| Floods occur frequently in medium and small watersheds,with short durations and large instantaneous flows,especially in mountainous basins.When heavy rain occurs,it is easy to cause damage to downstream and cause geological disasters and increase flood risk.Flood forecasting and decreasing are important measures for flood control and disaster mitigation in medium and small watersheds.In this paper,for the medium and small watershed,the flood forecasting intelligent networking model suitable for the data scarcity basin is proposed,and the model application and verification are carried out.Exploring the space driving factors of flood in the basin and comprehensively assessing the potential and impact of natural flood management(NFM)measures.The specific research contents and results are as follows:(1)The flood intelligent network forecasting model for medium and small watersheds with scarce data is researched and constructed.According to the convergence characteristics,the watershed is divided into several sub-basins.Based on BR-ANN,sub-basin flood forecasting models with runoff data is established.Based on GRNN,a sub-basin flood parallel prediction model with no runoff data is established.Taking the results of sub-basin prediction model as the boundary condition,the Muskingum-Cunge method is used to connect the catchment sub-basins to form a confluence network to calculate the downstream flood process.(2)The typical area is taken as the research object to evaluate the performance of flood intelligent network forecasting model in medium and small watersheds.Divide the Rujigou watershed into 13 catchment sub-basins,identify key forecasting factors and forecast the flood process.The results show that the forecast results of the flood intelligent network forecasting model are better,with KGE’ of 0.88,NSE of 0.982,and flood peak deviation of 7.14%,which is better than linear forecast and BP-ANN nonlinear forecasting model.The model can predict the storm flood in medium and small watersheds with relatively scarce data.(3)Using the Geographical Weighted Regression(GWR)model to study the driving forces of flood space in medium and small watersheds.The hydrological response unit of the basin is divided and used as spatial sample.The rainfall confluence coefficient is used as the dependent variable,and the natural factors such as rainfall,temperature,area,slope,elevation,leaf area index,potential evaporation and soil infiltration capacity are used as explanatory variables to conduct regression analysis.The results show that the mean regression coefficient of rainfall is 0.70.The leaf area index has a negative driving relationship for storm floods.The mean regression coefficient is-0.19.The larger the leaf area index,the smaller the rainfall convergence coefficient.Under the spatial small-scale analysis,other factors have less driving effect on floods.(4)Study and evaluate the reduce measures of natural flood management in medium and small watersheds.Based on the concept of natural flood management(NFM),the measures such as afforestation,contour planting,slope to terrace,river promotes infiltration and rainwater collection and infiltration enhancement to analyze in the study area.For the flood of 2006,afforestation can reduce the flood peak by3.39%;the contour planting can reduce the flood peak by 6.83%;the flood reduction ability of slope to terrace measures is strong,which can reduce the flood peak by16.27%;the river promotes infiltration can reduce the flood peak by 4.2%;The rainwater collection and infiltration enhancement can reduce the flood peak by 6.4%;the combined measures are better than the single measure,and the flood peak can reach more than 25%.And medium and small natural flood management can also improve the river basin ecological environment. |
