Recent advances in very high resolution PlanetScope imagery and deep-learning techniques have enabled effective mapping of small water bodies (SWBs), including ponds and ditches. SWBs typically occupy a minor proportion of remote-sensing imagery. This creates significant class imbalance that introduces bias in trained models. Most existing deep-learning approaches fail to adequately address this imbalance. Such an imbalance introduces bias in trained models. Most existing deep-learning approaches fail to adequately address the inter-class (water vs. non-water) and intra-class (SWBs vs. large water bodies) simultaneously. Consequently, they show poor detection of SWBs. To address these challenges, we propose an area-based weighted binary cross-entropy (AWBCE) loss function. AWBCE dynamically weights water bodies according to their size during model training. We evaluated our approach through large-scale SWB mapping in the middle and east of Hubei Province, China. The models were trained on 14,509 manually annotated PlanetScope image patches (512 × 512 pixels each). We implemented the AWBCE loss function in State-of-the-Art segmentation models (UNet, DeepLabV3+, HRNet, LANet, UNetFormer, and LETNet) and evaluated them using overall accuracy, F1-score, intersection over union, and Matthews correlation coefficient as accuracy metrics. The AWBCE loss function consistently improved performance, achieving better boundary accuracy and higher scores across all metrics. Quantitative and visual comparisons demonstrated AWBCE’s superiority over other imbalance-focused loss functions (weighted BCE, Dice, and Focal losses). These findings emphasize the importance of specialized approaches for comprehensive SWB mapping using high-resolution PlanetScope imagery in low-latitude regions.

River–lake systems in the middle and lower reaches of the Yangtze River Basin function as critical ecological interfaces for maintaining regional water security and biodiversity. However, the complex interplay between environmental factors and biological communities in these systems remains poorly understood, limiting evidence-based management strategies essential for achieving sustainable development goals. This study investigated the spatiotemporal heterogeneity of environmental variables and their relationships with biological communities across 36 sampling sites (4 rivers, 5 lakes) from January to November 2022. Significant spatial differences were observed between river and lake systems, with lakes exhibiting higher concentrations of biochemical oxygen demand (BOD₅), chlorophyll-a, permanganate index (CODMn), loss on ignition (LOI), sediment total nitrogen (STN), total organic carbon (TOC), and turbidity, while rivers showed elevated total nitrogen levels. These patterns reflected fundamental differences in hydrodynamic mechanisms, particularly water retention time and sedimentation processes. Environmental parameters displayed distinct seasonal variations, with BOD₅ increasing markedly in autumn, and chlorophyll-a showing system-specific peaks in lakes (May) and rivers (September). Multivariate analyses revealed that water temperature, organic matter, and nutrients (particularly phosphorus) were key drivers shaping both phytoplankton and benthic communities. The structural equation model identified a strong cascade pathway from turbidity through phosphorus to phytoplankton richness, and uncovered a “phosphorus paradox” wherein total phosphorus exhibited contrasting effects on different biological components—positive for phytoplankton richness but negative for benthic richness. The positive relationship between phytoplankton and benthic macroinvertebrate richness provided evidence for benthic–pelagic coupling in this river–lake system. These findings advance our understanding of the complex mechanisms linking physical factors, nutrient dynamics, and biological communities across river–lake continuums, providing a quantitative framework for ecosystem-based management that supports sustainable development in the Yangtze River Basin and similar freshwater systems globally.

Phosphorus (P) release from reservoir sediments critically influences water quality and ecosystem stability. This study analyzed surface sediments from four representative zones to investigate phosphorus fraction distribution, key influencing factors, and implications for water quality. Results showed that total phosphorus (TP) content in sediments from main and tributary inflow zones was significantly higher than in open-water and transition zones. Inorganic phosphorus (IP) was the dominant form, with iron-bound phosphorus (Fe-P) accounting for 33.2–42.0% of IP. A strong correlation existed between P release and the Fe/P molar ratio; notably, when the ratio approached 10, phosphorus desorption increased significantly, indicating a shift from sink to source. Sediments with grain sizes <0.01 mm had the highest P release rates, suggesting particle size, Fe content, and hydrodynamics jointly regulate P mobilization. Using the Diffusive Gradients in Thin Films (DGT) technique, phosphorus release in inflow zones exceeded 1 g/m² in all hydrological periods, contributing substantially to internal loading. Sediment-derived P primarily influenced bottom water, while surface water was more affected by external inputs. These findings highlight the spatial heterogeneity of P release and underscore the need for zone-specific management strategies in reservoir systems.

The “Four Illegal Activities”, which involve occupation, extraction, and construction along shorelines, have become significant challenges in river and lake management in recent years. Due to the diverse and scattered nature of monitoring targets, coupled with the large volumes of data involved, traditional manual inspection methods are no longer sufficient to meet regulatory demands. Late fusion change detection methods in deep learning are particularly effective for monitoring river and lake occupation due to their straightforward principles and processes. However, research on this topic remains limited. To fill this gap, we selected eight popular deep learning networks—VGGNet, ResNet, MobileNet, EfficientNet, DenseNet, Inception-ResNet, SeNet, and DPN—and used the Jinshui River Basin in Wuhan as a case study to explore the application of Siamese networks to monitor river and lake occupation. Our results indicate that the Siamese network based on EfficientNet outperforms all other models. It can be reasonably concluded that the combination of the SE module and residual connections provides an effective approach for improving the performance of deep learning models in monitoring river and lake occupation. Our findings contribute to improving the efficiency of monitoring river and lake occupation, thereby enhancing the effectiveness of water resource and ecological environment protection. In addition, they aid in the development and implementation of efficient strategies for promoting sustainable development.

Reconciling hydropower development with aquatic biodiversity conservation is a central challenge for sustainable river management worldwide. Cascade dam configurations, in which multiple impoundments are arranged in series along a single channel, impose longitudinal environmental gradients that restructure biological communities across trophic levels. Whether the resulting multi-trophic responses are independently driven by shared abiotic gradients (environmental filtering) or mechanistically coupled through direct food-web interactions (trophic cascading) remains unresolved. We surveyed phytoplankton, zooplankton, and benthic macroinvertebrates simultaneously at seven stations along a 430 km gradient downstream of Danjiangkou Dam in the Hanjiang River, the largest tributary of the Yangtze River and the source of China’s South-to-North Water Diversion Middle Route, over eight seasonal campaigns (2015–2017). Variance partitioning, piecewise structural equation modeling, Mantel tests, and co-occurrence network analysis were applied to partition environmental and trophic pathways. Environmental filtering dominated community restructuring at all three trophic levels, while the biotic proxy for direct trophic interactions explained less than 0.4% of community variation, consistent with weak detectable trophic coupling at seasonal resolution. Distance from Danjiangkou Dam shaped downstream transparency and turbidity gradients that mediated trophic-level-specific responses along distinct environmental axes (pH and water temperature for phytoplankton, conductivity for zooplankton, and transparency for benthic macroinvertebrates). Benthic macroinvertebrates were systematically decoupled from the pelagic analytical framework, absent from the cross-trophic co-occurrence network and structured more by spatial configuration than by water-column variables. Hub species in the network were associated with downstream mineralized conditions, confirming that network architecture reflects shared environmental preferences rather than biotic interactions. These findings support a management shift from single-dam mitigation toward cascade-scale coordination of environmental flow regimes, sediment connectivity, and substrate restoration as integrated strategies for sustaining multi-trophic biodiversity in regulated rivers.

Shallow eutrophic lakes recover from nutrient loading on time scales ranging from less than one year to many decades, yet whether this range is set by the lake or by the biological response group has rarely been quantified within a single monitoring framework. We assembled a five-year (2020–2025) quarterly monitoring panel from three shallow Yangtze-floodplain lakes (Lake Changhu, Lake Liangzihu, and Lake Honghu; 15 stations, 21 quarters) and applied a panel mixed-effect distributed lag model (PME-DLM) to estimate the lag-response windows of phytoplankton and benthic macroinvertebrate densities against five water-quality drivers. Cross-lake consistency was tested with a station-resampled bootstrap, and the contributions of water quality, season, and lake identity to community variation were resolved by three-table variation partitioning. The PME-DLM resolved a 3-month temperature window for phytoplankton and 9–15 month chlorophyll a and temperature windows for benthic communities, while total nitrogen and total phosphorus were non-significant in either group. Cross-lake bootstrap intervals on window width overlapped substantially across the three lakes, whereas cross-group differences in window centre and shape were an order of magnitude greater. Variation partitioning further showed a mirror-image structure in which phytoplankton variation was dominated by the pure water-quality fraction (12.2%) and benthic variation by the water-quality × season joint fraction (5.8%). Within the resolution of this five-year, three-lake panel, group-level differences in lag-response time scale were more apparent than lake-level differences and provide a quantitative basis for matching restoration assessment cadence to pelagic versus benthic recovery.