[Objective] Traditional ventilation approaches for underground powerhouses lack adaptability to changing environmental conditions, leading to insufficient pollutant removal efficiency when concentrations are high, and excessive energy consumption when demand is low. This study proposes an intelligent energy-saving control method for underground powerhouse ventilation by integrating a highly accurate dynamic demand prediction model with a robust fan control strategy, seeking to ensure environmental safety compliance while minimizing energy waste during the construction phase. [Methods] The intelligent control framework comprises three main components:variable selection, predictive modeling, and control execution.(1) Input Parameter Selection: Utilizing Mutual Information (MI) analysis on 41 sets of field data, the study identified CO concentration, dust concentration, and ventilation time as the key input variables for the demand prediction model, effectively filtering out redundant environmental factors to enhance model efficiency.(2) IPSO-RBF Prediction Model: A Radial Basis Function (RBF) neural network was constructed to predict the required wind speed in real-time. An Improved Particle Swarm Optimization (IPSO) algorithm was introduced. The IPSO algorithm employs non-linear adjustment strategies for inertia weight (ω) and learning factors (c₁, c₂). (3) Fuzzy PID Control Strategy: Recognizing the large inertia and hysteresis of ventilation systems, a Fuzzy PID (Proportion Integration Differentiation) control algorithm was developed to replace traditional PID control. This method utilizes fuzzy logic rules to self-tune K_p, K_i, and K_d based on system error (e) and error change rate (e_c), ensuring rapid and stable fan frequency adjustment.(4) System Integration: The proposed methods were integrated into a Web-based intelligent energy-saving feedback control platform. The system was deployed at the Lushan Pumped Storage Power Station in Henan Province, utilizing a custom-built mobile integrated sensor box and a Siemens PLC-based control layer for field validation. [Results] The performance of the proposed method was validated through numerical experiments and on-site engineering applications.(1) Prediction Accuracy: The IPSO-RBF model significantly outperformed both the traditional RBF and standard PSO-RBF models. Specifically, on the test set, the IPSO-RBF model achieved a Mean Square Error (MSE) of 0.25, a Mean Absolute Error (MAE) of 0.38, and a Mean Absolute Percentage Error (MAPE) of 2.37%. The relative error of the proposed model was consistently kept within ±5%, demonstrating superior precision in predicting ventilation demand under complex working conditions.(2) Control Performance: Fuzzy PID algorithm significantly improved the dynamic response of the fan frequency control. Compared to traditional PID algorithm, the Fuzzy PID method reduced the system overshoot by 7% and shortened the settling time (time to reach stability) by 37.5% (from 45 s to 30 s), proving its strong robustness against system lag.(3) Field Application: Field tests following blasting operations at the Lushan project confirmed the system’s effectiveness. The platform successfully reduced CO and dust concentrations to regulated safety standards within the required timeframe (30 min). In terms of energy efficiency, the intelligent control mode consumed 121.9 kW·h of electricity during the test period, whereas the traditional “one wind blowing” mode consumed 154.7 kW·h, a significant energy saving of 21.2% per ventilation cycle. [Conclusion] The proposed IPSO-RBF model provides highly accurate, real-time predictions of ventilation demand by effectively handling small-sample data in complex environments. Fuzzy PID strategy further solves the control challenges associated with the large inertia of ventilation systems, offering faster response times and greater stability than conventional methods. The deployment of the Web-based intelligent platform at the Lushan Pumped Storage Power Station verifies the practical engineering value of this approach. By achieving a 21.2% reduction in energy consumption while guaranteeing air quality compliance, the model provides a generalized, efficient, and intelligent solution for ventilation management in the construction of large-scale underground hydropower facilities, contributing to the industry’s goals of smart construction and green development.

[Objective] Although deep learning-based semantic segmentation models, such as U-Net and DeepLabv3+, have improved water body extraction performance, they remain limited in delineating continuous boundaries and recognizing narrow water bodies. To address these limitations, this study proposes an improved model named DSP-UNet (U-Net enhanced with Dual Attention and Strip Pooling). The model is designed to improve both local boundary sensitivity and global contextual reasoning, achieving higher precision and robustness while maintaining computational efficiency. The objective is to develop a high-performance deep learning model capable of stable water body extraction under varying geographic and imaging conditions. [Methods] The proposed DSP-UNet model was constructed based on the classical U-Net architecture and integrated three key innovations: strip pooling module (SPM), dual attention mechanism (DAM), and SimAM attention module. The model was trained using the OpenWUSU512 dataset. Comparative experiments with baseline models, including U-Net, PSPNet, HRNet, and DeepLabv3+, were performed under the same training conditions. Ablation studies were conducted to quantify the contribution of each proposed module. Moreover, cross-domain generalization was evaluated on the Satellite Images of Water Bodies dataset to assess model transferability. [Results] Experimental results demonstrated that DSP-UNet achieved higher segmentation accuracy and robustness across all evaluation indicators. On the OpenWUSU512 dataset, DSP-UNet reached a mean intersection over union (MIoU) of 0.962 and an F1 score of 0.966, outperforming PSPNet by 2.4% in MIoU. The precision and recall values were 0.976 and 0.956, respectively, indicating a balanced trade-off between false positives and false negatives. The overall accuracy (OA) was 0.991. Compared with the baseline U-Net, the dual attention mechanism increased MIoU by 1.1%; the integration of SimAM improved edge recognition and reduced boundary noise; and the addition of strip pooling expanded spatial perception, increased MIoU by 2.3%, and reduced total training time by 11.5%. These results confirmed that the progressive introduction of modules effectively enhanced both feature representation capability and computational efficiency. Cross-domain experiments on the Satellite Images of Water Bodies dataset demonstrated strong generalization performance. DSP-UNet achieved an F1 score of 0.909, an IoU of 0.833, and an OA of 0.954, outperforming U-Net and SegNet by more than 5% in IoU. Visual assessment showed that the model accurately distinguished water bodies from shadows and reflections, maintained continuous boundaries, and effectively extracted narrow channels and small ponds, which were often misclassified by conventional models. [Conclusion] The proposed DSP-UNet provides an accurate and robust solution for water body extraction from high-resolution remote sensing imagery. By combining dual attention and strip pooling, the model effectively enhances both global contextual understanding and local boundary refinement. The parameter-free SimAM module further improves spatial discrimination without introducing additional computational burden. DSP-UNet achieves the highest accuracy among the compared models while maintaining stable convergence and reduced training time, demonstrating its efficiency for large-scale applications. The model’s strong generalization ability across datasets indicates its potential application in remote sensing tasks, including wetland mapping, flood detection, and shoreline monitoring. Future research should focus on integrating multi-temporal and multi-sensor data, developing temporal attention mechanisms for dynamic water monitoring, and optimizing the architecture for lightweight and real-time deployment in practical remote sensing applications.

[Objective] This study aims to address two core challenges in the digital governance of complex water networks: insufficient integration of physical topological relationships with multidimensional hydraulic characteristics, and limited multi-level routing analysis capabilities, which collectively weaken the support for water diversion decision-making. Focusing on the Beijing Section of the Middle Route of South-to-North Water Diversion Project and its supporting municipal projects, the research seeks to construct a knowledge graph that realizes the integrated expression of spatial topology and hydraulic attributes, and develop a scalable query system to meet diverse scheduling demands, thereby providing technical support for digital twin water network construction. [Methods] The research adopts a systematic approach for knowledge graph construction and application: (1) Knowledge modeling: A novel “topology-hydraulics-business” three-dimensional framework is proposed to integrate engineering entities, relational entities, and attribute entities, with core concepts aligned with the national standard SL/T 213—2020. (2) Knowledge extraction: Data from structured documents, semi-structured texts, and GIS spatial datasets are processed through standardization (unified units/coordinate system), missing value imputation, and synonym dictionary optimization to ensure data quality, followed by triple extraction using regular expressions. (3) Knowledge fusion: Entity alignment (resolving duplicates like “Miyun Reservoir Sluice” and “Miyun Reservoir Gate”), attribute integration (unified units: m³/s for flow, km for length), and relationship disambiguation (clarifying multi-semantic terms like “Yongding River”) are conducted manually. (4) Knowledge storage: The graph is stored in Neo4j for efficient visualization and querying. (5) Query system development: A three-level query system is built using Cypher language, integrated with the Dijkstra algorithm for path optimization. (6) Performance evaluation: Comparative experiments with Floyd-Warshall and SPFA algorithms are conducted to assess efficiency, convergence, and stability. [Results] The constructed knowledge graph achieves comprehensive semantic expression of the study area, encompassing 437 engineering entities, 149 relationship entities, and 6 attribute types. The three-level query system demonstrates strong applicability: (1) Basic topological analysis successfully retrieves direct downstream nodes of Guanting Reservoir (i.e., “Guanting Reservoir Sluice” and “Guanting Reservoir Spillway”) and the complete path of the Hexi Branch Line (from Daning Reservoir to Shimeng Gate Station via 3 pumping stations). (2) Conditional constraint filtering identifies 5 pipelines over 40km and generates an alternative path for water transfer from Guanting Reservoir to Zhuwo Reservoir when the spillway fails. (3) Multi-objective optimization yields optimal solutions: the shortest path between Huairou and Miyun Reservoirs is 30.33 km; the path with the highest water transmission efficiency (L/Q=3.658 5) from Tuancheng Lake to Huairou Reservoir; and the optimal path meeting dual constraints. Performance evaluation shows the Dijkstra-based algorithm outperforms competitors: average runtime of 0.05 s (vs. 1.655 s for Floyd-Warshall), average relaxation steps of 19 (vs. 34 for SPFA), and zero response time standard deviation, confirming high efficiency and stability. [Conclusions] This study makes two key innovations: (1) The “topology-hydraulics-business” three-dimensional modeling framework resolves knowledge fragmentation by integrating physical structure, hydraulic characteristics, and scheduling rules. (2) The three-level query system supports dynamic constraint filtering and multi-objective optimization, adapting to scenarios like emergency scheduling and daily management. The results validate the knowledge graph’s effectiveness in integrating multi-source data and supporting hierarchical decision-making, providing a replicable technical route for the digital governance of large-scale water diversion projects. Future research will focus on dynamic topology update mechanisms, logical modeling of complex scheduling rules, and human-machine interaction interface development, advancing the knowledge graph into an intelligent decision tool with spatiotemporal dynamic response capabilities to promote precision and intelligence in digital twin water network governance.

[Objective] The Mekong Delta is one of the globally vital biodiversity hotspots, while long-term coastal geomorphic evolution driven by natural sediment transport and intensive anthropogenic interference has triggered prominent spatial differentiation in shoreline progradation and retrogradation across the delta. This study aims to quantitatively characterize the long-term spatio-temporal evolution characteristics of sandbars and shorelines in the Mekong Delta, identify the spatial heterogeneous patterns of coastal accretion and erosion, and further clarify the coupling driving mechanisms of natural hydrological processes and human activities behind delta geomorphic changes. The research is expected to provide reliable long-term baseline data and decision-making references for coastal ecological conservation, integrated water resource regulation, coastal risk prevention and regional sustainable governance in the transboundary Mekong coastal zone. [Methods] We acquired 35-year continuous Landsat remote sensing images covering the period from 1988 to 2023. The Normalized Difference Water Index (NDWI) was adopted to enhance the contrast between water bodies and terrestrial surfaces, and the Otsu’s method (maximum between-class variance algorithm) was applied to realize the automatic threshold segmentation of water-land boundaries, so as to extract multi-phase vector datasets of shorelines and sandbars accurately. On this basis, multiple quantitative indicators including sandbar total area and patch number, total shoreline length, inter-annual shoreline change rate and net shoreline movement distance were calculated to systematically reveal the temporal variation trends of sandbar geomorphology and the spatial differential features of shoreline dynamics between river channels and estuarine regions. Combined with historical sediment data, tidal-hydrodynamic records and regional socio-economic engineering statistics, We further summarized the dominant natural and anthropogenic driving factors responsible for the spatially differentiated geomorphic evolution of the Mekong Delta. [Results] During 1988-2023, the total terrestrial area of the Mekong Delta showed a significant increasing trend, accompanied by an obvious decline in the number of sandbar patches, which indicated the geomorphic evolution characteristic of small scattered sandbars gradually merging into large contiguous tidal flats. The total shoreline length exhibited dramatic inter-annual fluctuations within the 35-year research period, with a net total growth of 63.39 km across the whole study time span. Obvious disparities existed in shoreline change rates between both banks of inland river channels and estuarine zones. The net horizontal migration distance of shorelines presented spatially asymmetric progradation and discontinuous dynamic features, where most estuarine segments advanced seaward remarkably, whereas the coastal zone in the southeastern Ca Mau Peninsula suffered continuous shoreline retreat. The spatio-temporal evolution of the Mekong Delta shoreline displayed prominent spatial differentiation: estuarine coasts experienced persistent seaward accretion, while the southeastern coast of the Ca Mau Peninsula underwent severe erosion and retreat. Such two opposite geomorphic processes were jointly controlled by the long-term coupling effects of natural sedimentation, tidal hydrodynamic evolution, sea level rise, land subsidence, mangrove degradation, hydraulic engineering construction, tidal flat reclamation and groundwater over-exploitation. [Conclusions] This study systematically supplements a long-term sequential geomorphic monitoring dataset for the Mekong Delta and clarifies the spatially divergent evolution law of delta sandbars and shorelines under the joint disturbance of natural processes and human activities. The reduction in sandbar quantity alongside the growth in total delta area verifies that sediment convergence and small sandbar amalgamation dominate the geomorphic accretion process of the Mekong estuary, while asymmetric shoreline migration reveals that regional hydrodynamic redistribution and anthropogenic engineering have reshaped the traditional natural sediment transport pattern of the delta. From the practical perspective, the severe coastal erosion occurring in the Ca Mau Peninsula reminds local management authorities to strengthen mangrove wetland protection, restrict excessive groundwater exploitation and optimize the layout of coastal hydraulic projects to mitigate land subsidence and wave-induced coastal retreat. The research innovatively integrates sandbar morphological metrics and multi-temporal shoreline dynamic indicators to interpret delta geomorphic evolution, which can provide methodological references for long-term coastal geomorphic monitoring in other large river deltas worldwide.

[Objective] This study aims to analyze research literature in the field of water ecological health assessment in the Yangtze River Basin using bibliometric software, systematically summarize development trends, frontier hotspots, knowledge clusters, and evolutionary pathways of research in recent years, and review research progress and main contents, thereby providing reference for the future research and protection of water ecological environment in the Yangtze River Basin of China. [Methods] Bibliometrics and knowledge graph methods are adopted. Using VOSviewer and CiteSpace software, this study systematically analyzes the research literature related to water ecological health assessment in the Yangtze River Basin from 2000 to 2024 in the Web of Science Core Collection (WoSCC) database (561 papers) and China National Knowledge Infrastructure (CNKI) database (364 papers), and identifies the clustering of research hotspots and the evolution patterns of burst keywords. [Results] The results show that: (1) the research development exhibits three-stage characteristics. In the early stage, the annual average number of publications was less than 10. In the middle stage, the annual number of publications gradually increased driven by policies. After 2017, the annual number of publications increased rapidly, showing a significant coupling relationship with the implementation of ecological protection policies in the Yangtze River Basin. (2) WoSCC research focuses on the pollution-driven ecological risk assessment mechanisms, while CNKI research emphasizes innovation in assessment methods and the diagnosis of biological integrity thresholds. The former mainly addresses micro-level mechanisms, while the latter focuses on macro-level management. (3) In the new era, research priorities have shifted toward multidimensional comprehensive assessment of water ecosystems, supporting precise health assessment of the river basin and facilitating the fundamental shift of the “14th Five-Year Plan” river and lake management objectives toward “water ecosystem health evaluation”. [Conclusion] Future research should focus on deepening interdisciplinary integration, further strengthening the capacity building of water ecological monitoring, promoting the deep integration of intelligent monitoring technology and multi-interface coupling models, developing comprehensive evaluation indicators of multi-group biological indicators, improving the monitoring indicators and standard system for water ecological health assessment, and providing scientific support for the protection and research of water ecological health in the Yangtze River Basin.

[Objective] This study aims to evaluate the long-term effects of the impoundment and operation of the Three Gorges Reservoir on the shoreline ecological environment of the middle and lower reaches of the Yangtze River, thereby providing scientific evidence and technical support for subsequent policy formulation and the implementation of the Yangtze River protection strategy. [Methods] The shoreline of the Zhicheng-Hukou section in the middle reaches of the Yangtze River was selected as the study area. Satellite images for four representative years (2004, 2010, 2016, and 2023) at the same water levels combined with UAV real-scene imagery from specific sampling sites during 2021-2023 and water level data from hydrological monitoring stations were utilized to systematically analyze the spatiotemporal dynamic evolution characteristics of the shoreline space after the impoundment of the Three Gorges Reservoir. Remote sensing images were used to calculate large-scale changes in shoreline water surface area for each section in the four representative years, and UAV data were employed to analyze shoreline slope and inundation at selected sampling sites. [Results] (1) After 20 years of impoundment and operation of the Three Gorges Reservoir, the water surface area and shoreline area in the Zhicheng-Hukou section of the middle reaches of the Yangtze River remained generally stable, with the standard deviation of the large-scale water surface area proportion controlled within 2.4%. Additionally, sedimentation occurred in local areas, particularly at river bends, requiring enhanced monitoring of these areas in the future. (2) Analysis of shoreline inundation and slope changes showed that from 2021 to 2023, slope variations along parts of the middle reaches of the Yangtze River shoreline were not significant. At the Zhijiang and Shishou sampling sites, the standard deviation of shoreline slope distribution remained within 2.0%, indicating overall stability. [Conclusion] The impoundment and operation of the Three Gorges Reservoir have a limited impact on the shoreline spatial patterns of the middle reaches of the Yangtze River, and overall changes in shoreline space in the middle and lower reaches of the Yangtze River remain minor. Overall water surface area and shoreline slope remain stable. However, shoreline area has changed in certain local sections, indicating that monitoring and management need to be continuously strengthened in the future. The innovations of this study lies in 1) integration of multi-source monitoring data to achieve high-precision monitoring of spatial dynamic changes in the shoreline of the middle reaches of the Yangtze River; 2) exploration of the influence mechanisms of the Three Gorges Reservoir impoundment on shoreline spatial patterns from two dimensions: large-scale water surface area changes and local shoreline morphological evolution.

[Objective] In response to the persistent high-temperature drought of the highest intensity since 1961 that occurred across most of the Yangtze River Basin during the summer and autumn of 2022, the Changjiang Water Resources Commission successively launched two rounds of special campaigns to combat drought, ensure water supply, and secure autumn grain harvest. A digital twin platform for drought relief scheduling in the Yangtze River Basin was developed and scenario simulations for drought defense were conducted, providing technical demonstration for the construction of a digital twin system for drought defense in the river basin. [Methods] The platform was developed in line with the operational requirements of early warning, simulation, and contingency planning. Using the platform, the severe drought in the Yangtze River Basin in 2022 were simulated. [Results] (1) Guided by the requirements of drought relief based on prediction, early warning, simulation, and contingency planning, and combining needs such as drought relief information management, emergency response, drought disaster verification and assessment, drought relief benefit evaluation, and drought relief contingency planning, a digital twin platform for drought relief and water replenishment scheduling was developed. Functions such as monitoring and alarm, prediction and early warning, scheduling simulation, contingency plan consultation, and user management were realized. (2) The severe drought in the Yangtze River Basin in 2022 was selected as a case study. Scenario simulations were conducted along the whole chain and entire process of “current drought diagnosis,future trend analysis,drought relief scheduling simulation,contingency plan consultation and decision-making”. “Current drought diagnosis” included drought monitoring data access, over-limit alarm for monitoring data, and monitoring of drought-related online public opinion, addressing the question of “where is the drought occurring?” “Future trend analysis” included drought prediction data access, prediction model calculation, and over-limit early warning for prediction data, addressing the question of “how will the drought evolve?” “Drought relief scheduling simulation” included the construction of a knowledge base of drought relief scheduling schemes and drought relief scheduling schemes based on knowledge base and scenario simulation, addressing the question of “how should the reservoirs be operated?” “Contingency plan consultation and decision-making” included drought relief emergency plan query, automatic generation of drought relief reports, and intelligent response of drought relief knowledge base, addressing the question of “what actions should be taken for drought relief?” [Conclusion] (1) A digital twin platform for drought relief and water replenishment scheduling is developed, and scenario simulation of the severe drought in the Yangtze River Basin in 2022 is conducted, forming a model case of a basin-level digital twin system for drought defense. (2) Future research should focus on water inflow and demand prediction models for the river basin, big data analysis of online public opinion, and intelligent matching and rolling optimization of drought relief scheduling scenarios, to effectively improve the intelligent level of drought relief and disaster reduction management in the Yangtze River Basin.

[Objective] To address the problems of insufficient 3D scene accuracy, unsatisfactory dynamic water effects, and imprecise hydrodynamic process simulation in existing river-lake digital twins, this study proposes an integrated method for 3D river-lake scene construction and hydrodynamic process simulation based on Unreal Engine 5 (UE5). The proposed method aims to construct a digital twin framework that combines high-fidelity scene representation with high-accuracy hydrodynamic process simulation, and to enhance the visualization, dynamism, and interactivity of river-lake digital twins. [Methods] UE5 was used as the research platform, and a real-scene 3D hydrodynamic process simulation method for river-lake scenarios was proposed and implemented by integrating terrain construction, water body simulation, and dynamic extraction of hydrological parameters. First, high-precision 3D river-lake terrain and environmental scenes were constructed using terrain height maps and high-precision photogrammetric models. Second, the Fluid Flux water simulation plugin in UE5 was modified by incorporating bottom friction factors influenced by the Manning coefficient, as used in engineering analysis, thereby establishing a hydrodynamic process model that better conformed to engineering practice. Finally, Blueprint programs were designed to dynamically extract and compute hydrological process parameters during simulation, enabling real-time calculation and dynamic extraction of key hydrological parameters such as flow velocity, water depth, watershed area, total water volume, and river cross-sections. An interactive user interface was also developed to support parameter visualization and scene interaction. [Results] A complete river-lake digital twin framework was constructed, and its functional effectiveness was verified through multiple experiments. First, a dam-break simulation experiment in a 90° bend was constructed to simulate the diffusion process of dam-break flow. The trends of water level variations at all measurement points showed good agreement with classical experimental data, validating the reliability of the hydrodynamic model. Subsequently, inundation simulation experiments under different vegetation cover conditions were conducted. These experiments reflected the influence of vegetation density on flow resistance and inundation processes in the simulated scenarios, demonstrated the capability of surface roughness variations to affect flow simulation, and verified that the proposed method could simulate the impacts of different vegetation environments on hydrodynamic processes. Finally, a complete 3D river-lake scene integrating 3D scenarios, dynamic water simulation, real-time hydrological parameter extraction, and an interactive interface was presented. Through the interface, users could obtain hydrological parameters such as water depth, flow velocity, cross-sectional morphology, watershed area, and total water volume at any location in real time, facilitating clear data acquisition and subsequent processing. [Conclusion] This study investigates methods for 3D scene construction of rivers and lakes and for hydrodynamic process simulation within such scenes, and successfully constructs a river-lake scene framework that integrates high-precision 3D scenes with hydrodynamic process simulation using UE5. The main innovations of this study lie in clarifying the method for constructing 3D river-lake scenes in the UE5 environment, generating terrain base surfaces using the terrain system and elevation data, and introducing high-precision photogrammetric models to enrich the surface environment, thereby improving the realism of 3D river-lake scene construction. From an engineering analysis perspective, the hydrodynamic model of the Fluid Flux plugin in UE5 is improved by adding a friction term influenced by the Manning coefficient, enabling hydrodynamic process simulation in 3D scenes to more accurately reflect the influence of environmental roughness. Simulation scenarios are also designed to verify the impacts of different terrain and vegetation roughness on flow simulation. In addition, Blueprint programs are designed to dynamically extract and compute various hydrological elements during the simulation of 3D river-lake scenes, forming a complete method for hydrodynamic process simulation in 3D river-lake scenes. The proposed method provides integrated capabilities for scene construction, hydrodynamic process simulation, and hydrological parameter extraction and computation, thereby improving the efficiency of data acquisition and processing during 3D scene simulation.

[Objective] To systematically review the development trends in terrestrial carbon sink research, this paper employs bibliometric methods to analyze Chinese and English literature from 1994 to 2024, aiming to reveal its evolution, research hotspot shifts, distribution of core research entities, and future directions, thereby providing a reference for grasping field frontiers and supporting relevant decision-making. [Methods] Data were sourced from the Web of Science Core Collection (SCI-E) and the China National Knowledge Infrastructure (CNKI) database, retrieving 8 431 relevant publications on terrestrial carbon sinks from 1994 to 2024 (4 727 in English, 3 704 in Chinese). Using CiteSpace software, knowledge graphs including literature co-citation, author collaboration, and keyword timeline maps were constructed. Combined with the burst detection algorithm, the analysis covered the temporal, disciplinary, journal, and country distributions of publications, identified high-impact institutions, prolific authors, high-centrality and highly-cited literature, and investigated the evolutionary stages and frontier hotspots of keywords. [Results] (1) Publication trends: The number of publications showed significant growth over the past three decades, with an average annual increase of approximately 12% after 2008 and 15% after 2019. The development process could be divided into three stages: slow inception (1994-2008), steady development (2008-2019), and rapid advancement (post-2019). (2) Distribution by country, journal, and discipline: China published the most papers (2 149 in WOS, 3 704 in CNKI), followed by the United States and Germany. Key journals included Global Change Biology, Science of the Total Environment, Agricultural and Forest Meteorology, Acta Ecologica Sinica, and Ecological Economy. Environmental science and technology and resources science and technology were core disciplines. Chinese literature emphasized forestry and agricultural economics, while English literature focused on ecology and geosciences. (3) Core research entities: The Chinese Academy of Sciences was the most prolific institution (832 papers), followed by the University of Chinese Academy of Sciences and the French National Centre for Scientific Research (CNRS). High-centrality and highly-cited literature concentrated on three areas: the dynamics of terrestrial carbon sinks, carbon flux monitoring models, and the coupling between climate change and carbon sinks. (4) Evolution of research hotspots: Keyword burst detection revealed three distinct stages. Stage 1 (1994-2008): focused on fundamental carbon cycle theory, with keywords such as “carbon cycle”,“carbon balance”, and “eddy covariance”. Stage 2 (2008-2019): research expanded to socio-economic dimensions, featuring keywords like “net primary production”,“low-carbon economy”, and “ecological compensation”. Stage 3 (2019-2024): closely aligned with global carbon reduction goals and ecosystem value realization, highlighted by keywords including “carbon neutrality”, “carbon peak”,“carbon emissions”, “temperature sensitivity”, “ecological product accounting”, and “carbon trading”. [Conclusion] This bibliometric analysis indicates that terrestrial carbon sink research is developing rapidly, with China being a major contributor. Research hotspots have evolved from fundamental mechanisms to socio-economic integration, and further toward a trajectory driven by carbon neutrality goals and market mechanisms. Future research directions mainly include: enhancing carbon sink monitoring, accounting, and assessment accuracy; studying carbon process mechanisms in coupled multiple ecosystems; evaluating regional emission reduction and sink enhancement potential along with technological applications; developing economic valuation and market trading mechanisms for carbon sinks; and strengthening international cooperation and data sharing. The findings can provide a basis for understanding the field’s development trajectory, predicting future trends, and supporting policies related to China’s “Dual Carbon” goals.

[Objective] To address the challenges of incomplete remote sensing imagery and insufficient measured terrain data in large-scale river numerical simulations, this study proposes a novel method for river channel terrain reconstruction that integrates Copernicus DEM 30 data with limited measured data, aiming to solve the problem of constructing large-scale river channel DEMs in data-scarce areas and to verify its applicability in MIKE 21 hydrodynamic-sediment numerical simulations. [Methods] This study took the Alar-Xinquman river section of the Tarim River as the study area and utilized relevant hydrological data and measured terrain elevation data from 2011 as the basic dataset. The specific reconstruction process was as follows: (1) Using ArcGIS software, combined with Google Earth historical imagery and Copernicus DEM 30, the inner and outer bank lines of the river channel were delineated by scaling overlapping imagery proportionally to determine the main channel boundary.(2) The river DEM was corrected stepwise. First, the quadratic interpolation method was applied to densify the elevation data of the main channel cross-sections. Second, the mean difference method was used to calculate the deviations in elevation between measured points and the Copernicus DEM at corresponding locations, enabling an overall vertical correction of the Copernicus DEM data. Next, a locally weighted regression (LOESS) algorithm was introduced to correct the longitudinal cross-section elevations of the main channel, smoothing the riverbed terrain and generating the main channel DEM. Finally, the main channel and floodplain DEM data were merged to construct the complete river channel DEM.(3) The reconstructed DEM was imported into the MIKE 21 hydrodynamic-sediment module. Measured data such as water level, flow, and sediment concentration were selected as boundary conditions for numerical simulation. The accuracy and reliability of the reconstructed DEM were evaluated by comparing the errors between the simulated and measured values. [Results] Comparison between numerical simulation results and measured data revealed that the results for flow velocity, flow-stage relationships, and river channel erosion-deposition tests all met the allowable deviation requirements specified in relevant technical standards. This indicated that the river channel DEM constructed in this study was reasonably suitable for two-dimensional hydrodynamic-sediment numerical simulations. [Conclusion] (1) The river channel terrain reconstruction method proposed in this study, integrating Copernicus DEM 30, quadratic interpolation, mean difference method, and locally weighted regression (LOESS) algorithm, can effectively address the lack of measured underwater terrain data for large-scale rivers. It demonstrates superior continuity and accuracy compared to traditional single-interpolation methods.(2) The hydrodynamic model established based on the reconstructed DEM achieves simulation accuracy within the allowable deviation limits specified by relevant technical standards. This method is characterized by low cost, high efficiency, and strong applicability, providing a practical new approach and technical support for the numerical simulation of large-scale rivers with scarce data, such as the Tarim River.

[Objective] With the development of deep learning and computer vision technologies, the application of machine vision in flood monitoring has gradually become a research hotspot. This study aims to overcome the limitations of traditional manual and satellite remote sensing methods in flood monitoring and early warning, including insufficient accuracy and high costs, and to explore the advantages and limitations of the Segformer model in extracting flood boundaries, while proposing future research directions and improvement strategies. [Methods] Using deep learning and machine vision techniques, a specialized “RiverDataset” for flood monitoring was constructed, and the performance of the Segformer model in extracting flood inundation range was evaluated based on this dataset. Additionally, the Segformer model was compared with the U-Net models based on ResNet50 and VGG16 to assess their performance in water body segmentation tasks. Taking Shashi District, Jingzhou City, Hubei Province as an example, UAV remote sensing imagery was used to accurately extract the flood contours in the region. [Results] The U-Net (VGG16) model demonstrated excellent performance on the training set but was slightly inferior to the Segformer model on the validation and test sets. The Segformer model achieved superior performance across most indicators, particularly outperforming the U-Net (ResNet50) model in complex scenarios. Although the U-Net (ResNet50) model achieved a slightly higher IoU, its higher loss value and lower mIoU and mAP indicated that its overall performance was inferior to that of the Segformer model and the U-Net model (VGG16). Consequently, the U-Net model based on VGG16 performed well across all evaluation indicators, demonstrating strong fitting capability during training. However, when processing complex water bodies, this model struggled to capture contextual information in narrow regions due to its limited receptive field, leading to frequent occurrences of information voids. Furthermore, the U-Net model failed to effectively eliminate gridding effects, compromising the local consistency of feature information. In contrast, the Segformer model did not exhibit this issue. This difference was mainly due to the relatively small receptive field of the convolutional kernels used in the U-Net model, which restricted its ability to interpret contextual information within narrow regions and hindered the establishment of long-range information continuity. The Segformer model, not constrained by the limited receptive field of conventional convolutional kernels, could better capture broader contextual information in the image. Additionally, the U-Net model failed to effectively eliminate the impact of the grid effect on feature information, resulting in extracted features that disrupted the local consistency of information. The absence of structures for perceiving local regional information made the model incapable of effectively eliminating the influence of interfering objects in complex aquatic environments, particularly in scenarios requiring precise boundary extraction. [Conclusions] In complex aquatic environments, the Segformer model demonstrates superior segmentation performance and robustness. This study validates the efficiency of the Segformer model in extracting flood inundation range, highlighting its potential for practical water body monitoring applications. Future research should further optimize the model, expand the dataset, and explore its potential for real-time application to enhance the efficiency and accuracy of flood early warning systems.

[Objective] Curtain grouting is a commonly used method for anti-seepage reinforcement in pumped storage projects, and injection rate is an important monitoring parameter that directly affects grouting quality. The fluctuation section of injection rate is a critical stage in the grouting process. At present, research on the analysis of injection rate monitoring data is scarce, scientifically sound criteria for reasonable injection rate are lacking, and it is difficult to guide the adjustment of grouting pressure. This paper proposes a method for analysing and calculating the injection-rate interval of the fluctuation section, providing scientifically sound intervals for different grouting scenarios. [Methods] By analysing the temporal evolution of injection rate during curtain grouting, the rate-time curve was divided into three stages—fluctuation, sharp-decline, and termination—and its patterns classified into four types: normal, sharp-decline, low-level injection, and non-convergent. Normal-pattern sections were selected as standard grouting segments, and two key parameters of the fluctuation section—unit-average injection rate and average slurry density—were calculated to indirectly represent the average geological conditions of the treated strata. Owing to the large scale of the grouting area and inherent geological variability of the treated strata, the unit-average injection rate exhibited high dispersion. Therefore, based on the intrinsic correlation between resource allocation decisions and geological information, the concept of “grouting similarity” and a dynamic geological-zoning approach were proposed. Correlation analysis between grouting parameters and unit-average injection rate selected GIN value, average slurry density, and hole sequence as energy, fracture, and sequence indices, respectively. Clustering was applied to standard grouting segments based on these indices to reduce dispersion. After de-noising the target parameters within each category using DBSCAN, based on the 3σ control principle in risk management, if the data followed a normal distribution, points falling outside the ±3σ range could, under the principle of controlling type Ⅰ and type Ⅱ errors, be identified as extreme outliers. The Shapiro-Wilk method was used to test the normality of each cluster, and for those that passed, the ±3σ interval was calculated as the injection rate interval for the fluctuation section. Based on the degree to which each segment's parameters deviated from the interval center, the specific interval of the fluctuation-section unit-average injection rate of completed segments was determined, and a preliminary post-grouting geological assessment was provided, thereby achieving data-driven quality management of curtain grouting. [Results] This method was applied to the curtain-grouting project of Wuyue Pumped-Storage Power Station for clustering analysis of completed grouting sections. Results showed that parameter distributions within most clusters satisfy the normality assumption. The calculated fluctuation-section injection rate intervals for each grouting-similarity pattern conformed to the similarity hypothesis and provided an effective control tool for injection rate process control and management. Based on these intervals, preliminary relative geological assessments precisely identified outlier sections with significant deviations from a large number of grouting segments and the proportion of outliers met quality-risk-management standards. [Conclusions] In summary, the proposed method for determining the injection rate interval of the fluctuation section is logically rigorous and reliable, significantly improving the scientific management of curtain-grouting construction in large-scale pumped-storage projects.

[Objective] Emergency response to breaches has high requirements for timeliness, information accuracy, overall situational awareness, and coordinated observation. The aim of this study is to meet the urgent practical need for dynamic real-time monitoring of breaches, providing technical support for rapid post-disaster assessment and decision-making for flood control emergency rescue. [Methods] Considering the constraints of time, space, frequency, and spectrum, we developed a multi-source remote sensing collaborative monitoring method by using the fuzzy multi-attribute decision-making method to evaluate the dynamic monitoring capabilities of different collaborative schemes in response to breaches and their secondary flood disasters. The breach progression can be traced by inverting indicators such as breach width and inundated area within the embankment area. We further selected the “breach in the Tuanzhou Embankment of the Dongting Lake on July 5” as a case study for experimental analysis and validation. [Results] (1) under the breach monitoring scenario, the derived collaborative monitoring scheme achieved a comprehensive fitness value of 0.654. The breach width variation curve exhibited a general consistency with the trend of in-situ hydrological monitoring data. The relative error between the breach width derived from collaborative inversion and hydrological measurement was 0.98%, demonstrating high accuracy suitable for emergency monitoring. (2) The coordination monitoring scheme under the flood monitoring scenario achieved a comprehensive fitness value of 0.591. The inundation area variation curve within the embankment area was overall consistent with the trend of values calculated by the flood re-simulation model, with a relative error of -2.39%. This error primarily stemmed from the fitting accuracy of discrete values derived from multi-source remote sensing inversion and systematic errors in the re-simulation model calculations. [Conclusions] The developed multi-source coordination combination method integrates both the coordination monitoring process and the coordination data inversion, covering the entire monitoring period before, during, after the breach. It accurately reflects the piping and seepage processes and their propagation trends before the occurrence of breaches, thereby addressing the practical needs for timeliness, comprehensiveness, and coordination in responding to breaches and secondary flood disasters. In addition, the effectiveness of the experiments and applications in this study depends on the types and quantities of satellite resources that can be scheduled under emergency monitoring modes, as well as their alignment with monitoring task requirements. The more satellite resources involved in coordination and evaluation, the higher the monitoring frequency and coverage cycle can be increased, potentially improving the comprehensive fitness value.

[Objective] In water conservancy and hydropower engineering, grouting is a key construction process to ensure the stability and leakage safety of hydraulic retaining structures. However, traditional grouting monitoring equipment exhibits deficiencies in accuracy, level of intelligence, and construction efficiency. This study aims to achieve fully automated operation and early warning feedback control of the grouting process by innovatively designing a new generation of intelligent sensing and efficient control equipment for grouting parameters, thereby improving engineering management efficiency, construction quality, and safety. [Methods] This study designed an intelligent centralized slurry station, an intelligent pressure regulation and flushing device, an intelligent grouting control coordination center, and a digital grouting recording unit, and integrated them into a new generation of intelligent sensing and efficient control equipment for grouting parameters. Additionally, a multi-channel, user-friendly human-computer interaction interface based on multi-touch control was developed to realize the fully automated operation and unified coordination management of the grouting process, including slurry preparation, slurry delivery, pressure regulation, and grouting. [Results] Through experiments and practical engineering applications, this equipment was verified to achieve fully automated operation of the grouting process, significantly improving the accuracy and intelligence of grouting construction. By implementing early warning feedback control and unified coordination management functions during the grouting process, construction safety was effectively ensured, and project management efficiency was enhanced. Compared to traditional grouting monitoring equipment, construction efficiency was increased by more than 30%, and construction quality was significantly improved. [Conclusions] The new generation of intelligent sensing and efficient control equipment for grouting parameters shows remarkable effectiveness in the grouting construction of water conservancy and hydropower projects, achieving automated, intelligent, and efficient grouting process, providing strong support for high-quality project construction and showing broad application prospects.

[Objective] To address the issues of insufficient real-time sensing capability and low level of intelligent decision-making in water resource scheduling and management in irrigation areas, this study focuses on the design and implementation of a digital twin water resource scheduling and management platform for irrigation areas. [Methods] By integrating the Internet of Things, digital twins, hydraulic models, and irrigation area water resource scheduling and management systems, a water resource sensing and control system for irrigation areas was established. This system gathered massive sensing data and combined it with digital twin visualization technologies such as full-factor digital representation, multi-source heterogeneous data fusion, and high-fidelity 3D visualization. It enabled synchronous simulation and diagnostic analysis of the operational status of various components of the irrigation area (canal headworks, canal sections, regulating gates, distribution gates, and irrigation area) and multiple factors (water level, flow, water quality, engineering safety, gate opening, etc.). Based on the canal system hydrodynamic model and water resource scheduling evaluation model, a water resource scheduling decision-making system of “forward simulation of scheduling impacts and backward deduction of scheduling plans” was constructed. The dynamic simulation and analysis of water allocation plans generated by the integrated model of “inflow prediction-water demand prediction-dynamic canal system water distribution” were performed, continuously iterating and optimizing scheduling plans to assist in the scientific development of joint water resource scheduling plans for the canal gate groups in the irrigation area. [Results and Conclusions] The practical application of the digital twin Dujiangyan Irrigation Project (canal head hub) demonstrates that through digital mapping and intelligent simulation of all elements of the physical watershed and water resources management activities in the irrigation area, the digital twin irrigation area water resource scheduling and management platform can timely perceive the supply and demand status of water resources, as well as the operational conditions of irrigation water conservancy projects and sensing devices. This enables intelligent and refined allocation of water resources, comprehensively enhancing the modernization level of water resource management in irrigation areas.

[Objectives] This study aims to explore the feasibility of employing DeepSeek, a large language model, to promote intelligent hydrological analysis through its natural language interaction and code generation functions. This research innovatively applies DeepSeek to engineering hydrological analysis, promoting intelligent development in the field of engineering hydrology. [Methods] First, based on the core concepts and characteristics of engineering hydrology discipline, it was concluded that DeepSeek’s application scenarios such as code generation, code rewriting, and code explanation were highly suitable for engineering hydrology, a field heavily dependent on data. Focusing on the typical task of frequency analysis of hydrological data, this study used a case-driven method and designed a two-stage experiment. During the data cleaning phase, daily water level data incorporating compound water level recording methods were fed into the system, and MATLAB cleaning code was iteratively generated using structured prompts. In the data analysis phase, the annual maximum water levels, 3-day and 5-day moving average maximum sequences during the flood season were generated, and the Pearson Type III (P-III) distribution was used to calculate key frequency design values such as 1% and 5%. Finally, a quantitative comparison was conducted between DeepSeek’s calculated results and conventional eye-fitting curve outcomes to evaluate the accuracy of the results. [Results] In terms of efficiency, the processing time for multiple prompts ranged from 33 to 109 seconds. Standardized tasks (such as moving average calculations) achieved “prompt as code”, substantially reducing programming time and significantly enhancing workflow efficiency. Additionally, the automated optimization of existing inefficient code notably improved efficiency. Regarding accuracy, DeepSeek could accurately identify user requirements and precisely interpret professional concepts. It achieved a 100% accuracy rate in the first attempt when interpreting key concepts such as the P-III distribution and flood season averages. However, for low-frequency terms (e.g., compound recording method), 2-3 rounds of prompt iteration were required. Additionally, DeepSeek’s calculated average and C_v parameters were consistent with those obtained using conventional methods, further demonstrating its high precision. [Conclusions] DeepSeek significantly lowers the technical barriers to engineering hydrological analysis. Its natural language interaction capability serves as an “intelligent bridge” between professional requirements and code implementation, while its automated data processing and model calculation alleviate practitioners’ workload, promoting the integration of AI technology from academic research into engineering practice. In the future, with in-depth research and expanded applications, DeepSeek is expected to evolve from an auxiliary tool into a core engine driving the transformation of engineering hydrology from “experience-based decision-making” to “knowledge-data collaborative decision-making,” thereby providing foundational support for intelligent water conservancy.

[Objective] In the context of big data from mobile internet, social media data with tags such as posting time and location has received widespread attention for its critical role in natural disaster response. In China, research on social attention and online public opinion regarding drought events remains limited, especially for the analysis of spatiotemporal and thematic characteristics of extreme drought events at the river basin scale, with no relevant reports yet. [Methods] This study used the 2022 extreme drought in China’s Yangtze River Basin as a representative case. Utilizing texts from Weibo, a mainstream social media platform in China, as data sources, this study used machine learning and artificial intelligence algorithms to collect Weibo text data throughout the drought progression process. The Latent Dirichlet Allocation (LDA) topic model was employed to perform term clustering and thematic characterization. Through this methodology, an in-depth mining of spatiotemporal and thematic characteristics of drought-related public opinion was conducted, along with sentiment analysis. [Results] (1) The temporal evolution of attention levels on social media was relatively synchronized with the progression of the drought event, with peak drought stage particularly prone to attracting heightened public attention. Across the entire Yangtze River Basin, drought-related discussions on social media remained relatively low in July 2022, rose dramatically in early August, peaked in mid-to-late August, gradually declined in mid-September, and returned to zero in early December. In terms of drought progression, an inverse correlation between the temporal variation characteristics of Weibo discussion level in severely affected provinces and municipalities including Sichuan, Chongqing, and Jiangxi and local hydrological flow data was observed. (2) The spatial characteristics of attention levels on social media basically matched the distribution of drought severity. The proportion of Weibo discussions in high-attention provinces and municipalities (e.g., Sichuan, Chongqing, and Jiangxi) exceeded 50%, reflecting widespread public concern about the drought and indirectly indicating severe socioeconomic impacts caused by the drought in these regions. In contrast, provinces and municipalities such as Yunnan, Tibet, Shanghai, and Qinghai showed relatively low levels of Weibo discussions. (3) The thematic characteristics of drought-related content on social media showed significant regional differences, with public attention levels being closely related to the severity of drought impacts. In Jiangxi and Hunan, key terms related to the drought were “shrinking of Poyang Lake” and “declining water levels” In Sichuan and Chongqing, key terms were secondary disasters such as “wildfires”, “earthquakes”, as well as drought-induced issues such as “reduced crop production by farmers” and “electricity supply shortages”. Other provinces primarily focused on “continuous high-temperature weather” and “meteorological drought”. As the drought progressed, the sentiment of public opinion on drought gradually transitioned from negative to positive. [Conclusion] Weibo texts serve as an effective data source for online public opinion analysis of sudden-onset disasters. The research findings can provide technical support for drought tracking analysis and mobilization efforts of the public for drought relief in river basins.

Irrigated area is the basic data required for effective agricultural water conservation, yet traditional survey and statistical methods no longer meet current monitoring needs. In this research, GF-1 and Sentinel-2 satellite images were fused to construct the sample spectrum of crop growth period. Based on the pixel-scale spectral matching method, the crop planting structure and actual irrigated area of Zaohe irrigation district in Suqian City, Jiangsu Province from 2017 to 2022 were synergistically extracted. Results show that the main planting pattern in Zaohe irrigation district is rice-wheat rotation. From 2017 to 2022, the actual irrigated area was 85.11 km², 91.91 km², 103.65 km², 95.85 km², 97.72 km² and 88.24 km². respectively. Validation using sample points and a confusion matrix yielded an overall accuracy of 89.71% and a Kappa coefficient of 0.80, indicating higher accuracy and better extraction effects compared to existing products like IrriMap_Syn and IWMI products. This method is suitable for extracting the irrigated area in south China, and can provide technical and data support for the daily supervision of management departments and the optimization of water resource allocation.

The Three Gorges Reservoir (TGR) area features a large number of tributaries, but the lack of basic hydrological station monitoring leads to the scarcity of hydrological and water-resources information for these tributaries, affecting regional water resource management and flood control safety. To address this issue, we selected 20 typical tributaries in the TGR area, among which 18 are ungauged, to build 3D digital river models for these tributaries by self-developed remotely-sensed hydrological station technology along with satellite and unmanned aerial vehicle (UAV) remote-sensing data. Based on this model, we calculated the river discharges, relative water levels, water-surface widths, and other information from January 2016 to July 2023 for the monitoring sections. Results revealed that: 1) The cross-sections of 20 typical tributaries in the TGR area display U-shape, demonstrating mountainous characteristics. The remotely-sensed hydrological station technology demonstrates high accuracy in calculating ungauged tributary discharges in the TGR area, with the average Nash-Sutcliffe efficiency coefficient (NSE) and coefficient of determination (R²) reaching 0.74 and 0.76, respectively. 2) During the study period, the relative water levels of the tributaries changed minimally. The percentages of months with monthly relative water-level fluctuations below 1.2 m in the upper, middle, and lower tributaries of the reservoir were 93.1%, 86.8%, and 87.4% respectively. 3) The average discharges of typical ungauged tributaries were generally stable. However, trend analysis indicated that the discharges of 13 ungauged tributaries have being decreasing, suggesting an overall downward trend in tributary discharges in the reservoir area. 4) The 20 tributaries have abundant annual average and total discharges, with the annual average inflow reaching 16.475 billion m³, accounting for approximately 4.8% of the total annual inflow of the reservoir area, offering substantial water resource support for regional economic development.

In the remote sensing monitoring for cyanobacteria blooms, the bloom area is a critical indicator for assessing the severity of the bloom and is crucial for relevant authorities in selecting preventive measures and determining emergency response levels. Traditional methods using medium-to-low-resolution imagery has limited precision in estimating bloom area. To address this issue, the cyanobacteria bloom areas of Taihu Lake as the study area extracted from Sentinel-2 and Sentinel-3 data were compared. Furthermore, the relationship between the NDVI from Sentinel-3 imagery and the cyanobacteria bloom area proportion within mixed pixels were analyzed. Based on these analyses, a corrected model for estimating bloom area was established using the NDVI density segmentation method to refine the bloom area extracted from Sentinel-3 images. The statistical results of bloom areas derived from Sentinel-3 with correction, Sentinel-3 without correction, and Sentinel-2 were compared and analyzed. The findings demonstrate that the corrected model significantly improves the accuracy and reliability of bloom area estimation using Sentinel-3 imagery compared to traditional methods, thereby enhancing its practical application value in cyanobacteria bloom monitoring.