On the use of temporal evolution of persistence for change point detection of streamflow datasets

Change points (CPs) in streamflow time series are indicative of abrupt shifts and the effective capturing of CPs are useful pre-requisite for accurate modeling of streamfows. The accurate detection of CPs is vital for the hydrological modelers, as the persistence properties are influenced by the shifts in the time series. This study proposed a framework for the detection of CPs through the temporal evolution of persistence of streamflow. The persistence of streamflow was quantified by Hurst exponent (H) estimates of 184 gauging station data located in 15 Indian river basins. Hurst exponent was estimated through Aggregated variance (AV) method and Rescaled range (RR) analysis. H estimates by these methods provided similar results for Indian streamflow and good agreement with the global streamflow persistence of 0.72. Pettit change point detection technique was then applied to identify the change points in the annual average streamflow data and Hurst exponent values of five major river basins, where significant shifts occur. By introducing new framework namely Sequential Pettitt Test (SQPT), CPs in the streamflow datasets were captured. This technique enabled the identification of CPs by analyzing the Hurst Exponent values of streamflows sequentially. The comparison with the CPs in annual streamflow datasets matched well for 86% of the cases with acceptable differences for the remaining series. The remarkable performance of the proposed method and the results of the persistence analysis provided valuable information for water resource managers, hydrologists. Identifying CPs in streamflow datasets can support decision making processes related to water allocation, flood risk management, and ecological conservation.