Mississippi River Climate and Hydrology Conference

Tuesday, 14 May 2002: 10:50 AM
Evaluation of seasonal predictive skill of a regional climate model
Francis Ochieng Otieno, Iowa State University Geology and Atmospheric Science, Ames, IA; and W. Gutowski, E. S. Takle, C. J. Anderson, Z. Pan, and R. W. Arritt
Regional climate models (RCMs) are among the tools used for downscaling GCM climate simulations to regional scales. The need for information on regional scales is spurred by the large variations in regional response to global climate variability and the fact that regional forcings often are nonexistent or represented poorly at GCM resolutions. To determine the bounds of confidence in regional climate projections, there is need to characterize the strengths and weaknesses of RCMs.

In this study we evaluated the skill of a nested regional climate model (RegCM2) in predicting seasonal precipitation using a number of statistical assessment tools including self-organizing maps (SOMs). The model data used consisted of output from the 10-year runs of RegCM2 for a U.S. continental domain (same as used for PIRCS experiment 1) driven at the lateral boundaries by the NCEP/NCAR reanalysis. For evaluation, observed data from the reanalysis, Higgins, CMAP and VEMAP precipitation and UNH/GRDC runoff estimates for the 1979-88 period were used.

Results showed that RegCM2 performs well for some regions but has large precipitation deficit in a region of the south-central United States (SCUS, 31-37oN, 85-95oW) comprising the lower Mississippi. This deficit had an annual cycle with the largest deficits occurring in late fall and early winter. The model identified SCUS rainy events, defined as a day or series of consecutive days with at least 1mm/day average precipitation. High spatial correlations between simulated and observed precipitation were found for most of the United States for the 10-year period. The main difference in the SCUS precipitation was that the magnitude of simulated precipitation events were usually smaller than observed. An atmospheric and terrestrial water budget analysis showed less than 10% deficit in precipitable water but substantial errors in model evapotranspiration during the fall season. The precipitation deficit was least when model evapotranspiration was almost balanced by model precipitation with little runoff and small (<1 mm) change in storage during the summer.

Supplementary URL: