Skip to content

eCLM-ParFlow: hk_l is never computed resulting in suppressed transpiration #128

Closed

Description

@s-poll

Summary

In ParFlow-coupled runs (soilwater_movement_method = 4, coupled_parflow), the soil hydraulic conductivity hk_l_col is never computed. It keeps its cold-start value of zero for the entire simulation.

use_hydrstress = .true. derives the soil-to-root conductance from hk_l. With hk_l = 0 that conductance falls to its numerical floor, the plant-hydraulic solve can extract no water from the soil, and canopy transpiration is nearly zero across the whole domain.

This bug seems to be the origin of the observed reduced latent heat flux, seen in #123.

Currently, I test the proposed solution code changes in the eCLM-ParFlow coupled code, which looks promising. In case of successful results, I will open a PR.

Mechanism

done by AI

  1. hk_l_col is assigned in exactly two places:

    • SoilWaterMovementMod.F90:771 — inside soilwater_zengdecker2009
    • SoilWaterMovementMod.F90:1624 — inside compute_hydraulic_properties, which is called only from soilwater_moisture_form (SoilWaterMovementMod.F90:1220)
  2. soilwater_parflow (SoilWaterMovementMod.F90:1447-1515) replaces the soil-water solver entirely and assigns neither. On a cold start hk_l therefore keeps the value set in SoilStateType.F90:345:

    this%hk_l_col(bounds%begc:bounds%endc,1:nlevgrnd) = 0._r8
    
  3. In PhotosynthesisHydraulicStress (PhotosynthesisMod.F90:2418), the soil conductance is built from hk_l:

    soil_conductance = min(hksat(c,j),hk_l(c,j))/(1.e3*r_soil)   ! :2841
    ...
    soil_conductance = max(soil_conductance, 1.e-16_r8)          ! :2849
    ...
    k_soil_root(p,j) =  1._r8/rs_resis                           ! :2861
    

    With hk_l = 0 the min is zero, the clamp pins soil_conductance at 1e-16, and k_soil_root follows.

  4. The plant-hydraulic solve (PhotosynthesisMod.F90:4740-4771) can then draw essentially no water from the soil column, so transpiration goes to zero regardless of how wet the soil actually is.

Note that hk_l is on the restart file (SoilStateType.F90:389). A run restarted from a non-ParFlow simulation will inherit non-zero values and can mask the problem.

Proposed solution

Calculate hk_l_col from soil water retention curve as part of soilwater_parflow in src/clm5/biogeophys/SoilWaterMovementMod.F90 based on ParFows h2osoi_liq.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Metadata

Metadata

Assignees

No one assigned

    Labels

    coupled-parflowIssues related to eCLM-ParFlow coupling

    Type

    No type

    Projects

    No projects

      Milestone

      No milestone

      Relationships

      None yet

      Development

      No branches or pull requests

      Issue actions