bcgen gives sea ice fractions that are too high and too extensive

[aflaim]

I am trying to use B-compset output to generate SST and ICE input for an F-case. I believe I need to use bcgen to convert the B-compset output to mid-month values expected as input for the F-compset.

However, sea ice fraction seems too high and too extensive after running bcgen.

I am using python to re-grid the POP output and correct SST and ice according to Hurrell 2008 prior to running bcgen. The pre-bcgen climatology attached below shows ice fraction after performing the re-gridding and Hurrell 2008 corrections.

I have attached the pre- and post-bcgen climatologies which illustrate the overly extensive sea ice after running ./bcgen.

Does bcgen typically have such a dramatic impact on sea ice extent? Does anyone have advice for addressing this issue?

Thank you,
Andrew

 

[asphilli]

It looks like there are some issues with the sea ice seasonal cycle as seen in the pre_bcgen.png attachment. There is minimal Arctic sea ice in November-January for example. So the first thing I would check is that the regridding that was done from the B-case POP output is correct.

Just to be clear, the procedure that you are using is:
1 - Regrid B-case SST and aice variables (on the POP grid) to a rectlinear grid using your own python code.
2 - Pass the output of 1) through regrid/bcgen.
3 - Use the output of 2) to force an F-case.
4 - The SST/ice that is output from the F-case should match the output of step 1) above. Note that the reason to use bcgen is to make sure that the original forcing dataset (=the output of 1 above) matches what comes out of your F-case.

 

[aflaim]

Hi Adam,

Thanks for the response. You're right, it turns out the default way that xarray was interpreting the time dimension was off by one month. Fixing that seems to help.