Version: dev

CalCldfra

WRF's grid-scale cloud-fraction diagnostic for icloud = 1, SUBROUTINE cal_cldfra1 of phys/module_radiation_driver.F (WRF 4.8), as a single-column component on the EarthSciDiscretizations grids/column_nonuniform_1d grid (layer index `lev`, layer 1 at the surface). The scheme is pointwise in the vertical -- no operator couples one layer to another -- and purely diagnostic: it answers the cloud fraction a radiation scheme should see, given the temperature, the pressure and the hydrometeor mixing ratios of the same layer. Three branches, which the observed `branch` reports as 1, 2 and 3 exactly as the Fortran's cldfra1_flag does: a layer whose total cloud condensate q_cld = qi + qc + qs falls below QCLDMIN is cloud-free (branch 1); a layer that holds condensate and is at or above the critical relative humidity RHGRID is fully cloudy (branch 2); otherwise the Randall (1994) / Zhao (1995) fit, modified for grid-scale saturation at RHGRID, gives the fraction (RHUM/RHGRID)^PEXP (1 - exp(-ALPHA0 q_cld / (RHGRID qvs - qv)^GAMMA)), snapped to zero below 0.01 (branch 3). The saturation mixing ratio is weighted between the over-water and over-ice Murray (1966) fits by the ice fraction (qi + qs)/q_cld of the condensate, so a glaciated layer is judged against the lower ice saturation and turns cloudy sooner. This component transcribes the mp_physics branch for schemes that carry cloud water, cloud ice and snow as separate species (F_QC .and. F_QI .and. F_QS: WSM6 and the other options 2, 4, 6, 7, 8), which is what the em_scm_xy reference run uses; the P3 (no snow), Kessler/WSM3 (cloud water only), option 5 (F_ICE_PHY-weighted) and Ferrier high-resolution branches are not transcribed, and neither is the mp-independent ELSE arm, which is dead code (EqWeFiC FORTRAN_BUGS.md B13). Consumer-supplied column fields (T, p, qv, qc, qi, qs) are rewrite-target ops `input_<name>` lowered by the consumer, in the inline tests through per-test expression_template_imports libraries generated by EqWeFiC tools/fill_tests.py (esm-spec 6.6 parameter_overrides admit scalars only). This is the component the EqWeather-SCM state document needs so that cloud fraction in a running column is diagnosed rather than prescribed from a dump: its `cldfra` feeds the coupling-target parameter `cldfra_in` of the RRTM longwave column stage (components/atmospheric_radiation/rrtm_lw/rrtm_column.esm, which reads it through `input_cldfra`), via EqWeFiC couplings/tests/scmstate_couple_rrtmcol_inputs.esm. Dudhia shortwave takes no cloud fraction (WRF's SWRAD has no CLDFRA argument). EqWeFiC, cited here for FORTRAN_BUGS.md, tools/ and couplings/, is https://github.com/EarthSciML/EqWeFiC, where this component was developed and is coupled into a WRF column.

Domain
atmospheric_radiation
Subdomain
cloud_fraction
Type
model
.esm schema
1.2.0
Source
components/atmospheric_radiation/cloud_fraction/cal_cldfra.esm
Tags
#atmospheric_radiation #cloud_fraction #model #wrf #radiation #cloud-fraction #cal_cldfra1 #icloud #column

Description

WRF’s grid-scale cloud-fraction diagnostic for icloud = 1, SUBROUTINE cal_cldfra1 of phys/module_radiation_driver.F (WRF 4.8), as a single-column component on the EarthSciDiscretizations grids/column_nonuniform_1d grid (layer index lev, layer 1 at the surface). The scheme is pointwise in the vertical – no operator couples one layer to another – and purely diagnostic: it answers the cloud fraction a radiation scheme should see, given the temperature, the pressure and the hydrometeor mixing ratios of the same layer. Three branches, which the observed branch reports as 1, 2 and 3 exactly as the Fortran’s cldfra1_flag does: a layer whose total cloud condensate q_cld = qi + qc + qs falls below QCLDMIN is cloud-free (branch 1); a layer that holds condensate and is at or above the critical relative humidity RHGRID is fully cloudy (branch 2); otherwise the Randall (1994) / Zhao (1995) fit, modified for grid-scale saturation at RHGRID, gives the fraction (RHUM/RHGRID)^PEXP (1 - exp(-ALPHA0 q_cld / (RHGRID qvs - qv)^GAMMA)), snapped to zero below 0.01 (branch 3). The saturation mixing ratio is weighted between the over-water and over-ice Murray (1966) fits by the ice fraction (qi + qs)/q_cld of the condensate, so a glaciated layer is judged against the lower ice saturation and turns cloudy sooner. This component transcribes the mp_physics branch for schemes that carry cloud water, cloud ice and snow as separate species (F_QC .and. F_QI .and. F_QS: WSM6 and the other options 2, 4, 6, 7, 8), which is what the em_scm_xy reference run uses; the P3 (no snow), Kessler/WSM3 (cloud water only), option 5 (F_ICE_PHY-weighted) and Ferrier high-resolution branches are not transcribed, and neither is the mp-independent ELSE arm, which is dead code (EqWeFiC FORTRAN_BUGS.md B13). Consumer-supplied column fields (T, p, qv, qc, qi, qs) are rewrite-target ops input_<name> lowered by the consumer, in the inline tests through per-test expression_template_imports libraries generated by EqWeFiC tools/fill_tests.py (esm-spec 6.6 parameter_overrides admit scalars only). This is the component the EqWeather-SCM state document needs so that cloud fraction in a running column is diagnosed rather than prescribed from a dump: its cldfra feeds the coupling-target parameter cldfra_in of the RRTM longwave column stage (components/atmospheric_radiation/rrtm_lw/rrtm_column.esm, which reads it through input_cldfra), via EqWeFiC couplings/tests/scmstate_couple_rrtmcol_inputs.esm. Dudhia shortwave takes no cloud fraction (WRF’s SWRAD has no CLDFRA argument). EqWeFiC, cited here for FORTRAN_BUGS.md, tools/ and couplings/, is https://github.com/EarthSciML/EqWeFiC , where this component was developed and is coupled into a WRF column.

Reference

  • Randall, D. A. (1994). Cloud parameterization for climate modeling: Status and prospects. Atmos. Res. 31, 1-26.
  • Hong, S.-Y., Juang, H.-M. H. and Zhao, Q. (1998). Implementation of prognostic cloud scheme for a regional spectral model. Mon. Wea. Rev. 126, 2621-2639.
  • wrf-model/WRF v4.8.0 phys/module_radiation_driver.F, SUBROUTINE cal_cldfra1.

Parameters

NameUnitsDefaultDescription
T_inK0.0Coupling target (assembly input path): a shaped parameter that receives the layer temperature (WRF t_phy) from the EqWeather-SCM state document (EqWeFiC couplings/eqweather_scm_dynamics.esm) through a document-level coupling.variable_map (param_to_var), when the mount-edge library EqWeFiC couplings/tests/scmstate_couple_cldfra_inputs.esm lowers input_T to it (esm-spec 9.7.10, 10.4). Unused by this component’s own inline tests, which lower input_T to dumped constants instead.
p_inPa0.0Coupling target (assembly input path): a shaped parameter that receives the layer pressure (WRF p_phy) from the column geometry (EqWeFiC components/atmospheric_dynamics/wrf_arw/column_geometry.esm, output p) through a document-level coupling.variable_map (param_to_var), when the mount-edge library EqWeFiC couplings/tests/scmstate_couple_cldfra_inputs.esm lowers input_p to it. Unused by this component’s own inline tests.
qv_inkg/kg0.0Coupling target (assembly input path): the layer water-vapour mixing ratio (WRF QV) from the state document’s ODE state, through input_qv. Unused by this component’s own inline tests.
qc_inkg/kg0.0Coupling target (assembly input path): the layer cloud-water mixing ratio (WRF QC) from the state document’s ODE state, through input_qc. Unused by this component’s own inline tests.
qi_inkg/kg0.0Coupling target (assembly input path): the layer cloud-ice mixing ratio (WRF QI) from the state document’s ODE state, through input_qi. Unused by this component’s own inline tests.
qs_inkg/kg0.0Coupling target (assembly input path): the layer snow mixing ratio (WRF QS) from the state document’s ODE state, through input_qs. Unused by this component’s own inline tests.

Observed

NameUnitsDefaultDescription
TKLayer temperature (WRF t_phy), through input_T.
pPaLayer pressure (WRF p_phy), through input_p.
qvkg/kgWater-vapour mixing ratio (WRF QV), through input_qv.
qckg/kgCloud-water mixing ratio (WRF QC), through input_qc.
qikg/kgCloud-ice mixing ratio (WRF QI), through input_qi.
qskg/kgSnow mixing ratio (WRF QS), through input_qs.
es_wPaSaturation vapour pressure over liquid water, Murray (1966) fit with the over-water coefficients (ESW).
es_iPaSaturation vapour pressure over ice, the same fit with the over-ice coefficients (ESI).
qvs_wkg/kgSaturation mixing ratio over liquid water (QVSW).
qvs_ikg/kgSaturation mixing ratio over ice (QVSI).
q_cldkg/kgTotal cloud condensate QCLD = qi + qc + qs. Rain and graupel are deliberately excluded: ‘Rain is not part of cloud, only cloud water + cloud ice + snow’.
has_cloud1Indicator of a layer holding at least QCLDMIN of condensate, 1 where q_cld >= QCLDMIN and 0 below it. The Fortran spells this threshold three times (the ice weight, the cloud fraction and the branch flag); it is written once here and reused.
ice_weight1Ice fraction of the condensate, WEIGHT = (qi + qs)/QCLD, zero in a cloud-free layer. The denominator is floored at QCLDMIN so the quotient stays finite where the indicator has already zeroed it; in every layer the indicator admits, the floor is inactive and the quotient is the Fortran’s exactly.
qvskg/kgPhase-weighted saturation mixing ratio QVS_WEIGHT = (1 - WEIGHT) QVSW + WEIGHT QVSI.
rh1Relative humidity against the phase-weighted saturation, RHUM = qv/QVS_WEIGHT.
subsatkg/kgSub-saturation SUBSAT = MAX(1.E-10, RHGRID QVS_WEIGHT - qv), the denominator of the Randall fit.
arg1Exponential argument ARG = MAX(-6.9, -ALPHA0 QCLD / SUBSAT^GAMMA).
cldfra_fit1The Randall (1994) fit itself, (MAX(1.E-10, RHUM)/RHGRID)^PEXP (1 - EXP(ARG)), before the 0.01 snap and before the two overriding branches. Evaluated in every layer; it is the branch-3 value.
cldfra1Grid-scale cloud fraction CLDFRA, the scheme’s output: 0 in a cloud-free layer, 1 in a saturated cloudy layer, and otherwise the Randall fit snapped to 0 below 0.01.
branch1Which of the three arms produced the fraction, the Fortran’s cldfra1_flag: 1 = no condensate (CLDFRA forced to 0), 2 = at or above RHGRID (forced to 1), 3 = the Randall fit. Reported so a test can pin the branch and not only the number.

Observed expressions

$$ T = \mathrm{input_T}\left(\right) $$

Layer temperature (WRF t_phy), through input_T.

$$ p = \mathrm{input_p}\left(\right) $$

Layer pressure (WRF p_phy), through input_p.

$$ qv = \mathrm{input_qv}\left(\right) $$

Water-vapour mixing ratio (WRF QV), through input_qv.

$$ qc = \mathrm{input_qc}\left(\right) $$

Cloud-water mixing ratio (WRF QC), through input_qc.

$$ qi = \mathrm{input_qi}\left(\right) $$

Cloud-ice mixing ratio (WRF QI), through input_qi.

$$ qs = \mathrm{input_qs}\left(\right) $$

Snow mixing ratio (WRF QS), through input_qs.

$$ es_{w} = par.pa_{per\,kpa} \cdot \mathrm{bolton_saturation_vapor_pressure}\left(T=T, svp1=par.svp1, svp2=par.svp2, svp3=par.svp3, svpt0=wrf.svpt0\right) $$

Saturation vapour pressure over liquid water, Murray (1966) fit with the over-water coefficients (ESW).

$$ es_{i} = par.pa_{per\,kpa} \cdot \mathrm{bolton_saturation_vapor_pressure}\left(T=T, svp1=par.svp1, svp2=par.svpi2, svp3=par.svpi3, svpt0=wrf.svpt0\right) $$

Saturation vapour pressure over ice, the same fit with the over-ice coefficients (ESI).

$$ qvs_{w} = \mathrm{saturation_mixing_ratio}\left(es=es_{w}, p=p, ep_{2}=wrf.ep_{2}\right) $$

Saturation mixing ratio over liquid water (QVSW).

$$ qvs_{i} = \mathrm{saturation_mixing_ratio}\left(es=es_{i}, p=p, ep_{2}=wrf.ep_{2}\right) $$

Saturation mixing ratio over ice (QVSI).

$$ q_{cld} = qi + qc + qs $$

Total cloud condensate QCLD = qi + qc + qs. Rain and graupel are deliberately excluded: ‘Rain is not part of cloud, only cloud water + cloud ice + snow’.

$$ has_{cloud} = \mathrm{ifelse}\left(\mathrm{<}\left(q_{cld}, par.qcldmin\right), 0.0, 1.0\right) $$

Indicator of a layer holding at least QCLDMIN of condensate, 1 where q_cld >= QCLDMIN and 0 below it. The Fortran spells this threshold three times (the ice weight, the cloud fraction and the branch flag); it is written once here and reused.

$$ ice_{weight} = has_{cloud} \cdot \frac{qi + qs}{\mathrm{max}\left(q_{cld}, par.qcldmin\right)} $$

Ice fraction of the condensate, WEIGHT = (qi + qs)/QCLD, zero in a cloud-free layer. The denominator is floored at QCLDMIN so the quotient stays finite where the indicator has already zeroed it; in every layer the indicator admits, the floor is inactive and the quotient is the Fortran’s exactly.

$$ qvs = \left(1.0 - ice_{weight}\right) \cdot qvs_{w} + ice_{weight} \cdot qvs_{i} $$

Phase-weighted saturation mixing ratio QVS_WEIGHT = (1 - WEIGHT) QVSW + WEIGHT QVSI.

$$ rh = \frac{qv}{qvs} $$

Relative humidity against the phase-weighted saturation, RHUM = qv/QVS_WEIGHT.

$$ subsat = \mathrm{max}\left(par.subsat_{floor}, par.rhgrid \cdot qvs - qv\right) $$

Sub-saturation SUBSAT = MAX(1.E-10, RHGRID QVS_WEIGHT - qv), the denominator of the Randall fit.

$$ arg = \mathrm{max}\left(par.arg_{floor}, \frac{\left(-1.0\right) \cdot par.alpha0 \cdot q_{cld}}{subsat^{par.gamma}}\right) $$

Exponential argument ARG = MAX(-6.9, -ALPHA0 QCLD / SUBSAT^GAMMA).

$$ cldfra_{fit} = \frac{\mathrm{max}\left(par.rh_{floor}, rh\right)}{par.rhgrid}^{par.pexp} \cdot \left(1.0 - e^{arg}\right) $$

The Randall (1994) fit itself, (MAX(1.E-10, RHUM)/RHGRID)^PEXP (1 - EXP(ARG)), before the 0.01 snap and before the two overriding branches. Evaluated in every layer; it is the branch-3 value.

$$ cldfra = has_{cloud} \cdot \mathrm{ifelse}\left(\mathrm{>=}\left(rh, par.rhgrid\right), 1.0, \mathrm{ifelse}\left(\mathrm{<}\left(cldfra_{fit}, par.cldfra_{floor}\right), 0.0, cldfra_{fit}\right)\right) $$

Grid-scale cloud fraction CLDFRA, the scheme’s output: 0 in a cloud-free layer, 1 in a saturated cloudy layer, and otherwise the Randall fit snapped to 0 below 0.01.

$$ branch = \mathrm{ifelse}\left(\mathrm{<}\left(has_{cloud}, 0.5\right), 1.0, \mathrm{ifelse}\left(\mathrm{>=}\left(rh, par.rhgrid\right), 2.0, 3.0\right)\right) $$

Which of the three arms produced the fraction, the Fortran’s cldfra1_flag: 1 = no condensate (CLDFRA forced to 0), 2 = at or above RHGRID (forced to 1), 3 = the Randall fit. Reported so a test can pin the branch and not only the number.

Raw .esm

Raw .esm JSON (this component)
{
  "expression_template_imports": [
    {
      "ref": "${ESD_ROOT}/grids/column_nonuniform_1d/grid.esm",
      "bindings": {
        "NLEV": "NLEV"
      }
    },
    {
      "ref": "../../../lib/wrf_thermo.esm"
    }
  ],
  "subsystems": {
    "wrf": {
      "ref": "../../../lib/wrf_constants.esm"
    },
    "par": {
      "ref": "./cal_cldfra_parameters.esm"
    }
  },
  "variables": {
    "T_in": {
      "type": "parameter",
      "units": "K",
      "shape": [
        "lev"
      ],
      "default": 0.0,
      "description": "Coupling target (assembly input path): a shaped parameter that receives the layer temperature (WRF t_phy) from the EqWeather-SCM state document (EqWeFiC couplings/eqweather_scm_dynamics.esm) through a document-level `coupling.variable_map` (param_to_var), when the mount-edge library EqWeFiC couplings/tests/scmstate_couple_cldfra_inputs.esm lowers `input_T` to it (esm-spec 9.7.10, 10.4). Unused by this component's own inline tests, which lower `input_T` to dumped constants instead."
    },
    "p_in": {
      "type": "parameter",
      "units": "Pa",
      "shape": [
        "lev"
      ],
      "default": 0.0,
      "description": "Coupling target (assembly input path): a shaped parameter that receives the layer pressure (WRF p_phy) from the column geometry (EqWeFiC components/atmospheric_dynamics/wrf_arw/column_geometry.esm, output `p`) through a document-level `coupling.variable_map` (param_to_var), when the mount-edge library EqWeFiC couplings/tests/scmstate_couple_cldfra_inputs.esm lowers `input_p` to it. Unused by this component's own inline tests."
    },
    "qv_in": {
      "type": "parameter",
      "units": "kg/kg",
      "shape": [
        "lev"
      ],
      "default": 0.0,
      "description": "Coupling target (assembly input path): the layer water-vapour mixing ratio (WRF QV) from the state document's ODE state, through `input_qv`. Unused by this component's own inline tests."
    },
    "qc_in": {
      "type": "parameter",
      "units": "kg/kg",
      "shape": [
        "lev"
      ],
      "default": 0.0,
      "description": "Coupling target (assembly input path): the layer cloud-water mixing ratio (WRF QC) from the state document's ODE state, through `input_qc`. Unused by this component's own inline tests."
    },
    "qi_in": {
      "type": "parameter",
      "units": "kg/kg",
      "shape": [
        "lev"
      ],
      "default": 0.0,
      "description": "Coupling target (assembly input path): the layer cloud-ice mixing ratio (WRF QI) from the state document's ODE state, through `input_qi`. Unused by this component's own inline tests."
    },
    "qs_in": {
      "type": "parameter",
      "units": "kg/kg",
      "shape": [
        "lev"
      ],
      "default": 0.0,
      "description": "Coupling target (assembly input path): the layer snow mixing ratio (WRF QS) from the state document's ODE state, through `input_qs`. Unused by this component's own inline tests."
    },
    "T": {
      "type": "unknown",
      "units": "K",
      "shape": [
        "lev"
      ],
      "description": "Layer temperature (WRF t_phy), through `input_T`."
    },
    "p": {
      "type": "unknown",
      "units": "Pa",
      "shape": [
        "lev"
      ],
      "description": "Layer pressure (WRF p_phy), through `input_p`."
    },
    "qv": {
      "type": "unknown",
      "units": "kg/kg",
      "shape": [
        "lev"
      ],
      "description": "Water-vapour mixing ratio (WRF QV), through `input_qv`."
    },
    "qc": {
      "type": "unknown",
      "units": "kg/kg",
      "shape": [
        "lev"
      ],
      "description": "Cloud-water mixing ratio (WRF QC), through `input_qc`."
    },
    "qi": {
      "type": "unknown",
      "units": "kg/kg",
      "shape": [
        "lev"
      ],
      "description": "Cloud-ice mixing ratio (WRF QI), through `input_qi`."
    },
    "qs": {
      "type": "unknown",
      "units": "kg/kg",
      "shape": [
        "lev"
      ],
      "description": "Snow mixing ratio (WRF QS), through `input_qs`."
    },
    "es_w": {
      "type": "unknown",
      "units": "Pa",
      "shape": [
        "lev"
      ],
      "description": "Saturation vapour pressure over liquid water, Murray (1966) fit with the over-water coefficients (ESW)."
    },
    "es_i": {
      "type": "unknown",
      "units": "Pa",
      "shape": [
        "lev"
      ],
      "description": "Saturation vapour pressure over ice, the same fit with the over-ice coefficients (ESI)."
    },
    "qvs_w": {
      "type": "unknown",
      "units": "kg/kg",
      "shape": [
        "lev"
      ],
      "description": "Saturation mixing ratio over liquid water (QVSW)."
    },
    "qvs_i": {
      "type": "unknown",
      "units": "kg/kg",
      "shape": [
        "lev"
      ],
      "description": "Saturation mixing ratio over ice (QVSI)."
    },
    "q_cld": {
      "type": "unknown",
      "units": "kg/kg",
      "shape": [
        "lev"
      ],
      "description": "Total cloud condensate QCLD = qi + qc + qs. Rain and graupel are deliberately excluded: 'Rain is not part of cloud, only cloud water + cloud ice + snow'."
    },
    "has_cloud": {
      "type": "unknown",
      "units": "1",
      "shape": [
        "lev"
      ],
      "description": "Indicator of a layer holding at least QCLDMIN of condensate, 1 where q_cld >= QCLDMIN and 0 below it. The Fortran spells this threshold three times (the ice weight, the cloud fraction and the branch flag); it is written once here and reused."
    },
    "ice_weight": {
      "type": "unknown",
      "units": "1",
      "shape": [
        "lev"
      ],
      "description": "Ice fraction of the condensate, WEIGHT = (qi + qs)/QCLD, zero in a cloud-free layer. The denominator is floored at QCLDMIN so the quotient stays finite where the indicator has already zeroed it; in every layer the indicator admits, the floor is inactive and the quotient is the Fortran's exactly."
    },
    "qvs": {
      "type": "unknown",
      "units": "kg/kg",
      "shape": [
        "lev"
      ],
      "description": "Phase-weighted saturation mixing ratio QVS_WEIGHT = (1 - WEIGHT) QVSW + WEIGHT QVSI."
    },
    "rh": {
      "type": "unknown",
      "units": "1",
      "shape": [
        "lev"
      ],
      "description": "Relative humidity against the phase-weighted saturation, RHUM = qv/QVS_WEIGHT."
    },
    "subsat": {
      "type": "unknown",
      "units": "kg/kg",
      "shape": [
        "lev"
      ],
      "description": "Sub-saturation SUBSAT = MAX(1.E-10, RHGRID QVS_WEIGHT - qv), the denominator of the Randall fit."
    },
    "arg": {
      "type": "unknown",
      "units": "1",
      "shape": [
        "lev"
      ],
      "description": "Exponential argument ARG = MAX(-6.9, -ALPHA0 QCLD / SUBSAT^GAMMA)."
    },
    "cldfra_fit": {
      "type": "unknown",
      "units": "1",
      "shape": [
        "lev"
      ],
      "description": "The Randall (1994) fit itself, (MAX(1.E-10, RHUM)/RHGRID)^PEXP (1 - EXP(ARG)), before the 0.01 snap and before the two overriding branches. Evaluated in every layer; it is the branch-3 value."
    },
    "cldfra": {
      "type": "unknown",
      "units": "1",
      "shape": [
        "lev"
      ],
      "description": "Grid-scale cloud fraction CLDFRA, the scheme's output: 0 in a cloud-free layer, 1 in a saturated cloudy layer, and otherwise the Randall fit snapped to 0 below 0.01."
    },
    "branch": {
      "type": "unknown",
      "units": "1",
      "shape": [
        "lev"
      ],
      "description": "Which of the three arms produced the fraction, the Fortran's cldfra1_flag: 1 = no condensate (CLDFRA forced to 0), 2 = at or above RHGRID (forced to 1), 3 = the Randall fit. Reported so a test can pin the branch and not only the number."
    }
  },
  "equations": [
    {
      "lhs": "T",
      "rhs": {
        "op": "input_T",
        "args": []
      }
    },
    {
      "lhs": "p",
      "rhs": {
        "op": "input_p",
        "args": []
      }
    },
    {
      "lhs": "qv",
      "rhs": {
        "op": "input_qv",
        "args": []
      }
    },
    {
      "lhs": "qc",
      "rhs": {
        "op": "input_qc",
        "args": []
      }
    },
    {
      "lhs": "qi",
      "rhs": {
        "op": "input_qi",
        "args": []
      }
    },
    {
      "lhs": "qs",
      "rhs": {
        "op": "input_qs",
        "args": []
      }
    },
    {
      "lhs": "es_w",
      "rhs": {
        "op": "*",
        "args": [
          "par.pa_per_kpa",
          {
            "op": "apply_expression_template",
            "args": [],
            "name": "bolton_saturation_vapor_pressure",
            "bindings": {
              "T": "T",
              "svp1": "par.svp1",
              "svp2": "par.svp2",
              "svp3": "par.svp3",
              "svpt0": "wrf.svpt0"
            }
          }
        ]
      }
    },
    {
      "lhs": "es_i",
      "rhs": {
        "op": "*",
        "args": [
          "par.pa_per_kpa",
          {
            "op": "apply_expression_template",
            "args": [],
            "name": "bolton_saturation_vapor_pressure",
            "bindings": {
              "T": "T",
              "svp1": "par.svp1",
              "svp2": "par.svpi2",
              "svp3": "par.svpi3",
              "svpt0": "wrf.svpt0"
            }
          }
        ]
      }
    },
    {
      "lhs": "qvs_w",
      "rhs": {
        "op": "apply_expression_template",
        "args": [],
        "name": "saturation_mixing_ratio",
        "bindings": {
          "es": "es_w",
          "p": "p",
          "ep_2": "wrf.ep_2"
        }
      }
    },
    {
      "lhs": "qvs_i",
      "rhs": {
        "op": "apply_expression_template",
        "args": [],
        "name": "saturation_mixing_ratio",
        "bindings": {
          "es": "es_i",
          "p": "p",
          "ep_2": "wrf.ep_2"
        }
      }
    },
    {
      "lhs": "q_cld",
      "rhs": {
        "op": "+",
        "args": [
          "qi",
          "qc",
          "qs"
        ]
      }
    },
    {
      "lhs": "has_cloud",
      "rhs": {
        "op": "ifelse",
        "args": [
          {
            "op": "<",
            "args": [
              "q_cld",
              "par.qcldmin"
            ]
          },
          0.0,
          1.0
        ]
      }
    },
    {
      "lhs": "ice_weight",
      "rhs": {
        "op": "*",
        "args": [
          "has_cloud",
          {
            "op": "/",
            "args": [
              {
                "op": "+",
                "args": [
                  "qi",
                  "qs"
                ]
              },
              {
                "op": "max",
                "args": [
                  "q_cld",
                  "par.qcldmin"
                ]
              }
            ]
          }
        ]
      }
    },
    {
      "lhs": "qvs",
      "rhs": {
        "op": "+",
        "args": [
          {
            "op": "*",
            "args": [
              {
                "op": "-",
                "args": [
                  1.0,
                  "ice_weight"
                ]
              },
              "qvs_w"
            ]
          },
          {
            "op": "*",
            "args": [
              "ice_weight",
              "qvs_i"
            ]
          }
        ]
      }
    },
    {
      "lhs": "rh",
      "rhs": {
        "op": "/",
        "args": [
          "qv",
          "qvs"
        ]
      }
    },
    {
      "lhs": "subsat",
      "rhs": {
        "op": "max",
        "args": [
          "par.subsat_floor",
          {
            "op": "-",
            "args": [
              {
                "op": "*",
                "args": [
                  "par.rhgrid",
                  "qvs"
                ]
              },
              "qv"
            ]
          }
        ]
      }
    },
    {
      "lhs": "arg",
      "rhs": {
        "op": "max",
        "args": [
          "par.arg_floor",
          {
            "op": "/",
            "args": [
              {
                "op": "*",
                "args": [
                  -1.0,
                  "par.alpha0",
                  "q_cld"
                ]
              },
              {
                "op": "^",
                "args": [
                  "subsat",
                  "par.gamma"
                ]
              }
            ]
          }
        ]
      }
    },
    {
      "lhs": "cldfra_fit",
      "rhs": {
        "op": "*",
        "args": [
          {
            "op": "^",
            "args": [
              {
                "op": "/",
                "args": [
                  {
                    "op": "max",
                    "args": [
                      "par.rh_floor",
                      "rh"
                    ]
                  },
                  "par.rhgrid"
                ]
              },
              "par.pexp"
            ]
          },
          {
            "op": "-",
            "args": [
              1.0,
              {
                "op": "exp",
                "args": [
                  "arg"
                ]
              }
            ]
          }
        ]
      }
    },
    {
      "lhs": "cldfra",
      "rhs": {
        "op": "*",
        "args": [
          "has_cloud",
          {
            "op": "ifelse",
            "args": [
              {
                "op": ">=",
                "args": [
                  "rh",
                  "par.rhgrid"
                ]
              },
              1.0,
              {
                "op": "ifelse",
                "args": [
                  {
                    "op": "<",
                    "args": [
                      "cldfra_fit",
                      "par.cldfra_floor"
                    ]
                  },
                  0.0,
                  "cldfra_fit"
                ]
              }
            ]
          }
        ]
      }
    },
    {
      "lhs": "branch",
      "rhs": {
        "op": "ifelse",
        "args": [
          {
            "op": "<",
            "args": [
              "has_cloud",
              0.5
            ]
          },
          1.0,
          {
            "op": "ifelse",
            "args": [
              {
                "op": ">=",
                "args": [
                  "rh",
                  "par.rhgrid"
                ]
              },
              2.0,
              3.0
            ]
          }
        ]
      }
    }
  ],
  "tests": [
    {
      "id": "scm_call1_clear_column",
      "description": "Reference: the real64 (-fdefault-real-8) standalone kernels/cldfra_driver of the ctessum-claude/WRF fork, which calls cal_cldfra1 EXTRACTED VERBATIM from phys/module_radiation_driver.F at build time (kernels/Makefile rule cldfra_kernels.F90), so the reference cannot drift from the model code. Inputs are the dumped t3d, p3d, qv3d, qc3d, qi3d and qs3d of the column named below (single-precision values widened), narrowed to a one-column tile by tools/prep_cldfra_dumps.py. The scheme is a pure diagnostic (no time step), so no small-dt replay is involved and no dt sensitivity applies. Why the real64 replay and not the CLDFRA the model itself wrote: the Randall branch divides the condensate by (RHGRID QVS_WEIGHT - QV)^GAMMA, a cancellation of two near-equal numbers, and near saturation that costs WRF's single-precision arithmetic up to 5.1e-5 absolute -- 1.2e-4 relative -- which is far coarser than the rel 1e-5 contract for real32 references. Evaluating this component's own expressions in binary32 reproduces the in-model CLDFRA to 6.0e-8 absolute, so the gap is WRF's arithmetic and not the transcription. Against the real64 kernel the residual is binary64 roundoff (measured over all five regimes: <= 4.3e-16 absolute in cldfra), so the column tolerances are abs 1e-12 -- about 1e-12 of the field's unit range and roughly a thousand times the measured error -- and the spot checks rel 1e-9, the double-precision contract. The `branch` column is the Fortran's cldfra1_flag and is exact. Assertions: Linf_error of the whole cldfra and branch columns against the inline reference, plus spot checks of cldfra at every layer the scheme returns a strictly fractional value for. Regime: the first radiation call of the em_scm_xy reference run (call 1, 19:00 UTC start), column 1 of the tile -- a COMPLETELY CLOUD-FREE column, and the one regime that pins the condensate guard rather than the fit. Not one of the 59 layers holds QCLDMIN of condensate, so every layer takes branch 1 and cldfra is identically zero. That is not a trivial answer: the Murray over-water fit is extrapolated far outside its range at stratospheric temperatures, and layers 55-59 come out at RHUM between 1 and 27.4, so without the QCLD < QCLDMIN test those five layers would be handed back as fully cloudy. Mutation-checked: dropping the guard from `cldfra` leaves the other five regimes green and fails here.",
      "parameter_overrides": {},
      "expression_template_imports": [
        {
          "ref": "./tests/scm_call1_clear_column_inputs.esm"
        }
      ],
      "time_span": {
        "start": 0,
        "end": 1
      },
      "assertions": [
        {
          "variable": "cldfra",
          "time": 0.0,
          "reduce": "Linf_error",
          "expected": 0.0,
          "reference": {
            "op": "faq",
            "output_idx": [
              "k"
            ],
            "args": [],
            "ranges": {
              "k": {
                "from": "lev"
              }
            },
            "expr": {
              "op": "index",
              "args": [
                {
                  "op": "const",
                  "args": [],
                  "value": [
                    0.0,
                    0.0,
                    0.0,
                    0.0,
                    0.0,
                    0.0,
                    0.0,
                    0.0,
                    0.0,
                    0.0,
                    0.0,
                    0.0,
                    0.0,
                    0.0,
                    0.0,
                    0.0,
                    0.0,
                    0.0,
                    0.0,
                    0.0,
                    0.0,
                    0.0,
                    0.0,
                    0.0,
                    0.0,
                    0.0,
                    0.0,
                    0.0,
                    0.0,
                    0.0,
                    0.0,
                    0.0,
                    0.0,
                    0.0,
                    0.0,
                    0.0,
                    0.0,
                    0.0,
                    0.0,
                    0.0,
                    0.0,
                    0.0,
                    0.0,
                    0.0,
                    0.0,
                    0.0,
                    0.0,
                    0.0,
                    0.0,
                    0.0,
                    0.0,
                    0.0,
                    0.0,
                    0.0,
                    0.0,
                    0.0,
                    0.0,
                    0.0,
                    0.0
                  ]
                },
                "k"
              ]
            }
          },
          "tolerance": {
            "abs": 1e-12,
            "rel": 0.0
          }
        },
        {
          "variable": "branch",
          "time": 0.0,
          "reduce": "Linf_error",
          "expected": 0.0,
          "reference": {
            "op": "faq",
            "output_idx": [
              "k"
            ],
            "args": [],
            "ranges": {
              "k": {
                "from": "lev"
              }
            },
            "expr": {
              "op": "index",
              "args": [
                {
                  "op": "const",
                  "args": [],
                  "value": [
                    1.0,
                    1.0,
                    1.0,
                    1.0,
                    1.0,
                    1.0,
                    1.0,
                    1.0,
                    1.0,
                    1.0,
                    1.0,
                    1.0,
                    1.0,
                    1.0,
                    1.0,
                    1.0,
                    1.0,
                    1.0,
                    1.0,
                    1.0,
                    1.0,
                    1.0,
                    1.0,
                    1.0,
                    1.0,
                    1.0,
                    1.0,
                    1.0,
                    1.0,
                    1.0,
                    1.0,
                    1.0,
                    1.0,
                    1.0,
                    1.0,
                    1.0,
                    1.0,
                    1.0,
                    1.0,
                    1.0,
                    1.0,
                    1.0,
                    1.0,
                    1.0,
                    1.0,
                    1.0,
                    1.0,
                    1.0,
                    1.0,
                    1.0,
                    1.0,
                    1.0,
                    1.0,
                    1.0,
                    1.0,
                    1.0,
                    1.0,
                    1.0,
                    1.0
                  ]
                },
                "k"
              ]
            }
          },
          "tolerance": {
            "abs": 1e-12,
            "rel": 0.0
          }
        }
      ]
    },
    {
      "id": "scm_call120_ice_cloud",
      "description": "Reference: the real64 (-fdefault-real-8) standalone kernels/cldfra_driver of the ctessum-claude/WRF fork, which calls cal_cldfra1 EXTRACTED VERBATIM from phys/module_radiation_driver.F at build time (kernels/Makefile rule cldfra_kernels.F90), so the reference cannot drift from the model code. Inputs are the dumped t3d, p3d, qv3d, qc3d, qi3d and qs3d of the column named below (single-precision values widened), narrowed to a one-column tile by tools/prep_cldfra_dumps.py. The scheme is a pure diagnostic (no time step), so no small-dt replay is involved and no dt sensitivity applies. Why the real64 replay and not the CLDFRA the model itself wrote: the Randall branch divides the condensate by (RHGRID QVS_WEIGHT - QV)^GAMMA, a cancellation of two near-equal numbers, and near saturation that costs WRF's single-precision arithmetic up to 5.1e-5 absolute -- 1.2e-4 relative -- which is far coarser than the rel 1e-5 contract for real32 references. Evaluating this component's own expressions in binary32 reproduces the in-model CLDFRA to 6.0e-8 absolute, so the gap is WRF's arithmetic and not the transcription. Against the real64 kernel the residual is binary64 roundoff (measured over all five regimes: <= 4.3e-16 absolute in cldfra), so the column tolerances are abs 1e-12 -- about 1e-12 of the field's unit range and roughly a thousand times the measured error -- and the spot checks rel 1e-9, the double-precision contract. The `branch` column is the Fortran's cldfra1_flag and is exact. Assertions: Linf_error of the whole cldfra and branch columns against the inline reference, plus spot checks of cldfra at every layer the scheme returns a strictly fractional value for. Regime: the em_scm_xy single-column reference run (WRF 4.8 GNU real32 build, 59 layers, mp = 6 lw = 1 sw = 1 sfclay = 1 sf_surface = 1 bl_pbl = 1, icloud = 1), radiation call 120, column 1 of the tile. A thin glaciated cloud at layers 51-57: seven layers hold condensate and every one of them is pure ice (the condensate weight (qi + qs)/QCLD is 1 to within 1e-3 everywhere), so the saturation is judged entirely against the over-ice Murray fit. All three arms fire -- 52 cloud-free layers (branch 1), three layers at or above RHGRID forced to exactly 1 (branch 2) and four on the Randall fit (branch 3), spanning 0.021 to 0.996.",
      "parameter_overrides": {},
      "expression_template_imports": [
        {
          "ref": "./tests/scm_call120_ice_cloud_inputs.esm"
        }
      ],
      "time_span": {
        "start": 0,
        "end": 1
      },
      "assertions": [
        {
          "variable": "cldfra",
          "time": 0.0,
          "reduce": "Linf_error",
          "expected": 0.0,
          "reference": {
            "op": "faq",
            "output_idx": [
              "k"
            ],
            "args": [],
            "ranges": {
              "k": {
                "from": "lev"
              }
            },
            "expr": {
              "op": "index",
              "args": [
                {
                  "op": "const",
                  "args": [],
                  "value": [
                    0.0,
                    0.0,
                    0.0,
                    0.0,
                    0.0,
                    0.0,
                    0.0,
                    0.0,
                    0.0,
                    0.0,
                    0.0,
                    0.0,
                    0.0,
                    0.0,
                    0.0,
                    0.0,
                    0.0,
                    0.0,
                    0.0,
                    0.0,
                    0.0,
                    0.0,
                    0.0,
                    0.0,
                    0.0,
                    0.0,
                    0.0,
                    0.0,
                    0.0,
                    0.0,
                    0.0,
                    0.0,
                    0.0,
                    0.0,
                    0.0,
                    0.0,
                    0.0,
                    0.0,
                    0.0,
                    0.0,
                    0.0,
                    0.0,
                    0.0,
                    0.0,
                    0.0,
                    0.0,
                    0.0,
                    0.0,
                    0.0,
                    0.0,
                    0.021190366713441787,
                    0.5614331164735359,
                    0.9764328684495708,
                    0.9957535527162134,
                    1.0,
                    1.0,
                    1.0,
                    0.0,
                    0.0
                  ]
                },
                "k"
              ]
            }
          },
          "tolerance": {
            "abs": 1e-12,
            "rel": 0.0
          }
        },
        {
          "variable": "branch",
          "time": 0.0,
          "reduce": "Linf_error",
          "expected": 0.0,
          "reference": {
            "op": "faq",
            "output_idx": [
              "k"
            ],
            "args": [],
            "ranges": {
              "k": {
                "from": "lev"
              }
            },
            "expr": {
              "op": "index",
              "args": [
                {
                  "op": "const",
                  "args": [],
                  "value": [
                    1.0,
                    1.0,
                    1.0,
                    1.0,
                    1.0,
                    1.0,
                    1.0,
                    1.0,
                    1.0,
                    1.0,
                    1.0,
                    1.0,
                    1.0,
                    1.0,
                    1.0,
                    1.0,
                    1.0,
                    1.0,
                    1.0,
                    1.0,
                    1.0,
                    1.0,
                    1.0,
                    1.0,
                    1.0,
                    1.0,
                    1.0,
                    1.0,
                    1.0,
                    1.0,
                    1.0,
                    1.0,
                    1.0,
                    1.0,
                    1.0,
                    1.0,
                    1.0,
                    1.0,
                    1.0,
                    1.0,
                    1.0,
                    1.0,
                    1.0,
                    1.0,
                    1.0,
                    1.0,
                    1.0,
                    1.0,
                    1.0,
                    1.0,
                    3.0,
                    3.0,
                    3.0,
                    3.0,
                    2.0,
                    2.0,
                    2.0,
                    1.0,
                    1.0
                  ]
                },
                "k"
              ]
            }
          },
          "tolerance": {
            "abs": 1e-12,
            "rel": 0.0
          }
        },
        {
          "variable": "cldfra",
          "time": 0.0,
          "coords": {
            "lev": 51
          },
          "expected": 0.021190366713441787,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "cldfra",
          "time": 0.0,
          "coords": {
            "lev": 52
          },
          "expected": 0.5614331164735359,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "cldfra",
          "time": 0.0,
          "coords": {
            "lev": 53
          },
          "expected": 0.9764328684495708,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "cldfra",
          "time": 0.0,
          "coords": {
            "lev": 54
          },
          "expected": 0.9957535527162134,
          "tolerance": {
            "rel": 1e-09
          }
        }
      ]
    },
    {
      "id": "scm_call360_ice_cloud",
      "description": "Reference: the real64 (-fdefault-real-8) standalone kernels/cldfra_driver of the ctessum-claude/WRF fork, which calls cal_cldfra1 EXTRACTED VERBATIM from phys/module_radiation_driver.F at build time (kernels/Makefile rule cldfra_kernels.F90), so the reference cannot drift from the model code. Inputs are the dumped t3d, p3d, qv3d, qc3d, qi3d and qs3d of the column named below (single-precision values widened), narrowed to a one-column tile by tools/prep_cldfra_dumps.py. The scheme is a pure diagnostic (no time step), so no small-dt replay is involved and no dt sensitivity applies. Why the real64 replay and not the CLDFRA the model itself wrote: the Randall branch divides the condensate by (RHGRID QVS_WEIGHT - QV)^GAMMA, a cancellation of two near-equal numbers, and near saturation that costs WRF's single-precision arithmetic up to 5.1e-5 absolute -- 1.2e-4 relative -- which is far coarser than the rel 1e-5 contract for real32 references. Evaluating this component's own expressions in binary32 reproduces the in-model CLDFRA to 6.0e-8 absolute, so the gap is WRF's arithmetic and not the transcription. Against the real64 kernel the residual is binary64 roundoff (measured over all five regimes: <= 4.3e-16 absolute in cldfra), so the column tolerances are abs 1e-12 -- about 1e-12 of the field's unit range and roughly a thousand times the measured error -- and the spot checks rel 1e-9, the double-precision contract. The `branch` column is the Fortran's cldfra1_flag and is exact. Assertions: Linf_error of the whole cldfra and branch columns against the inline reference, plus spot checks of cldfra at every layer the scheme returns a strictly fractional value for. Regime: the same em_scm_xy run at radiation call 360, four hours later. The ice cloud has thinned and spread: eight condensate layers (51-57 and 59), three saturated (branch 2), five on the Randall fit (branch 3) with fractions 0.013 to 0.185 -- an order of magnitude smaller than call 120, so the cancellation in the sub-saturation denominator is at its sharpest here. This is the regime that made the in-model real32 CLDFRA unusable as a reference: WRF's own value at layer 54 is 0.1851160 against the real64 kernel's 0.1851372, a 1.1e-4 relative gap.",
      "parameter_overrides": {},
      "expression_template_imports": [
        {
          "ref": "./tests/scm_call360_ice_cloud_inputs.esm"
        }
      ],
      "time_span": {
        "start": 0,
        "end": 1
      },
      "assertions": [
        {
          "variable": "cldfra",
          "time": 0.0,
          "reduce": "Linf_error",
          "expected": 0.0,
          "reference": {
            "op": "faq",
            "output_idx": [
              "k"
            ],
            "args": [],
            "ranges": {
              "k": {
                "from": "lev"
              }
            },
            "expr": {
              "op": "index",
              "args": [
                {
                  "op": "const",
                  "args": [],
                  "value": [
                    0.0,
                    0.0,
                    0.0,
                    0.0,
                    0.0,
                    0.0,
                    0.0,
                    0.0,
                    0.0,
                    0.0,
                    0.0,
                    0.0,
                    0.0,
                    0.0,
                    0.0,
                    0.0,
                    0.0,
                    0.0,
                    0.0,
                    0.0,
                    0.0,
                    0.0,
                    0.0,
                    0.0,
                    0.0,
                    0.0,
                    0.0,
                    0.0,
                    0.0,
                    0.0,
                    0.0,
                    0.0,
                    0.0,
                    0.0,
                    0.0,
                    0.0,
                    0.0,
                    0.0,
                    0.0,
                    0.0,
                    0.0,
                    0.0,
                    0.0,
                    0.0,
                    0.0,
                    0.0,
                    0.0,
                    0.0,
                    0.0,
                    0.0,
                    0.012960408215861549,
                    0.11689247711032254,
                    0.17699737616995914,
                    0.18513720847644013,
                    0.045361853513636814,
                    1.0,
                    1.0,
                    0.0,
                    1.0
                  ]
                },
                "k"
              ]
            }
          },
          "tolerance": {
            "abs": 1e-12,
            "rel": 0.0
          }
        },
        {
          "variable": "branch",
          "time": 0.0,
          "reduce": "Linf_error",
          "expected": 0.0,
          "reference": {
            "op": "faq",
            "output_idx": [
              "k"
            ],
            "args": [],
            "ranges": {
              "k": {
                "from": "lev"
              }
            },
            "expr": {
              "op": "index",
              "args": [
                {
                  "op": "const",
                  "args": [],
                  "value": [
                    1.0,
                    1.0,
                    1.0,
                    1.0,
                    1.0,
                    1.0,
                    1.0,
                    1.0,
                    1.0,
                    1.0,
                    1.0,
                    1.0,
                    1.0,
                    1.0,
                    1.0,
                    1.0,
                    1.0,
                    1.0,
                    1.0,
                    1.0,
                    1.0,
                    1.0,
                    1.0,
                    1.0,
                    1.0,
                    1.0,
                    1.0,
                    1.0,
                    1.0,
                    1.0,
                    1.0,
                    1.0,
                    1.0,
                    1.0,
                    1.0,
                    1.0,
                    1.0,
                    1.0,
                    1.0,
                    1.0,
                    1.0,
                    1.0,
                    1.0,
                    1.0,
                    1.0,
                    1.0,
                    1.0,
                    1.0,
                    1.0,
                    1.0,
                    3.0,
                    3.0,
                    3.0,
                    3.0,
                    3.0,
                    2.0,
                    2.0,
                    1.0,
                    2.0
                  ]
                },
                "k"
              ]
            }
          },
          "tolerance": {
            "abs": 1e-12,
            "rel": 0.0
          }
        },
        {
          "variable": "cldfra",
          "time": 0.0,
          "coords": {
            "lev": 51
          },
          "expected": 0.012960408215861549,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "cldfra",
          "time": 0.0,
          "coords": {
            "lev": 52
          },
          "expected": 0.11689247711032254,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "cldfra",
          "time": 0.0,
          "coords": {
            "lev": 53
          },
          "expected": 0.17699737616995914,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "cldfra",
          "time": 0.0,
          "coords": {
            "lev": 54
          },
          "expected": 0.18513720847644013,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "cldfra",
          "time": 0.0,
          "coords": {
            "lev": 55
          },
          "expected": 0.045361853513636814,
          "tolerance": {
            "rel": 1e-09
          }
        }
      ]
    },
    {
      "id": "supercell_warm_liquid",
      "description": "Reference: the real64 (-fdefault-real-8) standalone kernels/cldfra_driver of the ctessum-claude/WRF fork, which calls cal_cldfra1 EXTRACTED VERBATIM from phys/module_radiation_driver.F at build time (kernels/Makefile rule cldfra_kernels.F90), so the reference cannot drift from the model code. Inputs are the dumped t3d, p3d, qv3d, qc3d, qi3d and qs3d of the column named below (single-precision values widened), narrowed to a one-column tile by tools/prep_cldfra_dumps.py. The scheme is a pure diagnostic (no time step), so no small-dt replay is involved and no dt sensitivity applies. Why the real64 replay and not the CLDFRA the model itself wrote: the Randall branch divides the condensate by (RHGRID QVS_WEIGHT - QV)^GAMMA, a cancellation of two near-equal numbers, and near saturation that costs WRF's single-precision arithmetic up to 5.1e-5 absolute -- 1.2e-4 relative -- which is far coarser than the rel 1e-5 contract for real32 references. Evaluating this component's own expressions in binary32 reproduces the in-model CLDFRA to 6.0e-8 absolute, so the gap is WRF's arithmetic and not the transcription. Against the real64 kernel the residual is binary64 roundoff (measured over all five regimes: <= 4.3e-16 absolute in cldfra), so the column tolerances are abs 1e-12 -- about 1e-12 of the field's unit range and roughly a thousand times the measured error -- and the spot checks rel 1e-9, the double-precision contract. The `branch` column is the Fortran's cldfra1_flag and is exact. Assertions: Linf_error of the whole cldfra and branch columns against the inline reference, plus spot checks of cldfra at every layer the scheme returns a strictly fractional value for. Regime: the idealized supercell em_quarter_ss reference case (mp = 6, ra_lw = 1, 59 layers, ztop = 12800 m; the same case that supplies the RRTM warm and mixed-phase cloud-optics tests), longwave call 61, tile column 20. A PURE LIQUID cloud: six layers of cloud water at 10-15 with no ice and no snow at all, so the condensate weight is exactly 0 and the phase blend reduces to the over-WATER Murray fit -- the only case here that exercises that arm on its own. No layer reaches RHGRID, so all six are on the Randall fit (branch 3), 0.423 to 0.988, and the other 53 layers are cloud-free.",
      "parameter_overrides": {},
      "expression_template_imports": [
        {
          "ref": "./tests/supercell_warm_liquid_inputs.esm"
        }
      ],
      "time_span": {
        "start": 0,
        "end": 1
      },
      "assertions": [
        {
          "variable": "cldfra",
          "time": 0.0,
          "reduce": "Linf_error",
          "expected": 0.0,
          "reference": {
            "op": "faq",
            "output_idx": [
              "k"
            ],
            "args": [],
            "ranges": {
              "k": {
                "from": "lev"
              }
            },
            "expr": {
              "op": "index",
              "args": [
                {
                  "op": "const",
                  "args": [],
                  "value": [
                    0.0,
                    0.0,
                    0.0,
                    0.0,
                    0.0,
                    0.0,
                    0.0,
                    0.0,
                    0.0,
                    0.4233829708735316,
                    0.8554281179998233,
                    0.9726412944914508,
                    0.9883816411448191,
                    0.9749973429179064,
                    0.7476490222301596,
                    0.0,
                    0.0,
                    0.0,
                    0.0,
                    0.0,
                    0.0,
                    0.0,
                    0.0,
                    0.0,
                    0.0,
                    0.0,
                    0.0,
                    0.0,
                    0.0,
                    0.0,
                    0.0,
                    0.0,
                    0.0,
                    0.0,
                    0.0,
                    0.0,
                    0.0,
                    0.0,
                    0.0,
                    0.0,
                    0.0,
                    0.0,
                    0.0,
                    0.0,
                    0.0,
                    0.0,
                    0.0,
                    0.0,
                    0.0,
                    0.0,
                    0.0,
                    0.0,
                    0.0,
                    0.0,
                    0.0,
                    0.0,
                    0.0,
                    0.0,
                    0.0
                  ]
                },
                "k"
              ]
            }
          },
          "tolerance": {
            "abs": 1e-12,
            "rel": 0.0
          }
        },
        {
          "variable": "branch",
          "time": 0.0,
          "reduce": "Linf_error",
          "expected": 0.0,
          "reference": {
            "op": "faq",
            "output_idx": [
              "k"
            ],
            "args": [],
            "ranges": {
              "k": {
                "from": "lev"
              }
            },
            "expr": {
              "op": "index",
              "args": [
                {
                  "op": "const",
                  "args": [],
                  "value": [
                    1.0,
                    1.0,
                    1.0,
                    1.0,
                    1.0,
                    1.0,
                    1.0,
                    1.0,
                    1.0,
                    3.0,
                    3.0,
                    3.0,
                    3.0,
                    3.0,
                    3.0,
                    1.0,
                    1.0,
                    1.0,
                    1.0,
                    1.0,
                    1.0,
                    1.0,
                    1.0,
                    1.0,
                    1.0,
                    1.0,
                    1.0,
                    1.0,
                    1.0,
                    1.0,
                    1.0,
                    1.0,
                    1.0,
                    1.0,
                    1.0,
                    1.0,
                    1.0,
                    1.0,
                    1.0,
                    1.0,
                    1.0,
                    1.0,
                    1.0,
                    1.0,
                    1.0,
                    1.0,
                    1.0,
                    1.0,
                    1.0,
                    1.0,
                    1.0,
                    1.0,
                    1.0,
                    1.0,
                    1.0,
                    1.0,
                    1.0,
                    1.0,
                    1.0
                  ]
                },
                "k"
              ]
            }
          },
          "tolerance": {
            "abs": 1e-12,
            "rel": 0.0
          }
        },
        {
          "variable": "cldfra",
          "time": 0.0,
          "coords": {
            "lev": 10
          },
          "expected": 0.4233829708735316,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "cldfra",
          "time": 0.0,
          "coords": {
            "lev": 11
          },
          "expected": 0.8554281179998233,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "cldfra",
          "time": 0.0,
          "coords": {
            "lev": 12
          },
          "expected": 0.9726412944914508,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "cldfra",
          "time": 0.0,
          "coords": {
            "lev": 13
          },
          "expected": 0.9883816411448191,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "cldfra",
          "time": 0.0,
          "coords": {
            "lev": 14
          },
          "expected": 0.9749973429179064,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "cldfra",
          "time": 0.0,
          "coords": {
            "lev": 15
          },
          "expected": 0.7476490222301596,
          "tolerance": {
            "rel": 1e-09
          }
        }
      ]
    },
    {
      "id": "supercell_mixed_phase",
      "description": "Reference: the real64 (-fdefault-real-8) standalone kernels/cldfra_driver of the ctessum-claude/WRF fork, which calls cal_cldfra1 EXTRACTED VERBATIM from phys/module_radiation_driver.F at build time (kernels/Makefile rule cldfra_kernels.F90), so the reference cannot drift from the model code. Inputs are the dumped t3d, p3d, qv3d, qc3d, qi3d and qs3d of the column named below (single-precision values widened), narrowed to a one-column tile by tools/prep_cldfra_dumps.py. The scheme is a pure diagnostic (no time step), so no small-dt replay is involved and no dt sensitivity applies. Why the real64 replay and not the CLDFRA the model itself wrote: the Randall branch divides the condensate by (RHGRID QVS_WEIGHT - QV)^GAMMA, a cancellation of two near-equal numbers, and near saturation that costs WRF's single-precision arithmetic up to 5.1e-5 absolute -- 1.2e-4 relative -- which is far coarser than the rel 1e-5 contract for real32 references. Evaluating this component's own expressions in binary32 reproduces the in-model CLDFRA to 6.0e-8 absolute, so the gap is WRF's arithmetic and not the transcription. Against the real64 kernel the residual is binary64 roundoff (measured over all five regimes: <= 4.3e-16 absolute in cldfra), so the column tolerances are abs 1e-12 -- about 1e-12 of the field's unit range and roughly a thousand times the measured error -- and the spot checks rel 1e-9, the double-precision contract. The `branch` column is the Fortran's cldfra1_flag and is exact. Assertions: Linf_error of the whole cldfra and branch columns against the inline reference, plus spot checks of cldfra at every layer the scheme returns a strictly fractional value for. Regime: the same supercell at longwave call 268, tile column 20 -- the mature storm column, and the widest phase sweep available. 47 layers hold condensate and the ice weight runs continuously from 0 (pure liquid, layers 1-9) through the mixed-phase interior (0.0011 to 0.37) to 1 (the glaciated anvil above), so the (1 - w) QVSW + w QVSI blend is exercised across its whole range rather than at its two endpoints. All three branches fire (12 cloud-free, 12 saturated, 35 on the fit) and both guards of the Randall arm are active in a way no other regime here reaches: the exponential argument hits its -6.9 floor in 8 layers, and 10 layers produce a fit below 0.01 and are snapped to zero. 25 layers end strictly fractional.",
      "parameter_overrides": {},
      "expression_template_imports": [
        {
          "ref": "./tests/supercell_mixed_phase_inputs.esm"
        }
      ],
      "time_span": {
        "start": 0,
        "end": 1
      },
      "assertions": [
        {
          "variable": "cldfra",
          "time": 0.0,
          "reduce": "Linf_error",
          "expected": 0.0,
          "reference": {
            "op": "faq",
            "output_idx": [
              "k"
            ],
            "args": [],
            "ranges": {
              "k": {
                "from": "lev"
              }
            },
            "expr": {
              "op": "index",
              "args": [
                {
                  "op": "const",
                  "args": [],
                  "value": [
                    0.0,
                    0.0,
                    0.0,
                    0.0,
                    0.0,
                    0.0,
                    0.0,
                    0.0,
                    0.0,
                    0.0,
                    0.16889500245857517,
                    0.5422102059550059,
                    0.7425617920385581,
                    0.861384976928407,
                    0.9329480751258108,
                    0.965225732565437,
                    0.9818318820690424,
                    0.99107586099125,
                    0.9957968090360563,
                    0.9973172423210348,
                    0.9972868726831552,
                    0.997299602558455,
                    0.9973653442004204,
                    0.9974813698498667,
                    0.9976676264886069,
                    0.9978481555834664,
                    0.9984066278319322,
                    1.0,
                    1.0,
                    1.0,
                    1.0,
                    1.0,
                    1.0,
                    1.0,
                    1.0,
                    1.0,
                    1.0,
                    1.0,
                    1.0,
                    0.5186591111871108,
                    0.06009687555905586,
                    0.0,
                    0.0,
                    0.0,
                    0.0,
                    0.0,
                    0.0,
                    0.0,
                    0.0,
                    0.0,
                    0.1113346218027536,
                    0.19959465820245134,
                    0.09191298117360905,
                    0.0584780642351787,
                    0.030769620992115804,
                    0.010691416884162306,
                    0.0,
                    0.0,
                    0.0
                  ]
                },
                "k"
              ]
            }
          },
          "tolerance": {
            "abs": 1e-12,
            "rel": 0.0
          }
        },
        {
          "variable": "branch",
          "time": 0.0,
          "reduce": "Linf_error",
          "expected": 0.0,
          "reference": {
            "op": "faq",
            "output_idx": [
              "k"
            ],
            "args": [],
            "ranges": {
              "k": {
                "from": "lev"
              }
            },
            "expr": {
              "op": "index",
              "args": [
                {
                  "op": "const",
                  "args": [],
                  "value": [
                    1.0,
                    1.0,
                    1.0,
                    1.0,
                    1.0,
                    1.0,
                    1.0,
                    1.0,
                    1.0,
                    1.0,
                    3.0,
                    3.0,
                    3.0,
                    3.0,
                    3.0,
                    3.0,
                    3.0,
                    3.0,
                    3.0,
                    3.0,
                    3.0,
                    3.0,
                    3.0,
                    3.0,
                    3.0,
                    3.0,
                    3.0,
                    2.0,
                    2.0,
                    2.0,
                    2.0,
                    2.0,
                    2.0,
                    2.0,
                    2.0,
                    2.0,
                    2.0,
                    2.0,
                    2.0,
                    3.0,
                    3.0,
                    3.0,
                    3.0,
                    3.0,
                    3.0,
                    3.0,
                    3.0,
                    3.0,
                    3.0,
                    3.0,
                    3.0,
                    3.0,
                    3.0,
                    3.0,
                    3.0,
                    3.0,
                    3.0,
                    1.0,
                    1.0
                  ]
                },
                "k"
              ]
            }
          },
          "tolerance": {
            "abs": 1e-12,
            "rel": 0.0
          }
        },
        {
          "variable": "cldfra",
          "time": 0.0,
          "coords": {
            "lev": 11
          },
          "expected": 0.16889500245857517,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "cldfra",
          "time": 0.0,
          "coords": {
            "lev": 12
          },
          "expected": 0.5422102059550059,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "cldfra",
          "time": 0.0,
          "coords": {
            "lev": 13
          },
          "expected": 0.7425617920385581,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "cldfra",
          "time": 0.0,
          "coords": {
            "lev": 14
          },
          "expected": 0.861384976928407,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "cldfra",
          "time": 0.0,
          "coords": {
            "lev": 15
          },
          "expected": 0.9329480751258108,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "cldfra",
          "time": 0.0,
          "coords": {
            "lev": 16
          },
          "expected": 0.965225732565437,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "cldfra",
          "time": 0.0,
          "coords": {
            "lev": 17
          },
          "expected": 0.9818318820690424,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "cldfra",
          "time": 0.0,
          "coords": {
            "lev": 18
          },
          "expected": 0.99107586099125,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "cldfra",
          "time": 0.0,
          "coords": {
            "lev": 19
          },
          "expected": 0.9957968090360563,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "cldfra",
          "time": 0.0,
          "coords": {
            "lev": 20
          },
          "expected": 0.9973172423210348,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "cldfra",
          "time": 0.0,
          "coords": {
            "lev": 40
          },
          "expected": 0.5186591111871108,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "cldfra",
          "time": 0.0,
          "coords": {
            "lev": 41
          },
          "expected": 0.06009687555905586,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "cldfra",
          "time": 0.0,
          "coords": {
            "lev": 51
          },
          "expected": 0.1113346218027536,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "cldfra",
          "time": 0.0,
          "coords": {
            "lev": 52
          },
          "expected": 0.19959465820245134,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "cldfra",
          "time": 0.0,
          "coords": {
            "lev": 53
          },
          "expected": 0.09191298117360905,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "cldfra",
          "time": 0.0,
          "coords": {
            "lev": 54
          },
          "expected": 0.0584780642351787,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "cldfra",
          "time": 0.0,
          "coords": {
            "lev": 55
          },
          "expected": 0.030769620992115804,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "cldfra",
          "time": 0.0,
          "coords": {
            "lev": 56
          },
          "expected": 0.010691416884162306,
          "tolerance": {
            "rel": 1e-09
          }
        }
      ]
    },
    {
      "id": "supercell_ice_anvil",
      "description": "Reference: the real64 (-fdefault-real-8) standalone kernels/cldfra_driver of the ctessum-claude/WRF fork, which calls cal_cldfra1 EXTRACTED VERBATIM from phys/module_radiation_driver.F at build time (kernels/Makefile rule cldfra_kernels.F90), so the reference cannot drift from the model code. Inputs are the dumped t3d, p3d, qv3d, qc3d, qi3d and qs3d of the column named below (single-precision values widened), narrowed to a one-column tile by tools/prep_cldfra_dumps.py. The scheme is a pure diagnostic (no time step), so no small-dt replay is involved and no dt sensitivity applies. Why the real64 replay and not the CLDFRA the model itself wrote: the Randall branch divides the condensate by (RHGRID QVS_WEIGHT - QV)^GAMMA, a cancellation of two near-equal numbers, and near saturation that costs WRF's single-precision arithmetic up to 5.1e-5 absolute -- 1.2e-4 relative -- which is far coarser than the rel 1e-5 contract for real32 references. Evaluating this component's own expressions in binary32 reproduces the in-model CLDFRA to 6.0e-8 absolute, so the gap is WRF's arithmetic and not the transcription. Against the real64 kernel the residual is binary64 roundoff (measured over all five regimes: <= 4.3e-16 absolute in cldfra), so the column tolerances are abs 1e-12 -- about 1e-12 of the field's unit range and roughly a thousand times the measured error -- and the spot checks rel 1e-9, the double-precision contract. The `branch` column is the Fortran's cldfra1_flag and is exact. Assertions: Linf_error of the whole cldfra and branch columns against the inline reference, plus spot checks of cldfra at every layer the scheme returns a strictly fractional value for. Regime: the supercell's ice-and-snow anvil, longwave call 396, tile column 40. 38 condensate layers, every one pure ice, and the deepest fractional structure of the five: 30 strictly fractional layers running from 0.012 at the anvil edge to 0.998 in its core, five saturated layers and one layer that hits the -6.9 exponential floor, with three more snapped to zero by the 0.01 threshold. Together with supercell_warm_liquid this pins the two endpoints of the phase blend against a column that is not the SCM's.",
      "parameter_overrides": {},
      "expression_template_imports": [
        {
          "ref": "./tests/supercell_ice_anvil_inputs.esm"
        }
      ],
      "time_span": {
        "start": 0,
        "end": 1
      },
      "assertions": [
        {
          "variable": "cldfra",
          "time": 0.0,
          "reduce": "Linf_error",
          "expected": 0.0,
          "reference": {
            "op": "faq",
            "output_idx": [
              "k"
            ],
            "args": [],
            "ranges": {
              "k": {
                "from": "lev"
              }
            },
            "expr": {
              "op": "index",
              "args": [
                {
                  "op": "const",
                  "args": [],
                  "value": [
                    0.0,
                    0.0,
                    0.0,
                    0.0,
                    0.0,
                    0.0,
                    0.0,
                    0.0,
                    0.0,
                    0.0,
                    0.0,
                    0.0,
                    0.0,
                    0.0,
                    0.0,
                    0.0,
                    0.011588979733451011,
                    0.043468105713251395,
                    0.07512078925042692,
                    0.08595183053656705,
                    0.0828320671357669,
                    0.08590946403806397,
                    0.0969568720490712,
                    0.08482129959329207,
                    0.030340678165119563,
                    0.0,
                    0.0675991351136989,
                    0.3011500511658455,
                    0.448477865040188,
                    0.5133557585767013,
                    0.5637245531197136,
                    0.6175470050974778,
                    0.6857329979483836,
                    0.7868963735079537,
                    0.9002085677245053,
                    0.9982921959559804,
                    1.0,
                    1.0,
                    1.0,
                    1.0,
                    1.0,
                    0.9968709297924274,
                    0.99185523931125,
                    0.9824924033543033,
                    0.9021965292141958,
                    0.7480220794558997,
                    0.6044734226654151,
                    0.45804764743197995,
                    0.3098247063053506,
                    0.1803606145552516,
                    0.0851310657029454,
                    0.029286770468905696,
                    0.0,
                    0.0,
                    0.0,
                    0.0,
                    0.0,
                    0.0,
                    0.0
                  ]
                },
                "k"
              ]
            }
          },
          "tolerance": {
            "abs": 1e-12,
            "rel": 0.0
          }
        },
        {
          "variable": "branch",
          "time": 0.0,
          "reduce": "Linf_error",
          "expected": 0.0,
          "reference": {
            "op": "faq",
            "output_idx": [
              "k"
            ],
            "args": [],
            "ranges": {
              "k": {
                "from": "lev"
              }
            },
            "expr": {
              "op": "index",
              "args": [
                {
                  "op": "const",
                  "args": [],
                  "value": [
                    1.0,
                    1.0,
                    1.0,
                    1.0,
                    1.0,
                    1.0,
                    1.0,
                    1.0,
                    1.0,
                    1.0,
                    1.0,
                    1.0,
                    1.0,
                    1.0,
                    1.0,
                    1.0,
                    3.0,
                    3.0,
                    3.0,
                    3.0,
                    3.0,
                    3.0,
                    3.0,
                    3.0,
                    3.0,
                    3.0,
                    3.0,
                    3.0,
                    3.0,
                    3.0,
                    3.0,
                    3.0,
                    3.0,
                    3.0,
                    3.0,
                    3.0,
                    2.0,
                    2.0,
                    2.0,
                    2.0,
                    2.0,
                    3.0,
                    3.0,
                    3.0,
                    3.0,
                    3.0,
                    3.0,
                    3.0,
                    3.0,
                    3.0,
                    3.0,
                    3.0,
                    3.0,
                    3.0,
                    1.0,
                    1.0,
                    1.0,
                    1.0,
                    1.0
                  ]
                },
                "k"
              ]
            }
          },
          "tolerance": {
            "abs": 1e-12,
            "rel": 0.0
          }
        },
        {
          "variable": "cldfra",
          "time": 0.0,
          "coords": {
            "lev": 17
          },
          "expected": 0.011588979733451011,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "cldfra",
          "time": 0.0,
          "coords": {
            "lev": 18
          },
          "expected": 0.043468105713251395,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "cldfra",
          "time": 0.0,
          "coords": {
            "lev": 19
          },
          "expected": 0.07512078925042692,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "cldfra",
          "time": 0.0,
          "coords": {
            "lev": 20
          },
          "expected": 0.08595183053656705,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "cldfra",
          "time": 0.0,
          "coords": {
            "lev": 25
          },
          "expected": 0.030340678165119563,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "cldfra",
          "time": 0.0,
          "coords": {
            "lev": 28
          },
          "expected": 0.3011500511658455,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "cldfra",
          "time": 0.0,
          "coords": {
            "lev": 29
          },
          "expected": 0.448477865040188,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "cldfra",
          "time": 0.0,
          "coords": {
            "lev": 30
          },
          "expected": 0.5133557585767013,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "cldfra",
          "time": 0.0,
          "coords": {
            "lev": 33
          },
          "expected": 0.6857329979483836,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "cldfra",
          "time": 0.0,
          "coords": {
            "lev": 35
          },
          "expected": 0.9002085677245053,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "cldfra",
          "time": 0.0,
          "coords": {
            "lev": 36
          },
          "expected": 0.9982921959559804,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "cldfra",
          "time": 0.0,
          "coords": {
            "lev": 42
          },
          "expected": 0.9968709297924274,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "cldfra",
          "time": 0.0,
          "coords": {
            "lev": 45
          },
          "expected": 0.9021965292141958,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "cldfra",
          "time": 0.0,
          "coords": {
            "lev": 47
          },
          "expected": 0.6044734226654151,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "cldfra",
          "time": 0.0,
          "coords": {
            "lev": 49
          },
          "expected": 0.3098247063053506,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "cldfra",
          "time": 0.0,
          "coords": {
            "lev": 50
          },
          "expected": 0.1803606145552516,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "cldfra",
          "time": 0.0,
          "coords": {
            "lev": 51
          },
          "expected": 0.0851310657029454,
          "tolerance": {
            "rel": 1e-09
          }
        },
        {
          "variable": "cldfra",
          "time": 0.0,
          "coords": {
            "lev": 52
          },
          "expected": 0.029286770468905696,
          "tolerance": {
            "rel": 1e-09
          }
        }
      ]
    }
  ]
}