Statistical and dynamical downscaling of global model output for U.S. National Assessment hydrological analyses

Analyses performed for the US National Assessment require accurate projections of climate at scales below those produced by global models. We compare two approaches, statistical and dynamical downscaling, for producing hydroclimate input to a number of river basins targeted for the National Assessment. Global, driving "models" for both approaches are the NCEP/NCAR reanalysis and the Hadley Centre general circulation model. This project builds on previous intercomparisons, namely the Project to Intercompare Regional Climate Simulations (PIRCS) and two recent statistical downscaling intercomparison projects.

Our comparisons use seven-year time series of output from dynamical and statistical downscaling models. Time series of current-climate precipitation and daily min/max temperature from both approaches are compared with observations and are shown to reproduce climatology better than the raw reanalysis output. As a consequence, both downscaling methods introduce gains to hydrological simulations for basins that are subgrid to the driving global model. Differences between the statistical and dynamical approaches are attributable in part to the way in which each assimilates orographic information. Further analysis shows the ranges of hydrological response that can arise from analyzing climate change impacts with different, but physically plausible, downscaling approaches.