A Comparison of Statistical and Dynamical Downscaling of Winter Precipitation over Complex Terrain

Abstract Statistical downscaling is widely used to improve spatial and/or temporal distributions of meteorological variables from regional and global climate models. This downscaling is important because climate models are spatially coarse (50–200 km) and often misrepresent extremes in important meteorological variables, such as temperature and precipitation. However, these downscaling methods rely on current estimates of the spatial distributions of these variables and largely assume that the small-scale spatial distribution will not change significantly in a modified climate. In this study the authors compare data typically used to derive spatial distributions of precipitation [Parameter-Elevation Regressions on Independent Slopes Model (PRISM)] to a high-resolution (2 km) weather model [Weather Research and Forecasting model (WRF)] under the current climate in the mountains of Colorado. It is shown that there are regions of significant difference in November–May precipitation totals (>300 mm) between the two, and possible causes for these differences are discussed. A simple statistical downscaling is then presented that is based on the 2-km WRF data applied to a series of regional climate models [North American Regional Climate Change Assessment Program (NARCCAP)], and the downscaled precipitation data are validated with observations at 65 snow telemetry (SNOTEL) sites throughout Colorado for the winter seasons from 1988 to 2000. The authors also compare statistically downscaled precipitation from a 36-km model under an imposed warming scenario with dynamically downscaled data from a 2-km model using the same forcing data. Although the statistical downscaling improved the domain-average precipitation relative to the original 36-km model, the changes in the spatial pattern of precipitation did not match the changes in the dynamically downscaled 2-km model. This study illustrates some of the uncertainties in applying statistical downscaling to future climate.