| description abstract | This paper presents the results of a regionalization study of the precipitation climate of the western United States using principal component analysis. Past eigen-based regionalization studies have relied on rain gauge networks, which is restrictive because rain gauge coverage is sparse, especially over complex terrain that exists in the western United States. Here, the use of alternate data products is examined by conducting a comparative regionalization using nine precipitation datasets used in hydrometeorological research. Five unique precipitation climates are identified within the western United States, which have centers and boundaries that are physically reasonable and that highlight the relationship between the precipitation climatology and local topography. Using the congruence coefficient as the measure of similarity between principal component solutions, the method is found to be generally stable across datasets. The exception is the National Centers for Environmental Prediction?Department of Energy (NCEP?DOE) Reanalysis 2, which frequently demonstrates only borderline agreement with the other datasets. The loading pattern differences among datasets are shown to be primarily a result of data differences in the representation of (i) precipitation over the Rocky Mountains, (ii) the eastward wet-to-dry precipitation gradient that occurs during the cold season, (iii) the magnitude and spatial extent of the North American monsoon signal, and (iv) precipitation in the desert southwest during spring and summer. Sensitivity tests were conducted to determine whether the spatial resolution and temporal domain of the input data would dramatically affect the solution, and these results show the methodology to be stable to differences in spatial/temporal data features. The results suggest that alternate data products can be used in regionalization studies, which has applications for rain gauge installation and planning, climate research, and numerical modeling experiments. | |
