5B.1 An effective approach to improve regional climate prediction using a proper combination of dynamical and statistical technique

Li, Rong, Jiming Jin, Shih-Yu Wang, Robert R. Gillies, Utah State University

There are often large biases associated with climate predictions and these are problematic when it comes to their application in the future assessments of water resources and ecosystems; this has been especially so in mountainous and drought-prone regions of the western United Sates. We have developed a hybrid approach that significantly reduces biases in regional climate projections; the method uses a combination of dynamical and statistical techniques. A statistical technique was employed first to correct biases in the CCSM (Community Climate System Model) data that are subsequently used to drive the Weather Research and Forecasting (WRF) model to produce estimates of precipitation. The biases associated with the model physics in WRF were reduced by identifying an optimal combination of physical parameterization schemes; this was achieved by conducting various sensitivity tests. Next, WRF downscaled simulations were performed for 41 simulation years (1969-1999 and 2001-2010) over the western United States; these were driven by three datasets: a) NCEP reanalysis data, b) the original CCSM data, and c) the bias-corrected CCSM data. The data analyses revealed that the bias-corrected CCSM data led to significantly improved WRF predictions of precipitation in all four seasons in comparison to the simulation forced with the original CCSM data. Future predictions (2056-2065 and 2090-2099) were also carried out, and significant differences were found in precipitation magnitude, trend, and variability between the WRF simulations forced with (a) the original and (b) the bias-corrected CCSM data.