It’s not clear to me what soil moisture in the soil column under ice tiles actually means. We typically set tiles to ice when there’s a very thick ice layer, far deeper than our soil levels. I don’t see liquid water permeating the ice in the same way it does soils, so I can’t reason about what this term physically represents.
For context, I’m investigating an issue with AM3, regarding the reconfiguration stage. When reconfiguring to a different resolution, we lose all information about the CABLE soil moisture over ice tiles due to this UM recon code, which is called if any of use_smc_stress, v_int_active or h_int_active are true (see here). The means the soil moisture in ice tile columns is 0 going into CABLE.
CABLE seems to detect that the soil moisture is not a meaningful value, so sets it to what I think is the saturation point, based off some work @lishx has done running AM3 at higher resolutions (and then doesn’t change over time?). The soil moisture obtained this way is not far off what the base n96e restart had anyway. How important is getting our soil moisture right in ice soil columns?
@lachlanswhyborn - introductory context: There are differences between what is a physically realistic representation of the water dynamics on ice tiles, how these processes are represented in CABLE and how they are represented in JULES. In this respect CABLE is an older model and this area of the code has not been looked at in detail for sometime. The UM recon code is designed to ensure that the initial values provided to UM-JULES are sensible for the expectations of JULES (and may not be physically realistic, nor what CABLE will need).
In practice ice tiles in CABLE are viewed as
a soil matrix whose thermodynamic and hydraulic properties relate to solid ice, plus soil moisture (liquid and frozen) within the matrix that is subject to time evolution.
vertical water movement in CABLE’s soil column (soil_snow module) has to be of liquid water (but is expected to be slow).
to satisfy any surface water flux, any infiltration of melt or rain, and internal numerics CABLEs ice tiles carry a minimum amount of water as liquid - irrespective of temperature.
There are unrealistic elements to this representation - notably the need for a liquid fraction. This liquid is still subject to gravitational drainage and (in ESM1.6) this leads to a net loss of water from the ice tile soil columns over time - which is reset at every restart leading to a slow addition of water to the global Earth system in time.
See also github #301 and github #610. #610 in particular addresses the reset of the soil moisture (and frozen fraction) that will be occurring between the restart of the reconfigured restart and CABLE in side AM3.
In short - yes there are weaknesses here, no I don;t think it’s important given the internal resetting, and I suggest not focussing on this (as long as AM3 functions) given the current ice sheet work.