WRF Ensemble Model Performance During Atmospheric River Events In California

Abstract

During the third week of January 2010 a series of heavy-rainfall-producing systems apparently related to an atmospheric river (AR) swept into California during the Hydrometeorological Testbed (HMT) winter exercise. As part of this exercise, an experimental regional ensemble Weather Research and Forecasting (WRF) numerical modeling system developed at NOAA/GSD was in operation. This system included a web-based quantitative precipitation forecast (QPF) verification system (developed by the Developmental Testbed Center and funded by the USWRP) that produced both near-real-time and retrospective verification statistics. These statistics provided an opportunity to assess the performance of the WRF ensemble with respect to other available models and also to diagnose some features of the weeklong event. Prominent among these features were several precipitation pulses that appeared related to a striking mid- to low-level periodic turning of onshore (and upslope) winds from strong westerlies to extreme southerlies with a 24h periodicity. Analysis of equitable threat scores suggests the usefulness of verification of precipitation (and ultimately of other derived meteorological variables) to assess capabilities of model forecasts to capture this process. We also discuss the relevance of several other QPF verification metrics to diagnoses of the rainfall event, and present comparisons that assess the impact of model scale on the performance of ensemble members and baseline lower-resolution model forecasts. Finally, we make use of options in the verification system to illustrate impacts of regionalization of verification domains (e.g., within individual river basins and watersheds) and selection of verification data.