Principle of correlation coefficient-based classification of hydrological trend and its verification
YuXi ZHAO,Ping XIE,YanFang SANG,HaiTing GU,ZiYi WU,Xu LEI
DOI: https://doi.org/10.1360/N972016-01369
2017-01-01
Chinese Science Bulletin
Abstract:Under the great influence of climate change and human activitiesduring the recent decades, climatic and hydrological processes inmany basins and regions worldwide are changing significantly. Analyzingthe variability of hydroclimatic variables, including temperature,evaporation, precipitation and runoff, is an important topic in thehydrology studies, which contributes to understanding the changesin hydrological regime, water resources management, water disasterscontrol and many other issues. Trend is one of those important formsof hydrological variability, and thus has received much attentionover the last several decades. Many techniques and methods have beendeveloped for trend detection, such as the least squares linear regression,Sen’s robust slope estimator, Mann-endall non-parametric test,Spearman rank correlation, Student’s t -testand others. Present studies about the issue of trend mainly focusedon the improvement, comparison and applicability of trend identificationmethods. However, there lacks effective approach for the classificationof significance degree of trends in hydrological time series. In thisarticle, by employing the index of correlation coefficient, we mainlyproposed a method for the trend identification and classificationof its significance degree. Its basic idea is to calculate the correlationcoefficient between the hydrological time series analyzed and itstime order, based on which the significance degree of trend can beclassified as five ranks: no, weak, mid, strong and very strong. Bydeducing the relationship between the correlation coefficient andthe trend’s slope, their positive correlation is mathematicallyformulated. Results of the statistical tests verified the effectivenessof the method, and also clarified the influences of mean values andvariance of a time series on the classification practice. Trend variabilityof water level series over different time scales and measured at theGaoyao hydrological station was analyzed by the proposed method, andthe index of weight was used to investigate the relationship amongthe significance degrees of trends on different time scales, whichwere evaluated by the proposed approach. Results indicated that thetrends of all water level series presented downward trends, but theyshowed different significance degrees at multi-time scales. Thosetrends with bigger weights at certain time scales had bigger influenceon the total variability of trends. By considering the physical formationmechanisms of the variability of runoff regimes in the basin, reasonabilityof the results and effectiveness of the proposed method were verified,which would be helpful for the evaluation of the influence of climatechange and human activities on the changes in hydrological process.