Characteristics Analysis And Prediction Of Hydrological Time Series

It respectively studies the characteristics and the prediction of precipitation data of Nanning for 55 years.Thereinto,the characteristics analysis includes annual analysis and analysis of the year.The analysis of interannual variation is mainly studied from the three aspects of tendency,mutation and period.For the tendency,using the linear tendency method,anomaly analysis method,moving average method,cumulative filter method,rescaled range analysis,Kendall rank correlation test and Spearman rank correlation test to research the tendency of the precipitation time series.The results show that there is a slight upward trend of the precipitation time series of Nanning for 55 years,but not obvious.For the mutation,using the ordered cluster analysis,moving T test,Lee-Heghinian test,Pettitt test,Mann-kendall nonparametric statistics test method to examine the mutation point of the precipitation time series,and 2008 is the most likely mutation point in the series.For the period,using the maximum entropy spectrum analysis,Morlet wavelet analysis and EMD-HHT to ensure the period of the precipitation time series,and finally getting three obvious periods of different time scales,respectively are 3-6a,12-14a,27-29a.In the precipitation years variation analysis,mainly using the allocation proportion method,annual variation analysis,nonuniform coefficient,concentration period,concentration degree,Gini coefficient and Lorentz asymmetrical coefficient to research the characteristics.It turns out that the precipitation is mainly concentrated in 5-9 month,and more than 70%of the total;during in 1961-1995 years,the precipitation time series is more volatile,in 1996-2015 years,the sequence is more stable.In addition,there are 27 years'nonuniform coefficient above the average and 28 years under the average,and the fitted curve indicates the nonuniform coefficient is decreasing.From the Gini coefficient and Lorentz asymmetrical coefficient,the distribution of precipitation is uneven during the year,mainly due to the heavy rainfall months.Using the gray prediction model,the moving improved gray model,weighted markov chain prediction model,rank set pairs analysis,least squares support vector machine prediction model and BP artificial neural network prediction model to predict the precipitation of Nanning.And the results show that all model is qualified to predict the precipitation of Nanning.The average relative fitting error of gray model is 7.8%,after improved,the precision is up to 3.4%.The weighted markov chain model predict that the next two years is normal years,and the rainfall quantity are respectively 1384.6mm,1221.6mm;The average relative fitting error of rank set pairs analysis model is 6.06%,and predict the rainfall quantity of next two years are 1246.3mm,1300.1mm.Besides,the average relative fitting error of the least squares support vector machine prediction model and BP artificial neural network prediction model are respectively 5.46%,4.89%,the NSE are 0.57 and 0.75,just for the research,the BP-ANN is better than LS-SVM.The research can provide scientific guidance for the future water resource planning and reasonable allocation of flood control and disaster reduction in Nanning.