Improving the accuracy and stability of daily runoff prediction is crucial for effective water resource management and flood control. This study proposed a novel stacking ensemble learning model based on attention mechanism for the daily runoff prediction. The proposed model has a two-layer structure with the base model and the meta model. Three machine learning models, namely random forest (RF), adaptive boosting (AdaBoost), and extreme gradient boosting (XGB) are used as the base models. The attention mechanism is used as the meta model to integrate the output of the base model to obtain predictions. The proposed model is applied to predict the daily inflow to Fuchun River Reservoir in the Qiantang River basin. The results show that the proposed model outperforms the base models and other ensemble models in terms of prediction accuracy. Compared with the XGB and weighted averaging ensemble (WAE) models, the proposed model has a 10.22% and 8.54% increase in Nash–Sutcliffe efficiency (NSE), an 18.52% and 16.38% reduction in root mean square error (RMSE), a 28.17% and 18.66% reduction in mean absolute error (MAE), and a 4.54% and 4.19% increase in correlation coefficient (r). The proposed model significantly outperforms the base model and simple stacking model indicated by both the Friedman test and the Nemenyi test. Thus, the proposed model can produce reasonable and accurate prediction of the reservoir inflow, which is of great strategic significance and application value in formulating the rational allocation and optimal operation of water resources and improving the breadth and depth of hydrological forecasting integrated services.

Accurately estimating the reservoir evaporation loss is crucial for water resources management. The existing research on reservoir evaporation loss estimates primarily focuses on large spatiotemporal scales and neglects the rapid dynamic changes to reservoirs’ surface area. For reservoirs essential for frequent flood control and regular water supply, high spatiotemporal evaporation data are crucial. By integrating remote sensing and the evaporation model, this study proposes a new method for the high spatiotemporal estimation of the evaporation losses from reservoirs. The proposed method is applied to the largest artificial freshwater lake in Asia, i.e., Danjiangkou (DJK) Reservoir. The daily reservoir water surface area is extracted at a spatial resolution of 30 m during the period 2014–2018 based on the Enhanced Spatial and Temporal Adaptive Reflectance Fusion Model (ESTARFM). The daily evaporation rate is estimated at a spatial resolution of 100 m using the generalized complementary relationship (GCR). The results show that the water surface area of the DJK Reservoir exhibits rapid and frequent fluctuations from 2015 to 2018, with a multi-year average area of 731.9 km² and a maximum and minimum difference of 304 km². Significant seasonal variations are observed in both the evaporation rate and volume, with a multi-year average evaporation rate of 806 mm and evaporation volume of 595 million m³. The estimated results align well with three other independent estimates, indicating that the GCR is capable of water surface evaporation estimation. Further analysis suggests that the data resolution has a great influence on the evaporative water loss from the reservoir. The estimated mean annual evaporation volume based on the 1000 m resolution water surface area data is 14% lower than that estimated using the 30 m resolution water surface area data. This study not only provides a new method for the high spatiotemporal estimation of reservoir evaporation by integrating remote-sensing data and the GCR method but also highlights that reservoir evaporation water loss should be quantified using the volume rather than the rate and that the estimated loss is noticeably affected by the estimation spatial resolution.

Evapotranspiration (ET) estimations at high spatiotemporal resolutions in urban areas are crucial for extreme weather forecasting and water management. However, urban ET estimation remains a major challenge in current urban hydrology and regional climate research due to highly heterogeneous environments, human interference, and a lack of observations. In this study, an urban ET model, called the PT-Urban model, was proposed for half-hourly ET estimations at a 10 m resolution. The PT-Urban model was validated using observations from the Hotel Torni urban flux site during the 2018 growing season. The results showed that the PT-Urban model performed satisfactorily, with an R² and root-mean-square error of 0.59 and 14.67 W m⁻², respectively. Further analysis demonstrated that urban canopy heat storage and shading effects are essential for the half-hourly urban energy balance. Ignoring the shading effects led to a 38.7% urban ET overestimation. Modeling experiments further proved that flux footprint variations were critical for the accurate estimation of urban ET. The setting source areas either as an invariant 70% historical footprint or as a circle with a 1 km radius both resulted in poor performances. This study presents a practical method for the accurate estimation of urban ET with high spatiotemporal resolution and highlights the importance of real-time footprints in urban ET estimations.

Flash-flood and floodplain inundation simulations are highly sensitive to the spatiotemporal variability of convective rainfall, particularly during the initial runoff generation stage. However, coarse-resolution numerical weather prediction (NWP) forcing tends to smooth localized rainfall extremes, limiting its ability to accurately represent hydrological responses in low-relief floodplains. In this study, we couple a diffusion-enhanced radar nowcasting model, Diff_ConvLSTM, with a spatial resolution of 1 km and a temporal resolution of 6 min, to assess the hydrological value of high-resolution rainfall forcing over the middle Yangtze River floodplain. We introduce a monotone piecewise cubic Hermite interpolation scheme to ensure a stable transition from discrete high-frequency rainfall inputs to continuous hydrodynamic integration. Evaluation using a radar dataset from 2023 to 2024 shows that Diff_ConvLSTM better preserves intense convective echoes and rainband structures compared to the baseline ConvLSTM, increasing the Probability of Detection at the 40 dBZ threshold by 65.8%. A forcing-replacement experiment for the flood event on 30 June 2023 demonstrates that AI-based nowcasting rainfall forcing reduces peak-discharge underestimation, improves volumetric consistency, and produces inundation patterns that are closer to the observation-driven reference than those generated by low-resolution forecast forcing, although positive biases in inundation area and water depth persist. An additional event in 2024 confirms that the improvements are primarily reflected in discharge magnitude and flood volume representation, while enhancements in peak timing remain limited. Overall, the results illustrate both the added value and the remaining limitations of AI-enhanced nowcasting for hydrologically informed flood forecasting.