不同类型水文模型的参数无资料区域化对比OA

Flood prediction in ungauged basins has become a key focus and challenging issue in current hydrological research.The core of this problem lies in the determination of hydrological model parame-ters,and parameter regionalization is the mainstream method to address this issue.Affected by multiple factors such as the complexity of basin underlying surface conditions and the diversity of hydrological model types,there are significant differences in the applicability of different parameter regionalization methods across various basins.Therefore,in practical applications,the rational selection of hydrological models and adapted parameter regionalization methods are crucial for improving the accuracy of flood pre-diction in ungauged basins.For this purpose,six sub-basins of the Dongjiang River Basin in Guangdong Province were selected as the research objects.Based on the lumped hydrological model(Xin'anjiang Model)and the distributed hydrological model(Liuxihe Model),this study investigated the applicabili-ty,advantages and disadvantages of three parameter regionalization methods(spatial proximity method,physical attribute similarity method,and random forest-multivariate regression fusion method).The re-sults showed that the spatial proximity method was optimal in the Xin'anjiang Model,while the random forest-multivariate regression fusion method was superior in the Liuxihe Model.Moreover,the Liuxihe Model achieved superior overall performance to the Xin'anjiang Model under all scenarios,its Nash-Sutcliffe efficiency(NSE)and Kling-Gupta efficiency(KGE)increased by 0.05-0.19 and 0.06-0.20,respectively.In summary,the distributed hydrological model possesses distinct practical advanta-ges in the application to ungauged basins;however its parameter quantification method should be deter-mined according to model characteristics.