eCLM-ParFlow: Multiphase water dynamics

[s-poll]

Description

The current implementation of the coupled eCLM–ParFlow system appears to have limitations in representing multiphase water dynamics, particularly under conditions where both liquid water and ice coexist in the subsurface. I try to summarize the current state and the model behavior and interaction in the following issue to better understand.

DISCLAIMER: The following text is my current understanding and its not a documentation. Please correct in case something is wrong!

The soil hydrology of eCLM is explained at the CTSM user guide. The ParFlow model equations are explained in the parflow documentation.

General Sketch of eCLM-ParFlow with vertical discretization

eCLM hydrology details:

volumetric soil moisture calculation:

eCLM/src/clm5/biogeophys/SoilHydrologyMod.F90

Lines 814 to 815 in 98054f5

	h2osoi_vol(c,k) = h2osoi_liq(c,k)/(dz(c,k)*denh2o) &
	+ h2osoi_ice(c,k)/(dz(c,k)*denice)

update of eCLM hydrology based on ParFlow input.

eCLM/src/clm5/biogeophys/SoilWaterMovementMod.F90

Lines 1502 to 1507 in 98054f5

	do j = 1, nlevgrnd
	h2osoi_liq(c,j) = pfl_h2osoi_liq(c,j)
	if (pfl_psi(c,j) <= 0) then
	smp_l(c,j) = pfl_psi(c,j)
	end if
	end do

Subsurface ice

Standard ParFlow does not explicitly simulate ice as a separate phase and does not include thermodynamics checks by default, whereas eCLM accounts for ice and liquid water phases. To my understanding this leads to two problems:

1. Lack of feedback in ParFlow: ParFlow does not respond to freezing or melting processes in the subsurface. Changes in phase therefore do not influence hydraulic properties or flow dynamics.
2. Inconsistency in water accounting in eCLM: The water state passed between models can become inconsistent. For example, if ice content increases, the corresponding reduction in liquid water may not be properly reflected in the exchanged variables, leading to non-physical or biased total water content.

There is an experimental branch for parflow/dev-soil-ice and eCLM/dev-soil-ice using the ice impedance 2.7.48. This branches still are subject to discussion as pointed out in the eCLM-ParFlow Coupling Notes. There is also pointed out that no thermodynamic checks in ParFlow to guarantee the energy balance of the eCLM-ParFlow coupled system.

The branches uses ice-impendence calculation in eCLM and coupled it to ParFlow, where a FlowBarrier is used to mimic the reduction of flow due to ice parflow code.

To my understanding this implementation accounts for the reduction of flow in ParFlow, but does not account for eCLM ice-content. The volumetric water content on the eCLM side is untouched, which might lead to an bias of water content as mentioned earlier.

There is also the 'ice_fraction' available in eclm, instead of usage of the ice impedance.

In TSMP1 the soil moisture was decoupled, when h2osoi_ice>0.1 see also.

Snow

There are multiple snow layers in eCLM. The interesting interface for the coupling is the snow/soil interface. The snow infiltrate to h2osoi_icq and h2osoi_liq see also. As the transition from H2OSOI_ICE to H2OSOI_LIQ is not accounted in the fields from ParFlow, this water seems to get lost in the coupling.

eCLM/src/clm5/biogeophys/SoilHydrologyMod.F90

Lines 845 to 859 in 98054f5

	! Renew the ice and liquid mass due to condensation
	if (snl(c)+1 >= 1) then
	! make consistent with how evap_grnd removed in infiltration
	h2osoi_liq(c,1) = h2osoi_liq(c,1) + (1._r8 - frac_h2osfc(c))*qflx_dew_grnd(c) * dtime
	h2osoi_ice(c,1) = h2osoi_ice(c,1) + (1._r8 - frac_h2osfc(c))*qflx_dew_snow(c) * dtime
	if (qflx_sub_snow(c)*dtime > h2osoi_ice(c,1)) then
	qflx_sub_snow(c) = h2osoi_ice(c,1)/dtime
	h2osoi_ice(c,1) = 0._r8
	else
	h2osoi_ice(c,1) = h2osoi_ice(c,1) - (1._r8 - frac_h2osfc(c)) * qflx_sub_snow(c) * dtime
	end if
	end if
	end do

Overland-Flow

Treatment of overland-flow in eCLM in coupled eCLM-ParFlow simulation should be checked also with respect to q_melt from snow.

Another subject for checking should be the irrigation in the coupled system.