According to the Ministry of Water Resources of the People's Republic of China, the number of casualties in flash flood disasters accounts for nearly 70% of flood disaster casualties in recent years and flash flood has become one of the major natural disasters causing casualties in China. Flash flood regionalization is fundamental for implementing the local measures of the flash flood prevention and mitigation according to local conditions. Using results from existing flash flood relevant physiographical regionalization schemes, and in accordance with the principles of comprehensive, dominant factors, the relative consistency within units, the relative difference between units, the integrity of spatial distribution and the hierarchical divisions, this paper firstly constructs an index system of the flash flood trigger factors and its underlying surface environment for the flash flood regionalization in China. In the combination of top-down and bottom-up regionalization, the hybrid self-organizing-map-based spatial clustering algorithm is then built. Finally, nine homogeneous regions at the first-grade and thirty-three subregions at the second-grade are delineated for flash floods in China. The historical flash flood events and the Geodetector method are adopted to evaluate the developed regionalization schemes. Results show that the first-grade and the second-grade flash flood regionalization schemes in China can respectively provide the determinant power of 66.4% and 75.4% for the spatial distribution of historical flash flood events in the whole country, indicating the developed flash flood regionalization has a good spatial consistency with the distribution of historical flash flood events with different densities. The developed flash flood regionalization not only effectively delineates the regional differentiation pattern of flash floods, but also provides a scientific basis for the implementation of localized flash flood prevention and mitigation measures in China, which benefits for the sound development of flash flood disasters prevention and management in China.

Mass mortality is one of the major threats posed by natural disasters to the population system, which consists of interactions between natural and socio-economic systems. Based on records of deaths extracted from the Integrated Natural Disaster Information System of the Qing Dynasty, events with mass deaths caused by natural disasters are identified, and the decadal deaths index series from 1644 to 1911 is reconstructed. The spatio-temporal characteristics and causes of these extreme events with mass deaths are also analyzed. The main results show the following. First, from 1644 to 1911, historical China suffered a high frequency of extreme events with mass deaths that occurred in 91 years, with events occurring once every three years on average. Second, the deaths index series varies significantly in three phases: Phase Ⅰ is from 1644 to the 1720s with a frequency and intensity close to the average over the entire Qing Dynasty; Phase Ⅱ is from the 1730s to the 1800s with a low frequency of events occurring in two years and in two provinces per decade; and Phase Ⅲ is from the 1810s to 1911 with a high frequency of events occurring in five years and in five provinces per decade. The maximum frequency and intensity appeared in the 1870s, whereas no extreme events with mass deaths occurred in the 1750s. Third, a total of 18 provinces experienced extreme events, and the spatial distribution of extreme events during the three phases differs. During Phase Ⅰ, extreme events were widely distributed but of relatively low intensity; such events occurred in fewer provinces during Phase Ⅱ. During Phase Ⅲ, extreme events occurred in 18 provinces with high intensity and their spatial distribution showed a significant pattern of floods in southern China and droughts in northern China. Finally, epidemics, famine, and flood were the main contributors to extreme events with mass deaths throughout the entire Qing Dynasty, with drought, heavy rain, and storm surges being the main indirect factors.

Under the background of urbanization and climate change, flood disasters are getting more and more serious. Floods tend to shift from single site to multi-sites and from urban area to suburb area, however, the mechanism of flood formation and transformation at different scales under changing environment is still unclear. In this study, we proposed a process-based flood event identification and scale division method. We then detected the characteristics, changes and driving mechanisms of local, regional and basin-scale floods at the Taihu Plain region during the past 50 years (1971-2020). The results indicate that floods in the study area showed distinct scales, with regional floods dominated (69.34%), and extraordinary floods were at basin sclae or regional scale. Floods in the study area showed a significant increase trend, mainly due to the increase of regional floods (accounting for 80.8%) in 1971-2020. Flood characteristics at different scales are significantly correlated with rainfall, tidal level, water level of the Taihu Lake, and upstream water amount. The antecedent conditions of water level were the main driving factor of local and regional floods (contributing 53.2% and 54.98%, respectively). Total rainfall played a dominant role in the rise of water level for local, regional and basin-scale floods, with the relative contributions of 64.91%, 65.01% and 78.72%, respectively. The main driving factors of flood characteristics at different scales and their influencing degree also have some spatial differences. The results of this study show that the spatial scale of floods has changed significantly under the background of changing environment in the Taihu Plain region, which would provide a feasible idea and reference for the identification of floods and mechanism analysis of its changes at different scales in this region or a similar estuary deltas area. This can help to ensure the regional flood control safety under the changing environment.