Accurate reservoir outflow simulation is crucial for water resource management. However, traditional machine learning-based simulation methods have not sufficiently considered the physical constraints of reservoir operation, which may lead to unrealistic issues such as negative outflows or water levels exceeding the reservoir’s own limitations. This study integrates physical constraints into the random forest (RF) model using the Sigmoid function, constructing a physics-constrained random forest model (PC-RF) for cascade reservoir outflow simulation. A stratified sampling strategy based on hydrological year types is used to create the training and validation datasets. The coefficient of determination (R2) and root mean square error (RMSE) are used to evaluate the model’s performance for medium- to long-term predictions of reservoir outflows on a 10-day time scale. Additionally, the mean decrease in impurity method is used to assess the importance of input features, thereby enhancing the model’s interpretability. The application the Yalong River cascade reservoir indicates that (1) compared to traditional RF, the PC-RF achieved significant breakthroughs, with an increase of 37.13% in the R2 and a decrease of 60.04% in the RMSE when simulating outflows from the Lianghekou Reservoir, with all reservoirs maintaining an R2 above 0.95, with no instances of unrealistic outcomes; (2) PC-RF effectively integrated historical operational patterns with top three features being previous period outflow, current inflow, and previous period inflow, providing interpretable insights for operational decision-making. The PC-RF model demonstrates high accuracy and practical potential in cascade reservoir outflow simulation, providing valuable applications for cascade reservoir management and water resource optimization.
Under accelerating global climate change, the increasing frequency and intensity of extreme precipitation events (EPEs) pose severe threats to socioeconomic and ecological security, highlighting the critical importance of satellite precipitation products (SPPs) for EPE monitoring. However, comprehensive multi-scale, multi-characteristic evaluations of different SPP types during EPEs remain limited. This study systematically evaluated five SPPs from three categories—satellite-derived products (IMERG-Early, IMERG-Late, IMERG-Final), reanalysis products (ERA5-Land), and merged products (MSWEP-NRT)—during an EPE in Guangdong Province, China (August 16–21, 2024), across three temporal scales (3-hour, 12-hour, 24-hour) and four precipitation characteristics (amount, frequency, intensity, duration). All SPPs exhibit significant scale dependence and systematic biases in reproducing EPEs. The IMERG near-real-time products (Early/Late) provide the best overall multi-scale performance, demonstrating superior spatial fidelity and preservation of dynamic features like intensity gradients and duration. In contrast, ERA5-Land and MSWEP-NRT suffer from excessive smoothing, while the bias-corrected IMERG-Final overly suppresses heavy rainfall intensity. A key limitation across all products is a severe underestimation of precipitation peaks. This study provides critical guidance for SPP selection in EPE monitoring and identifies that future algorithmic improvements must focus on enhancing the identification and quantitative retrieval of convective precipitation to improve reliability.
Reservoirs play a critical role in global water resource management systems; however, reservoir operation simulation faces significant challenges under different hydrological conditions. This study compares the performance of four machine learning models (Random Forest, Gradient Boosting, Long Short-Term Memory, and Bidirectional LSTM) in reservoir outflow simulation under varying hydrological conditions for the Lianghekou reservoir in the Yalong River Basin, China. Results showed that Random Forest achieved the best overall performance (R 2 = 0.6940), followed by Gradient Boosting (R 2 = 0.6794), BiL-STM (R 2 = 0.6588), and LSTM (R 2 = 0.6247). Model performance exhibited clear year-type dependency, with accuracy generally declining from wet to dry years while maintaining consistent ranking. Dry years presented the most significant simulation challenges, with all models showing reduced accuracy. This study demonstrates that model selection should be tailored to specific application scenarios and hydrological conditions, providing valuable insights for intelligent water resource management systems and reservoir operation optimization.
Climate change and rapid urbanization are reshaping the spatial relationship between flood risk and socioeconomic development, yet basin-scale evidence on where risk pressure and development capacity diverge remains limited. Taking 54 cities in China’s Pearl River Basin as the study area, this study adopts a dual-system evaluation framework integrating game-theory weighting, coupling coordination analysis, and spatial autocorrelation. Flood risk is characterized through hazard, exposure, vulnerability, and Resilience, while socioeconomic development is represented by economic scale, urban construction, and population agglomeration. This design extends previous Pearl River Delta-centered analyses to the entire basin and enables city-scale identification of risk–development imbalance. The results reveal a pronounced downstream–upstream gradient. The Pearl River Delta has the highest flood exposure (mean exposure index = 0.2603, 117% higher than the basin average of 0.1208), but strong Resilience and socioeconomic capacity prevent high exposure from translating into the highest overall risk. By contrast, east-central Guangxi emerges as a critical hotspot where moderate to high hazard overlaps with high vulnerability and weak Resilience. Coupling coordination remains generally low, shifting from basic coordination in parts of the Pearl River Delta to severe imbalance in upstream Guizhou, Guangxi, and Yunnan. The global Moran’s I is 0.4261 and significant at the 1% level, indicating marked spatial clustering, with high-high clusters concentrated in core Pearl River Delta cities and low–low clusters in eastern Yunnan and southwestern Guizhou. By distinguishing high exposure cities from high priority intervention areas, this study provides a potentially transferable framework for differentiated flood governance, infrastructure investment, and Resilience enhancement across large river basins.