| Accurately deducing the spatial distribution of precipitation is an important scientific research target in the field of hydrometeorology.At the same time,it is also an important basis for hydrological analysis and prediction,natural disaster prevention and other work.Traditional ground-based rainfall station network observation and surface rainfall radar observations have limited their application in the hydrometeorological field due to their respective defects.At present,the ground rainfall station network measured data,radar rain products and satellite telemetry precipitation information are merged into hydrology.One of the research hotspots in the field.This paper mainly studies the fusion method of precipitation data in urban areas.Due to the small spatial area of urban areas and the low spatial resolution of satellite telemetry rainfall products,it is not suitable for precipitation data fusion research in small-area urban areas.Therefore,this paper focuses on ground rainfall stations.The fusion data of two kinds of sources,net measured data and radar rain data,are combined to study the fusion method to improve the accuracy of spatial precipitation estimation in urban areas,and provide more accurate data support for urban hydrometeorological research.This study takes Zhengzhou City,Henan Province as an example,and obtains measured precipitation data and radar rainfall data from Zhengzhou City Ground Rainfall Station.Firstly,it is based on geographic weighted regression(GWR)and geographically weighted regression Kriging(GWRK).Source precipitation information for fusion studies.Secondly,based on the GWR model,based on an improved Kriging method,namely the improved geographically weighted regression Kriging(GWROK),the precipitation data from two different sources are merged.Finally,the cross-validation and comparison analysis of the estimation results of the three fusion methods shows that the improved geo-weighted regression Kriging method(GWROK)has the highest accuracy and the smallest error.The main research contents are as follows:(1)The fusion of precipitation data from different sources based on geographic weighted regression(GWR).The radar rain data is used as the precipitation background field,and the ground measured data is used as the precipitation observation field.Under the “addition model” framework,the precipitation background error can be approximated as the difference between the precipitation observation value and the precipitation background value.Based on the GWR model,the regression relationship between the precipitation background error and the latitude and longitude of the observation point is established.The precipitation background error and the precipitation background value are added to obtain the GWR model precipitation data fusion result.(2)Interpolation estimation of GWR regression residuals based on improved Kriging method.In this study,the Gaussian model is used to fit the experimental variogram in the Kriging method.Based on this theory,the model parameters are introduced to replace the parameter 2 in the original Gaussian model,the value of the change,and the nugget value is calculated.And the root mean square error RMSE,and finally choose the value that meets the requirements of the variogram.The results show that when taking between 0.5 and 1.5,the RMSE value is the smallest,and Nugget is close to 0,which is consistent with the principle that the variogram requires less error and the nugget value is smaller.(3)Accuracy analysis and evaluation of the precipitation estimation results of the three fusion methods.For the comparative analysis of the precipitation data fusion results of the above three models and the ground measured precipitation data,the cross-validation(CV)method is used to generate the accuracy evaluation index parameters: average absolute error MAE,root mean square error RMSE and system.Deviation BIAS.The final results show that the precipitation data fusion result of GWROK model has the smallest error and the highest precision. |
PDF Full Text Request