Downscaling of reanalysis precipitation data using the analog ensemble method for global and regional reanalysis

Abstract:

The analog ensemble method is a statistical post-processing technique which has been proven to provide added value to numerical weather forecasts. It determines the best matches for the current time in a training data set and takes the corresponding observations, i.e. the analogs, as realizations of the analog ensemble. We have applied this technique to a global (ERA-Interim) as well as a regional (COSMO-REA6) reanalysis data set in order to downscale the reanalyzed precipitation. Further, we use the method to also estimate analog ensembles for a high-resolution reanalysis (at 2km) from the coarser grids. We will present the methodology and an evaluation of the downsclaed data sets in comparison to independent observations. We further show an intercomparison of the results from global and regional forcing data sets with a focus on the general potential of such data sets for downscaling purposes.