nctxck incorrectly complains about wrong time step #6
Description
When running nctxck on some (but not all!) Eyr datasets from the EC-Earth3-ESM-1 model we get the following errors:
Units:
IN FILE -- days since 1850-01-01
REF -- days since 1850-01-01
Calendar:
IN FILE -- proleptic_gregorian
REF -- proleptic_gregorian
Start:
IN FILE -- 1850 = 1850-12-31T12:00:00 = 364.5
REF -- 1850 = 1850-01-01T00:00:00 = 0.0
End:
FILENAME -- 1850 = 1850-01-01T00:00:00 = 0.0
IN FILE -- 1850 = 1850-12-31T12:00:00 = 364.5
REBUILT -- 1850 = 1850-01-01T00:00:00 = 0.0
Length: 1
MIP table: Eyr
Frequency: yrPt = 1 years
Is instant: True
Is climatology: False
Has bounds: False
Status: Error 001 -- Incorrect time axis over one or several time steps
Status: Error 010b -- Start date in file is later than start date from filename
Status: Error 007b -- End date in file is late than end date from filename
Status: Error 009b -- End date in file is later than theoretical end date
Wrong time step at index 1: IN FILE -- 1850-12-31T12:00:00 = 364.5 vs. REBUILT -- 1850-01-01T00:00:00 = 0.0
Obviously nctxck finds a timestep at the end of the year, thinks this is wrong and wants reset the time step to Jan 1 00Z. However , the dataset in thsi case is Eyr/fracLut, and the CMIP6 data request for fracLut says:
description[i]: End of year values (not annual mean)
The time step in the file is at the end of the year and therefore correct for this variable, and nctxck shouldn't throw an error.
For comparison I also checked the timestep in fracLut_Eyr_AWI-ESM-1-1-LR_piControl_r1i1p1f1_gn_1855-1855.nc from the AWI-ESM model that is published in the CMIP6 archive and found 1855-12-31 23:52:00, i.e. a value at the end of the year.
I therefore think the nctxck error is wrong, it should accept end of the year time steps for these variables. Could this be fixed in the source code?