A method to verify National Weather Service gridded precipitation forecasts

Precipitation is often an important component of high-impact weather events. As part of the National Weather Service National Digital Forecast Database, weather forecast offices now produce high-resolution (in both time and space) gridded precipitation forecasts. Such gridded forecasts provide a potential source of additional forecast information beyond numerical model gridded precipitation forecasts. However, verification of a single high-impact event can be problematic with traditional verification techniques, where small differences in timing, location, or intensity can lead to unexpected precipitation verification scores. The techniques presented here produce reasonable verification scores on small grids and over short time periods, while also scaling well to large grids and over long time periods. The techniques provide a way to measure long-term biases and trends, as well as short-term measures of forecast skill and added information for single events. The techniques also allow for the investigation of the spatial distribution of errors, and the measurement of skill on different spatial scales.