Statistical Reliability of Quantitative Precipitation Forecasts From The HIGH-RESOLUTION Rapid Refresh

Abstract

Quantitative precipitation forecasts (QPF) from the 3-km High-Resolution Rapid Refresh (HRRR) model fill a variety of needs within the user community, from short-term flash flood forecasting to day-ahead probability of precipitation outlooks. Comparison with Stage IV quantitative precipitation estimates (QPE) reveals that HRRR QPF is relatively reliable for many 6-h precipitation thresholds, but exhibits a high bias in occurrence of heavy rainfall rates. In conjunction with the development of a HRRR “time-lagged ensemble” (HRRR-TLE), a realtime bias correction has been implemented, using the last several weeks of forecasts, to permit statistically reliable forecasts of the probability of exceedance of precipitation thresholds. In a novel application, a five-year dataset of HRRR QPF is also being interrogated to derive an estimate of the “probable maximum precipitation” (PMP) quantity at each gridpoint, a value that is important for safe dam engineering.