There exist conflicts of interest between upstream and downstream regions in flood control and drainage; how to balance these conflicts and achieve collaborative flood management remains an important scientific problem. To explore a balanced governance strategy, this study took the Demonstration Zone of Green and Integrated Ecological Development of the Yangtze River Delta, which consists of three separate administrative regions, as the research domain. Using evolutionary game theory, the study conducts a comparative analysis of the interests between upstream and downstream areas. It introduces external drivers, such as the intervention of higher-level administrative bodies and incentive-constraining policies, along with internal balancing mechanisms like bidirectional compensation. The goal is to explore collaborative strategies and cooperation mechanisms that can balance the conflicts of interest between upstream and downstream areas. Results indicate that: (1) The final collaborative strategy was closely related to factors such as the cost of conflict, the amount of two-way compensation, additional benefits of flood control and drainage, and the intensity of incentive constraints. (2) Incorporating a reasonable two-way compensation and reward and punishment mechanism into the evolutionary game theory model can promote the model to a stable strategy. (3) The external driving mechanisms aim to coordinate the conflicts between upstream and downstream regions through incentive or constraint policies, which help motivate and encourage proactive collaboration in flood control and drainage management. The internal balancing mechanism is responsible for compensating for economic losses caused by imbalances, thereby creating pressure that fosters regional cooperation in flood control and drainage governance. In a word, the collaborated management mechanism helps provide a more balanced strategy across different administrative regions.

To explore the evolution of precipitation in megacities under changing environments and human activities, Shenzhen was chosen as the study area. Using daily precipitation data from 10 stations (1960–2019), the TFPW-MK and EOF methods analyzed the spatiotemporal characteristics of precipitation. The findings are: (a) Precipitation in Shenzhen shows a slight upward trend with spatial variations. Western stations exhibit a downward trend. The oscillation period of precipitation indicators is about 10–20 years, with an abrupt change in 1991. (b) Different precipitation indicators have similar distribution patterns. The first mode shows global consistency, indicating overall high or low precipitation, and a shift in the precipitation center from northwest (before 1990) to southeast (after 1990). The second mode shows an “East-West” reverse distribution. (c) Urbanization significantly affects precipitation distribution and has an enhancing effect on precipitation. The smaller the time scale, the stronger the effect of urbanization on extreme precipitation.

Static flood evacuation plans frequently fail during levee breaches because dynamic floodwaters rapidly submerge pre-designated routes, trapping affected populations. This study aimed to develop a dynamic evacuation route planning model by coupling hydrodynamic flood simulation with road network reconstruction for levee breach scenarios. A coupled one- and two-dimensional hydrodynamic model simulated breach flood dynamics, while road network topology and travel-time weights were dynamically updated based on water-depth-dependent vehicle speed attenuation. Using the Dijkstra algorithm to calculate stage-specific optimal routes, the framework was evaluated in the Huayang River Flood Retention Area (HRFRA) along the Yangtze River under a 300-year flood scenario. Results show that: (a) Static routes failed at a rate of 51.5% within 24 h, with 22.7% failing within 6 h. Dynamic route failure rates were 4.5% and 31.8% over the corresponding periods. (b) For 66 villages requiring external relocation, the model dynamically adjusted shelter destinations as roads became inundated. For Tongxing, switching shelters from Qianling (0–21 h) to Chengling (21–57 h) and Liangquan (57–63 h) extended the available window to 63 h, compared with 21 h under static planning. (c) Ten villages, including Qiaobei, scheduled for evacuation after 24 h, lost all viable routes within 9 h, highlighting the need to schedule evacuation batches based on route cutoff times rather than flood arrival times. By integrating network accessibility with flood dynamics, the model improves evacuation planning and supports emergency management.