UM crashes dues to sea ice ancillary

Hi everyone,

I’ve been working on running the ESM1.5 AMIP configuration with some modified SST and sea ice ancillaries. The SST’s I’m using are based on ERSST V5 and the sea ice is from the Merged Hadley-OI dataset. I’ve regridded both datasets and applied a diddling procedure.

Running the model with these ancillaries results in a bi_linear_h crash in July of the first year. After applying a couple of different perturbations to the start dump, the model still crashes in the same month.

Running the model with only the new SST’s while leaving the sea ice as the original AMIP data runs through the whole first year without crashing, so it seems like the sea ice file is causing some issues.

The following plots show the change in the sea ice boundary condition from June to July, for the new sea ice file on the left, and the original AMIP file on the right

The following show the changes from July to August:

In both datasets there are grid cells where the sea ice will go from 0 to full coverage rapidly, though this looks to occur on a bigger area of the new dataset. The most rapid change actually occurs in June to July of the original AMIP data, though only for one cell.

I was wondering if anyone knows whether these types of rapid changes could be the cause of the crashes, and whether you have advice about working around these crashes?

One suggestion has been to try reducing the model time step. Does anyone have any experience doing this with Payu?