Contrail domain not properly tracked #81
Description
I noticed long compute times for some instances of APCEMM and found a possible issue with the contrail tracking mask or something related to it. These instances are repeatable and there are no stochastic variables implemented (i.e. no temperature perturbations, and as a precaution, have the same seed). The same behavior occurs at transport timesteps of 5 minutes and 10 minutes. I've included one example here to illustrate the issue, but this is not the only case in which this happens.
Conditions are 110% RHi 218K, 2.5 K/m lapse rate, and a 500m supersaturated layer below contrail spawn point.
X and Y axes are reported in meters.
You can see in the first figure how the domain explodes and shrinks between 5 hours and 12 hours, but even before that the domain exists well past the end of the superstaurated layer.
An easier visualization is shown here as a True/False mask of whether a cell contains any ice crystals (and is therefore tracked by APCEMM). We see crystals being tracked well beyond the point at which they should have evaporated into the surrounding subsaturated air.
APCEMM theoretically tracks if a cell is contrail containing
with a threshold of 0.90 particles per m3
but I don't see this in practice.
Not sure what the path forward is specifically but wanted to put this up in case anyone has run into something similar.