Verification of the National Blend of Models

The National Weather Service (NWS) is developing the National Blend of Models (NBM) to provide a nationally consistent and skillful suite of calibrated forecast guidance based on a blend of NWS and non-NWS deterministic, ensemble, and statistically post-processed model output. The NBM contains blended guidance for many National Digital Forecast Database (NDFD) elements over the conterminous United States (CONUS), Alaska, Hawaii, and Puerto Rico. NBM guidance is based on output from the GFS deterministic, GFS ensemble, Canadian ensemble, Gridded MOS (GMOS) and Ensemble Kernel Density MOS (EKDMOS). The NBM is intended to serve as valuable guidance for NWS forecasters at local Weather Forecast Offices (WFOs) and national centers.

The National Digital Forecast Database (NDFD) contains official NWS forecasts produced by NWS forecasters on a fine-resolution grid. The Meteorological Development Laboratory (MDL) routinely evaluates the NDFD and compares the skill of the human forecast to guidance for the same weather element. This guidance includes the NBM, as well as Weather Prediction Center (WPC) forecasts, model forecast grids, and statistically post-processed model output. MDL evaluates forecasts and guidance both on grids and at stations.

In this presentation, we review the performance of the NBM compared to the model blend components, WPC forecasts, and NDFD forecasts for forecast periods out to seven days in advance. We will show the value of post-processing model output to generate improved guidance. We also investigate improvements to the NBM resulting from recent upgrades and issues associated with the use of an analysis to verify guidance forecasts tuned to that analysis.