论文部分内容阅读
地理加权回归克里金(GWRK)是在地理加权回归(GWR)基础上扩展得到的一种既能考虑回归关系的空间非平稳性又能考虑回归变量空间自相关性的降水数据融合方法。以赣江流域为例,在评价TRMM卫星数据精度的基础上,分别以GWRK和GWR方法构建了站点-卫星降水数据融合模型,然后采用降水融合数据驱动GR4J水文模型进行水文预报。根据站点尺度降水融合数据精度及水文预报表现,对GWRK和GWR构建的降水融合模型效果进行评价,结果表明:较之GWR方法,GWRK方法能较明显的提高降水融合数据在站点尺度上的精度,但是由于输入到水文模型中的数据为面降水数据,受空间均化的影响,对水文预报精度的提高不如对站点尺度降水融合数据精度的提高明显。
Geographic Weighted Regression Kriging (GWRK) is a precipitation data fusion method based on the geo-weighted regression (GWR), which can not only consider the spatial inhomogeneity of regression relationship but also consider the autocorrelation of regression variables. Taking Ganjiang river basin as an example, based on the evaluation of the accuracy of the TRMM satellite data, a data fusion model of station-satellite precipitation was constructed by using GWRK and GWR methods, respectively, and then the precipitation data were used to drive the GR4J hydrological model for hydrological forecasting. According to the accuracy of precipitation fusion data and the performance of hydrological forecast, the effect of precipitation fusion model constructed by GWRK and GWR is evaluated. The results show that compared with GWR method, GWRK method can significantly improve the accuracy of precipitation fusion data at the site scale, However, since the data input into the hydrological model is surface precipitation data, the accuracy of hydrological forecasting is not as good as that of the site-scale precipitation fusion data due to the influence of spatial averaging.