Investigating the spatial-temporal changes of flood events across the Yangtze River Basin, China: Identification, spatial heterogeneity, and dominant impact factors

Flood is one of the most devastating natural hazards and is responsible for sizeable social-economic losses and substantial fatalities. Flood events with diverse behavior characteristics vary greatly in spatial patterns and bring great challenges for flood management. In this study, the Peak Over Threshold (POT) approach coupled with an event-start-and-end detecting approach was proposed to separate flood events during 2008 -2018 across the Yangtze River Basin. Ten flood behavior metrics, comprehensively characterizing magnitude, duration, timing, rate of change, and flood forms of flood events, were used to fully describe flood events and then to identify flood event classes. Subsequently, the spatial heterogeneity of flood events was revealed based on behavior similarity classification of numerous events. Furthermore, a Contribution-Based Impact Factor Analysis (CBIFA) method was constructed to investigate the contributions of natural and anthropogenic factors to flood event variations. Results show that: (1) six flood event classes were identified including long-duration and small-rate-of-change flood events (Class 1, 31.00%), sharp-thin and short-tail flood events (Class 2, 20.58%), sharp-thin and early -peak flood events (Class 3, 5.12%), dwarf-fat flood events (Class 4, 18.37%), small and sharp-thin flood events (Class 5, 4.24%), and conventional events (Class 6, 20.67%). (2) Class 1 and Class 4, characterized by small rate of change, long duration and large magnitude, occurred across the whole Yangtze River Basin and mainly in the Middle-Lower Yangtze plains. Class 2 with small magnitude and short duration was mainly distributed in the Qinghai Tibet Plateau. Class 3, Class 5, and Class 6 mainly occurred in mountainous areas and hills in the middle and upper reaches, wherein Class 3 and Class 5 were characterized by large rate of change and short duration. (3) The meteorological factors were the most important impact factors and explained 30.23%-68.10% of the total flood event variations, followed by geographical (9.19%-32.10%), human activities (6.56%-23.17%), soil (9.19%-19.34%) and vegetation (2.56%-5.31%) factors. Identifying the spatial pattern of flood events and their driver factors is a crucial step toward accurate flood prediction and will finally provide essential information for reliable flood management.