Long-term seasonal streamflow forecasting with lead times up to 12 months is crucial for water security but remains a persistent challenge, plagued by the high uncertainty of General Circulation Model (GCM) climate projections and a lack of consensus on optimal modeling paradigms. To overcome these limitations, a systematic evaluation of more practical and reliable approaches is urgently needed. This study addresses this critical gap by developing and evaluating a comprehensive framework of autoregressive, data-driven, process-driven, and hybrid modeling approaches. As a key innovation, we evaluate these paradigms not only with standard Coupled Model Intercomparison Project (CMIP) projections but also with novel climate forcings derived from historical data using a newly developed comprehensive similarity evaluation index. The study focuses on simultaneous prediction of the monthly streamflow process for the next 12 months in three snow-dominated catchments in the Yellow River’s source region. Our results reveal that simple autoregressive models provide a robust baseline, consistently achieving a Normalized Nash–Sutcliffe Efficiency (NNSE) exceeding 0.75 across both simulation and forecasting periods. Critically, climate data from hydrological similarity years offered a superior alternative to GCM-driven forecasting, outperforming CMIP projections with NNSE improvements of up to 26.28

To improve the accuracy of watershed runoff prediction and address the lack of interpretability in data-driven hydrological models, we propose a spatiotemporal prediction model, named the multigraph attention mechanism with gated recurrent unit (MGAT-MIC-GRU). This model leverages a multigraph attention mechanism, combining graph attention network (GAT) and gated recurrent neural network (GRU), for runoff forecasting. Firstly, employing hydrological data from the Qishui River Basin, a graph neural network, is utilized to extract the topological structure of the watershed site and generate feature vectors. Secondly, according to the characteristics of hydrological time series data, a multigraph attention runoff spatiotemporal prediction model based on MIC mechanism is established to predict the runoff of the basin. Finally, we conduct comparative experiments over multiple forecasting horizons to evaluate the predictive performance of various models, including the MGAT-MIC-GRU model. The research findings demonstrate that our proposed model improves the accuracy of runoff prediction.

Runoff simulation is a critical component of flood prevention and mitigation strategies, as well as water resource management in watersheds. This study proposes a rolling forecasting model based on Convolutional Neural Networks (CNN), Long Short-Term Memory networks (LSTM) and attention mechanisms. The model integrates the feature extraction capabilities of CNN, the strengths of LSTM in time series modeling, and the weighted focus of attention mechanisms to enhance the accuracy of runoff forecasting. Rolling forecasting enables the model to continuously update and adapt based on the latest data, making the forecasting process more dynamic and responsive. The model is further verified and applied in the Wei River basin, the largest tributary of the Yellow River basin of China. Results indicate that the CNN-LSTM-Attention (CLA) coupling model achieves R2 greater than 0.80 and NSE greater than 0.70, demonstrating greater precision in daily runoff forecasting with respect to traditional models and effectively capturing complex features in time series data. The rolling forecasting results of CLA coupling model exhibits high accuracy and stability across various hydrological stations, with both R2 and NSE exceeding 0.90. This study proposes a novel technical approach for runoff forecasting and offers a scientific basis for watershed water resource management and scheduling.

Scientific flood control emergency plans serve as the primary support for enhancing the efficiency of flood emergency management. This study proposes a hybrid entity recognition model based on Bidirectional Long Short-Term Memory (BiLSTM), Conditional Random Fields (CRF), and the Ratcliff/Obershelp algorithm. A basin flood control knowledge graph is established, and a dynamic emergency plan generation framework based on hybrid reasoning is proposed. The case study indicated that the entity recognition model achieves a precision of 0.985, a recall of 0.990, and an F1-score of 0.987. The framework-generated plan reaches an 0.810 similarity with the 2023 historical flood control emergency plan of Tongchuan city of Shaanxi Province of China. This research offers an efficient entity recognition solution and an intelligent decision-making framework for the field of basin flood control, which can help to improve the precision of flood control management and emergency response efficiency.

The acceleration of the global water cycle has led to a remarkable increase in the frequency of extreme disasters such as droughts and floods under changing environment. Drought-flood abrupt alternation (DFAA) events, as compound events more destructive than a single flood and drought event, pose severe threats to the sustainable development of ecosystem and society. This study applied the short-cycle drought-flood abrupt transition index (SDFAI) to identify extreme DFAA events in the Wei River Basin (WRB), the largest tributary of the Yellow River Basin of China, extracting their spatiotemporal evolution and non-stationary characteristics. By integrating meteorological factors and teleconnection indices, the drivers and inherent mechanism of DFAA events were ascertained. The results indicate that: (1) Flood-to-drought (FTD) were more severe than drought-to-flood (DTF) during the rainy season. DTF events are projected to dominate after 2016, and prominent mutations in the SDFAI, Min-FTD, and Max-DTF occurring around 1980 and 1995-2001, respectively. (2) Under the main oscillation period, the periods of extreme FTD and DTF events in the WRB were concentrated at 18-30 years and 10-30 years, respectively. (3) Both meteorological factors and teleconnection indices significantly affected the DFAA events. The teleconnection indices mainly associated with the sub-basins were Southern Oscillation Index, Arctic Oscillation, and North Atlantic Oscillation, which exhibited significant resonance periods with extreme FTD and DTF events. The combined contribution of meteorological factors to DFAA events was more susceptible, and the joint effects of teleconnection indices exerted a greater influence on the contribution rate of meteorological factors.

Water resource management in irrigation districts is of pivotal importance for safeguarding agricultural production, promoting regional economic development, and maintaining social stability. The development of a scientific and reasonable water allocation model has emerged as a pivotal research area in the context of sustainable development of irrigation districts, particularly in the context of climate change. This study explores the use of a combination of methodologies for agricultural water management under uncertainty conditions, including Bayesian network, interval parameter programming, and a bi-level multi-objective programming model approach. The Bayesian network has been demonstrated to quantify the nonlinear effects of precipitation, temperature, evaporation, and other factors on water diversion. The bi-level multi-objective programming model is designed to balance economic efficiency and social equity between the macro and micro decision-making levels. A case study conducted in Jiaokou Irrigation District, Shaanxi Province of China, demonstrated the adaptability of the model under different scenarios. The findings indicate that the aggregate benefits of the irrigation district under the normal scenario can amount to 3.30 x 109 yuan. The fluctuation in the mean value of water diversion is less than about 15 %. The Gini coefficient is maintained within the range of 0.02-0.29. This study rationally and dynamically allocates water resources through multi-model coupling, achieving scientific management of agricultural water resources. It offers novel concepts and specialized technical assistance for the harmonized implementation of the Sustainable Development Goals at the irrigation district level, with the objective of addressing the progressively intricate challenges posed by water scarcity and uncertainty.

IntroductionAccurate identification of environmental issues in river and lake ecosystems is essential for the protection, management, and sustainable use of water resources. Traditional inspection-based approaches are limited by their extensive spatial scope, high labor demands, prolonged execution time, and increased likelihood of overlooking hazards.MethodsTo overcome these limitations, this study investigates intelligent methods for detecting environmental hazards in river and lake settings. Images representing 12 common types of water-related hazards were collected. Using image augmentation techniques, including rotation, transformation, and annotation, a dataset comprising over 1,500 samples of river and lake environmental hazards was constructed. An intelligent recognition model was then developed based on the YOLOv11 algorithm, incorporating transfer learning techniques to enable the detection of pollution categories, pollutant types, sewage outfalls, and shoreline encroachments.ResultsThe experimental results demonstrate that, with adequate training data, appropriate categorization, and accurate annotation, the proposed method achieves reliable performance, yielding a balanced F1 score of 0.72.DiscussionThis approach can be deployed on devices such as smartphones, cameras, and unmanned aerial vehicles, offering practical tools for water pollution surveillance, shoreline monitoring, and the broader management of aquatic ecosystems.

Mine water is both wastewater and a valuable unconventional water resource, and its recycling is crucial for the sustainable development of coal-resource-based cities. In response to the complex interactions among multiple stakeholders in the process of mine water recycling, this study innovatively develops a four-party evolutionary game model involving local government, coal mining enterprises, mine water operators, and water users. For the first time, key variables—mine water pricing, water volume, water rights trading, water resource taxation, and objective utility of water resources—are systematically integrated into a multi-agent game framework, extending the analysis beyond conventional policies, such as penalties and subsidies, to explore their impact on recycling behavior. The results show the following: (1) There are 10 possible evolutionary stabilization strategies in the system. The current optimal strategy includes supply, input, use, active support, while the ideal strategy under the market mechanism includes supply, input, use, passive support. (2) Local governments play a leading role in collaborative governance. The decisions of coal mining enterprises and mine water operators are highly interdependent, and these upstream actors significantly influence the water users’ strategies. (3) Government subsidies exhibit an inverted U-shaped effect, while punitive measures are more effective than incentives. The tax differential between recycled and discharged mine water incentivizes coal enterprises to adopt proactive measures, and water rights trading significantly enhances the users’ willingness. (4) Mine water should be priced significantly lower than fresh water and reasonably balanced between stakeholders. Industries with lower objective utility of water tend to prioritize its use. This study provides theoretical support for policy optimization and a market-based resource utilization of mine water.

Traditional methods for identifying hydrologically similar years, such as Euclidean Distance (ED) and Dynamic Time Warping (DTW), often fail to balance magnitude consistency with temporal pattern alignment. To address this, we propose a comprehensive Similarity Index (SI) that dynamically integrates six methodologies: ED, DTW, Fréchet Distance, Hausdorff Distance, Longest Common Subsequence Similarity, and Cosine Similarity (CS). Using monthly data from 15 Yellow River Basin stations, we systematically compared ten identification schemes, including weighted SI variants and a hybrid CS + ED model. Performance was assessed via a dual-dimensional framework: pattern similarity (based on Kendall’s τ and Theil-Sen slope) and magnitude similarity (based on Normalized Root Mean Square Error (NRMSE), Nash-Sutcliffe Efficiency (NSE), and Q-Q plot correlation). Results indicate that while the CS + ED hybrid achieved the highest similarity score (0.8347), the SI family demonstrated superior cross-station stability, yielding the lowest coefficient of variation (4.2415). Practical value analysis using a simple linear runoff prediction model demonstrated that the proposed similarity indices achieved significant improvements in runoff forecasting. The SI method showed a 13.47

The construction of large-scale, dynamic datasets for specialized domain models often suffers with problems of low efficiency and poor consistency. This paper proposes a method that integrates multi-role collaboration with automated annotation to address these issues. The framework introduces two new roles, data augmentation specialists and automatic annotation operators, to establish a closed-loop process that includes dynamic classification adjustment, data augmentation, and intelligent annotation. Two supporting tools were developed: an image classification modification tool that automatically adapts to changes in categories and an automatic annotation tool with rotation-angle perception based on the rotation matrix algorithm. Experimental results show that this method increases annotation efficiency by 40% compared to traditional approaches, while achieving 100% annotation consistency after classification modifications. The method’s effectiveness was validated using the WATER-DET dataset, a collection of 1500 annotated images from the water conservancy engineering field. A model trained on this dataset achieved an F1-score of 0.9 for identifying water environment problems in rivers and lakes. This research offers an efficient framework for dynamic dataset construction, and the developed methods and tools are expected to promote the application of artificial intelligence in specialized domains.

Studying the ameliorative effects of psammitic soft rock and combined water measures in Mu Us Sandy Land (northern Shaanxi, China) is of great significance for improving soil quality and ensuring food security in sandy regions. The 0–30 cm soil layer was filled with a mixture of soft rock and sand, with four volume ratios: 0 : 1, 1 : 5, 1 : 2, and 1 : 1 (soft rock: sand). The irrigation method used drip irrigation, with two irrigation regimes during the maize growth period: full irrigation (4350 m3/ha) and incomplete irrigation (2925 m3/ha), divided into 10 irrigation sessions. The soil nutrient content was measured annually after the maize harvest. Field trials were conducted, with each experimental plot covering 60 m2 and three field replicates for each treatment. The experimental data from four years (2020 to 2023, planting years 11 to 14) were selected for analysis. The results showed that the nutrient content of the soil increased significantly under incomplete irrigation. The average content of SOM (soil organic matter) and TN (total nitrogen) increased by 4.29 and 0.86 g kg–1, respectively, after 14 years of planting compared with 11 years, particularly in the 1 : 2 compound soil. The average content of AP (available phosphorus) under incomplete irrigation increased by about 71

Extension-based streamflow naturalization methods struggle with identifying mutations and selecting key features, while routing methods overlook the contribution of interstation runoff. This study proposes a Combined Extension and Routing (CER) approach to address these issues. The CER approach employs multiple change detection techniques to identify the earliest significant mutation and a multiple linear factors reconstruction method to select key features influencing natural flow. The CER models, implemented using extreme gradient boosting, Long Short-Term Memory (LSTM), Bidirectional LSTM (BiLSTM), Convolutional Neural Networks, and multiple linear regression, were evaluated in two snow-dominated catchments in the Yellow River, China. Results show that CER models effectively captured both peak and low flow events, achieving Nash-Sutcliffe efficiency of about 0.9 when comparing the estimation results from a water balance model. This study highlights the importance of stable land conditions for the CER approach's effectiveness, providing a reliable framework for natural streamflow estimation.

Urban flooding has become a recurrent issue, leading to significant losses for both people and the economy. Enhancing emergency response capability is crucial for effectively managing flood events. This paper taking Xi’an city of China for an example, applying the prevention-preparation-response-recovery (PPRR) theory to structure the assessment into four criterion layers including prevention, preparation, response and recovery. Assessment indicators are selected accordingly. Analytic Hierarchy Process (AHP) method is used to determine the weights of these indicators. Fuzzy Petri Net is used to evaluate the emergency response capacity of flood disasters. The results indicate that Xi'an City’s capacity is at a medium level. Key factors contributing to this rating include the arrangement of disaster prevention projects, implementation of emergency plans, the level of professional skills in disaster relief, and the disaster insurance system. The use of Fuzzy Petri Net proves to be effective in this assessment. The results offer valuable insights for improving urban flood disaster response capacity in Xi'an City, providing guidance for urban planners and policymakers to enhance flood preparedness and response strategies in rapidly urbanizing areas.

Due to the uncertainties of future climate change, the response of traditional stormwater management systems to climate change has become increasingly complex. This study proposes an optimization method for configuring urban Low-Impact Development (LID) facilities to adapt to future climate uncertainty and promote the achievement of sustainable development goals (SDGs). Design storms for climate change scenarios under shared socioeconomic pathways were extrapolated. A spatial optimization layout model for LID facilities was proposed using benefits, life cycle cost and hydrological functions as objective functions. The optimal layout of LID facilities under different scenarios was solved using the gamultiobj algorithm. An adaptive spatial optimization layout model was established based on adaptive decision theory. The Sobol method was used for sensitivity analysis of the model. The results show that the spatially optimal layout of LID facilities under different application scenarios varies, forming Pareto sets. The performance of the adaptive spatial optimization layout solutions for LID facilities mainly maintains a relatively stable intermediate level for three objective functions. The optimization layout model of urban LID facilities exhibits significant differences under different decision preferences. Green roofs dominate the model results and are extremely sensitive to all three objective functions. The findings can guide the development of a sustainable urban stormwater system.

The strong volatility of wind power presents persistent challenges to the stable operation of power systems, highlighting the critical need for accurate wind power forecasting to ensure system reliability. This study proposes a wind power prediction approach based on graph convolutional networks, incorporating ramp feature recognition and error correction mechanisms. First, an improved ramp event definition is applied to detect and classify wind power ramp events more accurately, thereby reducing misjudgments caused by short-term fluctuations. Then, a GCN-based model is developed to construct graph representations of various ramp scenarios, allowing for the effective modeling of their coupling relationships. This is integrated with a bidirectional long short-term memory network to enhance prediction performance during power fluctuation periods. Finally, a dynamic error feedback correction mechanism is introduced to iteratively refine the prediction results in real time. Experiments conducted on wind power data from a Belgian wind farm show that the proposed method significantly improves prediction stability and accuracy during ramp events, achieving an approximate 28% improvement compared to conventional models, and demonstrates strong multi-step forecasting capability.

Soil total nitrogen (TN) serves as both a fundamental indicator of agricultural fertility and a critical marker for ecological balance and environmental security. Rapid monitoring of soil TN is essential for evaluating soil fertility and guiding precision agriculture in ecologically fragile agro-pastoral transitional zones like Northwest China. A total of 116 farmland soil samples were collected from Jingbian County, China. Six spectral transformations, combined with Correlation Analysis (CA), Competitive Adaptive Reweighted Sampling (CARS) were applied to extract TN-sensitive spectral bands and elucidate spectral mechanisms. PLSR, RF and GBDT were used to establish a prediction model. The study showed that: 1) Soil TN content of the surface soil of farmland in Jingbian County ranged from 0.003 to 0.781 g kg-1, with an average content of 0.266 g kg-1. 2) Soil spectral reflectance decreased gradually with the increase of soil TN content, and soil spectral reflectance was negatively correlated with soil total nitrogen content. However, the trend of spectral reflectance with wavelength is consistent with an overall upward trend; 3) Derivative variability effectively improves spectral sensitivity to target information, and the CA, CARS band screening methods achieve characteristic band screening and reduce data redundancy; 4) The R2 of the calibration and validation sets of the soil TN prediction model built based on Log1/R-CARS-GBDT were 0.92 and 0.89, RMSE was 1.09 and 1.33 g kg-1, and the RPD and RPIQ were 2.01 and 2.62, respectively, which allowed the model to carry out the estimation of TN better. Integration of preprocessing, feature extraction, and modeling significantly improved prediction accuracy, enabling rapid hyperspectral quantification of TN in arid agro-pastoral soils. This framework provides a scientific basis for hyperspectral-based TN monitoring and precision agriculture in ecologically fragile regions, supporting data-driven land management and policy formulation.

Real time remote data collection and precise processing are the foundation for smart water conservancy to carry out business applications. This paper studies real-time denoising algorithms for hydrological monitoring data in the context of higher requirements for real-time and accurate data processing in smart water conservancy. Based on the analysis of current data noise types and characteristics, as well as the main denoising algorithms and their advantages and disadvantages, a real-time denoising algorithm based on suspicious data caching mechanism is proposed. Through theoretical analysis and practical application verification, this algorithm is progressiveness in terms of time consumption, space consumption, accuracy and adaptability.

Controlling the total amount of river pollutant discharge is an important means of water resource protection and management, and it is also a necessary condition for ensuring the normal functioning of water areas. The total amount of pollutant discharge is closely related to the water environmental capacity (WEC). Shifting from the traditional method of calculating WEC to dynamic analyses and calculations, concerning practical applications, in this paper, a dynamic adaptive calculation method is proposed for the river WEC that considers the changes in adaptive demand and hydrological conditions. In this method, the dynamic WEC is represented by intervals based on dynamic changes in different spatial and temporal scales, various calculation methods, hydrological conditions, and parameters. According to the calculation results for the WEC, a variable interval was formed. Taking the Shaanxi section of the main stream of the Wei River as the research object, with the support of an integrated platform, the dynamic adaptive calculation of the WEC in the Shaanxi section of the Wei River was realized, and a corresponding simulation system was constructed. The verification results show that (1) the dynamic calculation of WEC can be realized by freely combining different model methods and calculation conditions; (2) the WEC is described using a variable interval, which has strong applicability and operability; and (3) the simulation system can quickly adapt to the changing needs of practical applications and provide managers with visual and credible decision support. The research results provide a theoretical basis for river water environment pollution prevention and environmental management decision-making and help in the high-quality development of the river basin.

Reducing river flows significantly affects livelihoods, ecology and industrial production. Therefore, allowing for ecological flow requirements has been applied to limiting human water consumption and is incorporated into water allocation planning. However, the unavailability of ecological data greatly limits e‐flow studies and this lack of sufficient ecosystem data is mostly addressed by using hydrological metrics as surrogates of the river ecosystem. This study employed the nine indicators of the ‘Flow Health’ model to assess river hydrological health and explore optimal ecological flow regimes. The modelling estimated the required flow and the extra flood flows of ecosystem conservation under default and custom thresholds. Ecological flows were calculated for five stations of the Weihe River through coupling with three hydrological methodologies. The index of flow deviation indicated that the flow regimes in Weihe River have partially changed from the reference period. Fish and other aquatic habitats have changed by reduced flows, but the health of the floodplain ecosystem has not been seriously affected. The ecological flow modelling results showed that the extra flow required for the ecology restoration in five stations is concentrated from July to September. The optimal ecological flows in dry seasons by coupling three hydrological methods were less than the average current flow, which is reasonable. This coupling provides ideas on how environmental flow assessments are undertaken in similar regions.