A Study On The Ensemble Assessment Techniques Of Runoff And Sediment Changes In The Yellow River Basin

Research on the runoff and sediment changes in the Yellow River Basin has always been a hot issue.In recent decades,due to the combined effects of human activities and climate change,the measured runoff and sediment transport in the middle reaches of the Yellow River have a significant reduction.Among them,the impact of climate change is mainly reflected in changes in precipitation,temperature,evaportranspiration and other factors;the impact of human activities mainly includes some hydraulic engineerings and some soil and water conservation measures.Most of the studies believe that human activities are the main driving forces for the reduction of water and sediment in the Yellow River,but there is no consensus on the quantitative study of the contribution rate of human activities to water and sediment changes.Taking the Huangchuanchuan watershed,a tributary of the middle reaches of the Yellow River,as an example,the Mann-Kendall trend test was used to analyze the trend of the annual precipitation,annual runoff and annual sediment transport in the basin,and the precipitation sequence was not significantly changed,while the runoff sequences and sediment transport sequences showed a significant decreasing trend.The variation point of the two series tested by the Pettitt test method was 1989.Taking 1954–1989 as the based period and 1989–2015 as human impacted period,the contribution of human activities and climate change were analyzed by hydrological method,double mass curve method,non-parametric elastic coefficient method and hydrothermal coupled equilibrium equation method which is based on the Budyko hypothesis.The results of the above methods indicate that human activities are the dominant factors in the reduction of runoff and sediment in Huangfuchuan.In the meanwhile,human activities account for 80–93% to runoff reduction,and 80% to the reduction of sediment transport.In addition,models based on machine learning are used to study the variation of water and sediment change in Huangfuchuan basin,which include three methods: multiple linear regression,k-nearest neighbours regression and support vector regression.After selecting different combinations of independent variables and different time scales,the optimal models in various cases are chosen respectively.When studying runoff,the multivariate linear regression model and the k NN regression model of annual scale are the better models for studying the runoff variation in the Huangfuchuan Basin.When studying sediment transport,the monthly scale k NN regression model is superior to the other two models.The vector regression model is not suitable for studying the monthly sediment transport in the Huangfuchuan Basin.Based on the research of empirical models and machine learning models,and referring to the results of a SWAT model in Huangfuchuan basin,a complete indexes system is constructed to evaluate the applicability of various types of models in Huangfuchuan.The results show that the monthly k-nearest neighbours regression model performs better in both regular and verification periods.It is a better choice to study the change of runoff and sediment transport in Huangfuchuan.The SWAT model is second,and the empirical models performed the worst during the verification period and is not suitable for simulating future runoff and sediment changes.