New Directions in Statistical Post-Processing (Keynote Presentation)

Statistical post-processing involves using past forecast information and associated observations/analyses to adjust for deficiencies in the current forecast. This talk will discuss some very recent developments in statistical post-processing, will review some of the more substantial remaining challenges, and will suggest some fruitful paths for future inquiry. Recent developments will include "ensemble copula coupling" as a way of preserving space-time covariance information in post-processed guidance and novel parametric methods of post-processing. Substantial remaining challenges include how to provide high-quality post-processed guidance with a limited training sample size of past forecasts and observations and incorporating information from ensembles in the post-processing. New directions may touch upon ideas such as the use of wavelets decompositions in statistical post-processing.