search

ACCESS-NRI / dev_coupling

0 stars 0 forks source link

Sea ice problems #31

Open MartinDix opened 1 month ago

MartinDix commented 1 month ago

Comparing the cm3-run-30-07-2024 run with the CM2 PD control u-bz687. These should start with the same ocean initial condition and sea ice area is initially the same.

cm3_icearea

Why does it fail to form in the SH winter? Does the ocean not get cold enough or does the ocean cool and then something go strange with ice formation (e.g. something keeps melting it).

MartinDix commented 1 month ago

Even more dramatic SH difference for sea ice volume icevol

ice_depth

MartinDix commented 4 weeks ago

Averages over a part of the Weddell Sea show that the net surface flux from the UM is more negative in CM3, so why doesn't the ice increase?

weddell_map

weddell

There seems to be some sort of weighting difference between the UM and CICE ice surface temperatures, but both show that it doesn't cool down despite the negative surface flux.

weddell_temp

MartinDix commented 4 weeks ago

The CICE radiation fluxes seem to be zero over the ice, including the upwards LW, FLWUP. E.g. downwards SW from GMOM_JRA.cice.h.1981-07.nc

cice_fswdn

Sensible heat flux is also zero, though latent seems ok

cice_flat

aekiss commented 3 weeks ago

We don't see the sea ice vanishing in ACCESS-OM3 https://github.com/COSIMA/access-om3/issues/39#issuecomment-2232460831

MartinDix commented 3 weeks ago

The scale-ice-fluxes and fix-airtemp experiments don't make much difference to the area means, though there are some local differences

icearea

icevol

ice_depth_s

MartinDix commented 3 weeks ago

The scaled-flux run saves some extra flux diagnostics from MOM, though they're not quite identical (in name at least) to those saved from CM2. Now averaging over a smaller Weddell sea region at the edge of where ice forms in the CM3 run

weddell2

CM3 cools similarly to CM2, suggesting there's no problem with the open ocean fluxes. However when at freezing ice fails to form at the same rate for some reason. Why is June so different?

weddell_comparison

MartinDix commented 2 weeks ago

With fix to scaling of snow temperature

icearea icevol

anton-seaice commented 2 weeks ago

Did you save fsurf and fcondtop in CICE ?

MartinDix commented 2 weeks ago

fsurf_ai and fcondtop_ai are saved.

anton-seaice commented 2 weeks ago

If its useful for me to have a look, send me the path to the output :)

MartinDix commented 2 weeks ago

Monthly means gathered to a single directory /scratch/tm70/mrd599/cylc-run/u-di851/cice_output

ofa001 commented 2 weeks ago

Had a quick look at fsurfn_ai and fcondtopn_ai in this u-di851 run first 2 years against the 1st year of the CM2 run di702 where I your run has only 1 year of data, the runs quickly diverge, and the CM2 retains ice in thicker categories, has higher fluxes in these categories fluxes by a up to 2~5Wm-2 depending on month, but its not clear which comes first. Looking at monthly data. The daily CM2 data doesn't have the data split by categories with the flux data, but we could run that way if we thought it would help.