Assessing the non-stationarity of low flows and their scale-dependent relationships with climate and human forcing

It is necessary to assess the non-stationarity of a hydrological series under changing environments. This study aimed to determine the validity of the stationarity of low flow series in terms of trends and possible change points, as well as the time-scale that is responsible for the production of trends and change points in low flow series. Further, we investigated how climatic variables affect low flow variations by studying their scale-dependent relationships. The modified Mann-Kendall trend test, heuristic segmentation method, discrete wavelet transform, and Pearson correlation coefficient were co-utilized to achieve these objectives. The Wei River Basin (WRB), a typical Loess Plateau region in China, was selected as the case study. Results showed significantly decreasing trends and change points in the low flow series, indicating that its stationarity assumption is invalid. The 2-year and 4-year events were the most important time-scales contributing to the trend of the original low flow series, and the 8-year periodic scale was the most influential frequency component for change point generation. Additionally, the strongest scale-dependent relationships among high frequency components (2-year and 4-year scales) of the low flow series and climatic variables (precipitation, potential evaporation, and soil moisture) demonstrated the importance of climatic factors for driving the trends of a low flow series. In contrast, human activities, including water withdrawals and water and soil conservation projects showed strong influences on the non-stationarity of low flows via affecting the low frequency component (8-year frequency and approximate components). These findings contribute to a better understanding temporal variations of low flow and their responses to changing environments, and the results also would be helpful for local water resources management as well as agricultural and ecological sustainable development. (C) 2019 Elsevier B.V. All rights reserved.