STATISTICAL DOWNSCALING OF GENERAL-CIRCULATION MODEL OUTPUT - A COMPARISON OF METHODS

A range of different statistical downscaling models was calibrated usi ng both observed and general circulation model (GCM) generated daily p recipitation time series and intercompared. The GCM used was the U.K. Meteorological Office, Hadley Centre's coupled ocean/atmosphere model (HadCM2) forced by combined CO, and sulfate aerosol changes. Climate m odel results for 1980-1999 (present) and 2080-2099 (future) were used, for six regions across the United States. The downscaling methods com pared were different weather generator techniques (the standard ''WGEN '' method, and a method based on spell-length durations), two differen t methods using grid point vorticity data as an atmospheric predictor variable (B-Circ and C-Circ), and two variations of an artificial neur al network (ANN) transfer function technique using circulation data an d circulation plus temperature data as predictor variables. Comparison s of results were facilitated by using standard sets of observed and G CM-derived predictor variables and by using a standard suite of diagno stic statistics. Significant differences in the level of skill were fo und among the downscaling methods. The weather generation techniques, which are able to fit a number of daily precipitation statistics exact ly, yielded the smallest differences between observed and simulated da ily precipitation. The ANN methods performed poorly because of a failu re to simulate wet-day occurrence statistics adequately. Changes in pr ecipitation between the present and future scenarios produced by the s tatistical downscaling methods were generally smaller than those produ ced directly by the GCM. Changes in daily precipitation produced by th e GCM between 1980-1999 and 2080-2099 were therefore judged not to be due primarily to changes in atmospheric circulation. In the light of t hese results and detailed model comparisons, suggestions for future re search and model refinements are presented.