Summary of Results and Issues emanating from tests with EOBS19
Overall timing is very favourable compared to CLIMPACT2, here is a summary extracted from the slurm log:
climix -s -e -x cdd EOBS19/pr 9300.0490 seconds.
climix -s -e -x cwd EOBS19/pr 9207.9771 seconds.
climix -s -e -x r10mm EOBS19/pr 335.3579 seconds.
climix -s -e -x r20mm EOBS19/pr 324.6040 seconds.
climix -s -e -x r30mm EOBS19/pr 366.9416 seconds.
climix -s -e -x sdii EOBS19/pr 403.3451 seconds.
climix -s -e -x prcptot EOBS19/pr 387.7817 seconds.
climix -s -e -x su EOBS19/tasmax 444.3942 seconds.
climix -s -e -x id EOBS19/tasmax 328.0380 seconds.
climix -s -e -x txge30 EOBS19/tasmax 336.1205 seconds.
climix -s -e -x txge35 EOBS19/tasmax 331.1915 seconds.
climix -s -e -x txx EOBS19/tasmax 326.3819 seconds.
climix -s -e -x txn EOBS19/tasmax 330.7312 seconds.
climix -s -e -x txm EOBS19/tasmax 330.0749 seconds.
climix -s -e -x tr EOBS19/tasmin 495.1577 seconds.
climix -s -e -x fd EOBS19/tasmin 335.2957 seconds.
climix -s -e -x tnlt2 EOBS19/tasmin 338.7568 seconds.
climix -s -e -x tnltm2 EOBS19/tasmin 329.6039 seconds.
climix -s -e -x tnltm20 EOBS19/tasmin 335.5620 seconds.
climix -s -e -x tnx EOBS19/tasmin 328.9269 seconds.
climix -s -e -x tnn EOBS19/tasmin 401.6072 seconds.
climix -s -e -x tnm EOBS19/tasmin 327.6455 seconds.
climix -s -e -x tmge5 EOBS19/tasmidpoint 441.7109 seconds.
climix -s -e -x tmlt5 EOBS19/tasmidpoint 323.7007 seconds.
climix -s -e -x tmge10 EOBS19/tasmidpoint 311.0065 seconds.
climix -s -e -x tmlt10 EOBS19/tasmidpoint 321.8500 seconds.
climix -s -e -x tmn EOBS19/tasmidpoint 303.0792 seconds.
climix -s -e -x tmx EOBS19/tasmidpoint 303.8381 seconds.
climix -s -e -x tmm EOBS19/tasmidpoint 320.8438 seconds.
climix -s -e -x hddheat18 EOBS19/tasmidpoint 428.0735 seconds.
climix -s -e -x cddcold18 EOBS19/tasmidpoint 356.9644 seconds.
climix -s -e -x gddgrow10 EOBS19/tasmidpoint 344.1421 seconds.
-
A identified before (#129 (closed)) the spell based calculations (
cdd
andcwd
) are comparatively much slower. By monitoring the progress, the stepINFO:root:Setting up aggregation
was identified as the time consuming part. Suggestion is to update the logging to facilitate this (#143 (closed)) -
For some indices (i.e.
cdd
,cwd
,txx
,txn
,tnx
,tnn
,tmn
,tmx
) it appears that masking is not correctly implemented or the mask is not correctly implemented in the input file (#144 (closed)). Cf. excerpt from the slurm log:
INFO:root:saving empty cube
/home/sm_lbarr/.conda/envs/climix-devel-3/lib/python3.7/site-packages/iris/fileformats/netcdf.py:2038: UserWarning: Cube 'air_temperature' contains unmasked data points equal to the fill-value, 1e+20. As saved,
these points will read back as missing data. To save these as normal values, please specify a 'fill_value' keyword not equal to any valid data points.
warnings.warn(msg.format(cube.name(), fill_value))
INFO:root:opening
INFO:root:Starting with DimCoord([1950-07-02 12:00:00], bounds=[[1950-01-01 00:00:00, 1951-01-01 00:00:00]], standard_name='time', calendar='gregorian', long_name='time', var_name='time')
-
Preliminary comparisons with the reference data (produced by
CLIMPACT2
) indicate that there are some differences, but this is complicated by the masked data are included.