search

plesager / ece3-postproc

Suite of processing tools for EC-Earth3 output
5 stars 8 forks source link

EC-mean not working with PRIMAVERA output #33

Open oloapinivad opened 6 years ago

oloapinivad commented 6 years ago

This is to let you know EC-mean is crashing with PRIMAVERA output.

A zero order inverstigation tells me that there is no vertical levels selection neither in hiresclim or EC-mean, causing a crash of the code when a comparison is performed between variables on pressure levels (climatology has only 11 levels, while Primavera has many)

this is a more general issue, I will work on it to fix it.

plesager commented 6 years ago

I have a fix. The filters in hiresclim are not adapted for the default PRIMAVERA output. I will commit ASAP.

oloapinivad commented 6 years ago

It seemed that the original PRIMAVERA flag was not correctly working. I updated the code, it is working on PRIMAVERA data on Marconi. We did it in the same time.

Actually the selection of the 11 vertical levels for comparison to the climatology perhaps should be a default. I wrote a simple fix that I have just committed in my branch 4a61ea8

In the meantime I have noticed that the partial PIs are equal to zero, which is strange. I will check if it is the same on CCA.

plesager commented 6 years ago

Oh, I see, different fix for different problem. I am still struggling with cdfmoc (#32). Need to get over it to test ecmean on my couple primavera run.

oloapinivad commented 6 years ago

This cdfmoc thing is very strange, especially since it is machine dependent. I also have some crash for hiresclim with PRIMAVERA output but they are related to ocean variables name. I will check it as soon as I can.

You can check what Alessia did in our fork (I am not completely aware of the changes) since the IFS part is working fine in PRIMAVERA configuration with both hiresclim and ECmean now.

plesager commented 6 years ago

See my comment in 4e39a98a595ead4d44e020c3d4d429522f38ab85, maybe that's the issue you have with nemo?

oloapinivad commented 6 years ago

Unfortunately it is not the case, it seems to be a nco-related issue on Marconi. I am investigating this further.

This is the error I am getting:

ncrename: nc_hashmap.c:106: rehashVar: Assertion `count == hm->count' failed.
./nemo_post.sh: line 152:  4328 Killed               (core dumped) ncrename $rename_str ${froot}_grid_T.nc
plesager commented 6 years ago

@oloapinivad wrote:

Actually the selection of the 11 vertical levels for comparison to the climatology perhaps should be a default. I wrote a simple fix that I have just committed in my branch 4a61ea8

I tried your fix on my primavera output, but I have no luck with this line:

$cdo setzaxis,$TMPDIR/axis.txt -invertlev -zonmean -sellevel,1000,5000,10000,20000,30000,40000,50000,70000,85000,92500,100000 $field $TMPDIR/new_${vv}.nc

even if I convert to hPa. I will look closer to that next week.