Analysis of Extreme Rainfall Forecasts from the HRRR Model

Despite decades of improvements in numerical weather prediction models and operational forecasting practices, extreme rainfall remains one of the most significant threats to life and property in the United States. Due to the infrequent and often localized nature of these events, and the difficulty of amassing a comprehensive forecast database for a frozen configuration, relatively little work has been done to establish model precipitation climatologies and objectively verify extreme rainfall forecasts from the latest operational, high-resolution models.

One such model is the convection-allowing High-Resolution Rapid Refresh (HRRR), an hourly updating, dynamic modeling system that uses a specially configured version of the Advanced Research, Weather Research and Forecasting (WRF-ARW) community model. HRRR assimilates many novel and conventional observation datasets, including radar reflectivity and satellite-derived cloud attributes, to produce weather forecasts for lead times of 18 to 36 hours, with a horizontal grid spacing of 3 km. HRRR version 3 was implemented operationally at the National Centers for Environmental Prediction (NCEP) in July 2018 and incorporates recent enhancements of the HRRR model physics suite, including improved land-surface and boundary layer prediction using the updated Mellor-Yamada-Nakanishi-Niino (MYNN) parameterization scheme, aerosol-aware Thompson microphysics and an upgraded Rapid Update Cycle land-surface model.

Precipitation forecasts from HRRR version 3, run experimentally in a frozen configuration at the Earth System Research Laboratory for approximately one year prior to implementation, are examined to determine the frequency, location and characteristics of extreme precipitation rates over intervals ranging from 1 to 18 hours. The HRRR forecast climatology is then compared to NCEP Stage-IV quantitative precipitation estimates for intervals of 6 to 18 hours to infer systematic biases in occurrence and areal coverage of extreme events, and evaluate changes in forecast performance with respect to increasing lead time. Potential applications of this analysis include real-time prediction of flash flooding, safe dam and levee design, and coupled hydrological modeling.