Mid-long Term Hydrological Forecast Of Snowmelt Runoff In Western Mountainous Area Of Tianshan Mountains

In order to rationally guide the sustainable development and utilization of water resource in glacier snowmelt area for the sake of better support of development of industry and agriculture snowmelt watershed,medium and long-term hydrological forecasting research of snowmelt runoff has been explored in Kashi river basin where located at west Tianshan mountains.Based on the rainfall,temperature,runoff data from the Wulasitai Hydrological Station,the Kashi River's hydrological characteristics were analyzed by wavelet analytical and wavelet correlation analytical method;This correlation coefficient method,PCA and mutual information are selected to get the optimized predictors,respectively;and using BP neural network model,RBF neural network model and combination wavelet BP neural network model to study the runoff forecast in study basin;Based on the meteorological data of CMIP5 under different emission scenarios,the runoff changes in the future study area are estimated.Based on the wavelet analysis of the runoff,precipitation and temperature of the Wulasitai Hydrological Station,the periodic characteristics of each sequence are obtained.It is found that the main cycle of the runoff occurs at 10 and 26 years.The main cycle of precipitation is 22 And 26 years;the main cycle of temperature occurred in 22 years.The correlation among temperature-runoff series,and precipitation-runoff series are both well by analyzing wavelet correlation respectively.Furthermore,rainfall has greater impact on runoff and directly determines river runoff volume,meanwhile temperature plays a role of peak load regulation and frequency modulation on runoff.This correlation coefficient method,PCA and mutual information are selected to get the optimized predictors,respectively,and using BP neural network model,RBF neural network model and combination wavelet BP neural network model to study the runoff forecast in study basin for getting 12 kinds of prediction results.(1)In the prediction results of BP neural network model,the average relative error of the mutual information method is 0.18% lower than that of the whole factor and 0.98% relative to the correlation coefficient,which is 4.85% lower than that of the principal component analysis method.(2)In the prediction results of the combination wavelet BP neural network model,the average relative error of the mutual information method is 4.82% lower than that of the whole factor method,which is 0.64% lower than the correlation coefficient method and 4.31% less than the principal component analysis method.(3)RBF neural network prediction results: mutual information method to predict the average relative error relative to the total factor reduction of 7.53%,relative to the correlation coefficient method to reduce 2.29%,compared with the principal component analysis decreased by 26.52%.The relative error is regarded as the standard of the accuracy of the model,and the best forecasting result is the model predictive result based on the mutual information method which is selected as the input data of RBF neural model.Based on the measured meteorological data and the simulated meteorological data of three different models in CMIP5,the mutual information method was used to select the forecasting factor and the RBF neural network model was used to forecast the runoff from 2006 to 2100.The results showed that the future runoff was in 2006~2030 Year will continue to rise,stabilized in 2060;future runoff in the non-flood season has increased substantially in the flood season to reduce the runoff.The main factors affecting the runoff in different scenarios under different modes are analyzed.The future changes of the relevant factors are analyzed.It is found that the main reason for the decrease of runoff in non-flood season and the decrease of runoff in flood season is: The precipitation in the flood season increased and the evaporationdecreased or increased slightly.The precipitation decreased with the increase of the temperature and the runoff of the flood season decreased.