Study And Analysis On Terrain Features Extraction And Runoff Prediction In Jinghe River Basin Based On GIS

Geographic Information System (GIS) plays an important role in the hydrologicalanalysis in many areas, because it makes geographical factors and hydrological analysisintegration. On the basis of summing up the various research results,this paper focus on theextraction and analysis of topographical features and river network in Jinghe River basinbased on the ArcGIS software platform and SRTM_DEM data. According to the data ofhydrologic stations, several interpolation methods were used to study on spatial interpolationof the average rainfall of many years. On this basis, the forecast model was established topredict the runoff series with GA-BP artificial neural network and grey system model. Thepaper do analyze and research from the view of watershed.The main conclusion are asfollows:(1) From the view of watershed, the microscopic and macroscopic terrain featureextraction and visual expression of Jinghe River basin were in ArcGIS platform.The analysisshows that most parts of the basin in low altitude below1800m.On aspects of the terrain, thenorthwest is higher than other place, and the southeast region is lowest. This is consistent withthe watershed actual terrain.The slope of most parts is less than25°, while steep slopes andsteeper slopes are few, and most of basin is gentle.But mountain and valley is a lot,so someareas display rugged terrain. From the data of Relief Amplitude, the hills is more, followed bythe small fluctuation mountain,while plains and terraces are slightly less. Obviously, Basinterrain with geomorphic characteristics of the Loess Plateau.(2) SRTM_DEM data can be used to extract drainage network, and when the thresholdvalue is5000(40.5km2) will reflect the regional water system better. Then sub-basin and thebasin boundary were obtained. The results show that the data is applied to extract watershedcharacteristics with high precision and accuracy for the digital watershed, water resourcesplanning and development, to build a distributed hydrological model application ofgeographic information.(3) Using the inverse distance weighting method, radial basis function method, krigingand co-kriging, these interpolation methods were respectively used to carry out spatialinterpolation of the average rainfall of many years based on20weather stations which in and around Jinghe River Basin.After the results cross-validation,the error analysis shows thatkriging has the smallest root mean square error of interpolation and the best interpolationeffect.Compared to other methods,the result can reflect the spatial distribution characteristicsof the average annual precipitation better. The precipitation of spatial distribution changes alot,and the south and southeast obviously more than the north and northwest. Distribution isobvious gentle gradient-like.(4) Based on the above study and analyze, the Jinghe River Basin is divided into foursub-basin, and using kriging to spatial interpolation of the average annual precipitation from1961to1996. The GA-BP neural network and gray system model are respectively establishedbased on the results of the above study and Measured data of rainfall and runoff. The relativeerror of the GA-BP model predicted values were-11.07%,0.19%,-3.39%,and the GM (1,6)model relative error were-0.96%10.66%-5.86%. The result demonstrates that the twomodels have high prediction accuracy. The former applies to the medium-and long-termsequence of forecast, while the latter advantage is more obvious in the short time seriesprediction, both of which provide a good way to predict runoff and improve accuracy. Butrunoff is extremely complex, so the influencing factors of runoff changes should be deepanalysis for further grasp sequence variation and intrinsic properties. The applicability of themodel under different conditions is subject to further examination.