 7.543865196067424e-14,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "xarg",
          "time": 0.0,
          "expected": 2.571678613853819e-14,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "pexp_m1",
          "time": 0.0,
          "expected": 1.2858393069269208e-14,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsx",
          "time": 0.0,
          "expected": 0.22530075234018287,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dw",
          "time": 0.0,
          "expected": 1.5046296296296297e-12,
          "tolerance": {
            "rel": 1e-09,
            "abs": 1e-30
          }
        },
        {
          "variable": "sndens_out",
          "time": 0.0,
          "expected": 0.22530075234134853,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsndens_dt",
          "time": 0.0,
          "expected": 1.1685324477131357e-06,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "pack_present",
          "time": 0.0,
          "expected": 1.0,
          "tolerance": {
            "abs": 0.0,
            "rel": 0.0
          }
        }
      ]
    },
    {
      "id": "cold_tiny_caseC_step1888",
      "description": "Reference: the same cold-season run and the same real64 kernels/noah_driver, replayed at ESM_DT = 1e-6 s, at rel 1e-9. THE POINT OF THIS REGIME is that it is exactly where a naively written compaction component fails: the density increment is 7.5e-17 of the density, so both sndens_out - sndens_in and sndens (pexp - 1) return pure round-off, and only the cancellation-free form -- sndens_in pexp_m1 + dw (1 - dsx), referenced against the series sum published before WRF adds one to it -- still carries the signal. `dsndens_dt` is the dt -> 0 compaction rate here to twelve digits. Step 1888 of case C, a thin pack over soil that thaws and refreezes. SNOWPACK's melt-water retention branch RUNS -- the snow surface is at or above freezing; neither density limit fires. The dt -> 0 compaction rate is 2.7267e-09 per second, and the 60 s regime's tendency for the same step exceeds it by xarg/3 = 4.80e-07 relative.",
      "parameter_overrides": {
        "esd": 0.003304766025327157,
        "sndens_in": 0.2270514846566476,
        "tsnow": 289.16088334733814,
        "tsoil": 285.25687927201136,
        "dtsec": 1e-06
      },
      "time_span": {
        "start": 0.0,
        "end": 0.0
      },
      "assertions": [
        {
          "variable": "bfac",
          "time": 0.0,
          "expected": 7.267779128621976e-14,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "xarg",
          "time": 0.0,
          "expected": 2.4018309543851714e-14,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "pexp_m1",
          "time": 0.0,
          "expected": 1.2009154771925953e-14,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsx",
          "time": 0.0,
          "expected": 0.22705148465665032,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dw",
          "time": 0.0,
          "expected": 1.5046296296296297e-12,
          "tolerance": {
            "rel": 1e-09,
            "abs": 1e-30
          }
        },
        {
          "variable": "sndens_out",
          "time": 0.0,
          "expected": 0.22705148465781333,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsndens_dt",
          "time": 0.0,
          "expected": 1.165727934784274e-06,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "pack_present",
          "time": 0.0,
          "expected": 1.0,
          "tolerance": {
            "abs": 0.0,
            "rel": 0.0
          }
        }
      ]
    },
    {
      "id": "cold_tiny_caseC_step1978",
      "description": "Reference: the same cold-season run and the same real64 kernels/noah_driver, replayed at ESM_DT = 1e-6 s, at rel 1e-9. THE POINT OF THIS REGIME is that it is exactly where a naively written compaction component fails: the density increment is 7.5e-17 of the density, so both sndens_out - sndens_in and sndens (pexp - 1) return pure round-off, and only the cancellation-free form -- sndens_in pexp_m1 + dw (1 - dsx), referenced against the series sum published before WRF adds one to it -- still carries the signal. `dsndens_dt` is the dt -> 0 compaction rate here to twelve digits. Step 1978 of case C, a thin pack over soil that thaws and refreezes. SNOWPACK's melt-water retention branch RUNS -- the snow surface is at or above freezing; neither density limit fires. The dt -> 0 compaction rate is 1.9936e-09 per second, and the 60 s regime's tendency for the same step exceeds it by xarg/3 = 3.41e-07 relative.",
      "parameter_overrides": {
        "esd": 0.0029972025658478296,
        "sndens_in": 0.23331962970999673,
        "tsnow": 287.2600513118683,
        "tsoil": 284.3804945164556,
        "dtsec": 1e-06
      },
      "time_span": {
        "start": 0.0,
        "end": 0.0
      },
      "assertions": [
        {
          "variable": "bfac",
          "time": 0.0,
          "expected": 5.701509947004215e-14,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "xarg",
          "time": 0.0,
          "expected": 1.7088580242367955e-14,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "pexp_m1",
          "time": 0.0,
          "expected": 8.544290121184028e-15,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsx",
          "time": 0.0,
          "expected": 0.2333196297099987,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dw",
          "time": 0.0,
          "expected": 1.5046296296296297e-12,
          "tolerance": {
            "rel": 1e-09,
            "abs": 1e-30
          }
        },
        {
          "variable": "sndens_out",
          "time": 0.0,
          "expected": 0.23331962971115228,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsndens_dt",
          "time": 0.0,
          "expected": 1.1555635522009617e-06,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "pack_present",
          "time": 0.0,
          "expected": 1.0,
          "tolerance": {
            "abs": 0.0,
            "rel": 0.0
          }
        }
      ]
    },
    {
      "id": "cold_tiny_caseC_step2003",
      "description": "Reference: the same cold-season run and the same real64 kernels/noah_driver, replayed at ESM_DT = 1e-6 s, at rel 1e-9. THE POINT OF THIS REGIME is that it is exactly where a naively written compaction component fails: the density increment is 7.5e-17 of the density, so both sndens_out - sndens_in and sndens (pexp - 1) return pure round-off, and only the cancellation-free form -- sndens_in pexp_m1 + dw (1 - dsx), referenced against the series sum published before WRF adds one to it -- still carries the signal. `dsndens_dt` is the dt -> 0 compaction rate here to twelve digits. Step 2003 of case C, a thin pack over soil that thaws and refreezes. SNOWPACK's melt-water retention branch RUNS -- the snow surface is at or above freezing; neither density limit fires. The dt -> 0 compaction rate is 1.7770e-09 per second, and the 60 s regime's tendency for the same step exceeds it by xarg/3 = 3.02e-07 relative.",
      "parameter_overrides": {
        "esd": 0.002933668671137215,
        "sndens_in": 0.23505094109091298,
        "tsnow": 286.2669097330707,
        "tsoil": 283.7584908909183,
        "dtsec": 1e-06
      },
      "time_span": {
        "start": 0.0,
        "end": 0.0
      },
      "assertions": [
        {
          "variable": "bfac",
          "time": 0.0,
          "expected": 5.153972082611458e-14,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "xarg",
          "time": 0.0,
          "expected": 1.512004643067306e-14,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "pexp_m1",
          "time": 0.0,
          "expected": 7.560023215336569e-15,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsx",
          "time": 0.0,
          "expected": 0.23505094109091476,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dw",
          "time": 0.0,
          "expected": 1.5046296296296297e-12,
          "tolerance": {
            "rel": 1e-09,
            "abs": 1e-30
          }
        },
        {
          "variable": "sndens_out",
          "time": 0.0,
          "expected": 0.23505094109206573,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsndens_dt",
          "time": 0.0,
          "expected": 1.1527420097633448e-06,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "pack_present",
          "time": 0.0,
          "expected": 1.0,
          "tolerance": {
            "abs": 0.0,
            "rel": 0.0
          }
        }
      ]
    },
    {
      "id": "cold_tiny_caseC_step2104",
      "description": "Reference: the same cold-season run and the same real64 kernels/noah_driver, replayed at ESM_DT = 1e-6 s, at rel 1e-9. THE POINT OF THIS REGIME is that it is exactly where a naively written compaction component fails: the density increment is 7.5e-17 of the density, so both sndens_out - sndens_in and sndens (pexp - 1) return pure round-off, and only the cancellation-free form -- sndens_in pexp_m1 + dw (1 - dsx), referenced against the series sum published before WRF adds one to it -- still carries the signal. `dsndens_dt` is the dt -> 0 compaction rate here to twelve digits. Step 2104 of case C, a thin pack over soil that thaws and refreezes. SNOWPACK's melt-water retention branch RUNS -- the snow surface is at or above freezing; neither density limit fires. The dt -> 0 compaction rate is 7.2135e-10 per second, and the 60 s regime's tendency for the same step exceeds it by xarg/3 = 1.19e-07 relative.",
      "parameter_overrides": {
        "esd": 0.002814299426972834,
        "sndens_in": 0.24200385594087023,
        "tsnow": 275.7970575572843,
        "tsoil": 275.6493858302569,
        "dtsec": 1e-06
      },
      "time_span": {
        "start": 0.0,
        "end": 0.0
      },
      "assertions": [
        {
          "variable": "bfac",
          "time": 0.0,
          "expected": 2.1182667727236767e-14,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "xarg",
          "time": 0.0,
          "expected": 5.961436964651838e-15,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "pexp_m1",
          "time": 0.0,
          "expected": 2.9807184823259244e-15,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsx",
          "time": 0.0,
          "expected": 0.24200385594087093,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dw",
          "time": 0.0,
          "expected": 1.5046296296296297e-12,
          "tolerance": {
            "rel": 1e-09,
            "abs": 1e-30
          }
        },
        {
          "variable": "sndens_out",
          "time": 0.0,
          "expected": 0.24200385594201143,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsndens_dt",
          "time": 0.0,
          "expected": 1.141224802862572e-06,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "pack_present",
          "time": 0.0,
          "expected": 1.0,
          "tolerance": {
            "abs": 0.0,
            "rel": 0.0
          }
        }
      ]
    },
    {
      "id": "cold_tiny_caseC_step2119",
      "description": "Reference: the same cold-season run and the same real64 kernels/noah_driver, replayed at ESM_DT = 1e-6 s, at rel 1e-9. THE POINT OF THIS REGIME is that it is exactly where a naively written compaction component fails: the density increment is 7.5e-17 of the density, so both sndens_out - sndens_in and sndens (pexp - 1) return pure round-off, and only the cancellation-free form -- sndens_in pexp_m1 + dw (1 - dsx), referenced against the series sum published before WRF adds one to it -- still carries the signal. `dsndens_dt` is the dt -> 0 compaction rate here to twelve digits. Step 2119 of case C, a thin pack over soil that thaws and refreezes. dry branch -- the snow surface is below freezing and only compaction acts; neither density limit fires. The dt -> 0 compaction rate is 5.6448e-10 per second, and the 60 s regime's tendency for the same step exceeds it by xarg/3 = 9.30e-08 relative.",
      "parameter_overrides": {
        "esd": 0.0028143154922875783,
        "sndens_in": 0.24289350120351655,
        "tsnow": 272.62433875257983,
        "tsoil": 273.0668121189063,
        "dtsec": 1e-06
      },
      "time_span": {
        "start": 0.0,
        "end": 0.0
      },
      "assertions": [
        {
          "variable": "bfac",
          "time": 0.0,
          "expected": 1.651532992769428e-14,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "xarg",
          "time": 0.0,
          "expected": 4.647934887575071e-15,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "pexp_m1",
          "time": 0.0,
          "expected": 2.323967443787539e-15,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsx",
          "time": 0.0,
          "expected": 0.24289350120351708,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dw",
          "time": 0.0,
          "expected": 0.0,
          "tolerance": {
            "rel": 1e-09,
            "abs": 1e-30
          }
        },
        {
          "variable": "sndens_out",
          "time": 0.0,
          "expected": 0.24289350120351708,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsndens_dt",
          "time": 0.0,
          "expected": 5.644765891045419e-10,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "pack_present",
          "time": 0.0,
          "expected": 1.0,
          "tolerance": {
            "abs": 0.0,
            "rel": 0.0
          }
        }
      ]
    },
    {
      "id": "cold_tiny_caseC_step2141",
      "description": "Reference: the same cold-season run and the same real64 kernels/noah_driver, replayed at ESM_DT = 1e-6 s, at rel 1e-9. THE POINT OF THIS REGIME is that it is exactly where a naively written compaction component fails: the density increment is 7.5e-17 of the density, so both sndens_out - sndens_in and sndens (pexp - 1) return pure round-off, and only the cancellation-free form -- sndens_in pexp_m1 + dw (1 - dsx), referenced against the series sum published before WRF adds one to it -- still carries the signal. `dsndens_dt` is the dt -> 0 compaction rate here to twelve digits. Step 2141 of case C, a thin pack over soil that thaws and refreezes. dry branch -- the snow surface is below freezing and only compaction acts; neither density limit fires. The dt -> 0 compaction rate is 4.2832e-10 per second, and the 60 s regime's tendency for the same step exceeds it by xarg/3 = 7.05e-08 relative.",
      "parameter_overrides": {
        "esd": 0.002814929699525834,
        "sndens_in": 0.24289463687606072,
        "tsnow": 268.83101045249964,
        "tsoil": 269.9544853583889,
        "dtsec": 1e-06
      },
      "time_span": {
        "start": 0.0,
        "end": 0.0
      },
      "assertions": [
        {
          "variable": "bfac",
          "time": 0.0,
          "expected": 1.2528912639208097e-14,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "xarg",
          "time": 0.0,
          "expected": 3.526800829087147e-15,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "pexp_m1",
          "time": 0.0,
          "expected": 1.7634004145435755e-15,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsx",
          "time": 0.0,
          "expected": 0.24289463687606117,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dw",
          "time": 0.0,
          "expected": 0.0,
          "tolerance": {
            "rel": 1e-09,
            "abs": 1e-30
          }
        },
        {
          "variable": "sndens_out",
          "time": 0.0,
          "expected": 0.24289463687606117,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsndens_dt",
          "time": 0.0,
          "expected": 4.2832050335765673e-10,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "pack_present",
          "time": 0.0,
          "expected": 1.0,
          "tolerance": {
            "abs": 0.0,
            "rel": 0.0
          }
        }
      ]
    },
    {
      "id": "cold_tiny_caseC_step2166",
      "description": "Reference: the same cold-season run and the same real64 kernels/noah_driver, replayed at ESM_DT = 1e-6 s, at rel 1e-9. THE POINT OF THIS REGIME is that it is exactly where a naively written compaction component fails: the density increment is 7.5e-17 of the density, so both sndens_out - sndens_in and sndens (pexp - 1) return pure round-off, and only the cancellation-free form -- sndens_in pexp_m1 + dw (1 - dsx), referenced against the series sum published before WRF adds one to it -- still carries the signal. `dsndens_dt` is the dt -> 0 compaction rate here to twelve digits. Step 2166 of case C, a thin pack over soil that thaws and refreezes. dry branch -- the snow surface is below freezing and only compaction acts; neither density limit fires. The dt -> 0 compaction rate is 4.0954e-10 per second, and the 60 s regime's tendency for the same step exceeds it by xarg/3 = 6.74e-08 relative.",
      "parameter_overrides": {
        "esd": 0.002815851476044067,
        "sndens_in": 0.2428957064342471,
        "tsnow": 268.2079244189253,
        "tsoil": 269.44896955996074,
        "dtsec": 1e-06
      },
      "time_span": {
        "start": 0.0,
        "end": 0.0
      },
      "assertions": [
        {
          "variable": "bfac",
          "time": 0.0,
          "expected": 1.1975614420121365e-14,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "xarg",
          "time": 0.0,
          "expected": 3.3721551541433357e-15,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "pexp_m1",
          "time": 0.0,
          "expected": 1.6860775770716696e-15,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsx",
          "time": 0.0,
          "expected": 0.24289570643424754,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dw",
          "time": 0.0,
          "expected": 0.0,
          "tolerance": {
            "rel": 1e-09,
            "abs": 1e-30
          }
        },
        {
          "variable": "sndens_out",
          "time": 0.0,
          "expected": 0.24289570643424754,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsndens_dt",
          "time": 0.0,
          "expected": 4.0954100418576693e-10,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "pack_present",
          "time": 0.0,
          "expected": 1.0,
          "tolerance": {
            "abs": 0.0,
            "rel": 0.0
          }
        }
      ]
    },
    {
      "id": "cold_tiny_caseC_step2880",
      "description": "Reference: the same cold-season run and the same real64 kernels/noah_driver, replayed at ESM_DT = 1e-6 s, at rel 1e-9. THE POINT OF THIS REGIME is that it is exactly where a naively written compaction component fails: the density increment is 7.5e-17 of the density, so both sndens_out - sndens_in and sndens (pexp - 1) return pure round-off, and only the cancellation-free form -- sndens_in pexp_m1 + dw (1 - dsx), referenced against the series sum published before WRF adds one to it -- still carries the signal. `dsndens_dt` is the dt -> 0 compaction rate here to twelve digits. Step 2880 of case C, a thin pack over soil that thaws and refreezes. dry branch -- the snow surface is below freezing and only compaction acts; neither density limit fires. The dt -> 0 compaction rate is 3.5820e-10 per second, and the 60 s regime's tendency for the same step exceeds it by xarg/3 = 5.90e-08 relative.",
      "parameter_overrides": {
        "esd": 0.002840091707185486,
        "sndens_in": 0.24292904682425578,
        "tsnow": 266.5171266345389,
        "tsoil": 267.59058241409787,
        "dtsec": 1e-06
      },
      "time_span": {
        "start": 0.0,
        "end": 0.0
      },
      "assertions": [
        {
          "variable": "bfac",
          "time": 0.0,
          "expected": 1.0383363949518005e-14,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "xarg",
          "time": 0.0,
          "expected": 2.9489705845714822e-15,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "pexp_m1",
          "time": 0.0,
          "expected": 1.4744852922857423e-15,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsx",
          "time": 0.0,
          "expected": 0.24292904682425617,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dw",
          "time": 0.0,
          "expected": 0.0,
          "tolerance": {
            "rel": 1e-09,
            "abs": 1e-30
          }
        },
        {
          "variable": "sndens_out",
          "time": 0.0,
          "expected": 0.24292904682425617,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsndens_dt",
          "time": 0.0,
          "expected": 3.581953066113596e-10,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "pack_present",
          "time": 0.0,
          "expected": 1.0,
          "tolerance": {
            "abs": 0.0,
            "rel": 0.0
          }
        }
      ]
    },
    {
      "id": "cold_tiny_caseD_step1",
      "description": "Reference: the same cold-season run and the same real64 kernels/noah_driver, replayed at ESM_DT = 1e-6 s, at rel 1e-9. THE POINT OF THIS REGIME is that it is exactly where a naively written compaction component fails: the density increment is 7.5e-17 of the density, so both sndens_out - sndens_in and sndens (pexp - 1) return pure round-off, and only the cancellation-free form -- sndens_in pexp_m1 + dw (1 - dsx), referenced against the series sum published before WRF adds one to it -- still carries the signal. `dsndens_dt` is the dt -> 0 compaction rate here to twelve digits. Step 1 of case D, a trace pack over frozen ground in a dry column. dry branch -- the snow surface is below freezing and only compaction acts; the pack is below the esdcx floor, so the compaction argument uses the floored 0.01 cm while the depth diagnostic keeps the true water equivalent; neither density limit fires. The dt -> 0 compaction rate is 1.8382e-11 per second, and the 60 s regime's tendency for the same step exceeds it by xarg/3 = 3.68e-09 relative.",
      "parameter_overrides": {
        "esd": 5.000000237483419e-05,
        "sndens_in": 0.2,
        "tsnow": 262.90190014862435,
        "tsoil": 262.9471643284575,
        "dtsec": 1e-06
      },
      "time_span": {
        "start": 0.0,
        "end": 0.0
      },
      "assertions": [
        {
          "variable": "bfac",
          "time": 0.0,
          "expected": 1.8381948174992332e-14,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "xarg",
          "time": 0.0,
          "expected": 1.8381948174992333e-16,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "pexp_m1",
          "time": 0.0,
          "expected": 9.190974087496167e-17,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsx",
          "time": 0.0,
          "expected": 0.2,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dw",
          "time": 0.0,
          "expected": 0.0,
          "tolerance": {
            "rel": 1e-09,
            "abs": 1e-30
          }
        },
        {
          "variable": "sndens_out",
          "time": 0.0,
          "expected": 0.2,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsndens_dt",
          "time": 0.0,
          "expected": 1.8381948174992335e-11,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "pack_present",
          "time": 0.0,
          "expected": 1.0,
          "tolerance": {
            "abs": 0.0,
            "rel": 0.0
          }
        }
      ]
    },
    {
      "id": "cold_tiny_caseD_step26",
      "description": "Reference: the same cold-season run and the same real64 kernels/noah_driver, replayed at ESM_DT = 1e-6 s, at rel 1e-9. THE POINT OF THIS REGIME is that it is exactly where a naively written compaction component fails: the density increment is 7.5e-17 of the density, so both sndens_out - sndens_in and sndens (pexp - 1) return pure round-off, and only the cancellation-free form -- sndens_in pexp_m1 + dw (1 - dsx), referenced against the series sum published before WRF adds one to it -- still carries the signal. `dsndens_dt` is the dt -> 0 compaction rate here to twelve digits. Step 26 of case D, a trace pack over frozen ground in a dry column. dry branch -- the snow surface is below freezing and only compaction acts; the pack is below the esdcx floor, so the compaction argument uses the floored 0.01 cm while the depth diagnostic keeps the true water equivalent; neither density limit fires. The dt -> 0 compaction rate is 1.5827e-11 per second, and the 60 s regime's tendency for the same step exceeds it by xarg/3 = 3.17e-09 relative.",
      "parameter_overrides": {
        "esd": 4.996308780390635e-05,
        "sndens_in": 0.20000037571748935,
        "tsnow": 261.02665625536906,
        "tsoil": 261.08120979942294,
        "dtsec": 1e-06
      },
      "time_span": {
        "start": 0.0,
        "end": 0.0
      },
      "assertions": [
        {
          "variable": "bfac",
          "time": 0.0,
          "expected": 1.5826935683751008e-14,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "xarg",
          "time": 0.0,
          "expected": 1.5826935683751007e-16,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "pexp_m1",
          "time": 0.0,
          "expected": 7.913467841875504e-17,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsx",
          "time": 0.0,
          "expected": 0.20000037571748935,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dw",
          "time": 0.0,
          "expected": 0.0,
          "tolerance": {
            "rel": 1e-09,
            "abs": 1e-30
          }
        },
        {
          "variable": "sndens_out",
          "time": 0.0,
          "expected": 0.20000037571748935,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsndens_dt",
          "time": 0.0,
          "expected": 1.5826965416033705e-11,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "pack_present",
          "time": 0.0,
          "expected": 1.0,
          "tolerance": {
            "abs": 0.0,
            "rel": 0.0
          }
        }
      ]
    },
    {
      "id": "cold_tiny_caseD_step51",
      "description": "Reference: the same cold-season run and the same real64 kernels/noah_driver, replayed at ESM_DT = 1e-6 s, at rel 1e-9. THE POINT OF THIS REGIME is that it is exactly where a naively written compaction component fails: the density increment is 7.5e-17 of the density, so both sndens_out - sndens_in and sndens (pexp - 1) return pure round-off, and only the cancellation-free form -- sndens_in pexp_m1 + dw (1 - dsx), referenced against the series sum published before WRF adds one to it -- still carries the signal. `dsndens_dt` is the dt -> 0 compaction rate here to twelve digits. Step 51 of case D, a trace pack over frozen ground in a dry column. dry branch -- the snow surface is below freezing and only compaction acts; the pack is below the esdcx floor, so the compaction argument uses the floored 0.01 cm while the depth diagnostic keeps the true water equivalent; neither density limit fires. The dt -> 0 compaction rate is 1.5096e-11 per second, and the 60 s regime's tendency for the same step exceeds it by xarg/3 = 3.02e-09 relative.",
      "parameter_overrides": {
        "esd": 4.9951497203431465e-05,
        "sndens_in": 0.20000064964772196,
        "tsnow": 260.43398841500215,
        "tsoil": 260.49147503268733,
        "dtsec": 1e-06
      },
      "time_span": {
        "start": 0.0,
        "end": 0.0
      },
      "assertions": [
        {
          "variable": "bfac",
          "time": 0.0,
          "expected": 1.509572244377127e-14,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "xarg",
          "time": 0.0,
          "expected": 1.509572244377127e-16,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "pexp_m1",
          "time": 0.0,
          "expected": 7.547861221885635e-17,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsx",
          "time": 0.0,
          "expected": 0.20000064964772196,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dw",
          "time": 0.0,
          "expected": 0.0,
          "tolerance": {
            "rel": 1e-09,
            "abs": 1e-30
          }
        },
        {
          "variable": "sndens_out",
          "time": 0.0,
          "expected": 0.20000064964772196,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsndens_dt",
          "time": 0.0,
          "expected": 1.5095771478279756e-11,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "pack_present",
          "time": 0.0,
          "expected": 1.0,
          "tolerance": {
            "abs": 0.0,
            "rel": 0.0
          }
        }
      ]
    },
    {
      "id": "cold_tiny_caseD_step392",
      "description": "Reference: the same cold-season run and the same real64 kernels/noah_driver, replayed at ESM_DT = 1e-6 s, at rel 1e-9. THE POINT OF THIS REGIME is that it is exactly where a naively written compaction component fails: the density increment is 7.5e-17 of the density, so both sndens_out - sndens_in and sndens (pexp - 1) return pure round-off, and only the cancellation-free form -- sndens_in pexp_m1 + dw (1 - dsx), referenced against the series sum published before WRF adds one to it -- still carries the signal. `dsndens_dt` is the dt -> 0 compaction rate here to twelve digits. Step 392 of case D, a trace pack over frozen ground in a dry column. SNOWPACK's melt-water retention branch RUNS -- the snow surface is at or above freezing; the pack is below the esdcx floor, so the compaction argument uses the floored 0.01 cm while the depth diagnostic keeps the true water equivalent; neither density limit fires. The dt -> 0 compaction rate is 6.9998e-11 per second, and the 60 s regime's tendency for the same step exceeds it by xarg/3 = 1.32e-08 relative.",
      "parameter_overrides": {
        "esd": 4.588760566372142e-05,
        "sndens_in": 0.2121168851409987,
        "tsnow": 282.1061051446816,
        "tsoil": 282.06111154540895,
        "dtsec": 1e-06
      },
      "time_span": {
        "start": 0.0,
        "end": 0.0
      },
      "assertions": [
        {
          "variable": "bfac",
          "time": 0.0,
          "expected": 6.599906703919676e-14,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "xarg",
          "time": 0.0,
          "expected": 6.599906703919676e-16,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "pexp_m1",
          "time": 0.0,
          "expected": 3.299953351959839e-16,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsx",
          "time": 0.0,
          "expected": 0.21211688514099875,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dw",
          "time": 0.0,
          "expected": 1.5046296296296297e-12,
          "tolerance": {
            "rel": 1e-09,
            "abs": 1e-30
          }
        },
        {
          "variable": "sndens_out",
          "time": 0.0,
          "expected": 0.21211688514218424,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsndens_dt",
          "time": 0.0,
          "expected": 1.1855422768843508e-06,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "pack_present",
          "time": 0.0,
          "expected": 1.0,
          "tolerance": {
            "abs": 0.0,
            "rel": 0.0
          }
        }
      ]
    },
    {
      "id": "cold_tiny_caseD_step417",
      "description": "Reference: the same cold-season run and the same real64 kernels/noah_driver, replayed at ESM_DT = 1e-6 s, at rel 1e-9. THE POINT OF THIS REGIME is that it is exactly where a naively written compaction component fails: the density increment is 7.5e-17 of the density, so both sndens_out - sndens_in and sndens (pexp - 1) return pure round-off, and only the cancellation-free form -- sndens_in pexp_m1 + dw (1 - dsx), referenced against the series sum published before WRF adds one to it -- still carries the signal. `dsndens_dt` is the dt -> 0 compaction rate here to twelve digits. Step 417 of case D, a trace pack over frozen ground in a dry column. SNOWPACK's melt-water retention branch RUNS -- the snow surface is at or above freezing; the pack is below the esdcx floor, so the compaction argument uses the floored 0.01 cm while the depth diagnostic keeps the true water equivalent; neither density limit fires. The dt -> 0 compaction rate is 7.3668e-11 per second, and the 60 s regime's tendency for the same step exceeds it by xarg/3 = 1.38e-08 relative.",
      "parameter_overrides": {
        "esd": 4.222644201633197e-05,
        "sndens_in": 0.21389331851641966,
        "tsnow": 283.1070103613335,
        "tsoil": 283.06199859771243,
        "dtsec": 1e-06
      },
      "time_span": {
        "start": 0.0,
        "end": 0.0
      },
      "assertions": [
        {
          "variable": "bfac",
          "time": 0.0,
          "expected": 6.888284727644278e-14,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "xarg",
          "time": 0.0,
          "expected": 6.888284727644278e-16,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "pexp_m1",
          "time": 0.0,
          "expected": 3.4441423638221396e-16,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsx",
          "time": 0.0,
          "expected": 0.21389331851641974,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dw",
          "time": 0.0,
          "expected": 1.5046296296296297e-12,
          "tolerance": {
            "rel": 1e-09,
            "abs": 1e-30
          }
        },
        {
          "variable": "sndens_out",
          "time": 0.0,
          "expected": 0.21389331851760254,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsndens_dt",
          "time": 0.0,
          "expected": 1.1828730729139808e-06,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "pack_present",
          "time": 0.0,
          "expected": 1.0,
          "tolerance": {
            "abs": 0.0,
            "rel": 0.0
          }
        }
      ]
    },
    {
      "id": "cold_tiny_caseD_step521",
      "description": "Reference: the same cold-season run and the same real64 kernels/noah_driver, replayed at ESM_DT = 1e-6 s, at rel 1e-9. THE POINT OF THIS REGIME is that it is exactly where a naively written compaction component fails: the density increment is 7.5e-17 of the density, so both sndens_out - sndens_in and sndens (pexp - 1) return pure round-off, and only the cancellation-free form -- sndens_in pexp_m1 + dw (1 - dsx), referenced against the series sum published before WRF adds one to it -- still carries the signal. `dsndens_dt` is the dt -> 0 compaction rate here to twelve digits. Step 521 of case D, a trace pack over frozen ground in a dry column. SNOWPACK's melt-water retention branch RUNS -- the snow surface is at or above freezing; the pack is below the esdcx floor, so the compaction argument uses the floored 0.01 cm while the depth diagnostic keeps the true water equivalent; neither density limit fires. The dt -> 0 compaction rate is 6.1199e-11 per second, and the 60 s regime's tendency for the same step exceeds it by xarg/3 = 1.11e-08 relative.",
      "parameter_overrides": {
        "esd": 1.2073031024383523e-05,
        "sndens_in": 0.22124005810918224,
        "tsnow": 282.27871277137297,
        "tsoil": 282.26730000822283,
        "dtsec": 1e-06
      },
      "time_span": {
        "start": 0.0,
        "end": 0.0
      },
      "assertions": [
        {
          "variable": "bfac",
          "time": 0.0,
          "expected": 5.532394504352602e-14,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "xarg",
          "time": 0.0,
          "expected": 5.532394504352602e-16,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "pexp_m1",
          "time": 0.0,
          "expected": 2.7661972521763015e-16,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsx",
          "time": 0.0,
          "expected": 0.2212400581091823,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dw",
          "time": 0.0,
          "expected": 1.5046296296296297e-12,
          "tolerance": {
            "rel": 1e-09,
            "abs": 1e-30
          }
        },
        {
          "variable": "sndens_out",
          "time": 0.0,
          "expected": 0.22124005811035405,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsndens_dt",
          "time": 0.0,
          "expected": 1.1718064823016544e-06,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "pack_present",
          "time": 0.0,
          "expected": 1.0,
          "tolerance": {
            "abs": 0.0,
            "rel": 0.0
          }
        }
      ]
    },
    {
      "id": "cold_tiny_caseD_step553",
      "description": "Reference: the same cold-season run and the same real64 kernels/noah_driver, replayed at ESM_DT = 1e-6 s, at rel 1e-9. THE POINT OF THIS REGIME is that it is exactly where a naively written compaction component fails: the density increment is 7.5e-17 of the density, so both sndens_out - sndens_in and sndens (pexp - 1) return pure round-off, and only the cancellation-free form -- sndens_in pexp_m1 + dw (1 - dsx), referenced against the series sum published before WRF adds one to it -- still carries the signal. `dsndens_dt` is the dt -> 0 compaction rate here to twelve digits. Step 553 of case D, a trace pack over frozen ground in a dry column. SNOWPACK's melt-water retention branch RUNS -- the snow surface is at or above freezing; the pack is below the esdcx floor, so the compaction argument uses the floored 0.01 cm while the depth diagnostic keeps the true water equivalent; neither density limit fires. The dt -> 0 compaction rate is 5.5715e-11 per second; this is the one step whose pack does not survive a full 60 s step, so it has no companion there.",
      "parameter_overrides": {
        "esd": 1.2517150421502591e-06,
        "sndens_in": 0.22348673836568872,
        "tsnow": 281.563297708391,
        "tsoil": 281.562242990759,
        "dtsec": 1e-06
      },
      "time_span": {
        "start": 0.0,
        "end": 0.0
      },
      "assertions": [
        {
          "variable": "bfac",
          "time": 0.0,
          "expected": 4.98593802792149e-14,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "xarg",
          "time": 0.0,
          "expected": 4.98593802792149e-16,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "pexp_m1",
          "time": 0.0,
          "expected": 2.4929690139607455e-16,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsx",
          "time": 0.0,
          "expected": 0.22348673836568878,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dw",
          "time": 0.0,
          "expected": 1.5046296296296297e-12,
          "tolerance": {
            "rel": 1e-09,
            "abs": 1e-30
          }
        },
        {
          "variable": "sndens_out",
          "time": 0.0,
          "expected": 0.22348673836685715,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "dsndens_dt",
          "time": 0.0,
          "expected": 1.1684205758067073e-06,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "pack_present",
          "time": 0.0,
          "expected": 1.0,
          "tolerance": {
            "abs": 0.0,
            "rel": 0.0
          }
        }
      ]
    }
  ]
}