The effects of statistical post-processing on resolution (Invited)

Statistical post-processing schemes are widely recognised as effective ways of improving the "skill" of dynamical model-based predictions. Depending upon the statistical procedure used, the "skill" improves only because of improvements in the calibration of the forecasts, so the reliability component of "skill" can improve fairly considerably, but usually at the loss of some resolution. Since ultimately, it is only the resolution that is the usable part of the forecast, any loss in resolution is problematic. In this presentation the reasons for loss of resolution are considered, and suggestions are made for ensuring that resolution is minimised while improvements in reliability are still made.