Abstract: The regression equations of the precipitation and the topographical data in the upper-middle reaches of Yangtze basin were established based on the 30 years seasonal and annual average precipitation of 697 meteorological stations, from 1971 to 2000. The topographical data include the elevation, aspect and slope. The regression equations had passed 0.05 F confidence tests. The effect of topographyon precipitation is extract from the total annual and seasonal precipitation. Then, the improved Stepwise Interpolation Approach （SIA）was combined with the GIS technique and the multi-variable successive regression equations. SIA considerably improves the accuracy of the annual and seasonal spatial interpolation. The results show that the relative error of the seasonal precipitation interpolation is 6.86%, the absolute error is 13.07 mm, the average coefficient of variation is 0.070, the average correlation coefficient is 0.9675; the absolute error of the annual average precipitation is 72.1 mm,the relative error is 7.34%,the average coefficient of variation is 0.092，the correlation coefficient is 0.9605.The analysis of the annual mean precipitation in terms of SIA showes that the SIA calculation with 3—5 steps considerably improves the interpolation accuracy, such as decreasing the absolute error form 211.0 mm to 62.4 mm, the relative error from 20.74% to 5.97%, the coefficient of variation from 0.2312 to 0.0761, and increasing correlation coefficient from 0.5467 to 0.9619. The SIA iterative results after 50 steps identically converges at the observed precipitation.