Study of Spatial Patterns of Daily Precipitation over the Western United States

The spatial patterns of daily precipitation over the western United States (30^o to 50^o N and 110^o to 130^o W) is investigated. EOF analysis on the U.S. daily precipitation product of Climate Prediction Center (CPC) reveals that there exist evident spatial patterns, which are represented by the dominant EOFs, in precipitation over the studied region. The analysis are carried out on two sets of data with resolutions of 1^o x 1^o and .25^o x .25^o respectively. However, the results show that these two data sets yield extremely similar dominant EOFs. This means that the spatial patterns are consistent across the spatial resolution of precipitation data. Furthermore, the EOF analysis is also carried out on precipitation of several consecutive periods in order to study the temporal properties of the spatial patterns. Results indicate that the spatial patterns also possess temporal consistency. Utilizing the consistency of spatial patterns, a empirical procedure is developed to downscale precipitation in the western United States. An experiment is designed to validate this procedure. In the experiment, the dominant EOFs of daily precipitation field over the 10-year period of 1998-2007 are calculated. The low-resolution daily precipitation of 2008 is projected to the dominant EOFs to get the principal components (PCs). Then, based on the consistency, the PCs of corresponding high-resolution dominant EOFs can be obtained, and the high-resolution precipitation can be retrieved. The result from this empirical downscaling procedure is compared with the high-resolution real-time precipitation analysis. The downscaled precipitation demonstrates a high correlation with the CPC analysis, with a correlation coefficient of 0.8857. Furthermore, compared with pixel-wise downscaling techniques, this procedure's major merit includes: 1) it maintains the precipitation field's spatial and temporal coherence; and 2) the dominant EOFs exclude the noisy components in the precipitation field, and, therefore, reduce the uncertainty of the downscaling. The spatial patterns revealed in this study can help evaluate and correct precipitation simulations by models. Moreover, the empirical procedure provides a novel and effective way of downscaling precipitation projections of GCMs for a variety of studies and applications.