Study On Runoff Change Rule And Forecast Of Xilin River Basin

As the main source of water resources, River flow plays an important role in the regulation of water resources. Xinlin River is one of the most typical inland river located in the arid and cold north areas in our country, whose water resources situation is critical to the grassland ecological construction of the whole valley, the agricultural in the use of water. So, the runoff change rule analysis of Xilin River basin and establishing of the reasonable runoff forecast model provide a scientific basis for carrying out river basin planning and the rational allocation of water resources. Based on the climate and runoff data during 46 years, in this paper we analyzed the trend, mutability and periodicity of the river runoff, etc. and discussed the factors causing the change of runoff, on this basis, through the establishment of BP neural network model and ARMA model to study on the runoff forecast. The main results of this study are embodied in the following points:(1) The analysis of the inter annual variability and departure of the Xilin River Basin showed that annual runoff changed a lot from the past to the present. It presents a phenomenon of wet year, normal year and dry year appear alternately; and using the mathematical statistical method of moving average analysis, accumulated filter, etc. and Kendall rank correlation test methods test out the annual runoff in the basin is a not obvious declining trend.(2) Analyzing the temporal change and factors of annual runoff of the Xilin River Basin, the conclusion draws mutation test is that runoff had mutations in 1990 year. The annual rainfall is the main factor in the basin runoff changes, and as rainfall-runoff cumulative curve demonstrated that the annual runoff of the Xilin River Basin affected by human activity since the 1979 year.(3) Analyzing the cycle change of annual runoff sequence. There mainly existed three cycle times, i.e.4,9 and 28 years of Xilin River annual runoff series. Among them, 28-year time scale of peak is the first master of annual runoff series cycle; corresponding to the second and third master of annual runoff series cycle respectively are 9-year and 4-year time scales.(4) On the basis of analysis of runoff change rule, the establishment of the BP neural network model and the time series ARMA model for the Xilin River basin gave a good forecast on the annual runoff time series of the basin. We can conclude that to climate data as the input samples of the BP neural network model is more suitable for the Xilin River basin annual runoff time series prediction, which provides more accurate prediction results.