Using global and regional ensembles simultaneously for fv3-lam GSI EnVar

[TingLei-daprediction]

Up to now, in the EMC regional (FV3-LAM) DA parallel runs, the pre-set global ensembles (80) are used as the sampling covariances in the GSI EnVAR. This is mainly for the resources required for a self-cycled FV3-LAM regional ensembles. With the emerging computational resources, EMC and collaborators are developing self-cycled FV3-LAM (regional) ensembles based GSI EnVAR with smaller ensemble size (less than 40). Hence, using global and regional ensembles simultaneously for fv3-lam GSI EnVar is expected to be potentially helpful for increase the actual sampling ensemble size and extract larger scale information from the global ensembles .
The GSI EnVar approach (all ensemble members are mapped to one regular ensemble grid) make implementation of such such mixed ensemble function rather straightforward.