Evaluation of Several Spatial Filtering Methods For Probabilistic Cpm Ensemble Forecasts

Abstract

To support the goals of a collaborative project through the US Weather Research Program (USWRP), which is focused on developing high-resolution ensemble-based hazard detection guidance tools, several convection-allowing model (CAM) ensemble post-processing methods are currently being evaluated with the operational version of the High Resolution Ensemble Forecast (HREF) system. Owing to challenges associated with small ensemble size and lack of gridscale agreement at convection-allowing scales, it has become common practice to conduct spatial, i.e. neighborhood, post-processing to generate useful probabilistic guidance. Currently a variety of methods are typically applied through field-specific search radii and spatial smoothing methods. Such post-processing approaches include fractional coverage and neighborhood exceedance methods. A new approach based upon the average agreement scale (Dey et al. 2016) is of particular interest since it adaptively determines neighborhood radii by finding the minimum neighborhood size over which the group of ensemble forecasts are deemed suitably similar at a grid point. Several CAM ensemble neighborhood probability methods will be presented with example applications along with discussion regarding their respective computational expense.