Mapping dynamic changes in hydrological time series using the average directional index

摘要

The current trend detection methods have mostly concerned with testing the monotonic change in the mean of a hydrologic variable concerned (hypotheses testing) but limited in detecting overtime change in time series. To the best of the authors' knowledge, this study made the first use of the average directional index (ADX), a trading index, and its components (positive and negative directional indices, DI +/-) in mapping the dynamic change in the streamflow, specifically the strength of a trend. The calculation of ADX depends on three parameters, i.e. differences between high and minimum streamflows relative to the range in a given time interval. Monthly datasets of 17 worldwide streamflow were collected and analyzed (1900-2018). The results were verified against two classical trend methods (the Mann-Kendall, MK, and the Partial Trend Index, PTI). While, MK and PTI indicated that all rivers progressed towards non-stationary conditions (presence of a trend); however, based on the ADX, the trend's strength at 35 of the tested rivers was unclear due to the cyclic nature of streamflow and the remaining unveiled unequivocally decreasing and increasing signals at 30 and 35 of the rivers, respectively. The ADX also was more sensitive in detecting shifts in the regime compared to the Regime_Shift method based on the annual mean of the streamflow. The ADX components (DI +/-) have also managed to detect the historical flood and drought events at 77-100 and 36-100 of the cases respectively, with a magnitude of error of +/- 1-year. Individually, the Zambezi is found to be the most sensitive worldwide river to change in high and minimum streamflows while Fraser and Mekong rivers are the least sensitive rivers worldwide.