Study About Some Issues Of Hydrological Time Series

Hydrological time series is a complicated uncertainty research system.Its main research contents focus on hydrological simulation,hydrological frequency calculation and hydrological forecast.This paper tries to make an analysis,comparison and summary of it,basing on the research fruit of overseas and domestic scholars about the hydrology and statistics theory.Besides,in the aspect of hydrological frequency calculation,the paper applies two methods---parameter estimation and non-parametric estimation.Relevant time series model is applied in this paper,as well as hydrological simulation,to make a hydrological forecast.The contents and conclusions of this paper are as follows:Comparative analysis and application of several kinds of methods about parameter estimation: The parameter estimation methods of hydrology and water resources system include moments method,weight function method,probability weighting method,linear moment method and optimum line method,etc.The non-biased test shows that the best non-biased is weight function method,and then is followed by linear moment method,probability weighting method and regular moments method;And the rank of their robust performance is linear moment method,weight function method,probability weighting method,and regular moments method.In addition,the compatible line method apply to the machine simulation of hydrological engineering;Therefore,we use the linear moment method of both good non-biased and robustness to estimate Pearson ? type distribution,and then find out the frequency corresponding to the table.The estimation and application of non-parametric kernel density:Non-parametric estimation in Hydrological frequency calculation requiresno assuming the overall distribution types,and it only needs to use least squares method of cross validation(LSCV method)to calculate the appropriate window width and proper kernel function,thus we can find hydrological frequency density function.Then,by the non-parametric kernel density estimation for the Yangtze River Anqing station 52 years' maximum flood peak flow,we can get its flood frequency function.And compared with the results of parameter estimation,we can find that nonparametric method is more suitable for uncertain general distributions.When in certain general distributions,there is minor deviation between them.But the parametric method is more targeted.Time series analysis model: According to the monthly historical water flow data of Yangtze River Anqing hydrological station from 1959 to 2010,we make a seasonal adjustment and trend decomposition.And through descriptive time series analysis,judgment of stationarity test and model of its whole water flow data,we through descriptive time series analysis,judgment of stationarity test and model of its whole water flow data,we eventually establish ARMA model of hydrological forecast.Finally,by use the Monte Carlo method in hydrological simulation to test the robustness of hydrological forecast model,we can ensure this model is reliable.Therefore,it can be used to forecast the hydrological environment of Anqing.