【作者】 谢俊；

【导师】 袁晓辉；

【作者基本信息】 华中科技大学 ， 水利水电工程， 2012， 硕士

【摘要】 河川径流是水资源中最主要的组成部分，是水资源科学管理、综合利用等方面的重要依据。人类社会、经济的发展以及周边生态环境的变化都离不开径流的影响。河川径流的形成是一个十分复杂的过程，其影响因素（如降雨、气候、气温、下垫面、蒸发、植被覆盖以及人类活动等）往往具有不确定性，因此径流时间序列也表现出明显的季节性、滞后性、自相关性等非线性特征。如何在现有的资料基础上，充分挖掘和研究河川径流的演化规律有着十分重要的理论意义和现实意义。本文以河川径流的演变规律为核心，并以黄河上游龙羊峡水文站月径流数据为例，对径流时间序列的长记忆性进行了深入的研究，取得的主要成果有：1.对径流时间序列的做了去季节性处理，并对一些基本统计信息进行了分析。运用滑动平均法对原始径流序列进行了去季节性处理，并从峰度、偏度等角度对序列的正态性进行了检验。介绍了时间序列的平稳性，异方差性，并分别通过ADF单位根法以及ARCH-LM法对研究对象进行了相关检验，最终得出结论：经过去季节性处理后的龙羊峡站月径流时间序列不服从正态分布，具有平稳性和明显的异方差性。2.河川径流的长记忆性分析。从序列自相关函数的角度对长记忆性进行了定义，分析了引起长记忆性的主要原因——系统结构性突变。基于此，分别通过经典R/S分析法（重标级差分析法）和修正R/S分析法对其进行了长记忆性检验，并运用ICSS（累积平方和）算法对径流序列做了结构突变分析，有效的证实了径流时间序列结构突变和长记忆性的存在。3.河川径流时间序列的长记忆性建模。分别介绍了基于短记忆特性的GARCH（广义滑动平均条件异方差）模型和基于长记忆性的FIGARCH（分整增广滑动平均条件异方差）模型。利用长短记忆模型对月径流序列进行建模分析，分别从样本拟合优度、极大似然函数和模型残差去异方差性的角度对两种模型的优劣进行了比较，并得出了具有长记忆的FIGARCH模型能够更好的反映径流的变化规律，为研究径流时间序列的动力学特性提供了一种新的方法。更多还原

【Abstract】 River runoff is the most important component of water resources.It is one of the mostimportant basis in the scientific management and comprehensive utilization of waterresources. The human society, the development of economy and the change of thesurrounding the ecological environment are inseparable from the influence of runoff.The formation of runoff is a complicated process, and the factors which effect therunoff (such as rainfall, climate, temperature, the underlying surface, evaporation,vegetation coverage and human activities and so on) are often uncertain. Because of this,runoff time series also show seasonal, hysteresis, the correlation and nonlinearcharacteristics, etc. Based on the existing material, how to research the rivers runoffevolution law has important theoretical and realistic significance.In this paper, we focus on the rivers runoff evolution law, take month runoff data ofLongYangxia hydrological station which is the upper reaches of the Yellow River as anexample, research the long memory of the runoff time series. The main achievements areas belowed:1. After seasonal processing, there are some basic statistical information of therunoff time series being analyzed. Using the moving average method to seasonalprocessing the runoff series, and check the normal characteristics of the runoff series fromangel of the kurtosis, partial degree. Introduced the stability, heteroscedastic characteristicof time series. Test the characteristic above by ADF unit root method and ARCH-LMmethod to study, and get the final conclusion: after seasonal processing,the runoff timeseries don’t obey the normal distribution,and have stability and obvious heteroscedasticcharacteristic.2. The analysis of long memory characteristic in Rivers runoff. Define long memoryfrom the view of autocorrelation function, analyze the cause of long memory-the structural mutations. Based on this, Test the long memory characteristic through theclassic R/S method and modified R/S method, and use ICSS algorithm to analysisstructure mutations of runoff series, effective confirmed existence of structure mutationsand long memory in the runoff time series.3. The runoff time series long memory modeling. Introducing the short memoryGARCH model and long memory FIGARCH model. After modeling, compared with this2models by the sample goodness-of-fit, maximum likelihood function and residualheteroscedastic. At last get the conclusion that long memory FIGARCH model canbetter reflect the changing law of runoff, provides a new method of runoff dynamicproperties.更多还原