Abstract: Climate change is a global issue of common concern to the international community today. Over the past century, great changes characterized by climate warming have taken place in both global climate and eco-environment. In recent years, with the implementation of NSFC major research plan, integrated study on eco-hydrological processes of Heihe River Basin(HRB)has become a hot spot of arid and semi-arid area research in China and great achievements have been made. Future climate change forecasting is one of the foundations of the research, but currently no forecasting has been made based on Global Climate Models(GCMs)and statistical downscaling methods. Such forecasting is in urgent need. Monthly meteorological monitoring data of the reaches and NCEP reanalysis data from 1960 to 2014, together with CNRM-CM5 data from 2006 to 2060 were used for the study. The research process of the paper is as follows:Firstly, statistical downscaling models for east watershed of HRB based on multivariable linear regression using the NCEP reanalysis data and meteorological monitoring data of air temperature and precipitation were built. Secondly, since the models cannot meet forecasting requirements, new complimentary statistical downscaling models were built based on residual analysis to improve forecasting results. Thirdly, regional applicability evaluation was employed to determine the models used for future climate forecasting based on GCM of Coupled Model Intercomparison Project Phase 5(CMIP5). Finally, future climate change prediction for HRB was made based on the chosen downscaling models and CNRM-CM5 model of CMIP5, and both time variations and regional distribution of future air temperature and precipitation of HRB were analyzed. The results show as follows:(1)The complementary downscaling models simulated better than the multivariable regression model overall, and the simulation results of air temperature were better than that of precipitation for both the two models.(2)Regional applicability evaluation of downscaling models based on the CNRM-CM5 model show that, the applicability of the complementary regression models were better than that of the multivariate regression models in general under both RCP4.5 and RCP8.5 pathways.(3)The future average annual precipitation of upstream and midstream of HRB were 324.94 mm and 330.15 mm under both RCP4.5 and RCP8.5 pathways respectively, and inhomogeneity enhancement was confirmed in the future.(4) The future average annual air temperatures of midstream and downstream of HRB under the two pathways were 10.25℃ and 10.77℃, respectively. It should be clear that, because of the uncertainty of the data, downscaling procedures, as well as the diversity of human emissions, uncertainty exists in the prediction of future climate change in HRB.

Key words: climate change, statistical downscaling model, CMIP5, meteorological elements, HRB