Abstract:Abstract: The consistency hypothesis cannot fully meet the current hydrological series in recent years, due to the dual impacts of climate change and human activities. The hydrological frequency has been also questioned under the consistency condition. Much effort has been made into the spatiotemporal evolution characteristics and leading factors of drought-flood complex events between adjacent seasons on the Loess Plateau. Therefore, it is of great significance to regional food security and the prevention of drought and flood disasters. Taking the Loess Plateau as the research object, this study aims to determine the evolution characteristics and dynamic changes of drought-flood complex events, in terms of non-consistency. Firstly, the non-consistencies of the single-season standardized precipitation index (SPI) were diagnosed to construct the two-dimensional joint distribution model of SPI sequences between adjacent seasons using the generalized additive model (GAMLSS model). Secondly, eight events were defined, including from the drought to the flood, from the flood to the drought, persistent drought, persistent flood, from the normal to the drought, from the drought to the normal, from the normal to the flood, and from the flood to the normal. The moderate, severe, and extreme scenarios were identified, according to the classification criteria of drought and flood. A systematic analysis was implemented on the spatiotemporal distribution of the combined events. Thirdly, the occurrence probability of combined drought-flood events under different scenarios was calculated using the consistent/inconsistent two-dimensional joint distribution model. Finally, the 5 year sliding window was combined with the Mann-Kendall test to explore the dynamic evolution characteristics of drought-flood complex events. The important criterion of variable projection was then used to reveal the leading factors of dynamic changes in complex events. The results showed as follows: 1) The occurrence frequencies of the normal to the drought, the drought to the normal, the normal to the flood, and the flood to the normal events were higher than 22 times between adjacent seasons. In addition, the occurrence frequencies of persistent drought (persistent flood) events were greater than that of alternating drought and flood events. 2) The combined events of drought and flood were more likely to occur, with a frequency of 28.88 and 27.40, respectively, from autumn to winter, and from winter to spring. To be specific, the events of the normal to the flood, and the flood to the normal were tended to occur in spring and summer. The events of the normal to the drought, and the normal to the drought were more likely to occur from summer to autumn, and from winter to spring. The probability of the normal drought (flood) events was higher in autumn-winter. 3) Spatially, the events of the normal to the drought, the drought to the normal, the normal to the flood, and the flood to the normal were evenly distributed over the whole basin. In addition, Inner Mongolia, Qinghai, Ningxia, and Shanxi regions tended to sustain the drought events, while Shaanxi and Gansu regions tended to the flood events. 4) There was a significant increase in the occurrence probability of spring-summer drought to flood, summer-autumn sustained flood, autumn-winter from flood to drought, autumn-winter sustained drought, and winter-spring sustained drought. At the same time, an increasing trend was found in the occurrence probability of spring-summer sustained drought in Inner Mongolia, summer-autumn sustained drought in Qinghai, autumn-winter sustained drought in Ningxia, winter-spring sustained drought in Shanxi, and summer-autumn sustained flood in Shaanxi (Gansu), indicating the adverse effects on the social economy and ecology in the region. 5) The leading factors of dynamic change in the occurrence probability of composite events were determined as the Arctic oscillation and sunspot index. The finding can provide scientific and technological support for the precise prevention of drought-flood complex events in the Loess Plateau.