Abstract: At present,general circulation models (GCMs) can represent the main features of the global atmospheric circulation reasonably well,but their capability in reproducing regional scale climatic details is rather limited owing to their low-resolution.As a result,there is a need to develop tools of downscaling GCM predictions of climate change to regional and local scales.Statistical downscaling are the main tools for downscaling.Statistical downscaling method is to build relation between large-scale climate elements (atmospheric circulation index) and regional climate (temperature,precipitation).Then,this relationship is tested by independent observation data.Finally,this relation is applied in the large-scale climate model to output information and predict the change trend of regional climate.The dynamic extended range forecast (DERF) has good prediction capability which provided the monthly average of ensemble forecast.Many studies have been made to statistically downscale large scale climatic information to regional level in East China by use of DERF's data.This paper focuses on the research of statistical downscaling method and its application over Shaanxi Province,China.Statistical downscaling method based on stepwise linear regressions and common empirical orthogonal functions (EOF) is applied in this paper.It has two predictors in this research.One is the monthly temperature data obtained at 96 observation stations in January and July from 1982 to 2015,which is provided by Shaanxi Meteorological Administration.The other is climatic predictors which are derived from the NCEP/NCAR reanalysis data in January and July from 1961 to 2015,and the dynamic extended range forecast (DERF) product from National Climate Center of China were applied to predict temperature by using the forecast model and compared with the observed temperature.It is built prediction model between the large scale climate predictor and the observed temperature by use of statistical downscaling.The main results of this research can be summarized in the following items:EOF as for large scale climate main modal method and sea level pressure and altitude difference field as predictors are feasible;In January,the consistent anomaly symbol rate of simulated values and observed values greater than 60% reached 50 counties of the province's 96 counties,and the rate of most of the northern Shanbei,Guanzhong and the southern Shannan has reached 70% or so.In July,the consistent anomaly symbol rate of simulated values and observed values greater than 60% reached 60 counties of the province's 96 counties,and the symbol rate of Western Guanzhong and the southern Shannan reached 75% or so.

Key words: monthly temperature, statistical downscaling, climate change, prediction model