[Objective] Developing new quality productive forces of water conservancy is an inevitable direction for promoting high-quality development of the water conservancy sector at present, and how to cultivate such forces according to local conditions has become a key issue faced by both the theoretical community and practical departments. Although related studies have continued to expand, there remain shortcomings in the water conservancy field, including insufficient systematic attention to differences in water resource endowment and an inadequate understanding of the formation mechanisms of new quality productive forces of water conservancy. Therefore, it is urgently necessary to carry out normative research focusing on the theoretical logic and practical pathways of development tailored to local conditions. [Methods] This study was based on resource endowment theory, comparative advantage theory, and new structural economics, and constructed an analytical framework for developing new quality productive forces of water conservancy according to local conditions. Starting from water factor endowments and functional attributes, the key roles of innovative allocation of water-related factors, upgrading of water-related industries, and the water-related science and technology talent system were systematically examined in promoting the development of new quality productive forces of water conservancy. Furthermore, by integrating regional water resource characteristics, the study identified five core dimensions—water science and technology, resource allocation, ecological value, industrial upgrading, and disaster prevention and control—and established a system of regionally differentiated development pathways. [Results] The results showed that developing new quality productive forces of water conservancy according to local conditions was a dynamic evolutionary process that started from water factor endowments and regional differences, progressed through the formation of comparative advantages and coordinated regional division of labor, and ultimately led to a spiral improvement of comprehensive capabilities. This development relied on three key driving forces: innovative allocation of water-related factors, which promoted the overall optimization of water engineering systems, water ecological patterns, and water resource utilization structures; in-depth development and transformational upgrading of water-related industries expanded the water economic value chain and strengthened the industrial foundation of the water conservancy modernization system; and improvements in the water-related scientific and technological innovation and talent cultivation system provided sustained momentum for the modernization of water conservancy. The study further identified five major manifestations of new quality productive forces of water conservancy from the perspective of the multiple values of water factors: water science and technology reflected innovation in governance capacity, resource allocation reflected improvements in water resource efficiency, ecological value reflected ecosystem improvement and enhanced services, industrial upgrading reflected industrial structure reshaping and value chain extension, and disaster prevention and control reflected enhanced regional water security resilience. These dimensions were interrelated and jointly constituted a diversified structural system of new quality productive forces of water conservancy. [Conclusion] Based on the characteristics of water factor endowments in different regions, this study proposes five representative types of differentiated development pathways. The northwestern arid regions are suited to adopt a “technology-based water compensation” pathway, improving water resource utilization efficiency through technological innovation. North China is suited to follow an “institutional adjustment” pathway, alleviating the contradiction between water supply and demand through institutional provision. Southwest China is suited to adopt an “ecological transformation” pathway, converting rich water ecological advantages into development momentum. The southeastern coastal regions are suitable for an “integrated water economy” pathway, strengthening the linkage between the water economy and regional industries. Typical high-risk regions need to adopt a “resilience enhancement” pathway, reinforcing flood control, disaster reduction, and comprehensive risk governance. These pathways reflect the heterogeneity of the formation mechanisms of new quality productive forces of water conservancy across regions and demonstrate the central role of the principle of adapting measures to local conditions in the modernization of water conservancy. Overall, developing new quality productive forces of water conservancy according to local conditions must be based on differences in water factor endowments and regional functions, forming an advantage-oriented and endogenous driving system, and achieving coordinated improvement in water science and technology, resource allocation, ecological value, industrial development, and disaster resilience. The results and pathway system of this study provide important theoretical foundations and practical guidance for regions to construct differentiated water conservancy development models.

[Objective] This study aims to analyze regional differences in agricultural and industrial water use efficiency in the Yangtze River Basin and their key influencing factors, reveal the spatial differentiation patterns of water use efficiency within the river basin, and provide scientific guidance for formulating differentiated and precise water resource management policies. [Methods] Three provinces (municipalities) from the upper, middle, and lower reaches of the Yangtze River were selected as sample regions. Authoritative data from sources such as the China Water Resources Bulletin, the Water Resources Bulletin of the Yangtze River Basin and Southwest Rivers, and the Statistical Bulletin on National Economic and Social Development from 2014 to 2023 were collected. Core water use efficiency indicators included: effective utilization coefficient of farmland irrigation water, actual irrigation water use per mu of farmland, and water use per 10 000 yuan of industrial added value. Using methods such as descriptive statistics and comparative analysis, the temporal changes and spatial differences in water use efficiency for the entire river basin and between regions were systematically evaluated. Additionally, combined with data on topography, per capita GDP, precipitation, and industrial structure of the sample provinces (municipalities), key factors affecting water use efficiency were qualitatively analyzed and quantitatively identified. [Results] (1) Temporal changes showed that in the past decade, both agricultural and industrial water use efficiency in the Yangtze River Basin gradually improved, but both indicators remained consistently below the national average level, indicating that the overall water-saving potential of the river basin still needed to be tapped. (2) Spatial differences were significant. Agricultural water use efficiency (characterized by the effective utilization coefficient of irrigation water) followed the order of downstream area > midstream area > upstream area. Per capita GDP and topography were key factors affecting coefficient changes. In general, regions with higher economic development level and flatter terrain exhibited higher agricultural water use efficiency. Industrial water use efficiency (characterized by water use per 10 000 yuan of industrial added value) followed the order of upstream area > midstream area > downstream area, with the rationality of industrial structure being its key influencing factor. The downstream area, due to the concentration of high water-consuming industries and a relatively large proportion of water use for direct-current thermal (nuclear) power, had relatively low industrial water use efficiency. (3) Actual irrigation water use per mu of farmland was affected by multiple factors such as cropping structure, climate variability, and irrigation methods. This resulted in poor cross-regional comparability, making it unsuitable as a reliable indicator for evaluating spatial differences in agricultural water use efficiency. [Conclusion] Economic development level, irrigation infrastructure conditions, and the rationality of industrial structure are key factors affecting agricultural and industrial water use efficiency in the Yangtze River Basin. In the upstream area, complex terrain, outdated irrigation facilities, and insufficient financial and technical support lead to significant irrigation water losses during water conveyance and use, resulting in relatively low agricultural water use efficiency. In the downstream area, although water-saving management measures are relatively well developed, the large proportion of high water-consuming industries in the industrial structure results in relatively low industrial water use efficiency. Based on these findings, to improve the overall water use efficiency of the river basin, differentiated and precise zonal management strategies should be implemented. In the upstream area, priority should be given to modernizing irrigation infrastructure and promoting advanced technologies. In the downstream area, efforts should focus on strengthening industrial structure optimization and formulating stricter water-saving standards and incentive-constraint mechanisms for high water-consuming industries. Future research could focus on smaller spatial scales and enhance the quantitative analysis of influencing factors to support the precise implementation of water-saving measures.

[Objective] To improve the accuracy of groundwater level time series prediction and explore the application effects of 17 decomposition techniques—EMD, EEMD, CEEMD, ICEEMD, LMD, RLMD, ITD, ESMD, WT, WPT, EWT, VMD, SSA, TVF-EMD, FDM, SGMD, and SVMD—in the decomposition of groundwater level time series data, a love evolution algorithm (LEA) - fast learning network (FLN) prediction model based on these 17 decomposition techniques is proposed. [Methods] Firstly, 17 decomposition techniques including EMD were used to decompose the groundwater level time series data, and several decomposition components were obtained. Secondly, based on the training set of each decomposition component, a fitness function was constructed, and LEA was used to optimize the fitness function to obtain the optimal FLN input layer weight and hidden layer threshold for FLN. Seventeen models, including EMD-LEV-FLN, were established to predict and reconstruct each decomposition component. Finally, the daily water level time series prediction of the Caoba groundwater monitoring well in Yunnan Province from 2019 to 2023 was used as an example to verify each model. [Results] (1) WPT-LEV-FLN, EWT-LEV-FLN, FDM-LEV-FLN, TVF-EMD-LEV-FLN models achieved the highest prediction accuracy, with average absolute percentage error (MAPE), average absolute error (MAE), and root mean square error (RMSE) ranging 0.000%-0.001%, 0.002-0.020 m, and 0.002-0.032 m, respectively. The determination coefficients (R²) were all 1.000 0. The SSA-LEV-FLN, WT-LEV-FLN, and VMD-LEV-FLN models came in second place, with predicted MAPE, MAE, RMSE, and R² ranging 0.003%-0.007%, 0.041-0.087 m, 0.063-0.131 m, and 0.999 5-0.999 9, respectively. Other models had relatively poor prediction accuracy, with predicted MAPE, MAE, RMSE, and R² ranging 0.017%-0.033%, 0.221-0.417 m, 0.385-0.705 m, and 0.985 3-0.995 6, respectively. Among them, WPT-LEV-FLN model had high prediction accuracy and small computational scale, demonstrating the greatest practical value and significance. (2) WPT, EWT, FDM, and TVF-EMD showed the best decomposition performance, among which WPT not only had good decomposition performance, but also produced fewer decomposition components, making it the most advantageous. SSA, WT, and VMD showed relatively good decomposition performance, and increasing the number of decomposition components could further improve the decomposition effectiveness. The other models performed relatively poorly, among which SGMD and SVMD had the least decomposition components and the greatest potential. [Conclusion] This study compares the application performance of 17 current mainstream time series decomposition techniques for processing groundwater level time series decomposition and proposes 17 prediction models, providing reference and guidance for the selection of time series decomposition methods and research on groundwater time series prediction.

[Objective] This study aims to analyze the changes in water resources in the middle and lower reaches of the Yangtze River and to investigate the impact of El Niño events on regional floods. [Methods] GRACE gravity satellite data from 2003 to 2022, released by various research institutions, were used to derive the Terrestrial Water Storage Anomaly (TWSA) for the study area. Based on correlation coefficients and cross-correlation analysis, the CSR MASCON TWSA data series exhibiting strong correlations with the indices of both Eastern Pacific (EP) and Central Pacific (CP) El Niño events was selected. Wavelet analysis, Empirical Orthogonal Function (EOF) analysis, and the Flood Potential Index (FPI) were employed to investigate the influence of the two types of El Niño events on regional TWSA and to analyze their relationship with flood risk in the study area. [Results] The results were as follows: (1) the highest correlation between TWSA and the EP El Niño event was found at a time lag of 6 months, with a correlation coefficient reaching 0.630. TWSA peaks showed a positive response to EP El Niño events, while the response to CP El Niño events was unstable. (2) Cross wavelet transform revealed common resonance periods between TWSA and both types of El Niño events, and the impact of the EP El Niño event on water resource changes in the middle and lower reaches of the Yangtze River was found to be more significant. The EOF analysis showed that the southern part of the study area was susceptible to the influence of both El Niño types. (3) The spatial distribution of the grid-based Flood Potential Index showed a higher flood risk in the southern part of the study area following the occurrence of both El Niño types. The flood risk corresponding to EP El Niño events was greater, with high-risk areas concentrated at the junction of the Dongting Lake and Poyang Lake sub-basins. In contrast, the flood risk distribution corresponding to CP El Niño events was more dispersed. [Conclusion] The results of this study, based on wavelet analysis, EOF analysis, and the Flood Potential Index, show that flood risk in the middle and lower reaches of the Yangtze River is closely related to El Niño events. These findings contribute to the prediction and prevention of floods in the region.

[Objective] The simplified structures of traditional hydrological models often limit their adaptability under complex hydroclimatic and anthropogenic conditions. Meanwhile, data-driven models such as deep learning frameworks typically lack explicit physical interpretability. To address these challenges, this study compares the performance of three representative runoff prediction methods: two physically based models—the Xinanjiang (XAJ) model and Soil & Water Assessment Tool (SWAT) model—and one data-driven model, the Long Short-Term Memory (LSTM) model. The Xiangjiang River Basin, a major tributary of the Yangtze River in southern China, is selected as the study area. This study aims to evaluate model performance and adaptability across multiple temporal scales. [Methods] The dataset included daily runoff records from 1971 to 2020 at the Xiangtan hydrological station, along with concurrent meteorological observations from 32 meteorological stations. Land use data with a spatial resolution of 1 km × 1 km were obtained from the Chinese Academy of Sciences, and soil property data were derived from the Harmonized World Soil Database (HWSD). The employed XAJ model estimated evaporation from three soil layers, routed surface runoff using the unit hydrograph method, and modeled interflow and groundwater components through the linear reservoir approach. The SWAT model divided the river basin into 13 subbasins and 192 hydrological response units (HRU) using high-resolution spatial inputs. For both models, the SUFI-II algorithm was employed for parameter calibration based on the Nash-Sutcliffe Efficiency (NSE) coefficient and the total water balance error. The LSTM model was trained using climate indices as inputs under different strategies (areal averages and multi-station series). The input indices included precipitation, temperature, and relative humidity. To evaluate the effects of temporal dependencies, lag times from 0 to 5 days were tested. All three models were calibrated using data from 1971 to 2010 (with the first two years as a warm-up period) and were validated over 2011-2020. [Results] (1) For daily-scale prediction, the LSTM model achieved the highest accuracy, with NSE values reaching 0.99 during calibration and 0.87 during validation when using multi-station meteorological inputs. Incorporating spatially distributed meteorological inputs significantly improved LSTM performance compared to using areal averages, highlighting the importance of spatial heterogeneity in data-driven hydrological forecasting. The XAJ model performed robustly (NSE>0.76), especially during flood seasons, but tended to overestimate dry-season flows and underestimate flood peaks. The SWAT model (NSE≈0.6) reproduced the overall hydrograph patterns but showed systematic biases similar to those of the XAJ model. (2) At the monthly scale, the performance of the SWAT model improved significantly (NSE=0.92 during calibration, 0.83 during validation), while LSTM accuracy declined (NSE = 0.86 during validation). The reduced training sample size (480 months) likely caused overfitting in the LSTM model and limited its generalization ability. Incorporating temperature and humidity as input features enhanced the stability of the LSTM model, indicating that variables related to evapotranspiration became more influential at coarser temporal resolutions. (3) Both hydrological models showed distinct seasonal bias patterns: overestimation during dry seasons and underestimation during wet seasons. The main reason was the fixed parameterization that failed to represent temporal variability in infiltration and storage processes. In contrast, the flexibility of the LSTM model enabled better adaptation, though at the cost of physical interpretability. The observed discrepancies also reflected the impact of nine cascade hydropower projects along the mainstream of the Xiangjiang River, which regulated flow seasonality but were not explicitly modeled in the hydrological frameworks used in this study. [Conclusion] In summary, this study systematically evaluates the performance of the XAJ, SWAT, and LSTM models for runoff forecasting in the Xiangjiang River Basin at daily and monthly scales. The results demonstrate that the LSTM model achieves the highest forecasting accuracy and computational efficiency, particularly at the daily scale. In contrast, the XAJ model is more reliable during flood seasons, and the SWAT model is more suitable for representing the long-term spatiotemporal variability of hydrological processes. Although the data-driven LSTM model lacks explicit physical mechanisms, it offers substantial advantages in adaptability and predictive precision, whereas physically based models maintain interpretability and stability under limited data conditions. The innovation of this study lies in bridging process-based and deep learning methods under unified experimental conditions, emphasizing the potential of hybrid modeling frameworks that integrate physical constraints with deep neural architectures to enhance both accuracy and interpretability in future runoff forecasting applications.

[Objective] This study aims to develop a daily-scale precipitation fusion product (2001-2023) with higher spatiotemporal accuracy covering the Yangtze River Basin by utilizing multi-source data and machine learning techniques, to address the poor quality of existing single or fusion products and to provide reliable data support for related research and applications in this region. [Methods] Multiple types of fundamental geographic data and in-situ measured precipitation data were collected and processed. Based on the aforementioned data, eight machine learning models—RF, CatBoost, KNN, Lasso, DTREE, XGBoost, HGBR, and ETREE—were selected for preliminary training, and their comprehensive capabilities were quantitatively evaluated. Subsequently, nine different ensemble model combinations were constructed based on the single models, and through quantitative evaluation, the seasonal ensemble model ELM4-S with the best overall performance was identified to generate the final daily precipitation fusion product for the Yangtze River Basin at a 0.1° resolution. [Results] (1) Based on multiple evaluation metrics, among the four original precipitation products (ERA5, ERA5-Land, GPM, and CMORPH) in the Yangtze River Basin, GPM exhibited the best overall performance. In terms of the probability of detection (POD), the ERA5 series demonstrated particularly outstanding performance, reaching 0.96. (2) A comparison of the performance of the eight machine learning models (RF, CatBoost, KNN, Lasso, DTREE, XGBoost, HGBR, and ETREE) indicated that RF exhibited the best overall performance. After training, all machine learning models achieved satisfactory results and outperformed the original precipitation products in terms of correlation (R), root mean square error (RMSE), and mean relative bias (MRB). (3) Among the nine ensemble models constructed from combinations of different machine learning models, ELM4-S demonstrated the best overall performance. The fusion precipitation product obtained by ELM4-S was superior to the original precipitation products, incorporating the advantages of different original products. It was numerically reasonable and could reflect the detailed characteristics of precipitation variation with topography in its spatial distribution. [Conclusion] The precipitation fusion product generated based on the ELM4-S model is more accurate than the four original gridded precipitation products adopted. This product not only integrates the advantages of each original dataset but also finely captures the spatial distribution characteristics of precipitation variation with topography, exhibiting outstanding detail. This study successfully develops a high-precision daily precipitation fusion product for the Yangtze River Basin from 2001 to 2023 using an ensemble machine learning approach. This product effectively balances POD and false alarm rate (FAR). It outperforms the original data and single-model results in overall performance and captures more reasonable spatial details of precipitation. It can serve as a reliable data product to widely support various production applications and scientific research within the basin.

[Objective] To address the challenges faced by Horizontal Acoustic Doppler Current Profilers (H-ADCP) in online discharge monitoring applications—specifically, the difficulty in selecting index velocity (feature cells), the insufficient non-linear expressiveness of traditional calibration models, and the poor generalization ability and high computational complexity of existing machine learning models under complex hydrodynamic conditions such as tides and engineering regulations—this paper aims to develop a new H-ADCP online discharge monitoring model that can automatically optimize velocity features, integrate the advantages of multiple algorithms, and improve model accuracy. This model is designed to address the complex non-linear mapping problem between high-dimensional velocity data and cross-sectional discharge, thereby enhancing the accuracy, stability, and automation of discharge monitoring. [Methods] A Feature Adaptive Optimization (FAO) model for H-ADCP online discharge monitoring was developed. The technical framework of this model comprised three core components: (1) feature dimensionality reduction: Principal Component Analysis (PCA) was applied to conduct initial dimensionality reduction on the high-dimensional velocity data from up to 128 cells generated by the H-ADCP, reducing subsequent computational complexity while preserving the main velocity distribution characteristics. (2) Multi-model parallel mapping: five machine learning models—Backpropagation (BP) Neural Network, Elman Neural Network, Radial Basis Function (RBF) Network, Generalized Regression Neural Network (GRNN), and Support Vector Machine (SVM)—were constructed in parallel to establish the non-linear mapping relationship between the dimension-reduced feature velocities and the measured cross-sectional discharge. (3) Global optimization and adaptive selection: the Particle Swarm Optimization (PSO) algorithm was utilized as a global optimization engine, with the Root Mean Square Error (RMSE) as the fitness function, to search within the feature subspace and model space through iterative optimization and adaptively determine the optimal combination of velocity cells, the best machine learning model, and its corresponding parameters. To validate the model’s performance, the Luohu Hydrological Station, which is affected by both tides and backwater effects from confluence and has a complex hydrological regime, was selected as the study area. The model was calibrated and verified using measured H-ADCP velocity data and comparative discharge data from a moving-boat ADCP for the years 2019 and 2023. [Results] (1) The FAO model demonstrated superior performance: during the 2019 model verification period, the discharge predictions of the FAO model showed a high degree of agreement with the measured values, with a RMSE of 6.06 m³/s and a Coefficient of Determination (R²) reaching 0.93. This was significantly better than the traditional linear regression model and any single machine learning model. In simulating extreme discharges such as flood peaks, the FAO model also demonstrated a greater ability to capture them, with an annual maximum discharge error of 1.56%. (2) The feature optimization was effective: the model successfully and automatically selected an optimal combination of 11 feature cells ({5,9,12,15,17,19,21,24,26,28,35}) from 40 velocity measurement cells, eliminating invalid data affected by riverbanks and blind zones. The distribution pattern of the selected cells was highly consistent with hydraulic characteristics, demonstrating the physical interpretability of the model’s feature selection. (3) The model showed strong stability: when validated with data from the entire year of 2023, the FAO model performed stably, with an RMSE of 6.02 m³/s and an R² of 0.91, and effectively fitted the entire annual discharge process, especially for maximum and minimum values. [Conclusion] The proposed FAO model, by organically integrating PCA, multiple machine learning algorithms, and the PSO optimization algorithm, successfully addresses the key technical challenges in H-ADCP online discharge monitoring. The model exhibits powerful self-learning and self-adaptive capabilities, enabling it to automatically find the optimal velocity features and computational model based on data samples, while ensuring computational accuracy and significantly reducing data processing complexity. The application case under complex hydrological conditions demonstrates that the FAO model has high accuracy, good stability, and strong adaptability, providing an efficient and intelligent solution for H-ADCP online discharge monitoring.

[Objective] Improving the prediction accuracy of medium- and long-term hydrological forecast is of great significance for water resources scheduling, flood control and drought relief, and agricultural production. This study aims to select reliable, efficient, and practical hybrid machine learning models to improve forecasting performance for highly irregular, complex nonlinear, and multi-scale variable medium- and long-term hydrological forecasts, providing new approaches for enhancing hydrological forecast accuracy in changing environments. [Methods] To improve the accuracy of hydrological forecasts, based on the measured monthly runoff series at Wanxian Station in the Three Gorges Reservoir area, the mutual information method was used to screen forecasting factors. Then, Long Short-Term Memory (LSTM) models optimized by the Whale Optimization Algorithm (WOA), Grasshopper Optimization Algorithm (GOA), and Sparrow Search Algorithm (SSA) were established. Combined with Time-Varying Filtered Empirical Mode Decomposition (TVF-EMD), Complementary Ensemble Empirical Mode Decomposition (CEEMDAN), and Variational Mode Decomposition (VMD), multiple hybrid prediction models were established. Their prediction performance was evaluated using five indicators: mean absolute error (MAE), root mean square error (RMSE), Nash-Sutcliffe efficiency coefficient (NSE), mean absolute percentage error (MAPE), and correlation coefficient (R). [Results] The forecast factor scheme selected by the mutual information method provided optimal model input, with a lag of 15 months achieving the maximum mutual information value and minimum MASE, representing the best input configuration. Among the three single machine learning models, LSTM and SVM outperformed BP, with LSTM and SVM showing similar performance. LSTM was preferred due to its sensitivity to temporal sequences, enabling better handling of nonlinear runoff prediction, and was thus used in coupling with different methods for runoff forecasting. The hybrid models following the “decompose-reconstruct” strategy outperformed single LSTM models: the VMD-LSTM model improved the NSE of the test set by 0.12 compared with the single LSTM model, exceeding CEEMDAN-LSTM and TVF-EMD-LSTM. Further integration with robust optimization algorithms enhanced accuracy: the VMD-SSA-LSTM model outperformed VMD-LSTM, VMD-GOA-LSTM, and VMD-WOA-LSTM, showing superior adaptability, generalization, and overall predictive performance. [Conclusions] Machine learning models provide effective runoff forecasting methods for regions with limited hydrological and meteorological data. The approach of combining forecasting factor screening, data preprocessing, and integrating robust optimization algorithms with the model can further improve the accuracy of a single hydrological forecasting model. The established VMD-SSA-LSTM model achieved test period performance of MAE=32.65, RMSE=43.44, NSE=0.95, MAPE=12.9%, and R=0.98, representing the highest accuracy among compared models. This model meets practical production and daily life requirements and can provide a reference for water resource management and industrial and agricultural production in the studied basin.

[Objective] Against the background of rapid urbanization, changes in the urban underlying surface constitute a significant factor influencing runoff processes, yet their mechanisms remain inadequately studied. [Methods] Taking Qingshan District of Wuhan City as a representative study area, this paper used remote sensing technology, GIS analysis, and a BP neural network model to quantitatively assess urban underlying surface changes during the typical study period and analyze its impact on the runoff coefficient. [Results] (1) Under urban development, land use in the study area from 2002 to 2017 shifted overall from permeable to impermeable surfaces. Vegetation, rooftops, and other land-use types fluctuated, whereas water bodies shrank year by year. Construction of the sponge city demonstration zone in 2015 slowed this trend. (2) The runoff coefficient was jointly affected by underlying surface changes and rainfall. However, urban rainfall changed little over short timescales, the impervious surface ratio was the dominant factor. As the area ratio of high-runoff land use (e.g., hardened ground) increased and that of low-runoff land use (e.g., vegetation, green space) decreased, the runoff coefficient rose yearly—from 0.399 in 2009 to 0.535 in 2017—showing that land-use change directly altered the runoff coefficient to some extent. (3) After sponge city interventions, the annual runoff coefficient showed a decreasing trend; in 2017 it was 0.535, 0.051 lower than in 2014. [Conclusions] Sponge city construction reduces the runoff coefficient by expanding highly permeable surfaces and adding storage volume, thereby mitigating the adverse impacts of urban development on stormwater regulation capacity. The study offers scientific guidance for urban planning and flood-control drainage system design, and technical support for urban hydrological cycles and water-resource management.

[Objective] In response to the operational challenge caused by high penetration of wind and solar power in modern power systems, this study aims to propose a bi-level optimized scheduling model for a multi-energy complementary power generation system incorporating pumped storage. The model seeks to enhance renewable energy utilization, optimize system economic performance, and improve system stability. The novelty lies in the integration of a bi-level optimization framework with a deep peak shaving strategy, while introducing CO₂ emission intensity and thermal power output coefficient as evaluation indicators for multi-objective coordination of economy, environmental performance, and stability. [Methods] The upper-level model optimized the joint dispatch of wind, solar, hydro, and pumped storage with objectives of maximizing wind and solar output, minimizing net load fluctuation, and minimizing curtailed electricity. The lower-level model optimized the economic performance of the system, aiming at minimizing thermal power operational costs, pumped storage costs, and curtailed electricity penalties. Constraints included wind and solar output limits, hydro and pumped storage capacity limits, thermal unit ramping capabilities, and power balance requirements. The CPLEX solver combined with the YALMIP toolbox was employed to solve the high-dimensional nonlinear mixed-integer programming problem. CO₂ emission intensity and thermal power output fluctuation coefficient were adopted as additional evaluation metrics to quantify environmental performance and system stability. [Results] Simulation results indicated that integrating pumped storage reduced total cost by 46 000 CNY (1.02%) and CO₂ emission intensity by 6.4% in the summer scenario, while the thermal power output fluctuation coefficient decreased from 33.34% to 7.88%. In winter, thermal output stability improved to 7.67%. Increasing wind-solar penetration from 31.25% to 47.62% lowered system costs by 39.5% and reduced CO₂ emission intensity by 58.3%. Enhancing deep peak shaving from 50% to 70% reduced total cost by 19.2% and decreased thermal power output fluctuation coefficient by 44.2%. [Conclusions] The introduction of pumped storage power station significantly enhances system flexibility, increasing renewable energy utilization by over 12% and reducing thermal unit peak regulation pressure by 50%. The bi-level optimization model ensures low-cost operation while reducing CO₂ emission intensity by more than 0.1 kg/kWh and maintaining thermal power output fluctuation coefficient below 8%. A combination of high wind-solar penetration (>40%) and deep peak shaving (70%) achieves optimal comprehensive benefits, providing theoretical support for scheduling of high renewable energy penetration power systems. This study provides an innovative methodology for the design and optimization of wind-solar-hydro-thermal-pumped storage multi-energy systems, and the findings can be generalized to other clean energy bases.

[Objective] This study aims to conduct a comprehensive evaluation of the ecological benefits of the South-to-North Water Diversion Project (SNWDP) by systematically quantifying the ecological benefits in the water-receiving areas during the first phase of the Middle Route Project. [Methods] The water receiving area was divided according to administrative units and assessed using statistical and remote sensing data. Taking 2014 as the base year and 2018, 2020, and 2023 as evaluation years, we evaluated the ecological benefits brought by project-supplied water in Beijing, Tianjin, 11 counties (or cities) of Henan, and 6 counties (or cities) of Hebei. Ecological benefit index systems were established for forest land, urban green space, wetlands, water bodies, and groundwater ecosystems by integrating the function value method and the equivalent factor method. For forest land, urban green space, and groundwater ecosystems, multiple ecosystem service functions were quantitatively analyzed. The market value method, replacement cost method, and other valuation methods were used to estimate the unit prices of each function and calculate their total service value. For wetlands and water body ecosystems, ecological benefits were calculated using the equivalent factor method based on regional characteristics. A spatiotemporal precipitation adjustment factor was introduced to dynamically adjust the factor values in the basic equivalent factor table, thereby determining the value of one standard unit of ecosystem service equivalent factor. [Results] Cumulative ecological benefits generated by the water supply from the first phase of the Middle Route Project amounted to 44.859, 18.328, and 37.102 billion yuan in each evaluation period, respectively. Wetlands and water bodies accounted for the largest proportions, at 64.90%, 58.98%, and 46.98%, respectively. From 2015 to 2018, new ecological benefits from water bodies and wetlands reached 24.724 and 4.391 billion yuan, respectively; for 2019-2020, they were 9.100 and 1.709 billion yuan; and from 2021 to 2023, new ecological benefits from wetlands and water bodies were 11.079 and 6.352 billion yuan, respectively. The annual average new ecological benefits for each period were 11.215, 9.164, and 12.367 billion yuan, indicating that the project’s water supply generated approximately 10 billion yuan of ecological benefits per year in the water receiving areas. In addition, the ecological benefit value per cubic meter of water varied across provinces and cities. In Beijing, the values were 1.64, 1.38, and 3.01 yuan; in Tianjin, 3.34, 2.19, and 0.52 yuan; in Henan’s 11 counties, 8.16, 5.06, and 3.79 yuan; and in Hebei’s 6 counties, 6.12, 2.69, and 4.07 yuan, respectively. The benefit value ratios for Beijing∶Tianjin∶Henan∶Hebei in each evaluation period were 1∶2.10∶4.99∶3.74, 1∶1.59∶3.66∶1.95, and 1∶0.17∶1.26∶1.35, respectively. [Conclusion] This study provides a case reference for ecological benefit evaluation the follow-up projects of the SNWDP and other inter-basin water diversion projects. It provides technical support for the scheduling and utilization of ecological benefits of the Middle Route Project, and further provides a calculation basis for promoting the establishment of horizontal ecological compensation standards between the water receiving and source areas.

[Objective] This study aims to reveal the influence mechanisms of hydrothermal conditions on the spatiotemporal variability of hyporheic exchange and to develop more reliable estimation methods. We acknowledge the limitations of single methods and innovatively propose an estimation framework for hyporheic exchange that integrates hydraulic methods, environmental tracer methods, and numerical simulation technologies. The proposed method is expected to address the insufficient accuracy and scale mismatch in existing estimation methods and to enhance the capacity to quantify highly variable hyporheic exchange fluxes. [Methods] First, based on years of practical experiences, and combined with a systematic review and critical analysis of existing literature, we deeply analyze the intrinsic driving mechanisms of the spatiotemporal variability of hyporheic exchange from two core perspectives: hydraulics and thermodynamics. Second, we propose an integrated multi-method estimation framework to improve the accuracy and robustness of the estimation results. [Results] The mechanisms by which hydrothermal conditions drive the spatiotemporal variability of hyporheic exchange are summarized as follows.(1) Hydrological rhythm: the dynamic variations in river water level and discharge alter the hydraulic head difference between river water and groundwater, serving as the primary driver of the temporal changes in the rate and direction of hyporheic exchange.(2) Topography, geomorphology, and bed heterogeneity: local topographic features of rivers and lakes (such as sand bars, pools, and point bars) and the spatial heterogeneity of riverbed sediments shape the spatial distribution pattern of hydraulic head differences, which is the fundamental cause of significant spatial variations in hyporheic exchange.(3) Temperature variation: strong daily temperature differences can generate significant thermal gradients within riverbed sediments, inducing rapid flows and shaping the diurnal variation patterns of hyporheic exchange.(4) Seasonal freezing and thawing processes substantially alter the spatial structural characteristics of riverbed permeability, profoundly affecting both the intensity and spatial extent of hyporheic exchange at seasonal and spatial scales. These driving factors are often in a state of nonstationary variations and exhibit complex couplings. Collectively, their combined effects make the spatiotemporal variation patterns of the hyporheic exchange difficult to be accurately captured or predicted by simple methods. [Conclusion] This study systematically elucidates the mechanisms by which hydrothermal conditions jointly influence the complex spatiotemporal variations of hyporheic exchange through hydraulic and thermodynamic processes. It deepens the understanding of surface water-groundwater interactions, providing a theoretical basis and practical guidance for developing more accurate watershed hydrological models, assessing the health of river ecosystems, and formulating science-based ecological restoration strategies for rivers and lakes.

[Objective] To address the low accuracy of monthly runoff point prediction and the difficulty in describing the uncertainty of point prediction results, this study proposes a monthly runoff point prediction model and an interval prediction model based on the Crested Porcupine Optimizer (CPO), Convolutional Neural Network (CNN), Bidirectional Long Short-Term Memory (BiLSTM) network, and Nonparametric Kernel Density Estimation (NKDE). [Methods] First, a hybrid point prediction model (CPO-CNN-BiLSTM) was developed. CPO was used to optimize key model parameters such as the number of hidden layer nodes, initial learning rate, and regularization coefficient. Monthly runoff data and its influencing factors were input to the model to obtain point prediction results. Next, the point forecasts were sorted using a range segmentation method and divided into low, medium, and high flow segments. The relative error for each predicted value within these segments was calculated. The NKDE method, with window width optimized by CPO, was employed to estimate the error probability distribution function for each segment. Cubic spline interpolation was then applied to fit the probability distribution functions of the three segments and derive segment-specific quantiles, forming a monthly runoff interval prediction model (CPO-CNN-BiLSTM-NKDE) based on NKDE method and the CPO-CNN-BiLSTM model. Finally, the runoff point forecasts were combined with the corresponding quantiles of their flow segments to generate monthly runoff interval predictions. Case studies compared the proposed CPO-CNN-BiLSTM point prediction model with traditional models including Least Squares Support Vector Machine (LSSVM), Kernel Extreme Learning Machine (KELM), LSTM, and BiLSTM, using RMSE, MRE, and MAPE as evaluation metrics. [Results] The CPO-CNN-BiLSTM model’s prediction accuracy was significantly better than the other models, especially during flood and dry seasons. Compared with the best-performing among the other four models in terms of RMSE, MRE, and MAPE, the values decreased by 43.71%, 38.56%, and 24.38%, respectively. This indicated a superior ability to accurately predict peak and valley runoff values. Additionally, deep learning models (LSTM, BiLSTM, CNN-BiLSTM) outperformed machine learning models (LSSVM, KELM), with the BiLSTM model surpassing LSTM, and the CNN-BiLSTM hybrid outperforming both. The proposed CPO-CNN-BiLSTM-NKDE interval prediction model was compared with other interval prediction models at confidence levels of 95%, 90%, and 85%, and it exhibited the highest Prediction Interval Coverage Probability (PICP)and the lowest Prediction Interval Normalized Average Width (PINAW), indicating strong reliability and superior capability in capturing uncertainty. This demonstrated that the interval prediction results of the proposed model could help decision-makers better understand and respond to the uncertainty and variability in the data. [Conclusion] The proposed CPO-CNN-BiLSTM point prediction model and the CPO-CNN-BiLSTM-NKDE interval prediction model effectively address the challenges posed by the spatial-temporal complexity of monthly runoff sequences and the uncertainty of monthly runoff point predictions. This provides new ideas for monthly runoff prediction and offers useful reference for fields such as wind speed and solar irradiance forecasting.

[Objective] Most existing studies on the response relationship between runoff variations and meteorological drought and flood characteristics focus on annual, seasonal, monthly, or weekly scales. This study aims to clarify the quantitative response relationship between meteorological drought and flood characteristics at the daily scale and runoff in the Jialing River Basin, and to effectively evaluate the impact of extreme meteorological drought and flood events on the flow process. [Methods] Based on long-term daily precipitation and flow data from 1989 to 2022, this study employed the SWAP index method and run theory to identify meteorological drought and flood events at the daily scale in the Jialing River Basin. Traditional multiple regression and emerging machine learning models were compared to simulate the internal relationship between meteorological drought and flood characteristics and flow change, revealing the response of runoff variation to drought and flood characteristics. [Results] The results showed that from 1989 to 2022, a total of 68 meteorological drought events occurred in the Jialing River Basin, leading to an average reduction of 48.25% in flow at the Beibei station. Compared to drought duration and intensity, the timing of drought events had a more significant impact on runoff variation and was the primary controlling factor influencing runoff variation. The support vector regression model considering only this factor could more accurately evaluate the change rate of flow caused by drought. During the same period, 40 meteorological flood events occurred in the Jialing River Basin, leading to an average increase of 130.46% in flow at the Beibei station. The accumulated precipitation before the flood peak had the greatest impact on the change rate of flow, and the timing of maximum precipitation before the flood peak had the greatest impact on the timing of flood peak. Multiple regression models were recommended to evaluate the response relationships between flood characteristic factors and the change rate of flow, as well as the timing of flood peak. To evaluate the impact of flood events on peak flow, the random forest model was recommended. The accumulated precipitation before the flood peak was the primary controlling factor influencing peak flow variation. [Conclusion] This study innovatively explores the response relationship between meteorological drought and flood characteristics at the daily scale and runoff variation in the river basin. The findings indicate that emerging machine learning models, such as support vector machines and random forests, can effectively simulate the complex mechanisms through which meteorological drought and flood events affect runoff in the river basin. This has significant implications for the scientific evaluation and prediction of the impact of extreme climate events on runoff characteristics.

[Objective] This study aims to conduct an in-depth analysis of the spatiotemporal variation characteristics of green water resource consumption and utilization efficiency in Jiangxi Province over the past two decades (2001-2020) to provide scientific support for regional water resource management and agricultural policy making. [Methods] We calculated green water utilization efficiency using two key datasets: (1) the transpiration-to-evapotranspiration ratio dataset (2001-2020) from China’s National Ecological Science Data Center, and (2) annual net primary productivity estimates derived from the MYD17A3H.006 remote sensing product. The Mann-Kendall trend analysis method was then applied to quantitatively evaluate the spatiotemporal trends in green water consumption and utilization efficiency across Jiangxi Province. [Results] From 2001 to 2020, total green water consumption in Jiangxi Province ranged from 796 to 965 mm, with an average of 887.75 mm, showing a significant downward trend at a rate of 5.26 mm/a. Productive green water consumption ranged from 519.12 to 692.53 mm, averaging 614.95 mm, and also showed a clear decreasing trend at 1.73 mm/a. Non-productive green water consumption ranged from 230.64 to 356.30 mm, with an average of 272.79 mm, showing a relatively significant downward trend at 3.53 mm/a. The annual green water utilization efficiency ranged from 0.67 to 0.81 g C/(kg H₂O), with a multi-year average of 0.74 g C/(kg H₂O), demonstrating a significant increasing trend at 0.006 7 g C/((kg H₂O)·a). Spatially, both total green water consumption and productive green water consumption exhibited a distribution pattern of higher values in the south and lower values in the north, with the highest observed in Ganzhou. Productive green water consumption showed an increasing trend around the Poyang Lake area, while most other regions exhibited decreasing trends. Total green water consumption showed an extremely significant decreasing trend across all regions, except in parts of Jiujiang and Jingdezhen. Across 92.70% of the province, green water utilization efficiency exhibited varying degrees of increase, with Ji’an and Yichun showing extremely significant improvements. [Conclusion] The decline in total green water consumption and the improvement in its utilization efficiency demonstrate notable achievements in water resources management and water-saving practices in Jiangxi Province. Future efforts should focus on optimizing water allocation, promoting water-saving irrigation technologies, and adopting high-efficiency cultivation practices to further enhance green water utilization efficiency in response to challenges posed by climate change and human activities.

[Objective] Danjiangkou Reservoir is a core project of the flood control system in the Hanjiang River Basin and a strategic water source for the Middle Route of the South-to-North Water Diversion Project, and its flood season staging schemes directly affect its flood control safety and inter-basin water transfer efficiency. Since its operation, Danjiangkou Reservoir has maintained an annual full storage rate of only 11%. How to improve this rate while ensuring flood safety remains an urgent issue. Existing research on flood season staging mainly focuses on statistical methods, with insufficient attention to the temporal characteristics of typical rainy seasons such as the plum rainy season in the middle-lower Yangtze River and the autumn rainfall in West China. This study aims to provide a theoretical basis and technical support for the operation and management of Danjiangkou Reservoir. [Methods] Based on daily inflow data of Danjiangkou Reservoir from 1961 to 2023, this study conducted preliminary flood season staging analysis using runoff statistical analysis, annual maximum flood analysis, mean change-point analysis, and vector statistical methods. [Results] The results obtained from different methods were relatively consistent. For safety reasons, the flood season was preliminarily divided into three stages. A further comprehensive analysis integrating the temporal patterns of the plum rain and West China autumn rainfall was conducted. The results showed that both exhibited significant temporal and periodic characteristics. From 1951 to 2023, the plum rain in the middle and lower reaches of the Yangtze River began as early as May 25 and ended as late as August 8. From 1961 to 2023, autumn rainfall in West China began as early as August 21 and lasted until November 30. During the plum rain period, when water levels in the middle-lower Hanjiang River were relatively high, any coincidental occurrence between floods from Danjiangkou and downstream floods may trigger flooding in the Hanjiang River Basin. After the plum rainy season ended, the flood control pressure in the middle-lower Hanjiang River decreased, allowing the Danjiangkou Reservoir to gradually release reserved flood control storage capacity. The autumn rainfall in West China started in late August in Sichuan and gradually moved eastward toward the Qinling Mountains, aligning with the autumn flood season. The autumn rainfall in West China directly affected Danjiangkou Reservoir’s storage conditions. During flood events, reservoir regulation must balance flood safety with increasing the full storage rate to fully utilize flood resources. In full consideration of both the plum rain and West China autumn rainfall, the flood season of Danjiangkou Reservoir was ultimately divided into three stages: summer flood season from June 21 to August 10, transition period from August 11 to August 31, and autumn flood season from September 1 to October 10. [Conclusion] Based on meteorological forecasts of end dates of plum rain and West China autumn rainfall, the flood limit water level for the summer flood season can be gradually raised to the level of the autumn flood season. This approach enables early water storage and improves both water resource utilization efficiency and the reservoir’s full storage rate. The research findings provide a theoretical basis for operation and scheduling decisions for Danjiangkou Reservoir.

[Objective] This study focuses on the Yangtze River Economic Belt and constructs a water resources carrying capacity evaluation system that covers four subsystems: water resources, society, economy and ecological environment. The aim is to reveal the current status and future development trend of water resources carrying capacity in the Yangtze River Economic Belt, and to provide a scientific basis and decision-making reference for the rational planning and utilization of water resources, the adjustment and optimization of industrial structure, and ecological environment protection within the region. [Methods] The spatiotemporal evolution characteristics of water resources carrying capacity in the Yangtze River Economic Belt from 2012 to 2021 were analyzed using the TOPSIS method and the standard deviation ellipse method. Combined with the coupling coordination degree model, the coordinated development level among the subsystems within the water resources carrying capacity system was further investigated. To better understand the future development trend of water resources carrying capacity, the grey prediction model was applied to predict its trend over the next five years. [Results] The study revealed the dynamic changes in water resources carrying capacity in the Yangtze River Economic Belt during the study period, along with its spatial distribution characteristics and evolution trends. Between 2012 and 2021, the water resources carrying capacity showed an overall fluctuating upward trend, gradually improving from an alert state to a good state, indicating a significant enhancement in the region’s water resources carrying capacity. Spatially, the center of water resources carrying capacity shifted southwestward toward areas with relatively lower capacity, which may be related to regional economic development patterns, industrial restructuring, and differences in water use efficiency. Regarding system coupling and coordination, except for 2012, the coupling degree between subsystems reached a high coupling stage, and the coupling coordination evaluation gradually shifted from near disorder to good coordination, demonstrating continuously improving coordinated development and enhanced synergy among the subsystems. Analysis of influencing factors identified the proportion of tertiary industry, urbanization rate, and per capita daily domestic water consumption as the three factors most strongly correlated with water resources carrying capacity. Changes in these factors significantly affected its increase or decrease. The water resources carrying capacity was projected to show a positive development trend over the next five years. [Conclusions] It is recommended that the upstream areas develop water-saving irrigation, control fertilizer usage, and enhance urbanization levels; the midstream areas develop reclaimed water use, strengthen sewage treatment, and accelerate industrial transformation; and the downstream areas control population growth, promote water conservation and environmental protection, restore ecosystems, and increase forest coverage. The research findings provide a valuable scientific basis for the efficient management and utilization of water resources in the Yangtze River Economic Belt,as well as for promoting the coordinated development of the regional economy and society,and for achieving sustainable development goals in the region.

[Objective] Taking the Shiyang River Basin, a typical arid inland river basin, as the study area, this study aims to explore the spatial distribution and matching patterns of agricultural water and soil resources under different scenarios of future land use change, identify the supply-demand imbalance and its causes in the river basin, and provide support for the planning and decision-making of sustainable agricultural development at the river basin scale. [Methods] Using the FLUS model, this study simulated the spatial patterns of land use of the Shiyang River Basin in 2035 under three scenarios: cropland protection, natural development, and ecological conservation. By introducing the agricultural water and soil resource equivalent coefficient, this study established a matching assessment model for future agricultural water and soil resources in the Shiyang River Basin. Combined with the water yield module of the InVEST model, this study predicted spatiotemporal variations in water yield under three scenarios in 2035 and evaluated the spatiotemporal matching relationships of agricultural water and soil resources in the Shiyang River Basin. [Results] (1) All seven county-level administrative regions in the Shiyang River Basin showed varying degrees of severe water shortage, indicating an overall imbalance in agricultural water and soil resources, with water supply unable to meet the demand of cropland-based agricultural production. (2) The spatial pattern of water and soil resource matching in the Shiyang River Basin was generally better in the west than in the east of the basin. The Gini coefficient ranged from 0.2 to 0.3, showing a slightly increasing trend but still indicating a relatively balanced condition. The average water and soil resource matching coefficient was projected to be 797 m³/hm² in 2035 under different scenarios, compared to 640 m³/hm² in 2020, showing an overall improvement. (3) Although water yield increased to some extent within each county-level region under different scenarios in 2035, the imbalance in actual water utilization and distribution continued to deepen over time. The continuous expansion of cropland under the natural development and cropland protection scenarios made it more difficult to balance the supply and demand of agricultural water and soil resources in Gulang County, Minqin County, and Liangzhou District. [Conclusion] For regions with poor balance between precipitation and evaporation, such as Liangzhou District, a typical resource-based water-scarce region, it is recommended to alleviate water shortage through interregional water transfer and the introduction of external water sources. In regions where agricultural development is dominant, it is advisable to accelerate agricultural modernization, reduce water consumption per hectare, and adjust crop structures. For regions with poor matching between water and soil resources, such as Minqin County, where both resource-based and engineering-related water scarcity coexist, it is proposed to address the imbalance through interregional water diversion and transfer, systematically improving the efficiency of water resource development and utilization.

[Objectives] This study aims to analyze the applicability of existing precipitation, temperature, and runoff data in data-scarce regions, and to develop and evaluate a deep learning hybrid model driven by multi-source information for improving the accuracy of monthly runoff forecasting. [Methods] Based on historical precipitation, temperature, and runoff sequences from the Yulongkashi River, a Convolutional Neural Network-Bidirectional Gated Recurrent Unit-Attention (CNN-BiGRU-Attention) model was developed. An Improved Particle Swarm Optimization (IPSO) algorithm was used to optimize this model, forming the IPSO-CNN-BiGRU-Attention hybrid model. The performance of this model was compared with that of the Gated Recurrent Unit (GRU) model and the ABCD water balance model. [Results] The IPSO-CNN-BiGRU-Attention model that incorporated precipitation and temperature data overall outperformed the CNN-BiGRU-Attention and GRU models, showing better agreement with the observed values. As the predication period increased, the proposed model achieved a root mean square error (RMSE) of 2.11 m³/s, a mean absolute error (MAE) of 1.32 m³/s, a mean absolute percentage error (MAPE) of 73.76%, and a Nash-Sutcliffe efficiency (NSE) coefficient of 0.94. The highest forecast accuracy was observed in the first three months. [Conclusions] The IPSO-CNN-BiGRU-Attention model effectively integrates precipitation, temperature, and runoff information, significantly enhancing the accuracy of monthly runoff forecasts in data-scarce regions. The model demonstrates robust performance across different forecast horizons, particularly suitable for short-term predictions of 1-3 months. This approach offers a practical and reliable tool for hydrological forecasting and flood control/drought management in data-scarce basins.

[Objectives] Identification of runoff components is a key aspect of hydrological analysis and is crucial for understanding the evolution patterns of watershed water resources. Traditional runoff component models are often constructed based on the criterion of maximizing the extraction accuracy of deterministic components for the runoff series of a given length. However, a unified criterion for selecting model forms that adapt to variations in runoff series length over time is still lacking, making it difficult to determine the types of runoff components and the order of their separation during modeling. To address this, this study proposes a selection criterion for runoff component models based on the balance between benefits and risks. [Methods] Based on the diagnosis and quantitative description of evolution characteristics such as mutations, trends, and periodicities using time-series variability detection methods—the Mann-Kendall test, sliding T-test, Pettitt test, Standard Normal Homogeneity test, Buishand test, and periodogram—different forms of linear superposition models were developed by combinations and extraction sequences of the identified components, such as mutation, trend, and periodicity. These models were then employed to dynamically identify the components of runoff sequences with varying lengths. The accuracy of deterministic component identification was used to represent the “benefits” achieved by the model in runoff component recognition, while the magnitude of fluctuations in model accuracy under varying runoff sequences (i.e., stability) was regarded as the “risk”. A weighting coefficient representing the decision-maker’s preferences was introduced as a balancing variable to construct a benefit-risk balance indicator. Subsequently, runoff component models were optimized based on the criterion of minimizing this benefit-risk balance indicator. [Results] Using the runoff sequence from 1956 to 2010 at the Pingshan Station on the lower reaches of the Jinsha River as a case study, variable-length runoff sequences (with sample sizes ranging from 30 to 55) were constructed, starting from 1956 and ending in any year from 1986 to 2010. Runoff component identification was conducted under different model forms, and the proposed benefit-risk balance criterion was applied for model selection analysis. The results indicated mutual offsetting among components such as mutations, trends, and periodicities in the runoff sequence, and the same runoff sequence could be characterized by multiple models, each representing distinct compositional forms of runoff components. Runoff component identification was jointly influenced by both the model form and the sequence length; models with higher identification accuracy exhibited relatively lower stability when responding to changes in sequence length. For instance, models incorporating periodic components demonstrated superior fitting accuracy compared to those containing only trend or mutation terms, which in turn outperformed multi-year average models, while the stability of accuracy changes followed the opposite trend. If the decision-making objective was to achieve a more adequate fitting, models that sequentially separate mutations and periodic components are prioritized; conversely, if the objective was to maintain more stable accuracy with varying sequence lengths, models that identify only mutation or trend terms were more advantageous. [Conclusions] A novel approach is proposed in this study for selecting component models of variable-length runoff sequences by balancing identification accuracy (benefit) and stability (risk). Both the accuracy and stability indicators proposed in the criterion can be flexibly defined according to decision-making needs, facilitating decision-makers in comprehensively considering their preferences for model accuracy and stability under varying conditions to optimize model selection.