enscrps
Added by Florian Pappenberger over 11 years ago
Hello,
I am using enscrps:
cdo enscrps ref_file ensemble_files[1-50] outputfile
Is it possible to have a crps per grid cell rather than averaged?
Replies (1)
RE: enscrps - Added by Florian Pappenberger over 11 years ago
please ignore, just saw that that was a new requested feature.
In the meantime, for normal distributed variables: (Reference: Gneiting and Raftery (2007) : Strictly Proper Scoring Rules, Prediction and Estimation. Journal of the American Statistical Association)
cdo ensmean input_ens.[1-50] mean
cdo ensstd input_ens.[1-50] std
cdo div std -sub observations mean mm
#compute normal pdf
cdo mulc,0.3989422840 -exp -divc,2 -mulc,-1 -sqr mm normpdf.grib
#compute normal cdf (approximation accroding to “A cosine approximation to the normal distribution” by D. H. Raab and E. H. Green, Psychometrika, Volume 26, pages 447-450.)
cdo pow,-1 -addc,1 -exp -sub -mulc,-0.07056 -pow,3 mm -mulc,1.5976 mm normcdf.grib
cdo mulc,-2 normpdf.grib first_part
cdo mulc,-1 -mul mm -subc,1 -mulc,2 normcdf.grib second_part
cdo mul -mulc,-1 sigma -add -addc,0.56418958354 first_part second_part crps
#build appropriate mean