How Well Can the NARCCAP Models Capture the High-Resolution Spatial Patterns of the Precipitation Field in the Western United States

In the western U.S. where complex terrains present, accurate information on high-resolution spatial distribution of precipitation is critical to many important issues, such as flood/landslide warning, reservoir operation, and water system planning. For instance, due to the highly heterogeneous spatial distribution of precipitation in California, vast manmade infrastructures, such as the Central Valley Project and California Water Project, have been built to resolve the mismatch between where precipitation falls and where water is consumed. In the face of climate variability, prudent planning and proper operation of these water infrastructures are essential to the region's sustainability. The success of planning requires accurate precipitation projection from climate models. In particular, two questions: 1) how well climate models can simulate the precipitation spatial patterns, and 2) how these patterns will change will climate, are of great interest to many researchers and managers in environmental planning and management. Our study are designed to answer these questions based on the dynamical downscaling results from the NARRCAP program and our recent findings on the high-resolution spatial patterns of the daily precipitation field in the western U.S. Using EOF analysis on the U.S. daily precipitation product of Climate Prediction Center (CPC), we revealed that there exists dominant spatial patterns in the precipitation field over the western United States. The spatial patterns are consistent at different spatial resolutions and persistent over decades. Therefore, we intend to evaluate how well climate models can capture these spatial patterns of observed precipitation in this region. EOF analysis is applied to precipitation outputs from NARRCAP, and the derived spatial patterns are compared with those of observation for the same time period. The similarity between model simulations and observation is quantified with mutual information based on Shannon entropy. Moreover, using model simulations, the spatial patterns of historical period (1971-2000) and future period (2041-2070) with the SRES A2 emissions scenario are compared. The results provide insight into the impacts of climate change on the spatial distribution of precipitation, helping the western states design adaptation strategies. This work is our first step towards the development of accurate future precipitation projection for hydrological applications. Based on the skills of simulating the spatial patterns, an ensemble of future precipitation projections from NARRCAP output will be produced.