rgknox
commented
5 years ago The relative error in the patches is definitely too large, its not a case of "the error checker is too strict". I am noticing that the crown areas are really massive, EDIT, I thought is was printing the per-plant cohort area... redacting this statement.
like some are about the size of the patch itself ~5k m2, for instance the last cohort printed. It is possible that with such massive crowns, that when we pare off the number density for promotion/demotion the math precision gets compromised...
ckoven
rosiealice
In some of the ensembles I've been running, competing two PFTs at the BCI testbed, and initialized from inventory, I'm seeing a consistent error from a couple of the ensemble members. This happens both with and without the changes in #441.
I'm not sure what's going here. it seems to happen with two of the ensemble members. both ensemble memebrs aer generating unrealistically large trees, which may be what's happening. possibly something like the error in conservation of crown area when fusing trees that are large means that more iterations are required to shuffle the canopy around. so possibly its as simple as increasing the allowable number of iterations.
here's the relevant part of the cesm log file of one of the ensemble members that crashes:
The lnd log from the first ensemble member that crashes ends with this:
here's the lnd log from the second ensemble member that crashes: