Estimation of precipitation using altitude and prevailing wind direction effect index in Mountainous region

Spatial explicit data of precipitation contribute to water resource management as well as large-scale eco-hydrological processes modeling. In this paper, spatial interpolation model ANUSPLIN and Geographic Information System techniques were employed to simulate the spatial distribution of mean monthly precipitation in the mountainous region. Upper Stream of Minjiang (USM), where this study was carried out, is a critical area in Southwest China. USM locates between the east Tibet plateau and Sichuan basin, covers 22919km~2 watershed area. Precipitation data were acquired through 51 rain observation gauges for the period of 1988~2002. Two independent variables and two covariates were used in simulating surfaces of monthly precipitation. Independent variables x, y are position information of the gauges in a projected coordinate system of Beijing 1954. Topographic descriptor DEM and Prevailing Wind-direction Effect Index (PWEI) were extracted as covariates. The latter were established by linking the interacting effects of aspect and prevailing wind direction (i.e. wind direction of monsoon). All variables were calculated and discretized as raster surfaces with 500 m resolution. Thin plate smoothing spline surfaces for monthly precipitation were fitted on the basis the gauged precipitation points, which enable the regular gridded precipitation data set to be produced by coupling spline surfaces with DEM and PWEI. Generalised Cross Validation (GCV) is applied to determine smoothing parameter that can be used to minimize the prediction error. Mean monthly precipitation varied significantly throughout the year. The highest mean monthly precipitation 174.05mm occurred in August and the lowest 7.98mm occurred in December. Spatial distribution patterns of precipitation appeared to be high topographic relevant. The higher precipitation occurred in the southeast upslope that exposed to the prevailing moisture-bearing wind, and the lower occurred in the north plateau and in the middle dry valley. Scatterplots of observed versus predicted precipitation showed overestimate in the upper end of the range and underestimate in the lower end of the range in dry season (October to April). In contrast, wet season (May to September) appeared to be well predicted. Narrow range of the value in May and June indicate there were no remarkable annual change of precipitation at the beginning of wet season. Statistics showed the prediction errors of monthly precipitation vary between 15.02%~42.06%. The prediction errors in wet season varies between 15.02%~21.04% and dry season 17.71%~42.06%. Lower prediction errors in wet season indicated stronger effect of prevailing wind-direction of monsoon than dry season. Better performance occurred when simulating has been done with PWEI than without it. Annual mean prediction error can be reduced by 3.0 percent when applying PWEI.