River Runoff Characteristics Analysis And Prediction In The Upper Reaches Of The Yellow River

River runoff is not only represent water resource quantity, but also accounts for a large proportion in the water resources system. In2011, No.1Document from the national central government pointed out that the strictest water resource management system should be taken. Bulletin of first national census for water in2013showed that a half of rivers basin area of100square kilometers disappeared. Therefore, research of the runoff change law and prediction plays an important role in the human sustainable development and national economy. We can timely understand drought and flood of rivers, and prevent the decrease of river, which provids the evidence for the development and utilization, rational allocation of water resources.The upper Yellow River water amount accounts for54%in the whole Yellow River basin, and undertakes the task of water supply of several provinces in the north China, and has a great influence on the development of economic and social. So it is very important to study the change rules and runoff prediction of the Yellow River. Using the methods of mathematical statistics, the Hilbert Huang transform, the natural runoff data of the Guide station, Xiaochuan station and Lanzhou station in the upper Yellow River from1920to2003were studied to analyze its characteristics and put forward a kind of runoff combination forecast method based on empirical mode decomposition. To analyze the interannual variation and annual variance of the runoff, the non-uniformity coefficient, completely adjustment, the degree and time of concentration, variation coefficient and curve deduction mass had been calculated based on the runoff data. The results showed that annual runoff distribution was not uniform in the upper Yellow River from1920-1989, relatively uniform in the90’s, and inter annual change was not obvious, high flow year and dry year appeared alternately. Runoff mutation and trend of these three hydrologic stations were analyzed by Mann-Kendall rank correlation test, moving average method and Rescaled Range Analysis(R/S) method. The results showed that annual runoff in the upper Yellow River had a mutation point at1933, a no-significant downward overall trend and persistence.However, runoff is affected by human activities and climate change, and the runoff system is a kind of random, nonlinear, weak dependence and high non-stationary one. Therefore, Hilbert-Huang Transform theory was introduced to study multi-scale change law of runoff time series. During the process of decomposition, endpoint continuation method is adopted to deal with boundary problem. The cycle of each Intrinsic Mode Function (IMF) component transformed by Hilbert is2～3a、5～6a、20～21a、37～38a、43～44a.Runoff prediction has always been a research focus at home and abroad, and there’re many methods of runoff prediction. According to characteristic of each IMF component and applicable condition of prediction model, the components were predicted by differert models. The high frequency component was predicted by Least Squares Support Vector Machine (LS-SVM) model, the low frequency component by Auto Regression Moving Average (ARMA) model, the trend component by the Grey Model GM (1,1), and finally superposition of each component was the final prediction. The eligible rate for runoff forecast was90%, and achieved satisfactory prediction effect. The result shows that the method of runoff combination forecast based on the EMD decomposition is effective.