Should we evaluate the performance of GCM at specific region to reject the "bad model for that region" before we choose it to do downscaling? What parameters you recommend to evaluate?
We have just done this for CMIP6 in Southeast Asia, sub-selecting 8 GCMs for downscaling. The dimensions are multiple: model skill; model spread (e.g. making sure you cover most of the spread of the entire ensemble); model independence (e.g. choosing models that are not similar); spread in model climate sensitivity; skill on important small and large scale processes driving local climate.
