Physically based non-dynamical downscaling of precipitation in complex terrain

Climate modeling research is often at one of two ends of a spectrum, low-resolution climate simulations, or high-resolution numerical experiments. Low-resolution (50-200km grid spacing) simulations have the benefit of being able to run many different scenarios over regional or global scales, but lack important spatial variability in both the internal model processes and the resulting model output fields. High-resolution simulations (1-6km) are often constrained to small areas, single scenarios, and short time scales; however, they have a good representation of real spatial patterns and have been well validated against observations. Commonly, statistical downscaling methods are used to generate high-resolution information from low-resolution climate simulations; however, these methods rarely incorporate physical processes, and as such, they may not be valid under future climate scenarios.

We present a new non-dynamical downscaling procedure that combines large-scale forcing data from a low-resolution simulation with a simple, physically-based model that incorporates advanced microphysical processes, linear mountain wave theory, and a positive definite advection scheme all computed on a high resolution grid. This model allows important physical processes to be resolved such as the effect of updrafts on hydrometeor fall rates, as well as the role of variable wind speeds and direction. The output from this model is compared to high-resolution (2 and 4km) WRF (Weather Research and Forecasting) model simulations, individual station observations, and gridded precipitation products such as PRISM. Analysis is performed over narrow mountain ranges for which spill over effects are an important, and over wider mountain ranges for which upslope effects dominate. These comparisons show important areas of agreement between this model and WRF that are not captured by the original low-resolution model even after applying common statistical downscaling techniques.