| description abstract | statistical downscaling model was developed with reanalysis data and applied to forecast northern China summer rainfall (NCSR) using the outputs of the real-time seasonal Climate Forecast System, version 2 (CFSv2). Large-scale climate signals in sea level pressure, 850-hPa meridional wind, and 500-hPa geopotential height as well as several well-known climate indices were considered as potential predictors. Through correlation analysis and stepwise screening, two ?optimal? predictors (i.e., sea level pressure over the southwestern Indian Ocean and 850-hPa meridional wind over eastern China) were selected to fit the regression equation. Model reliability was validated with independent data during a test period (1991?2012), in which the simulated NCSR well represented the observed variability with a correlation coefficient of 0.59 and a root-mean-square error of 18.6%. The statistical downscaling model was applied to forecast NCSR for a 22-yr period (1991?2012) using forecast predictors from the CFSv2 with lead times from 1 to 6 months. The results showed much better forecast skills than that directly from the CFSv2 for all lead months, except the 3-month-lead example. The biggest improvement occurred in the 1-month-lead forecast, in which the hit rate increased to 77.3% from 45.5% in the CFSv2 forecast. In the forecast of rainfall at 15 stations, the statistical downscaling model also showed superior capability when compared with the CFSv2, with forecast skill being improved at 73% of stations. In particular, 13 of 15 stations obtained a hit rate exceeding 55%. | |
