search

plesager / ece3-postproc

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

Hiresclim2 IFS monthly script crashes on 2006-08 #36

Closed plesager closed 5 years ago

plesager commented 6 years ago

This is more for the record, since I doubt we'll spend time trying to fix this, and is likely dependent on my installation of cdo.

While processing 60+ years of PRIMAVERA output with hiresclim2, I got one year (2006) that crashed the hiresclim2/script/ifs_monthly.sh script with this error:

cdo cat (Abort): Input streams have different number of layers per timestep!

This is happening l.87 of the script. Turns out that one month, of one variable, had only 43 levels instead of 90 for all the other months and variables. Trying to reproduce the step and find out where it went wrong (l.61):

$cdo -b F64 -t $ecearth_table splitvar -sp2gpl \ -setdate,$year-$m-01 -settime,00:00:00 -timmean \ icmsh_${ym} icmsh_${ym}_ & Before, all data have 90 levels, but after that line, the variable v (and only that one) has only 43 levels. So 1 month out of 780 creates a problem, and I have no clue why.

plesager commented 5 years ago

This problem happened again on cca at ECMWF, same place in the code, same variable, with one member of PRIMAVERA AMIP low res for two consecutive years (2044 and 2045). This is happening with the current default cdo/1.8.2. I tried the older version of cdo (1.6.1) available on the system, and it works! Comited a fix 587c99a.