摘要 : Monthly scenarios of relative humidity (R H</sub>) were obtained for the Malaprabha river basin in India using a statistical downscaling technique. Large-scale atmospheric variables (air temperature and specific humidity at 925 mb... 展开 Monthly scenarios of relative humidity (R H</sub>) were obtained for the Malaprabha river basin in India using a statistical downscaling technique. Large-scale atmospheric variables (air temperature and specific humidity at 925 mb, surface air temperature and latent heat flux) were chosen as predictors. The predictor variables are extracted from the (1) National Centers for Environmental Prediction reanalysis dataset for the period 1978–2000, and (2) simulations of the third generation Canadian Coupled Global Climate Model for the period 1978–2100. The objective of this study was to investigate the uncertainties in regional scenarios developed for R H</sub> due to the choice of emission scenarios (A1B, A2, B1 and COMMIT) and the predictors selected. Multi-linear regression with stepwise screening is the downscaling technique used in this study. To study the uncertainty in the regional scenarios of R H</sub>, due to the selected predictors, eight sets of predictors were chosen and a downscaling model was developed for each set. Performance of the downscaling models in the baseline period (1978–2000) was studied using three measures (1) Nash–Sutcliffe error estimate (E f</sub>), (2) mean absolute error (MAE), and (3) product moment correlation (P). Results show that the performances vary between 0.59 and 0.68, 0.42 and 0.50 and 0.77 and 0.82 for E f</sub>, MAE and P. Cumulative distribution functions were prepared from the regional scenarios of R H</sub> developed for combinations of predictors and emission scenarios. Results show a variation of 1 to 6% R H</sub> in the scenarios developed for combination of predictor sets for baseline period. For a future period (2001–2100), a variation of 6 to 15% R H</sub> was observed for the combination of emission scenarios and predictors. The variation was highest for A2 scenario and least for COMMIT and B1 scenario. 收起