Entity-based Verification and Uncertainty Issues
Pertti NURMI and Christoph ZINGERLE
Finnish Meteorological Institute, Finland
Abstract
A common problem associated with many of the classical methods to verify forecasting systems of high spatial resolution is the dependence of the verification metrics on the scales of the forecasts and the observations (so-called "double penalty" issue). Entity-based verification methods are aiming at scale independent verification. Such methods consider specific objects (entities, "blobs") formed by given suitable meteorological quantities (e.g. precipitation, cloud pattern, pressure). The CRA verification scheme (Contiguous Rain Area; Ebert & McBride, 2000) was originally developed at the Bureau Of Meteorology (BOM) of Australia to verify their NWP precipitation forecasts against analyzed, gridded, surface observations. The scheme provides a methodology to decompose the total forecast error into contributions from the displacement (position), intensity (volume) and pattern (computed as a residual) between the forecast and observed entities. The original CRA software package from BOM has been adopted and implemented locally at the Finnish Meteorological Institute (FMI).
Some of the features of the CRA scheme are highlighted, as well as a couple of examples of the behavior of the scheme, and it’s output, for given precipitation events representing different spatial scales. One example (showcasing ECMWF forecasts) represents a flooding event in Southern Finland in summer 2004, the other (demonstrating HIRLAM forecasts) an intense MCS storm producing heavy precipitation during the Athletics World Championships in Helsinki in August 2005. Finally, the applicability of entity-based verification methodology for hydrological verification, reflecting COST Action 731 requirements, is discussed.