Repository navigation
eCLM-ParFlow: Comparing soil saturation from both components during a coupled run #103
Description
This issue is based on a discussion with @kvrigor and @skollet .
The goal was to demonstrate correct coupling between eCLM and ParFlow by comparing soil saturation from both components during a coupled run. Under consistent soil properties and correct coupling, the saturation values from both components should match. I performed a functional test for the coupled ICON–eCLM–ParFlow system using an idealised setup. Ideally, eCLM and ParFlow should output consistent saturation values (ignoring ice effects).
Experiment: pft00-sid02-sv06 ( Land unit: bare soil, Soil type: sandy loam, Initial condition: dry soil until layer 11, below saturated).
Soil characteristics (e.g., porosity, permeability) are identical in both eCLM and ParFlow. Values used for comparison are taken directly from the respective model outputs.
Observations and verification steps
- ParFlow soil liquid water (H2OSOI, kg/m²) is correctly passed to eCLM
- A unit transformation alone should not introduce differences between ParFlow saturation and eCLM saturation. The transformation is done with convert_phead2volsoilmoist.py script. The results obtained from pressure head–based conversion are identical to those using ParFlow’s relative saturation output.
- Ideally, eCLM and ParFlow should output consistent saturation values (ignoring ice effects).
- Interpolation could be a reason for differences between the models: To exclude this, a setup with the same grid was created. This can be find in the ibg3-gitlab. The saturation values between eCLM and ParFlow are still not the same.
- eCLM uses by Default the soil water retention curve after Clapp and Hornberg. A switch to van-Genuchten soil retuntion curve might be useful. A switch might be not that easy, as in some parts of the CLM the Clapp and Hornberg curve is indirectly assumed as pointed out in the CTSM issues. However, a branch with a plainly used van-Genuchten soil water retention curve can be found here.
Figures
Saturation comparison between ParFlow and eCLM with ice.
Saturation comparison between ParFlow and eCLM no ice
Saturation comparison between Parflow and eCLM with the same grid and no ice.
Saturation comparison between Parflow and eCLM van-genuchten with the same grid and no ice.
Preliminary conclusion
As a recommendation, volumetric soil moisture output should not be used in a coupled simulation. The recommendation is to use the ParFlow output during a coupled run.
The coupling during ice needs to be checked (additional issue).
Activity
kvrigor commentedon Mar 6, 2026
Thanks for the writeup @s-poll ! Just wanted to highlight #90 which is potentially connected to this issue
s-poll commentedon Apr 2, 2026
While investigating #105, some insights came in, which might be relevant for this issue as well. Thanks @DCaviedesV for discussion.
Soil water retention curve
Figure 8 of Zhang et al 2022 and figure 1 of Shao and Irannejad 1999 shows soil retention curves for Clopp and Hohenberger vs van Genuchten distributions. The figures show, for the same pressure head, the soil water content can be doubled. From the energy and mass perspective that the coupling in combination acting with two soil water retention curves might lead to a mass difference.
Saturation comparison
Saturation in real case (EUR-12) seems to not match between eCLM and ParFlow even for areas without snow.
Porosity (probably less important)
For eCLM porosity it might be better to use the eCLM output variable
watsatinstead of calculating the porosity based on %sand for the comparison. The porosity in eCLM is calculated as a weighted average of mineral soil and organic matter components, which might be also good to send to ParFlow. However, for this issue it might create further "noise".[-]eCLM ParFlow saturation in coupled run[/-][+]eCLM-ParFlow: Comparing soil saturation from both components during a coupled run[/+]