Estimation Of Runoff And Sediment Load And Analysis Of Driving Factors In Midstream Of Yellow River By Artificial Neural Networks

It is of great significance to investigate the variation of runoff and sediment load in Yellow River Basin for water resources management and water environment remediation.Both runoff and sediment load have significant changes in the past few years due to humen activitities.In general,it is difficult to construct an accurate sediment load estimation model because sediment concentration is affected by many complex factors.On the other hand,the artificial neural network(ANN)is widely used in hydrological prediction over the last two decades.Therefore,the ANN is selected to build the sediment load estimation model and to evaluate the influence of driving factors on sediment load based on its powerful learning ability.It is important and valuable to quantitatively evaluate the changes of runoff progresses and sediment load for the management of sediment load water resources in Yellow River Basin.The static backpropagation neural network(BPNN)-and dynamic recurrent neural network(RNN)are constructed to construct the sediment load estimation in the Huangfuchuan basin according to the basis of precipitation-streamflow-sediment load relations.Moreover,the performance of model outputs and the relative errors are further analyzed.The results indicate that:1)Both BPNN and RNN models perform well in sediment load estimations and the CE values are higher than 0.82;2)The accuracy of peak sediment discharge simulation obtained from RNN is more close to actual observations and the relative error of peak sediment discharge estimation is within 2%;3)The performance of BPNN is significantly better than that of RNN during recession periods.The impacts of sediment trapping dam and vegetation types on runoff and sediment load are also investigated based on hydrologic analysis,double mass curve and ANN methods in this study.The results of Mann-Kendall test and accumulative annual anomaly show that there are two significant turning points(1979 and 1999)for both runoff and sediment load during the periods of 1956-2012.Thus,it can be divided into three phases,namely 1956-1979,1980-1999 and 2000-2012.Taking the periods of 1956-1979 as benchmark,the effect of soil and water conservation on runoff and sediment load reduction reaches to 23.1%and 22.8%respectively in the periods of 1980-1999,and the values are about 66.7%and 71.7%respectively in the periods of 2000-2012 based on accumulative annual anomaly.Furthermore,two types of ANN structures with ensemble skill are also built to analyze the impacts of sediment trapping dam and vegetation cover on the runoff and sediment load according to the storage capacity of the dam and NDVI collected in 2010.The results indicate that the vegetation change is the main driving factor that significantly influence the variation of runoff and sediment load in Huangfuchuan basin.The impact of vegetation change on runoff and sediment load is about 70%?83%and 70%?82%,respectively;whereas the impact of sediment trapping dam on runoff and sediment load is about 17%?30%and 18%?30%,respectively.According to the analysis of one-way ANOVA,it can be detected that the effect of forestland and grassland vegetation on runoff and sediment reduction is significant,demonstrating that forest-grass protection have positive effects on water and soil erosion.