[MITgcm-devel] seaice adjoint and EVP

[Dimitris Menemenlis]

Martin and Jinlun, now that the snow and zmin stuff has been sorted out, I am 
rerunning a timestep test for the EVP solver and it looks like Jinlun's guess is 
correct.  Some figures comparing sea-ice fields after 2 months integration of 
LSR vs EVP with time steps of 1, 10, 20, 30, and 60 s time steps:
ftp://ecco2.jpl.nasa.gov/data1/arctic/output/tests/figs

ftp://ecco2.jpl.nasa.gov/data1/arctic/output/tests/figs/AREA4320.ps
shows that EVP solution tends to LSR as time step is decreased.

ftp://ecco2.jpl.nasa.gov/data1/arctic/output/tests/figs/UICE4320.ps
shows that striation goes away as time step is decreased

ftp://ecco2.jpl.nasa.gov/data1/arctic/output/tests/figs/maskedUICE4320.ps
is same figure but velocity where AREA=0 has been masked

so does this all mean that for 18-km resolution LSR solver is preferable?

> It seems there is some kind of instability shown in the sigma plots in the
> marginal ice zone. I wonder if a smaller EVP timestep, close to that of a
> fully explicit method, would get rid of this problem. The time step for FEX
> is 1 s or less.