Analysis of the spatial variability of rainfall in Wuding River Basin
无定河流域降雨量空间变异性研究

The Wuding river basin lies in the middle reaches of the Yellow River, the typical area of the Loess Plateau which is characterizedby serious soil erosion. So the analysis of hydrological factors such as precipitation is invaluable for understanding water cycle process and its impact on the watershed. Geostatistical methods, originating from geology in the 1950s, have been expanded to many other scientific fields for their unique statistical and interpolating advantages. In this paper, we tried and applied ordinary Kriging(OK) technique,one of the geostatistical methods, to analyze the spatial variability of annual precipitation at 67 rainfall stations for this watershedduring the period from 1990 to 1997. Because of the attributes of the rainfall data of which the frequencydistribution tallies with logarithmicnormal distribution, the original data were processed by logarithmictransformation, corresponding thelogarithmic Kriging technique was takenas the analyzing and interpolating method, which made the application of geostatistical technique more convenient and easier,also the analysis more reasonable. The result revealed that the semi-variograms of rainfall were best described by spherical variogram model, suitable to most natural variables, and subsequently it was chosento fit in the needed parameters for interpolation. Fromthe final parameters attained,it was indicated that the rangesof eachyear in the studied timeseries were not the same, even varied greatly. The range of 1991 was up to 20.17 km, whichwas the minimum, while that of 1994 appearedto be 68.65 km, with thecoefficient of variability of the year being the minimal. This consequence can best explain that the period of 1994 was the most evenperiod in the time series for the variability of rainfall in Wuding river basin. In the meantime, studies indicated the existence of thenugget effect in each year; which may be primarily caused by random factors at a spatial scale less than the range. After the analysis of rainfall data usingKriging technique, three-dimensional maps of different years were created for the further study and utilization, which made the spatial variation easy to be discerned. The northeast and northwestparts are characteristic of low variations with low rainfall compared with other regions, while the middle part of the south exhibitshigh fluctuation. If the statistical analysis is combined with other maps such as the relief map or DEM, some other meaningful information can also be acquired.