Abstract: Considering the lack of spatial and temporal accuracy, poor reliability of global climate models (GCM) at regional or local scale, a statistical downscaling model (SDSM) is used to generate daily weather data. With SDSM, the daily series of HadCM3 outputs under the A2 and B2 scenarios were downscaled to the stations. The future trends of annual precipitation, daily maximum (TMAX) and mininum temperatures (TMIN) in the upper-middle reaches of the Yellow River were analyzed on the basis of the reference period of 1961-1990. Daily TMAX and TMIN both show increasing trends under the A2 and B2 scenarios, but the increment is more distinct under the A2 scenario. The increasing trends of TMAX at Jingtai station and TMIN at Hequ station are the most remarkable. The increment in annual precipitation in the basin ranges from -18.2% to 13.3%. Under the A2 scenario, the areas with an increasing trend and a decreasing trend are almost equal, and the largest increment occurs at Baoji station. Under the B2 scenario, the annual precipitation shows a decreasing trend in most parts of the basin, and the largest reduction appears at Xifengzhen station.

Key words: the Yellow River basin, statistical downscaling, maximum temperature, minimum temperature, precipitation