基于相关系数的水文趋势变异分级方法

Hydrological Trend Variation Classification Method Based on Correlation Coefficient

基于Hurst系数法和Bartels检验的水文变异联合分级方法可以从整体上识别与检验时间序列变异及其变异程度,但无法判断序列的具体变异形式(趋势、跳跃、周期等).基于相关系数描述序列与时序的相关性大小以及表征序列趋势变异程度的特性,提出了基于相关系数的水文趋势变异分级方法.该方法首先计算序列的相关系数,然后作假设检验:依据统计学原理和经验选用不同的阈值作为不同变异程度的划分依据,并将相关系数划分成5个区间,对应于无变异、弱变异、中变异、强变异和巨变异5个等级.实际应用中,根据相关系数的大小判断其落在哪个区间,即可确定序列是否发生趋势变异及变异程度的大小.对5个实例序列42年资料的变异分析表明,枝城站年径流序列无趋势变异,兰州站和花园口站年径流分别为趋势弱变异和中变异,荆江三口分流量为趋势强变异,红崖山站年径流为趋势巨变异,上述结果与采用Hurst系数法所得整体变异分级结果是一致的;成因分析表明,不同强度的人类活动导致了序列发生不同程度的变异.

Joint analysis method for hydrological variation based on Hurst coefficient method and Bartels Test detects and indicates the variation on the whole,i. e. without identifying the specific variation form,such as trend,jump,periodic,etc. While the correlation coefficient could describe the tightness of the relation between series value and time,and represent the degree of trend variation. Based on that characteristic,a method specialized for trend variation was proposed to detect the variation and get the classification of series＇ variation state. Firstly,the correlation coefficient of hydrological series was calculated. Then hypothesis testing was conducted： According to the principles and application experience of Statistics,different threshold values were chosen to divide different variation degrees of hydrological series,and the correlation coefficient was divided into 5 intervals,namely no trend variation,weak trend variation,moderate trend variation,strong trend variation and giant trend variation. So in practical applications,it could be judged which interval the correlation coefficient fall in,according to the magnitude of the coefficient,and then decided whether the series had varied in trend form and its variation degree. Lastly the 42 years annual runoff series of 5 Stations in the different basins were used to verify this method. The results showed that the variation state of annual runoff series of Zhicheng Station was not significant（ no trend variation）,degree of the series in Lanzhou and Huayuankou Station was weak and moderate,respectively,series diverted from the three outfalls was strong trend variation,and series in Hongyashan Station was giant trend variation,which were accordant with those using Hurst coefficient method. The causal analysis showed that different intensities of human activities lead to those diverse variation states.