Runoff Characteristics Analysis And Prediction Model Of The Huangwei River Basin In The Dabie Mountain

Based on the long time rainfall-runoff series data of the Huangwei river basin, using theMann-Kendall trend test, the cumulative departure method, cramer mutation test and waveletanalysis method to reveal the character such as character within the year and interannual, periodicchange characters and so on; Using GIS and remote sensing technology, atmospheric circulationand climate variation theory of the driving mechanism of runoff change has carried on the deepdiscussion. Rainfall-runoff forecast model based on artificial neural network and support vectormachine(SVM) theory is established, forecasting or simulating the runoff change process.Themain research results are as follows:(1) Analysis about runoff change characteristics: runoff presents decreasing trend overall,mutations in the time domain is1991~1993. multiple time scale structure characteristic ofrunoff is obvious, there are34~40years,11~34years and10years following three cyclicalchange law of scale, and10years following the small scale of cycle changes significantly, centerscale in6years. Runoff distribution is uneven within the year, the summer runoff proportion isthe largest, is almost half of the annual runoff, minimum in the winter, spring runoff proportion islarger than autumn runoff.(2) Analysis of the runoff influence factors: pay attention on analyzing the influence ofprecipitation, land use/cover change, ENSO phenomenon, solar activity on watershed runoff.Results show that the changes of rainfall and runoff series basically show synchronous, almost notime lag effect;Watershed land use change way is given priority to with “woodland-farmland”conversion, construction land and water area is too small, so their influence of watershedhydrological processes negligible, woodland area increased and reduced farmland area willweaken the rainfall runoff correlation, and at the same time causes the decrease of runoff; Runoffchanges in response to ENSO events is statistically relationship, when ENSO events occur, thechange of characteristics of flood and dry must happen, the probability of the flood and thenormal is greater than the probability of the dry. When sunspot activity enhancement, runoff hasincreasing trend, on the other hand, showed a trend of decrease.(3) Analysis of the runoff forecast simulation: For the Huangwei river basin, the overallrelative error of the BP model is14.43%, the percent of pass is77.5%, the deterministiccoefficient is0.76, forecasting accuracy level is b; the overall relative error of the SVM model is12.41%, and the percent of pass, the deterministic coefficient and forecasting precision level isthe same as the BP model, the result is ideal.The SVM model and the BP model close to theaverage level on simulation, but the SVM simulation results compared with the BP model ismore focused on small error range, high precision of runoff simulation ability. BP model of thecumulative error is greater than the SVM model, and with the increase of error of degree offreedom, the gap has a tendency to expand, showing that the SVM model of the error range, smaller error interval are less than BP model, the simulation performance is more stable than BPmodel. Finally use the rainfall runoff series of Bailianya basin adjacent with the Huangwei riverbasin north to verify the performance of model, similar results were obtained.