Impacts Of Spatial Interpolation Methods For Precipitation On Hydrology Simulation At Different Time Scales

The areal precipitation is the main drive force for hydrology models.Accurate estimation of areal precipitation is the prerequisite for the reliable hydrology simulation and prediction.At present,the spatial distribution density of precipitation gauging stations is low and the distribution is not uniform in most basins of China,leading to the impossibility of obtaining accurate observed areal precipitation data in whole basins.The spatial interpolation method,which can estimate areal precipitation data based on limited observed rainfall data,is usually an effective method to solve this problem.While the different alternative interpolation methods may cause uncertainty in areal precipitation estimation at one time scale,and the accuracy of one spatial interpolation method may also vary at different time scales.Therefore,the applicability of spatial interpolation methods for precipitation at different time scales and their impact on the accuracy of hydrology simulation are worth further research.The upstream of the Changjiang river basin and the upstream of the Hanjiang river basin were selected as case studies.The results and conclusions are as follows:(1)The spatial statistical characteristic values of precipitation were firstly calculated to analyze the spatial distribution pattern of precipitation at daily,monthly and yearly scales.Then the spatial autocorrelation and high/low clustering of precipitation were detected using the global Moran index,the global G and local G~*indexes,respectively.The results indicate that:(1)The spatial variability of precipitation is reduced as the time scale increases at the upstream of the Hanjiang river basin,but the spatial variability for the daily precipitation is the highest while for the yearly precipitation is the lowest at the upstream of the Changjiang river basin.(2)Significant spatial positive autocorrelation were detected in two basins at three timescales.And the autocorrelation degree of precipitation is yearly>monthly>daily.(3)High value clusters were also detected in two basins at three timescales.The cluster degree of precipitation is increased as the time scale increases at the upstream of the Changjiang river basin,while the cluster degree of the yearly precipitation is not significant at the upstream of the Hanjiang river basin.(2)The cross-validation method was adopted to study the applicability of seven spatial interpolation methods in the spatial interpolation for the daily,monthly and yearly precipitation,respectively.The results of two study areas similarly show that:(1)The correlation coefficients between the measured precipitation and the estimated precipitation by seven spatial interpolation methods increase as time scale increases.(2)An spatial interpolation method may have different performance at different time scales,and there is no spatial interpolation method which can be superior to other spatial interpolation methods in all three timescales.(3)The best spatial interpolation method for daily precipitation is the Inverse Distance Weighting method,while for the monthly and yearly scales are both the Trend Surface Analysis Combined with Kriging method.(4)The correlation between auxiliary variables and precipitation has a large impact on the accuracy of spatial interpolation for precipitation.If the selected auxiliary variable is not highly correlated with the main variable,the accuracy of spatial interpolation for precipitation will be worse.(3)The SWAT model,"abcd"model and the Budyko-based water balance formula were used to analyze the impacts of different spatial interpolation methods on hydrology simulation at daily,monthly and yearly scales,respectively.The results show that:(1)The spatial interpolation method with the higher accuracy in the cross-validation will improve the accuracy of the hydrology simulation more greatly.(2)The performance differences of the hydrology models with different spatial precipitation inputs are large at the yearly scale,while the differences are smaller at the daily and monthly scales.(3)The total number of points to be interpolated under the same spatial interpolation method has little effect on the performance of the lumped hydrology model at the monthly and yearly scales.