Study On The Runoff Change Characteristic And Forecast Method Of Nuomin River Basin

Water plays an important role in nature. It is an irreplaceable resource, no matter for the structure of the organism, life activity, or ecosystems. And water problems in our country have been very prominent. Especially the water resources shortage, water pollution, drought and flood disasters, serious influence on the social and economic development in our country, and become an important restriction factor. As one of the main sources of water, runoff is the important basis of reasonable development and utilization, optimizing the allocation of water resources. The change of runoff plays a leading role in the changes of hydrology and water resources system. How to accurately analyze the change rules and the future development trend of the runoff, scientific and rational use of existing water resources become the most pressing problem. Further study of the change rules of water resources of river basin, and accurate prediction the future development trends, are of great significance to rational development and utilization of water resources of river basin.This paper based on the Nen River’s tributary Nuomin River, studies the change rules and the future development trends of runoff. Adopting the method of mathematical statistics to analyze the basic statistical characteristics, annual distribution and interannual variation rules of runoff at first through field investigation combined with mathematical model. Then analyzing the trend, cycle and mutation variation characteristics of runoff. A variety of methods were used to verify the accuracy and reliability. After that, applying the BP neural network and AFSA-BP neural network to predict the monthly runoff, applying EMD coupled harmonic model and EMD coupled AFSA-BP neural network to predict the basin change trends of annual runoff. The main research results are as follows:(1) Through the analysis of basic statistical characteristics, annual distribution nonuniformity, concentration, rangeability and the general characteristics of interannual changes and anomaly analysis of runoff, etc, it is concluded that annual runoff distribution are positively, volume of runoff distribution is dispersed, volume of runoff distribution is uneven within the year, the interannual change is larger. There are large variations of runoff within the year. It presents a trend of fluctuations up and down overall. Distribution curve within the year presents unimodal type. Runoff mainly concentrates in July and August, volume of runoff from January to March are minimal. (2) Nuomin River annual runoff presents fluctuant change, increase and decrease alternating. Annual runoff is roughly a downward trend as a whole. But the trend is not significant. The trend of annual runoff series in the future and the past is the same, which has a downward trend in the future. Through the cycle analysis and multiple time scale analysis, there are4years of short cycle and30years of long cycle in Nuomin River basin. the most obvious change point of annual runoff of Nuomin River occurred in1998. There is also a mutation in1963.(3) In cycle analysis, multi-time-scale analysis based on the EMD shows the changes on runoff series of multiple time scales, multilayered and complexity. And analyzing the main cycle through the calculation of variance contribution rate of each component, which is superior to the single cycle analysis method. In mutation analysis, Mann-Kendall mutation test and Pettitt mutation point test are more comprehensive. In addition, Mann-Kendall rank correlation analysis and rescaled range analysis can be applied to the analysis of multiple aspects of rules. And they have more practicability.(4) Fully considered the advantages and disadvantages of various mathematical models, coupled the multiple models together, a dynamic prediction model comfort to the characteristics of basin is established. And the basin runoff forecast, including using AFSA-BP neural network model to predict river basin runoff, EMD coupled harmonic method and EMD coupled AFSA-BP neural network model to predict annual runoff.