Check for NaN and Inf about vom HOT 4 CLOSED

I think the problem might be that a value becomes exactly zero, leading to a division by 0 error and NaN. It might have to do with the native precision of the infrastructure, which would be different on the cluster vs. your local computer. Need to check if model can run without a water table at all, and if so, we could create a threshold to make it disappear without creating a significant error, rather than becoming infinitely thin.

from vom.

Yes, I think that's what happens. It's however risky behaviour that you can still run the VOM in both cases and get completely different results. So we should think about some thresholds indeed.

from vom.

Not sure, do you get different results, or just a result vs. no result? Could you also compare successful model runs on both systems? Anyway, we should catch the NaN error and see what exactly happens there. Can you use gdb on the HPC to debug the critical run?

from vom.

Apparently, this issue has not re-occurred since, so I am closing this issue.

from vom.