NOAH-MP + RRTMGP

[AMLattanzi]

ERF RRTMGP-LSM Coupling Findings

This note summarizes issues found in the coupling between ERF's RRTMGP radiation interface and the land-surface-model exchange, compared against WRF.

Main Findings

1. Non-positive LSM inputs are treated as invalid in ERF

In ERF, the radiation wrapper only accepts an LSM-provided surface field if the value is strictly positive:

• t_sfc
• sfc_emis
• sfc_alb_dir_vis
• sfc_alb_dir_nir
• sfc_alb_dif_vis
• sfc_alb_dif_nir

Code:

• Source/PhysicsInterfaces/Radiation/ERF_Radiation.cpp

Relevant logic:

bool valid_lsm_data{false};
if (lsm_in_arr) { valid_lsm_data = (lsm_in_arr(i,j,k) > 0.); }

This is problematic for albedo in particular, because a valid 0.0 value is silently replaced by the ERF default 0.06.

WRF does not apply this filter. It passes the surface quantities directly through the surface driver into the LSM.

Relevant WRF call:

• phys/module_surface_driver.F

2. ERF only uses LSM surface properties over land

ERF only uses LSM-provided t_sfc, emissivity, and albedos when lmask says the point is land. Otherwise it falls back to hard-coded defaults:

• t_sfc = rad_t_sfc
• sfc_emis = 0.98
• albedos = 0.06

Code:

• Source/PhysicsInterfaces/Radiation/ERF_Radiation.cpp

This differs from WRF, where surface optical and thermal properties are coupled across all surface types, including water and fractional sea ice, with explicit blending logic in the surface driver.

Relevant WRF logic:

• phys/module_surface_driver.F

3. ERF Noah-MP receives only total downward shortwave

ERF Noah-MP reads:

• SWDOWN
• GLW
• COSZEN

from the shared LSM state:

• Source/LandSurfaceModel/Noah-MP/ERF_NOAHMP.cpp

WRF Noah-MP receives:

• SWDOWN
• SWDDIR
• SWDDIF
• GLW
• COSZEN

through the WRF surface driver:

• phys/module_surface_driver.F

ERF radiation does compute directional and diffuse surface shortwave components and writes them into m_lsm_fluxes, but those fields are not the generic lsm_output_ptrs used by ERF Noah-MP. Relative to WRF, this is a reduced coupling.

4. Longwave downward flux semantics look consistent

ERF exports surface lw_flux_dn directly to the LSM coupling field. WRF Noah/Noah-MP expects GLW as downward atmospheric longwave and multiplies by surface emissivity internally.

ERF side:

• Source/PhysicsInterfaces/Radiation/ERF_Radiation.cpp

WRF side:

• phys/module_sf_noahdrv.F

So this part of the coupling appears consistent.

Bottom Line

The main ERF-vs-WRF coupling concerns are:

1. ERF rejects non-positive LSM input values and silently substitutes defaults.
2. ERF effectively limits LSM-provided surface optical properties to land points only.
3. ERF Noah-MP does not receive the direct/diffuse shortwave decomposition that WRF Noah-MP receives.

The LW downward-flux handoff itself appears consistent with WRF.

[AMLattanzi]

The filtering of non-negative data was due to NOAH-MP returning -9999 when not over land. That logic will be modified to be >= 0.

However, the exporting of direct and diffusive fluxes to NOAH has not been done yet. We do export those to the SLM model. We should expose those through the NOAH interface and correct.

[AMLattanzi]

  - TSK over water may be populated, but not from an active open-water Noah-MP solve.
  - EMISS and ALBSFC* over water are not reliable from Noah-MP in ERF.
  - ERF’s radiation-side choice to only trust LSM surface properties over land is consistent with that
    behavior.

  This is also different from WRF, where the surface driver fills/blends water and sea-ice surface
  properties outside the Noah-MP land solve in module_surface_driver.F:1724 and
  module_surface_driver.F:2941.