Output inconsistent missing_value or _FillValue attributes and actual data between pr and tas variables
Both raw and reference data has variable attributes _FillValue=1e+20 and missing_value=1e+20, but the bias-adjusted output data is inconsistent with that:
- Output data has variable attribute only _FillValue=1e+20. The missing_value attribute has been removed.
- For pr variable, the data in the output has values of 1e+20 in empty areas, however for tas variable, the data in the output file actually contains 1e+36 values instead.
Possible reasons - something in the processing steps of a pr-type bias-adjustment is retaining the correct value, while in the tas-type is removing it and replacing with the netcdf default missing value 1e+36
Desired outcomes - retain the missing_value attribute if present in the raw data and ensure that the values in the output file are not reverting to default ones, double-check that the attribute values actually correspond to the values in the file?
Paths for the example files where the issue was identified:
- /nobackup/smhid20/proj/fouh/hydrologiska_klimatprojektioner/bcjobs/midas_reports/ssp126/r1i1p1f2/HCLIM43-ALADIN/v1/input_data - to check the attributes of raw data
- /nobackup/smhid20/proj/fouh/hydrologiska_klimatprojektioner/refdata/final/ - to check the attributes of the reference data
- /nobackup/smhid20/proj/fouh/hydrologiska_klimatprojektioner/bcjobs/midas_reports/ssp126/r1i1p1f2/HCLIM43-ALADIN/v1/output_data - bias-adjusted output data
- runscripts and so on are in the /nobackup/smhid20/proj/fouh/hydrologiska_klimatprojektioner/bcjobs/midas_reports/ssp126/r1i1p1f2/HCLIM43-ALADIN/v1 directory
This data was produced using Thomas' branch of midas-GCF2, although I am unsure if it is the latest version.