The Impact of Precipitation Dataset Choices On Analyses and Forecast Verification During The Hmt

Abstract

Over the continental United States, a variety of different sources of precipitation observations and estimates are available for use as verification datasets and as input to hydrologic forecast models. These range from radar and satellite estimates, generally available in continuous spatial grids, to gauge networks of varying spatial and temporal resolution. Under difficult conditions (e.g., extreme terrain and heavy rainfall) the performance characteristics of different precipitation observations can vary widely. As a result, analyses, streamflow forecasts, and model verification scores based on these different datasets will not generally be identical. Understanding how the choice of data impacts precipitation analysis and verification thus becomes an important objective. In the present work, we address this issue by evaluating several observational datasets (primarily independent gauge datasets and radar-derived estimates) that are used to compute basin-average precipitation statistics. An eventual use of these results will be to assess other national- and regional-scale precipitation analyses, and to verify numerical predictions of precipitation by high-resolution ensemble forecast systems. Here, we compare precipitation amount distributions computed from the so-called Stage IV gridded radar/gauge analyses with scores computed from independent hourly and daily gauge sites in a specific mountainous region of the western United States. We also briefly examine the effects of data quality. Finally, we discuss the most useful next steps that can increase our understanding of inherent uncertainties in precipitation analyses and numerical forecast verification.