水文时间序列高阶自相关性识别及其对水文分析的影响研究

High-order Autocorrelation Identification of Hydrological Time Series and Its Impact on Hydrological Analysis

以往水文序列分析中,对序列自相关性的研究多集中于一阶自相关性及其处理方法,而对于水文序列的高阶自相关性则鲜有表述。以陕西省典型流域实测年径流序列为研究对象,开展序列高阶自相关性的研究,探讨高阶自相关存在与否的问题。结果表明陕西省渭河、汉江、无定河三大典型流域中,仅有无定河流域风沙区的韩家峁、横山以及流域出口的白家川水文站的年径流序列检测到了高阶自相关,其余站点的径流序列均不存在高阶自相关性。为了验证高阶自相关性对水文分析的影响,通过BP神经网络预测模型对高阶自相关性的影响进行了评估,结果显示在输入变量中加入合适的高阶变量序列后,可在一定程度上提高模型的预测效果,从侧面验证了高阶自相关性客观存在的假设。同时,研究结果还表明对径流关系不佳的小样本资料地区,考虑高阶自相关性的影响,可有效提升小样本径流资料的展延效果,为区域水文设计的科学性与应用效果提供保障。

In the past,the study on the autocorrelation of hydrological series mainly focused on the first-order autocorrelation and its processing methods,while the high-order autocorrelation of hydrological series was rarely described.This paper intended to study the high-order autocorrelation of the observed annual runoff series in typical watersheds of Shanxi Province,and discussed the existence of high-order autocorrelation.The results showed that among the three typical basins of Weihe River,Hanjiang River and Wuding River,only Hanjiamao,Hengshan and Baijiachuan hydrology station at the outlet of the Wuding River basin had detected high-order autocorrelation in their annual runoff series,and there was no high-order autocorrelation in the runoff series of other stations.This paper used BP neural network prediction model to evaluate the impact of high-order autocorrelation on hydrological analysis,and the results showed that the prediction effect of the model could be improved to a certain extent by adding the appropriate sequence of higher-order variables into the input variables,and the hypothesis of objective existence of higher-order autocorrelation was verified from the side.At the same time,the research results also indicated that taking into account the influence of high order autocorrelation in small sample data areas with poor runoff relationship,the extension effect of small sample runoff data could be effectively improved,and the scientific and applied effect of regional hydrological design could be guaranteed.