Who's the King of PoP? Comparing the Accuracy of NWS and NAM/GFS MOS Precipitation Forecasts for Ten U.S. Cities, 2003–2012

Perceptions vary regarding the added value of the human element to probability of precipitation forecasts. In an effort to shed light on the added benefit of human forecasters, we examine National Weather Service (NWS) and Model Output Statistics (MOS) probability of precipitation forecasts across the United States. Specifically, we analyze North American Model (NAM) and Global Forecasting System (GFS) Model Output Statistics (MOS) 12-hour probability of precipitations (PoP) forecasts from 2003-2012 and compare them to subsequently issued National Weather Service forecasts for the same time period. Forecasts for ten cities are examined: Atlanta, Dallas, Seattle, New York City, Denver, San Francisco, Washington D.C., Kansas City, New Orleans, and Chicago. The forecasts are evaluated using the Brier Skill Score for each city, season, and period of day (00-12Z and 12-00Z). The results paint a more detailed picture of human-model interaction in the forecast process, as well as regional and temporal variations in both human and model forecasts. Time permitting, we will discuss the application of Bayes' Theorem to our results.>