“…Although observed changes in hydrological variables such as flooding are not widespread across the globe [Stocker et al, 2013], the so-called nonstationarity has been detected in many runoff series including annual runoff, low flow and flood [Xiong and Guo, 2004;Villarini et al, 2009;Vogel et al, 2011;Hall et al, 2014;Bender et al, 2014;Jiang et al, 2015a;Kim et al, 2016]. The basic stationarity assumption in traditional hydrological frequency analysis is being questioned, and therefore nonstationarity is increasingly recognized as a considerable challenge in water resources and flood management, In most current literature, identifying whether nonstationarity exists in runoff series is mainly treated as a pure statistical problem, which is then addressed using the trend/change-point tests, such as the Mann-Kendall (MK) test and Spearman test for trends, and Pettitt test for change points [Mann, 1945;Kendall, 1975;Pettitt, 1979;Yue et al, 2002;Burn and Hag Elnur, 2002;Xiong and Guo, 2004;Vogel et al, 2013;Cui and Li, 2016;Wei et al, 2016]. The nonstationarity of runoff series can be modeled by the time-varying moments model, in which the statistical parameters (which can be expressed through moments) of the runoff probability distributions are linked to time, climatic indices, or human activity indices [Villarini et al, 2009;L opez and Franc es, 2013;Du et al, 2015;Jiang et al, 2015a].…”