Towards a Gridded Best Estimate of Accumulated Precipitation from Quantitative Precipitation Estimates (QPE) and Forecasts (QPF) in Mountainous Regions: A Simplified Kalman Filter Approach

Quantitative precipitation estimation (QPE) in complex terrain regions such as the intermountain western United States remains a major challenge. Rain gauge coverage is sparse, ground-based remote sensing systems such as precipitation radar suffer from beam blockage and uncertain microphysical assumptions, and satellite-estimates are prone to errors due to variations in the characteristics of the underlying surface. In recent years, convection-allowing models (CAMs) have provided high-resolution quantitative precipitation forecasts (QPF) which are challenging to verify in these regions. In order to obtain gridded estimates of precipitation, we propose applying a simplified Kalman Filter (KF) framework in order to combine information from radar-based QPE and CAM QPF, taking into account the uncertainties inherent in both sources.

In this presentation, we describe our approach for optimally merging radar-based Multi-Radar Multi-Sensor (MRMS) QPE and QPF from the High-Resolution Rapid Refresh (HRRR), in which QPE uncertainties are quantified through the MRMS Radar Quality Index (RQI). We experiment with several methods for defining CAM QPF uncertainties, and describe approaches for addressing the non-Gaussian nature of precipitation. The KF framework is illustrated through an example precipitation case, and statistics are provided for the performance of the merged product over a longer period. While originally motivated by a need for improved heavy precipitation occurrence records in regions of complex terrain for the purpose of training machine learning prediction systems, the approach has promise for a broad range of applications, including for the inclusion of precipitation in a real-time mesoscale analysis or a precipitation analysis of record, and evaluation of QPF from other NWP systems.