(AUTRES) Série de simulations préparatoires à CMIP6, LMDZ-Orchidee, CTRL : CloudThVert3AAA


Metrics computation

RMSE (Root Mean Square Error) between models and reference datasets (observations or reanalyses depending on the variable; identified at the top of the columns). The RMSE are computed on a 12-month climatological annual cycle (spatio-temporal variability + mean); they mix the information on the spatio-temporal correlation and standard-deviation ratio, as well as the mean bias. The results are shown in % of the average RMSE obtained for the set of simulations chosen as reference. A result of -10 indicates that this error is 10% lower than the average of the errors of the reference simulations.  Basically these metrics provide an overall view on whether the simulations are closer to or further of a set of reference simulations (say, the AMIP simulations, an AR4 simulation...)

Metrics with respect to forced-by-SST AMIP multi model

Metrics with respect to coupled CMIP5 simulations (historical)

Metrics with respect to IPSLCM5A-LR (clim A verifier)