Closed pwaidelich closed 9 months ago
@pwaidelich Thanks for tracking this down.
@sidietz79 Our Excel model doesn't have this constraint, so I think we were just dropping those MCs. But this seems like an easy fix. Any objections?
No objections @jrising . Those runs would have been dropped if they returned an error somewhere, if not then potentially in the process of trimming the distribution, but this is not very likely given the small magnitude of SLR damages relative to temperature. Bill didn't have a constraint on this but then he didn't explore the extremities of the parameter space. Thanks @pwaidelich for picking this up.
The
VGIS
variable insrc/components/GIS.jl
denotes the "volume of the GIS relative to initial volume" in %, but it can currently reach negative values for outlier Monte Carlo runs. FlooringVGIS
at zero resolves this issue.Since the difference in VGIS between timesteps (
Diff_VGIS
) is a product of multiple terms includingvv.VGIS[tt-1] ^ pp.expvol
, the floor also ensures thatDiff_VGIS
will be equal to zero once the zero lower bound is reached