An application of multivariate regression model to predict precipitation
based on GIS in the Heihe river basin
-
-
Abstract
Based on precipitation data collecting at 21 stations from 1971 to 2000 and five topographic factors (altitude, slope, aspect, longitude and latitude) acquiring from three different resolution digital elevation model (DEM), the multivariate regression analysis, combined with GIS, was used to develop a precipitation prediction model for the Heihe river basin. The results of this study showed that the multivariate regression model explained 74.5% of the spatial variability of precipitation over the whole year, and this model had better explanation precipitation for wet season (MaySeptember) than the whole year and dry season. Precipitation during dry season was difficult to simulate owing to little rainfall and a different synoptic system. The 100 m resolution model in the three periods were better than other resolution model to explain the precipitation because the spatial distribution of precipitation was uneven. The 100 m resolution model predicted that the precipitation increased from below 200 at the northwest regions to 700 mm at southeast regions, indicating that a precipitation line exit was observed from northeast to southwest. The 500 m resolution model predicted that the rainfall was ribbon boundaries with a certain degree shift. The 1 000 m resolution model predicted rainfall distribution with a big error. The model established in this study could be potentially applied to other mountains; however, improving the model accuracy was necessary in the future.
-
-