def get_fc_diag_loop(data_fc, fc_veri, diag):
diag_dict={}
for veri in fc_veri:
diag_dict[veri]={}
for start_date in fc_veri[veri]:
diag_dict[veri][start_date]={}
for model in data_fc[start_date]:
print 'Getting '+diag+' for '+model
diag_dict[veri][start_date][model]=get_fc_diag(data_fc[start_date][model]['data'],fc_veri[veri][start_date][model],diag)
return diag_dict
if 'rmss' in args.diags_cont and 'rmse' in args.diags_cont:
args.diags_cont.insert(0, args.diags_cont.pop(args.diags_cont.index('rmss')))
diags_cont={}
for diag in args.diags_cont:
scores=get_fc_diag_loop(data_fc, fc_veri, get_score_name(diag))
#Convert to skill scores where appropriate
diags_cont[diag]={}
for veri in args.veri:
diags_cont[diag][veri]={}
for start_date in start_dates:
diags_cont[diag][veri][start_date]={}
for model in model_list:
if diag!=get_score_name(diag): #true only if a skill score can be sensibly computed
score_perf=get_fc_diag(fc_veri[veri][start_date][model], fc_veri[veri][start_date][model], get_score_name(diag)) #perfect skill score
diags_cont[diag][veri][start_date][model]=get_skill_score(scores[veri][start_date][model],scores[veri][start_date]['clim_'+veri][:len(scores[veri][start_date][model])],score_perf)
else:
diags_cont[diag][veri][start_date][model]=scores[veri][start_date][model]
#Plotting diagnostic as a function of lead time
if not (diag=='rmss' and 'rmse' in args.diags_cont): #don't plot rmss separately from rmse.
for veri in diags_cont[diag]:
for start_date in diags_cont[diag][veri]:
fig,ax=plt.subplots(1)
plt_count=0
for model in model_list: #to get lines plotted in order, with multi-model combinations put first in the legend
if not ((diag=='acc' or (diag!=get_score_name(diag))) and fnmatch.fnmatch(model,'clim*')): #don't plot anomaly correlation or skill scores for climatology, as these are just zero
xpts=1+np.arange(diags_cont[diag][veri][start_date][model].shape[0])
