Analysis of HREF Summertime Extreme precipitation forecasts over three years

Extreme precipitation events, though rare, are highly impactful and present many forecast challenges, particularly when they occur as a result of summertime convective activity. Ensemble forecast systems can provide forecasters with an idea of the uncertainty in an extreme precipitation forecast, though due to the infrequent nature of these events, forecasters may lack an understanding of how well the ensembles represent them. In this study we evaluate forecasts of extreme precipitation (defined as exceeding Annual Recurrence Interval thresholds at 2, 10, and 50 year periods) from the NCEP High Resolution Ensemble Forecast (HREF) system, using the Multi-Radar Multi-Sensor (MRMS) Pass 2 product as a reference.

The HREF is a 10-member ensemble of opportunity, consisting of five different convection-allowing models (CAMs), plus time-lagged forecasts from each of the five CAMs. High resolution (3-km grid spacing, 1-h) probabilistic forecasts out to 48h are available from version 3 of the HREF since January 2021, resulting in three North American summers over which to evaluate model performance using a consistent model configuration. Comparisons between HREF and MRMS will be performed at several accumulation periods (1h, 6h, and 24h) using both grid-point and neighborhood techniques. Given the diverse membership of the HREF, there is value in characterizing the contributions of the individual ensemble members to extreme precipitation forecasts. General characteristics of the individual members (e.g. PDF, frequency of extremes) will be discussed along with which individual members contribute most to forecasts of extreme precipitation at the aforementioned thresholds and durations. This analysis aims to provide forecasters with insight into the forecast skill and possible systematic errors in the HREF for a number of potentially high-impact precipitation scenarios.