As a major institutional innovation in China's water governance system, evaluating the effectiveness of the River and Lake Chief System (RLCS) is crucial for optimizing water governance policies and enhancing practical governance outcomes. To enable more precise assessment, this paper develops an integrated "Input–Process–Output" (IPP) evaluation framework tailored to the RLCS as a complex public policy, and introduces Data Envelopment Analysis (DEA) as a quantitative tool for measuring resource-use efficiency during policy implementation, thereby forming the IPP-DEA evaluation approach. Taking Guangdong Province as a case study, and drawing on policy process theory and efficiency assessment methods, we apply the IPP-DEA framework to comprehensively evaluate RLCS performance. The results show that, according to the composite index analysis, indicators such as social participation, patrol and duty performance by river and lake chiefs construction carry relatively large weights and are thus major drivers of performance. Among these, volunteer engagement, public complaints, and satisfaction are particularly prominent, underscoring the critical role of the "co-construction, co-governance, and sharing" philosophy in RLCS implementation. The composite index for Guangdong rose from 0.222 in 2018 to 0.763 in 2025, a cumulative increase of 2.4 times, indicating a substantial leap in river and lake protection and governance, with institutional effectiveness continuously unleashed and displaying an evolutionary pattern of "rapid launch, sustained optimization, and steady deepening". The process management of RLCS work exhibits strong standardization, with the process-level index consistently remaining in the relatively high range of 0.24–0.30, reflecting sustained coordination and compliance in policy implementation. Output levels grew significantly in the early period but leveled off after 2022. Given efficiency fluctuations and the remediation status of the "four disorders", future efforts should focus on enhancing the precision of river patrols and complaint response mechanisms, consolidating governance gains, preventing backsliding, and transitioning the RLCS from "effective operation" toward "long-term quality improvement". During 2018–2022, the RLCS in Guangdong achieved full DEA effectiveness, indicating an optimal input–output state. Starting in 2023, however, scale efficiency declined (reaching only 0.911 in 2024), while pure technical efficiency remained high (crste > 0.95), signaling scale diseconomies. This suggests that, although overall resource allocation in Guangdong's RLCS is broadly efficient, scale economy performance has fluctuated in recent years. Future work should therefore optimize the scale economy of RLCS inputs and strengthen refined policy implementation to further enhance overall resource utilization efficiency. The combined evaluation approach—integrating comprehensive assessment with efficiency analysis—provides more robust decision support than either method alone. This study has certain limitations regarding DEA sample size and relative efficiency measurement; future research should expand the sample based on prefecture-level city panel data and employ advanced methods such as network DEA for deeper efficiency analysis.
With the deepening of high-quality development, China's economic growth is shifting from a resource-driven mode to the one led by the social innovation and efficiency. Water-use efficiency has become an important indicator for assessing the level of green development and determines whether economic growth can shift from high consumption to high efficiency under resource constraints. Improving water-use efficiency helps reduce water consumption per unit of output, thereby alleviating the dependence of economic growth on water resources. Guangdong Province, one of the fastest-growing and most open regions in China, ensuring the coordination between economic growth and water use has become an urgent priority for it. Therefore, this study explores how water consumption and economic growth are decoupling over time across Guangdong's cities, and what drives this process. Such exploration deepens understanding of economic and water resource coordination. It also provides references for building a water-saving society and optimizing water allocation. Using panel data on economic development and water consumption for 21 prefecture-level cities in Guangdong Province, China, spanning the period 2003—2024 and derived from the Guangdong Water Resource Bulletin and the Guangdong Statistical Yearbook, this study applies the Tapio decoupling index and the Logarithmic Mean Divisia Index (LMDI) model to discover the temporal evolution of the decoupling relationship between economic growth and water consumption, as well as to identify the key driving factors underlying these patterns. The results show:①Overall, Guangdong Province shows a favorable trend in decoupling economic growth from water consumption over the study period, indicating that economic expansion has increasingly relied on improvements in water use efficiency rather than proportional increases in water demand. Nevertheless, pronounced regional disparities persist across cities. In particular, highly industrialized and rapidly urbanizing cities such as Shenzhen and Dongguan have experienced a gradual deterioration in the decoupling relationship in recent years, suggesting that intensified economic activities and rising resource pressures may offset efficiency gains in certain stages of development.② The primary industry maintained 'strong decoupling' throughout the study period and remained stable. The secondary industry shifted from 'weak' to 'strong decoupling' with water consumption in 2008, peaking in 2020. This shift highlights the role of industrial restructuring and technological progress in mitigating water consumption associated with manufacturing growth. In contrast, the tertiary industry was considerably volatile, suggesting a relatively higher sensitivity of water demand in service-oriented sectors to short-term economic fluctuations and structural changes. ③Decomposition results further reveal that economic scale expansion and population effects are the principal drivers of increases in total water consumption. Whereas improvements in water use efficiency in the primary and secondary industries serve as the main forces constraining water demand. However, owing to differences in regional economic development patterns and water resource endowments, the relative importance of these driving factors varies markedly across regions and development stages. These inter-city differences and temporal heterogeneity underscore the need to consider local conditions when interpreting decoupling outcomes and assessing the sustainability of economic-water relationships.
Against the backdrop of the increasingly prominent contradiction between socioeconomic expansion and limited water resources, this study aims to identify the water resources carrying thresholds of the Shangyoujiang Basin with full consideration of external water transfers. A comprehensive evaluation index system covering driving forces, pressure, state and response factors is established under the DPSR framework. The entropy weight method is adopted to objectively calculate indicator weights, and the TOPSIS model is further applied to systematically assess the temporal variation characteristics of the basin's water resources carrying status for the period 2021–2025. Following the principle that population, industrial, cultivated land and urban development should adapt and match the local water availability of the basin, this study fully takes into account available water resources corresponding to different hydrological frequencies as well as diversified water demand under typical water-saving scenarios. On this basis, water resources carrying thresholds are quantitatively quantified from four core dimensions, namely population scale, industrial development, cultivated land utilization and urban construction scale. The results indicate that under wet-year, normal-year, dry-year and water-saving scenarios, the population carrying thresholds are 10.65–12.80 million people, 5.92–7.09 million people, 3.51– 4.20 million people and 4.23–5.64 million people, respectively. The thresholds of industrial added value are CNY 327.3–344.7 billion, CNY 180.8–190.4 billion, CNY 98.0–103.2 billion and CNY 145.7– 153.9 billion. The thresholds of effective irrigated cultivated land are 1.13– 1.51 million mu, 0.58–0.77 million mu, 0.23–0.35 million mu and 0.48–0.55 million mu (1 mu ≈ 0.0667 hm²). The thresholds of urban construction land area are 871–1742 km², 513– 1025 km², 294–588 km² and 346–692 km², respectively.On the whole, the basin maintains a favorable water resources carrying level at present. For the planning year, indicators of population, industry, cultivated land and urban construction reach the lower bounds under dry-year scenarios while remaining within the upper limits. Under water-saving scenarios, only the urban construction indicator reaches the lower bound without exceeding the upper limit. Sufficient carrying surplus can be maintained under normal-year and wet-year scenarios.To cope with severe water stress in dry years, a cross-regional coordinated water management mechanism should be established alongside scientific annual water diversion plans. During dry seasons, all industrial, agricultural and domestic sectors shall continuously strengthen water conservation measures. Meanwhile, multiple emergency water sources should be deployed in a timely manner to steadily guarantee safe domestic water supply for residents.
To address the inherent limitations of conventional seawall inspection methods, including restricted spatial coverage, delayed response to potential hazards, inefficient data sharing mechanisms, and inadequate target recognition accuracy, this study integrates edge computing with cloud-edge collaborative architectures to develop an intelligent perception and monitoring system based on a networked unmanned aerial vehicle (UAV) swarm. Traditional inspection approaches, which rely heavily on manual patrols or single-platform monitoring systems, often struggle to meet the increasing demands for real-time, large-scale, and high-precision assessment of coastal defense infrastructure. In contrast, the proposed system is specifically designed to support continuous and automated monitoring of abnormal conditions, such as seawall deformation, surface erosion, and structural damage, in complex coastal environments. At the system level, a novel cooperative path planning strategy, referred to as CTCP-A* (Cooperative Trajectory Constraint Pruning-A*), is introduced to coordinate multi-UAV operations. By incorporating trajectory constraint pruning into the classical A* search framework, the proposed algorithm enables efficient task allocation and collision avoidance while minimizing redundant flight paths. This approach significantly reduces overall energy consumption and ensures stable, highly coordinated autonomous flight among UAVs operating within the same inspection region. To support reliable data transmission, a hybrid communication framework that integrates 5G cellular networks with microwave communication links is adopted, enabling low-latency and real-time transmission of high-resolution video data from UAV platforms to edge and cloud servers. To further improve the accuracy of visual inspection, the YOLOv11 object detection model is enhanced by introducing an EMA attention mechanism. This optimization allows the network to better capture salient structural features of seawalls under varying illumination and background conditions. Based on this improvement, a customized YOLOv11-EMA target detection model tailored specifically for seawall inspection tasks is developed, enabling automated and robust identification of seawall structures and damage features from aerial video streams. In addition, a dedicated seawall erosion detection dataset is constructed to support model training and performance evaluation. Experimental results demonstrate that the proposed system achieves a detection accuracy of 90.1%, representing a 5.6% improvement over the original YOLOv11 model, with detection outcomes showing strong consistency with in situ monitoring data. Furthermore, the inspection efficiency of the UAV swarm is approximately five times higher than that of traditional manual inspections, substantially enhancing operational automation and data reliability. Overall, the proposed method provides effective technical support for the safe operation, real-time monitoring, and intelligent management of seawall infrastructure.
Power distribution facilities in tidal urban river-network areas are exposed to compound flooding driven by intense rainfall, overloaded drainage systems, and elevated water levels in surrounding tidal rivers. Conventional area-based inundation assessments primarily characterize flood extent and depth, and therefore provide limited support for facility-level protection, mitigation prioritization, and emergency deployment. This study develops a refined flood-risk assessment method that couples hydrodynamic simulation outputs with surveyed characteristic elevations at individual power distribution rooms.A 118.4 km² tidal urban river-network area covering Haizhu District and adjacent areas of Guangzhou was selected as the study area. A coupled hydrodynamic model integrating a one-dimensional river network, two-dimensional overland flow, and an urban drainage network was constructed on the HydroMPM platform. Field investigations and real-time kinematic (RTK) surveys were conducted at 122 power distribution rooms with identified flood hazards to obtain road-surface elevations, flood-protection threshold elevations, water-ingress opening elevations, and equipment-base elevations. The simulated maximum inundation depth at each site was converted into a maximum water level using the surveyed road-surface elevation, and a "maximum water level–characteristic elevation" criterion was established for facility-level risk classification. Of the 122 surveyed rooms, 101 with complete modeled water-depth data were analyzed under 10 rainfall–tide scenarios. Model performance was evaluated using Typhoon Mangkhut in 2018 and the rainfall events of June 14, June 17, and August 3, 2025. For events with complete observational records, the maximum absolute tidal-level error did not exceed 0.12 m, the error in peak timing was within 1 h, and the maximum absolute error in inundation depth at representative sites was less than 0.30 m.Dominant water-ingress pathways were identified for 48 of the 101 assessed rooms, with rainfall-induced surface waterlogging and ingress through underground garages together accounting for 83.3% of these cases. Under the multi-year mean tide condition, the number of high-risk rooms ranged from 2 to 9 across the tested 1 h rainfall scenarios and from 5 to 13 across the tested 3 h rainfall scenarios. With the 3 h rainfall total fixed at 100 mm, replacing the multi-year mean tide boundary with a peak tide level of 3.5 m at Zhongda Station increased the number of high-risk rooms from 5 to 37. The extreme-tide stress-test scenario therefore produced 7.4 times as many high-risk rooms as the same rainfall scenario under the mean tide condition. The high-risk proportion was 58% for underground rooms and 27% for above-ground rooms; the difference was statistically significant according to a two-sided Fisher's exact test (p = 0.004).Within the range of scenarios examined, the increase in the tidal boundary produced a greater increase in the number of high-risk rooms (from 5 to 37) than did the variation across the tested 3 h rainfall scenarios (from 5 to 13), indicating a more pronounced amplification effect of tidal level on flood risk. Flood-risk assessment and emergency decision-making for power distribution facilities in tidal urban river-network areas should therefore not rely solely on rainfall forecasts. The proposed maximum water level–characteristic elevation criterion translates regional inundation outputs into site-specific risk information and can support graded protection, prioritization of pre-flood-season mitigation measures, and emergency-resource pre-positioning during extreme tide conditions. Since the 3.5 m tide scenario exceeds the highest tide level in the model validation events, the predicted number of high-risk rooms under this scenario should be further verified with additional high-tide observations.
Inner Mongolia lies at the margin of the East Asian monsoon zone and the climatic transition region, where extreme precipitation evolution is modulated by the interaction between the westerlies and the East Asian summer monsoon, exhibiting distinctly different characteristics from other regions. To elucidate the spatiotemporal evolution and future persistence of extreme precipitation and provide scientific underpinning for regional meteorological disaster mitigation and ecological security, this study employed the CHM_PRE V2 dataset. Five extreme precipitation indices (Prcptot, Rx5day, R10, CWD, and CDD) were selected for the period 1960–2024 across Inner Mongolia and its four climatic subregions. Theil–Sen slope estimation, Mann–Kendall trend test, Hurst exponent, and partial correlation analysis were comprehensively applied, with teleconnection factors (ENSO, NAO, and TNA) also investigated. The findings indicate that: (1) Extreme precipitation exhibits a pronounced "east–west" gradient, with the northeastern humid region being the high-value area for Rx5day, R10, and Prcptot, and the northwestern arid region being the low-value area. During 1991–2024 relative to 1960–1990, Rx5day, R10, and Prcptot increased, while CWD and CDD declined, reflecting a structural shift toward intensified precipitation and reduced dry spells. The humid region belongs to an area with continuously increasing and relatively stable precipitation intensity, whereas the semi-arid zone exhibits a composite characteristic of stronger precipitation fluctuation and alleviated drought. (2) Temporally, intensity indices showed upward trends, with Rx5day and R10 passing significance tests in the northeast; CDD decreased significantly across the entire region (-1.783 d/a), with the semi-arid zone registering the largest decline (-2.432 d/a), indicating that the risk of persistent drought has been significantly reduced. (3) Hurst exponents for all four indices exceeded 0.71, significantly higher than the random walk threshold, indicating strong positive persistence in historical trends that are highly likely to continue in the near future. (4) TNA correlated significantly and negatively with CDD across all subregions, peaking in the semi-humid zone (r = -0.2036), indicating that elevated TNA significantly suppresses persistent dry processes, consistent with the mechanism whereby North Atlantic sea surface temperature anomalies affect the East Asian summer monsoon through the Eurasian teleconnection wave train. NAO effects were spatially heterogeneous, with a weak negative correlation with CDD in the semi-arid region and a weak positive correlation with R10 in the humid region, possibly related to the regulatory effect of NAO on the intensity and position of the westerlies. ENSO correlations remained weak (|r| < 0.1), showing no significant association, indicating that the ENSO signal gradually attenuates during its northward propagation to mid- and high-latitudes. It is concluded that extreme precipitation in Inner Mongolia is characterized by increasing totals, intensifying extremes, and diminishing drought persistence, and this trend is highly likely to continue under the background of global warming. Elevated TNA exerts a significant suppressive effect on prolonged dry conditions.
To elucidate the hydrodynamic evolution mechanism in the downstream reach after the reserved units of existing large-scale hydropower hubs are put into operation, and to further optimize the multi-unit operation mode of power stations, this paper takes the construction project of No. 8 and No. 9 units of Longtan Hydropower Station as the research object. A two-dimensional hydrodynamic mathematical model of the Longtan–Tian'e County river channel was developed using the MIKE 21 hydrodynamic module, and subsequently calibrated and verified with measured water level data downstream of the dam. By simulating the hydrodynamic conditions of the downstream river channel under full-load conditions before and after the capacity expansion, this study focuses on analyzing the hydrodynamic evolution patterns induced by the increased discharge (from 3,882 m³/s to 4,990 m³/s) and quantifies the water level variation characteristics in the downstream reach. The results indicate the following: In terms of flow velocity, the disturbance caused by the increased discharge to the velocity and flow direction is concentrated within approximately 500 m downstream of the tailrace tunnel outlet. The primary impact manifests as a slight shift of the main stream axis toward the right bank in the local reach near the outlet, accompanied by a velocity increase of 0.61 m/s. Consequently, attention should be paid to riverbed erosion and deposition variations and the reinforcement of the right bank slope toe induced by such flow pattern changes. As the flow continuously adjusts during its downstream progression, the velocity variation gradually converges, with the direction of the main stream velocity vector varying within a range of approximately 0° to 0.56°. The main stream axes before and after the capacity expansion basically coincide, indicating that the overall flow pattern of the river channel does not change significantly due to the increased tailrace discharge, and no large-scale swinging of the main stream axis occurs. In terms of water level, the water levels at the cross-sections where the tailrace outlet is located rise by approximately 2.35 m and 2.16 m. The rise in tailwater level after expansion reduces the effective head of the hydraulic turbine-generator units, thereby constraining the power generation output and economic benefits under full-load conditions. This study effectively simulates the hydrodynamic variations in the downstream river channel after the reserved units of Longtan Hydropower Station are put into operation, providing a valuable reference for the operation mode and functional positioning of newly added units in subsequent large-scale hydropower hubs.
This study investigates the spatiotemporal distribution of water resources in the Dongjiang River Basin and explores the applicability of multi-timescale water resource assessment methods across different regions of the basin. Taking the Dongjiang River Basin as the study area, the spatiotemporal characteristics of water resource distribution are analyzed. The standardized runoff index (SRI) at multiple timescales was calculated for three hydrological stations—Longchuan, Heyuan, and Boluo—to determine and evaluate water resource conditions in the upper, middle, and lower reaches, respectively. The main steps for computing the SRI are as follows: ① collect and preprocess the runoff series data; ② fit various probability distribution functions to the runoff series and select the optimal one; ③ calculate runoff occurrence probabilities based on the selected distribution; ④ normalize the probability values to obtain the SRI. The study reveals the following spatiotemporal distribution characteristics of water resources in the Dongjiang River Basin: ① Across different timescales, the years with the most abundant water in the upper, middle, and lower reaches, as well as the driest years in the upper reach, are largely consistent, with differences only in months; however, the driest years differ between the middle and lower reaches. Water availability in the Dongjiang River Basin varies across timescales, and although connections exist among these scales, notable differences in water abundance and scarcity are also observed at each scale. Therefore, an integrated multi-timescale assessment is essential for evaluating the basin's water resources condition. ② Spatial differences in water resource distribution are evident within the basin, and these spatial patterns also vary across timescales. The upper reach is more prone to overall water deficits on medium- and long-term scales, the middle reach tends to experience water surplus on short timescales, and the lower reach is inclined toward water surplus on medium- and long-term scales, with a higher likelihood of extreme wet and dry events on short, medium, and long timescales. ③ The primary anthropogenic factor driving water resource changes in the Dongjiang River Basin is the operation of upstream reservoirs, while the seasonal distribution of precipitation also exerts a significant influence. The analysis reveals that water resources in the upper reach are less affected by human activities; in contrast, the middle and lower reaches experienced marked anthropogenic impacts from 1960 to 1990, and this influence has tended to weaken in recent years. The SRI provides a relatively accurate and straightforward means for comparative analysis of water resource conditions at different times and locations within the basin. Conducting water resource analysis by sub-basin regions on multiple timescales facilitates a comprehensive understanding of the spatiotemporal distribution characteristics of water resources in the Dongjiang River Basin and offers a reference for optimizing basin water resource management strategies.
With the deepening advancement of digital twin and smart water conservancy construction, hydraulic projects have placed higher demands on the accuracy and efficiency of 3D scene reconstruction. As an emerging 3D representation and rendering technology, 3D Gaussian Splatting (3DGS) provides an effective pathway for hydraulic scene reconstruction. This study systematically elaborates on the fundamental principles and technical approach of 3DGS, conducting a comparative analysis with oblique photography technology from dimensions such as data acquisition, data processing, and output results, revealing its significant advantages in modeling efficiency, lightweight capabilities, and visual effects. Building on this, a game engine-based 3DGS scene construction method is proposed, with practical applications demonstrating its feasibility in multi-source data fusion, real-time rendering, and immersive interaction. The research indicates that 3DGS technology can effectively address challenges in hydraulic projects, such as diverse landform types, complex observation environments, and high update frequencies, offering a feasible technical solution to enhance modeling efficiency and scene plasticity in digital twin water conservancy construction.
As a unique and important component of the reservoir ecosystem, the water level fluctuation zone plays a crucial role in blocking and purifying soil erosion and non-point source pollution in the reservoir area. Current research on the water level fluctuation zone primarily focuses on comprehensive utilization reservoirs, while studies on the effects of different bank slope control technologies in water supply reservoirs on soil pH, organic matter, and nitrogen, phosphorus, and potassium levels in the water level fluctuation zones remain relatively scarce.Based on this, this paper takes Xinfengjiang Reservoir as the study area, and takes "creating basic habitat conditions + cultivating species adapted to the habitat conditions" as the core concept, and the core technical idea is to reconstruct terrain slopes, retain and store surface runoff, stabilize soil, and realize water, soil and nutrient conservation, so as to create favorable conditions for the ecological protection and restoration of the water-level fluctuation zone. Targeted slope protection designs are formulated for gentle, moderate and steep bank slopes of the fluctuation zone in accordance with local site conditions. On this basis, uses blank as a control to compare and analyze the effects of four types of slope prevention and control technologies, namely, eco-bag retaining wall (multi-level platform), eco-gridding retaining wall (multi-level platform), eco-sheet-pile retaining wall (multi-level platform), and eco-sheet-pile retaining wall (slope) on soil pH, organic matter, nitrogen, phosphorus, and potassium in the water level fluctuation zone. The results of the study show that: After the implementation of the slope erosion control project, the pH value of the soil in the water level fluctuation zone was obviously improved, the soil organic matter content and the soil total nitrogen content were greatly improved, and the soil total phosphorus content and the soil total potassium content were mainly decreased. With the rise of elevation, the pH value of the soil in the water level fluctuation zone gradually increased, and the content of soil organic matter and total nitrogen also increased, while the content of soil total phosphorus and total potassium decreased. Comprehensive soil physicochemical properties, the overall effect of the ecological bag retaining wall (multi-level platform) is better than other slope prevention and control projects.The results of the study can provide important technical support for the ecological restoration of Xinfengjiang Reservoir's water level fluctuation zone, and can also provide useful reference for the ecological management of other reservoirs' water level fluctuation zones in the Pearl River Basin and further furnish solid support for advancing the theoretical development of ecological restoration technologies and facilitating interdisciplinary research.
This study investigates the dynamic response and impact of underwater rock plug blasting on adjacent hydraulic structures within a reservoir, with the aim of establishing a blasting impact assessment methodology for coupled dam–slope–structure systems.Taking the underwater rock plug blasting at the water intake of the Sanya Xishui Zhongdiao Water Diversion Project as the engineering background, an integrated approach combining theoretical analysis, field testing, and numerical simulation was adopted. Rational blasting parameters were determined through on-site breakthrough blasting tests. Based on the Sadovsky empirical formula and numerical analysis, the dynamic characteristics and potential risks imposed by blasting loads on critical structures—including the dam, slope, sump, and tunnel—were systematically evaluated.The findings demonstrate that: (1) the rock plug blasting exerts a negligible impact on the Dalong Reservoir dam, with a maximum peak particle velocity (PPV) of 0.234 cm/s recorded at the dam location, which satisfies the prescribed safety criteria; (2) the impact on the muck pit and tunnel is relatively pronounced, with a PPV of 9.54 cm/s observed near the blasting source, yet still compliant with safety standards; (3) the slope above the rock plug is the most critically affected, where the surface PPV exceeds the allowable vibration velocity (10 cm/s) at distances less than 24.5 m from the blasting center, and remains below the threshold at greater distances.Consequently, the removal of loose rocks and debris within a 30~40 m range above the slope is imperative prior to blasting to preclude their sliding or rolling into the water intake, thereby preventing obstruction. The findings of this study offer theoretical underpinnings and practical guidance for safety control and protective design in analogous underwater rock plug blasting engineering.
Under climate change, robust quantification of future precipitation and runoff evolution is critical for reservoir operation and water resources management. This study investigates the upstream basin of the Longtan Reservoir in Southwest China. A multi-model ensemble framework based on CMIP6 outputs is constructed, integrating bilinear interpolation downscaling, Delta bias correction, and Bayesian Model Averaging (BMA) to generate precipitation projections under SSP1-2.6, SSP2-4.5, and SSP5-8.5 scenarios for 2026–2100. A hybrid LSTM-Transformer model is then developed to simulate runoff responses and diagnose hydroclimatic nonstationarity. Evaluation against observations at 26 rain gauges during 2017–2025 demonstrates that the ensemble precipitation reproduces daily variability well, with a mean correlation of 0.84, confirming its suitability for driving hydrological simulations. The projected precipitation field maintains a stable spatial structure characterized by an east–west gradient, indicating persistent topographic and large-scale climatic controls. However, under high-emission forcing, both precipitation intensity and extreme-event frequency increase substantially, with SSP5-8.5 exhibiting the strongest amplification. All scenarios show increasing trends in annual precipitation, with accelerated growth under high forcing conditions. The LSTM–Transformer model achieves robust performance in runoff simulation, with NSE values of 0.89 (training) and 0.78 (testing), and successfully captures nonlinear rainfall–runoff dynamics and flood peak evolution, indicating strong generalization under nonstationary forcing conditions. Runoff projections reveal a consistent increase in annual mean inflows relative to the historical period (2010–2025), yet the response is non-monotonic with respect to emission intensity. The SSP2-4.5 scenario produces the highest annual runoff, despite not exhibiting the largest precipitation increase. This counterintuitive response is primarily attributed to more favorable hydroclimatic organization, characterized by a balanced precipitation structure and concentrated, sustained flood-season runoff contributions, which enhances rainfall–runoff conversion efficiency. In contrast, SSP5-8.5 exhibits stronger precipitation extremes but a more fragmented runoff regime, with reduced baseline flow contributions and enhanced intra-annual variability, reflecting intensified hydroclimatic instability. The SSP1-2.6 scenario remains constrained by insufficient precipitation input, limiting sustained runoff generation. Further analysis highlights that runoff responses exhibit strong nonlinearity across scenarios and cannot be explained solely by precipitation magnitude or extreme indices. Instead, runoff dynamics are jointly controlled by precipitation temporal–spatial organization and catchment-scale rainfall–runoff transformation efficiency, reflecting increasing system complexity under climate forcing. Despite scenario differences, the intra-annual runoff regime remains monsoon-dominated; however, high-emission conditions significantly intensify hydroclimatic contrasts, characterized by enhanced flood peaks, strengthened dry–wet separation, and increased seasonal asymmetry. Overall, future hydroclimatic conditions in the Longtan Reservoir upstream basin are projected to shift toward higher mean runoff, stronger extremes, and amplified temporal variability. These changes become progressively more pronounced with increasing emission levels, posing substantial challenges for reservoir flood regulation and adaptive water resources management under deepening climate nonstationarity.
Water‑surface area serves as a critical indicator that characterizes reservoir water storage status, watershed hydrological cycles and water‑resource conditions. Long‑term monitoring of its spatio‑temporal dynamics is essential for reservoir operation and watershed‑scale management. Taking Sentinel‑1 satellite imagery as the primary data source, together with topographic and meteorological datasets, this study extracts water‑body information and constructs a monthly time‑series dataset of Guishi Reservoir's water‑surface area spanning 2019‑2024 via the water‑index method on the Google Earth Engine (GEE) cloud‑computing platform. On this basis, this paper identifies the spatio‑temporal patterns of reservoir water‑surface variation and investigates its driving factors.The main results are shown as follows.①Obvious seasonal features exist in the inter‑monthly variation of Guishi Reservoir's water‑surface area during 2019–2024. The water‑surface area increases in spring and summer and peaks around July, before declining markedly in autumn and winter, and reached its five‑year minimum at the end of 2024.②The spatial distribution of reservoir inundation frequency presents a pattern of high‑value in the inner and southern reservoir areas and low‑value in peripheral and northern zones. The permanent water‑body covers approximately 30.93 km², accounting for 72.18% of the total water area; the seasonal water‑body occupies 5.71 km² (13.32%), and the ephemeral (temporary) water‑body accounts for the remaining 14.50%. Zones with frequent land‑water conversion are concentrated in the northern reservoir section and along reservoir shorelines.③The reservoir water‑surface area has an immediate geometric correlation with water‑level fluctuation. Its variation is strongly tied to reservoir operational scheduling and comparatively weakly affected by climatic conditions. Water‑surface area yields a high positive correlation with water level (R2=0.91). Positive correlations are also detected between water‑surface area and monthly precipitation as well as monthly average temperature, albeit with low coefficient‑of‑determination values. In addition, anthropogenic land occupation and land‑use activities are not the dominant cause for water‑surface‑area changes.In conclusion, the spatio‑temporal dynamics of Guishi Reservoir's water‑surface area are the combined outcome of man‑made reservoir regulation and natural environmental conditions.
Identifying propagation thresholds from meteorological drought to hydrological drought supports early warning and adaptive water resource management at the basin scale. Most existing threshold analyses rely on temporally defined drought events. They focus mainly on duration-based relationships and provide limited insight into spatial triggering conditions such as area. Consequently, a three-dimensional framework combining temporal, spatial, and intensity characteristics is needed to reveal meteorological–hydrological drought transformation mechanisms. This study proposes such a multidimensional threshold identification framework and applies it to the East River Basin in southern China. Monthly Standardized Precipitation Index and Standardized Runoff Index were used to represent meteorological and hydrological drought conditions, respectively. A three-dimensional spatiotemporal clustering algorithm was employed to identify drought events with temporal continuity, spatial extent, and intensity evolution. A minimum effective drought area threshold of 6 396 km² was applied to filter local noise while preserving regional signals. Cross-type event matching established meteorological–hydrological drought event pairs to quantify propagation patterns and time lags. Five candidate models, including linear, quadratic polynomial, exponential, power, and logarithmic functions, were fitted to describe nonlinear relationships between meteorological drought characteristics and hydrological responses. Optimal response-triggering threshold models were selected through statistical evaluation, significance testing, and Bootstrap uncertainty assessment. Results show that 26 meteorological and 29 hydrological drought events were identified during 1992–2013, yielding 27 valid matched pairs. Propagation time lags were predominantly synchronous, with the observed zero-month lag reflecting compression of soil moisture depletion, runoff generation, and channel routing processes within the same monthly statistical window rather than instantaneous physical transmission. Significant nonlinear threshold effects were found across all three dimensions. The optimal models for duration, area, and severity were exponential, power, and quadratic polynomial functions, respectively. The response-triggering threshold for duration was approximately 1 month. The spatial threshold for area was 17 515 km², equivalent to 50% of the basin area. This value marks the critical extent beyond which local buffering capacity is exceeded and systemic hydrological attenuation begins. Graded severity thresholds were identified as 2.86×104, 3.52×104, and 6.43×104 km².month for low, medium, and high levels. These correspond to three progressive hydrological stages: initial triggering buffered by soil water and shallow groundwater, stable propagation with passive consumption of reservoir storage, and strong response indicating near-exhaustion of both natural baseflow and artificial regulation capacity. The proposed framework extends traditional drought propagation analysis from single-dimensional temporal assessment to comprehensive three-dimensional characterization. The identified area threshold provides a quantitative measure of spatial critical conditions for drought propagation. The graded severity thresholds capture the progressive evolution of hydrological drought from initial triggering through stable propagation to intensified response. These findings improve understanding of meteorological–hydrological drought transformation mechanisms in humid basins under complex human regulation. They also provide quantitative indicators for drought early warning, flood–drought risk management and coordinated multi-reservoir operation in the East River Basin.
Under the dual influences of global climate change and human activities, the runoff dynamics in the Dongjiang River Basin have undergone significant changes, posing new challenges to water resource security. Based on measured runoff sequences from 1956 to 2023 at four hydrological stations (Longchuan, Heyuan, Lingxia, and Boluo) along the main stream of the Dongjiang River, along with precipitation data above the gauging sections, this study employs a hydrological variability diagnostic system to identify the variability characteristics of runoff and precipitation. Using an attribution analysis method based on the rainfall-runoff relationship, the contributions of climate change and human activities to runoff variability were quantitatively separated. Results indicate that runoff variability exhibits distinct seasonal distribution patterns. During the dry season, runoff generally shows a jumping upward trend, with variability points concentrated between 1968 and 1974, aligning with the operational water replenishment scheduling of the Three Major Reservoirs (Xinfengjiang, Fengshuba, and Bai Puzhu) in the basin during the dry season. In contrast, runoff during the wet season mostly exhibits a jumping downward trend, with variability points concentrated in 2008 and 2019, reflecting the combined effects of reservoir flood regulation and drought control. For annual runoff sequences, the Heyuan and Lingxia stations experienced a jumping downward variability in 2019, with significant reductions in annual runoff, reflecting the regulatory role of comprehensive basin management on runoff processes. Precipitation variability primarily occurs in May, June, and October, with consistent variability years across all sections, all showing a jumping downward trend, while the annual pan-precipitation sequence did not exhibit significant variability. Human activities are the dominant driver of runoff evolution, with contribution rates exceeding 90% during the dry season and reaching over 80% in most months during the wet season. The contribution rate of climate change significantly increased in March and April, closely related to the high variability of the pre-monsoon rainfall in South China and the strong modulation effects of ENSO events. These findings reveal the spatiotemporal differentiation patterns and driving mechanisms of runoff evolution in the Dongjiang River Basin, providing a scientific basis for optimized water resource management, water supply security, and climate change adaptation in the region.
Water diversion projects represent critical engineering interventions for mitigating the spatiotemporal mismatch of water resources in densely populated deltaic regions. However, long-distance raw water conveyance inevitably reshapes nutrient concentrations and trophic states in receiving reservoirs, triggering complex non-stationary shifts in aquatic water quality regimes. To systematically assess the temporal dynamics and spatial gradients of water quality across successive diversion nodes along the Pearl River Delta Water Resources Allocation Project, this study conducted year-round continuous monthly monitoring of key physicochemical water quality parameters in 2025 at four representative monitoring sites. These sites cover the full diversion cascade, including the Liyuzhou Pumping Station (the Xijiang River source water intake), Gaoxinsha Reservoir, Luotian Reservoir, and Gongming Reservoir (the terminal impoundment of this diversion corridor).Based on the field monitoring dataset, the Carlson trophic state index (TSI) was quantified for each sampling month to classify the eutrophication risk of each water body. Two multivariate statistical approaches, principal component analysis (PCA) and permutational multivariate analysis of variance (PERMANOVA), were further applied to disentangle the dominant environmental drivers of water quality differentiation and test the statistical disparities of nutrient regimes across different reservoir sites.The results showed that the water quality status of Gaoxinsha Reservoir, the first receiving impoundment downstream of the intake, was predominantly constrained by the nutrient loads carried by diverted inflow from Liyuzhou Pumping Station. Following the sequential water diversion gradient from Liyuzhou Pumping Station through Gaoxinsha Reservoir and Luotian Reservoir to Gongming Reservoir, the concentrations of total nitrogen (TN), total phosphorus (TP), nitrate nitrogen (NO3-N), as well as water electrical conductivity exhibited a steady declining trend, whereas underwater transparency continuously rose along the flow path. Correspondingly, the calculated TSI values dropped significantly at Gongming Reservoir, the terminal section of the entire project. Compared with the raw Xijiang River inflow at Liyuzhou Pumping Station, the three cascaded reservoirs achieved substantial nutrient retention and purification effects after sequential water storage and transport, which led to remarkable reductions in dissolved and particulate nutrient concentrations and obvious overall improvements in water quality.From a seasonal perspective, TN, chlorophyll-a (Chl.a) and TSI shared identical seasonal fluctuation patterns: their concentrations and index values peaked in summer and autumn, while decreasing to relatively low levels during spring and winter. In stark contrast, total phosphorus exhibited pronounced spatial heterogeneity, with divergent concentration magnitudes across the four monitoring nodes throughout the entire hydrological year. The outputs from PCA and PERMANOVA collectively corroborated that nitrogen nutrients, phosphorus nutrients, and integrated trophic status constituted the primary drivers governing spatial water quality variations. Statistically significant differences in the full suite of measured environmental variables were further confirmed among the three downstream cascade reservoirs.This study confirms that all water bodies along the Pearl River Delta Diversion Project currently stay in the oligotrophic to mesotrophic trophic state, with no severe eutrophication outbreaks recorded. However, significant spatial heterogeneity in nitrogen-phosphorus nutrient stoichiometric ratios was observed across the entire diversion corridor. Given the concurrent rise in chlorophyll-a (Chl.a) levels and Trophic State Index (TSI) values during the high-temperature summer–autumn transition, we recommend that targeted monitoring and precision water operation management prioritize key environmental drivers governing algal proliferation and nutrient retention in hot seasons. This proactive strategy mitigates latent algal bloom risks and sustains stable raw water quality to underpin regional water supply security.
Based on physical model experiments on sediment deposition behind a pile-supported wharf in a wave basin, this study proposes a sediment dredging design and site selection scheme for an offshore groin. The results show that sediment deposition behind the wharf increases towards the pile foundations beneath the wharf, and the most severe deposition area is located near the central part of the second berth. After the groin is installed in the most severe sediment deposition area, mean sediment deposition is reduced by about 29%, indicating that the groin can enhance the local flow velocity and improve the sediment dredging behind the wharf. Subsequently, a field demonstration is conducted in the basin behind a pile-supported wharf on the southeastern coast of China to evaluate the flow-diversion and sediment dredging potential of the device under realistic tidal current conditions. The offshore groin is designed as a movable flow-diversion structure composed of floating platforms, anchoring chains, diversion curtains, etc., which makes it be deployed fast in the restricted water area behind the wharf. Detailed flow measurements are conducted to examine the spatial distributions of the velocities as well as their dredging potential behind the wharf under the influence of the groin. The results of continuous 48 h field demonstration show that, the effective accelerating region exceeds 130% of the device characteristic length and the scouring-related velocity increases by at least 50% in 67% of the observation periods. The number of effective periods during ebb tide is 28% higher than that during flood tide. The area with enhanced dredging is also found to be more closer to the pile foundations due to the groin's orientation inclined to the wharf. Meanwhile, within 43% of the ebb-tide periods, the presence of the offshore groin enables local flow velocities to reach or exceed the critical velocity for sediment incipient motion, indicating its favorable flow diversion and sediment dredging capability under the tidal forcing. The findings provide a technical basis for the design and deployment of those devices in flow diversion and dredging activity behind the pile-supported wharves, with both benefits of reducing the dredging costs using mechanical equipment and improving the personnel safety with less operation on sea.
Monthly runoff series are widely recognized to exhibit pronounced nonlinearity, non-stationarity, and multi-scale fluctuation characteristics, which pose notable challenges to accurate single-station monthly runoff forecasting. Even after primary decomposition processing, the reconstructed high-frequency sub-components often retain intricate local fluctuating patterns, rendering them difficult to predict directly with satisfactory performance. To further enhance the prediction accuracy and operational stability for single-station monthly runoff series, this study proposes a novel hybrid forecasting framework that integrates secondary decomposition, intelligent parameter optimization, and deep learning techniques. The proposed hybrid model couples complete ensemble empirical mode decomposition with adaptive noise (CEEMDAN), variational mode decomposition (VMD), the crested porcupine optimizer (CPO), Transformer, and bidirectional long short-term memory (BiLSTM) into an integrated forecasting workflow. Monthly runoff records from the Gongshang Hydrological Station in the upper Xihe River Basin were adopted as the experimental dataset, covering a total of 552 consecutive months from 1975 to 2020. The full time series was chronologically partitioned into a training subset spanning 442 months (1975–2011) and a test subset covering 110 months (2012–2020). During the training phase, CEEMDAN was first implemented on the runoff series in the training set to generate a set of intrinsic mode functions (IMFs) and one residual term. Sample entropy was subsequently employed to quantify the complexity of each decomposed sub-component, and K-means clustering was adopted to reconstruct these sub-components into three grouped components: high-frequency, medium-frequency, and low-frequency. The reconstructed high-frequency component was further processed via secondary decomposition using VMD. All VMD-derived sub-modes, together with the pre-obtained medium- and low-frequency reconstructed components, were assembled into the final component set for subsequent component-wise forecasting. The CPO algorithm was introduced to automatically tune the critical hyperparameters of the Transformer-BiLSTM architecture, including the input time step, number of BiLSTM hidden units, initial learning rate, and number of attention heads. In the testing phase, a rolling forecasting strategy was applied to generate predictions for each individual component, and the final monthly runoff forecast was derived by summing the predicted values of all sub-components. The proposed hybrid CEEMDAN-VMD-CPO-Transformer-BiLSTM model demonstrates excellent forecasting performance on the test dataset. The corresponding determination coefficient (𝑅R), Nash-Sutcliffe Efficiency (NSE), root mean square error (RMSE), and mean absolute error (MAE) are quantified as 0.960, 0.951, 0.226 m³/s, and 0.171 m³/s, respectively. Compared with four benchmark models (Transformer-BiLSTM, CNN-BiLSTM, BiLSTM, and LSTM), the proposed model achieves RMSE reductions of 47.36%, 53.69%, 56.47%, and 59.59%, along with MAE reductions of 49.93%, 52.61%, 56.68%, and 57.65%, respectively. Over 20 independent repeated runs, the full proposed model yields a stable RMSE of 0.226±0.006. The Wilcoxon signed-rank tests between the full model and its corresponding ablation variants return p-values below 0.0001, confirming that the observed performance improvements are statistically significant. The proposed hybrid framework effectively enhances the forecasting accuracy and operational stability of monthly runoff at the Gongshang Hydrological Station. The results demonstrate that the integrated performance improvement stems from the synergistic effects of the CEEMDAN-VMD decomposition-reconstruction scheme, secondary VMD decomposition, CPO-driven parameter optimization, and the adopted Transformer-BiLSTM forecasting architecture. This developed model can serve as a reliable reference for analogous single-station monthly runoff forecasting investigations.
Against the backdrop of intensifying global climate change, the evolutionary dynamics of flood characteristics in small- and medium-sized watersheds have grown increasingly intricate. Accurately identifying these spatiotemporal patterns carries critical practical implications for the formal review and updating of design flood standards. Taking the Liuxihe Reservoir as the study case, this paper systematically investigates the evolutionary behaviors of its inflow flood peak discharge and flood volume series from three complementary dimensions: long-term trend, abrupt change point detection, and multi-timescale periodicity, followed by a quantitative assessment of how hydrological record extension influences the reassessment outcomes of design flood values.Using a 66-year flood record (1959–2024) of the Liuxihe Reservoir, this study applied the Mann–Kendall test and Sen's slope estimator to identify the trends and rates of change in the inflow flood peak and volume series. The Pettitt test was then employed to detect change points and their statistical significance. The Morlet wavelet transform was used to reveal the time-scale periodic characteristics of the flood series. Finally, based on these identified evolution patterns, the impact of extending the data series on the reassessment of design floods and characteristic water levels was quantitatively evaluated.The results showed that the annual maximum peak discharge, 24-hour flood volume, and 3-day flood volume series all exhibited slight decreasing trends, but none were statistically significant. The Sen's slope estimates were -2.33 m³/s per year, -0.001 1×10⁸ m³ per year, and -0.0007×10⁸ m³ per year, respectively, with 95% confidence intervals included zero, suggesting considerable uncertainty in the trends. The Pettitt test showed that the p-values were all above 0.05, indicating no statistically significant abrupt changes, thus satisfying the hydrological consistency assumption. The Morlet wavelet analysis revealed two distinct periodic components in the flood variables: a long period of approximately 33 years and a short period of 5–7 years. The peak discharge and 24-hour flood volume were dominated by the long period, while the 3-day flood volume was governed by the short period. This discrepancy reflects the differential responses of flood events with varying durations to multi-decadal atmospheric circulation regimes and interannual climate anomalies. Extending the series to 2024 effectively covers multiple cycles of the dominant periods, thereby improving the reliability of design flood parameter estimation. The reassessed design flood peak discharge and flood volumes were generally comparable to the 1999 results but slightly lower, which is expected because no larger flood events occurred in the extended period, and the reduction in frequency estimates follows statistical principles. Flood routing showed that the 100-year and 1 000-year maximum water levels were 237.24 m and 238.42 m, both below the original design values, thus meeting the regulatory requirements.The annual maximum flood series of Liuxihe Reservoir exhibits overall statistical stationarity, with well-defined multi-timescale periodic oscillations and no statistically significant monotonic long-term trend. These empirical findings deviate substantially from the widely documented rising tendencies of extreme flood events in small- and medium-sized watersheds under ongoing climate change, which underscores the regional specificity of flood regime evolution in this catchment. Follow-up investigations based on extended observational datasets and cross-watershed comparative analyses are warranted to fully disentangle the intricate hydrological response mechanisms behind such atypical flood behaviors. The outcomes of this study provide a robust scientific reference for design flood reassessment and reservoir operational management under the non-stationary background of a changing climate.
Rainwater harvesting is an important strategy for alleviating urban water shortages, reducing drainage pressure, and improving the resilience of urban stormwater management systems. Accurate land-cover recognition is essential for estimating surface runoff and rainwater harvesting potential, particularly in highly urbanized areas where impervious surfaces, vegetated land, roads, rooftops, and water bodies are spatially mixed. In this study, an urban rainwater harvesting potential assessment framework was developed by integrating high-resolution remote sensing interpretation, a U-Net land-cover recognition model, and hydrological estimation methods. The six urban districts of Xi'an, China, were selected as the study area. Five major land-cover types, including water bodies, bare land, vegetated land, building rooftops, and roads, were identified from remote sensing imagery. The U-Net model was constructed using ArcGIS Learn with a ResNet-34 backbone, and a support vector machine (SVM) model was introduced as a traditional supervised classification method for comparison. Classification performance was evaluated using overall accuracy, Kappa coefficient, precision, recall, F1-score, and confusion matrices. Based on the land-cover recognition results, runoff coefficients were estimated using both the composite runoff coefficient method and the Soil Conservation Service Curve Number (SCS-CN) model. The composite runoff coefficient method assigns empirical runoff coefficients according to land-cover types, whereas the SCS-CN model further incorporates hydrological soil groups and antecedent moisture conditions to represent differences in infiltration and retention capacity. A slope correction factor was also introduced to adjust runoff coefficients under different terrain conditions. Precipitation records from 1961 to 2023 were analyzed to establish rainfall scenarios corresponding to 10%, 50%, and 90% frequency levels, as well as the long-term mean annual rainfall. In addition, classification error propagation was considered using the U-Net confusion matrix to evaluate the influence of land-cover recognition uncertainty on rainwater harvesting potential estimates. The results show that the U-Net model achieved an overall accuracy of 0.780 and a Kappa coefficient of 0.725, outperforming the SVM model, which achieved an overall accuracy of 0.430 and a Kappa coefficient of 0.276. The U-Net model showed relatively good recognition performance for building rooftops and roads, while some confusion remained among water bodies, bare land, and vegetated land. Compared with the composite runoff coefficient method, the SCS-CN model produced higher runoff coefficients and stronger spatial heterogeneity, indicating that it can better reflect the effects of land-cover types and hydrological soil conditions on runoff generation. Under rainfall scenarios corresponding to the 10%, 50%, and 90% frequency levels and the long-term mean annual rainfall, the corrected rainwater harvesting potentials were estimated as 285, 220, 173, and 226 million m³·yr⁻¹, respectively. Building rooftops and roads were identified as the dominant contributors to rainwater harvesting potential. The proposed framework provides a technical reference for urban rainwater resource assessment and resilient stormwater management.
