peterlafollette
commented
1 year ago I believe I have solved the issue during LGARTO-C development. The issue was that infiltration was set to 0 when there was AET>0. Technically in the cases where f_p would be 0 (because for example the soil is completely saturated at the start of the time step) and AET>0, some water should be able to enter the soil during the time step. The solution was to set infiltration equal to AET in these rare cases. This issue only seems to occur when the maximum ponded head is greater than 0 and when the soil is completely saturated; in typical use cases for LGAR (arid / semi arid areas), the soil should almost never get completely saturated. Solution not yet pushed.
Short description explaining the high-level reason for the new issue.
Current behavior
If you run the current default Phillipsburg simulation, changing only the initial_psi from 2000 cm to 2 cm, then a bug will occur where ponded head will reach its maximum value and not infiltrate into the soil. Additionally, the final soil moisture is very dry, with maximized ponded head, and runoff is far larger than expected.
Expected behavior
Ponded head should infiltrate into soil that is not completely saturated. In this case, we expect the final cumulative infiltration to be larger, the final cumulative runoff to be smaller, and the final soil moisture profile to be wetter. I speculate that this issue was caused by the soil moisture profile reaching saturation while simultaneously having maximized ponded head, which might have also happened in the same time step that a wetting front exceeded the lower model boundary. I also notice that this issue is rare (or at least, I haven't encountered it before), and it is probably unlikely if LGAR is applied in semi arid or arid areas, but it's something we should ultimately fix.
Steps to replicate behavior (include URLs)
Screenshots