Comparison study on hydrological time series change-point testing methods
Identification of the change-point of hydrological time series is important to hydrological analysis and prediction. Nine commonly used methods were selected in this study to identify the change points of the sediment load time series at Toudaoguai and Longmen stations from 1960 to 2016, including cumulative anomaly, Mann-Kendall test (M-K), order cluster analysis, double mass curve, Pettitt test, BFAST, Regime Shift Index (RSI), Lee-Heghinian and Yamamoto method. The applicability and accuracy of various methods were compared and analyzed. Results show that cumulative anomaly, order cluster analysis, double mass curve and Lee-Heghinian can accurately identify abrupt change points in sediment load data. The RSI and Pettitt exhibited best performance (p<0.01), and M-K performed well (p<0.05), whereas the Yamamoto method wasn't so good (p<0.05). BFAST can be used to identify breakpoints, as well to analyze the seasonal changes of the monthly hydrological series.
hydrological serieschange-pointtesting methodsmiddle reach of Yellow River
国家科技期刊平台
NSTL
万方数据
维普
