Feature-Relative Forecast Evaluation

Traditionally, spatial forecast verification has been performed in an Eulerian sense, as features move through a fixed geographical region, which helps to identify errors in the forecast for specific geographic areas. However, forecasters and model developers alike may gain valuable insight by knowing the spatial error patterns relative to a particular type of weather feature, that is, in a semi-Lagrangian sense. For example, in this work, forecasts from the North American Mesoscale (NAM) and Global Forecast System (GFS) were analyzed to determine the forecast errors relative to extratropical cyclones during the 2013-14 winter season for portions of the United States east of the Rocky Mountains. The location of an extratropical cyclone was determined and then centered on a pre-defined grid which moved with the cyclone. Error fields were produced and tested for statistical significance after first adjusting for spatial correlation and serial dependence. Errors found to be statistically significant might then be accounted for in a forecast or an area of further development for modelers. Preliminary results from this ongoing work will be presented at the conference.