Search Results (52)

Under complex and changing environmental conditions, hydrological drought in the upper Hanjiang River (UHR) is becoming increasingly severe, so investigating the variation characteristics and influencing factors of hydrological drought in this basin can provide favorable support for drought prevention and water resources management. In this study, based on monthly runoff data from the Ankang Hydrological Station of the UHR, the mutation change year at the Ankang Station was first identified using the Pettitt mutation test and the B-G segmentation algorithm. Subsequently, the ABCD hydrological model coupled with eight machine learning algorithms was employed to simulate the runoff variation process in the Ankang Station. Finally, we used the Standardized Runoff Index to describe the hydrological drought conditions and quantitatively analyzed the impacts of human activities and climate change on the seasonal hydrological drought in the UHR. The results showed that (1) the coupled machine learning–hydrological model can effectively improve the simulation accuracy of the runoff change process. (2) The coupled ABCD–Random Forest model has the highest accuracy. (3) Hydrological drought exhibits a significant increasing trend in spring and autumn, a significant decreasing trend in winter, and a non-significant increasing trend in summer. (4) Climate change serves as the primary driver of hydrological drought variations across four seasons in the UHR. Full article

Climate change is reshaping hydrologic regimes in snow-dominated watersheds, with important implications for mine rock drainage quantity and contaminant mobilization. This study quantifies potential long-term changes in drainage quantity by coupling a previously published physics-informed machine learning model with a Monte Carlo framework driven by downscaled monthly climate projections from ClimateNA. The proposed methodology was applied to three drainage monitoring stations at a mine site in Western Canada to assess projected drainage responses over the 2011–2100 period. An ensemble of daily weather sequences was generated by sampling historical within-month variability and scaling the resulting series to match projected monthly climate statistics, which were then used as inputs for the drainage model. Trends were assessed using the Mann–Kendall test modified for serial correlation, and their magnitudes were summarized using the Theil–Sen slopes. The trend analysis results indicate scenario-dependent changes in annual drainage across stations, alongside consistent seasonal shifts toward higher spring (April–May) and lower early-summer (June–July) drainage. These patterns are consistent with earlier snowmelt and earlier snowpack depletion. Corresponding shifts in intra-annual flow timing suggest that a larger fraction of annual drainage occurs earlier in the year. Overall, these findings provide a physics-informed basis for changes in drainage quantity and for guiding monitoring, design, and mitigation strategies under a warming climate. Full article

Accurate estimation of reference evapotranspiration (ET₀) can be decisive in agricultural, hydrological and meteorological applications. Although different machine learning (ML)-based models have been successfully applied for ET₀ estimation under a wide spectrum of climatic conditions, most of these models present the crucial shortcoming of being site-specific. Hence, a thorough hyperparameter tuning would be necessary before translating such models to another domain with different data distributions. The hyperparameter tuning is a complex procedure that mainly depends on the operator’s experience. Automated ML might be a suitable approach to adapt the models’ architectures. The present study evaluated the performance of different automated ML algorithms, namely, neural architecture search (NAS), Optuna, enhanced grey wolf (EGWO), and quantum whale optimization (QWOA) algorithms coupled with random forest, neural networks, and light gradient boosting models for estimating daily ET₀ at three different climatic regions (Cairo, Singapore, and London). For local validation, the NN-NAS model provided the most accurate results in Cairo (R² = 0.969, RMSE = 0.432 mm/day) and Singapore (R² = 0.657, RMSE = 0.596 mm/day), while NN-Optuna provided the highest performance accuracy in London (R² = 0.941, RMSE = 0.370 mm/day). Hybrid AutoML models improved R² by 5–15% and reduced RMSE by 10–20% compared to standalone models. In external validation, NN-NAS and NN-Optuna presented superior generalizability, with R² values up to 0.899 and 0.680 in London and Cairo, respectively. Nonetheless, the performance of the hybrid models depended on the climatic conditions of the studied sites, where NN-NAS was the best model for the arid site, while NN-Optuna provided the highest accuracy in the temperate climate. Further, the analysis of variance confirmed significant differences among the performance accuracies of the developed model. The Shapley additive explanations (SHAP) analysis was performed to identify the variables’ effect on ET₀ estimation, which suggested that solar radiation showed the highest impact in all three studied climatic contexts, although the degree of importance was climatic dependent. Finally, an external modeling scenario was conducted using exogenous data for estimating ET₀ at the target sites, which confirmed the models’ ability. Full article

Flood susceptibility mapping in high-altitude ungauged basins faces a structural dichotomy: physically based models often suffer from systematic biases due to uncertain satellite precipitation, whereas data-driven models are prone to overfitting and lack physical consistency in data-scarce regions. To resolve this, this study proposes a Physically constrained Particle Swarm Optimization–Random Forest (P-PDRF) framework, validated in the Lhasa River Basin. The core innovation lies in coupling a hydrological model with statistical learning by utilizing the maximum daily runoff depth as a “Relative Hydraulic Intensity Index.” This approach leverages the topological correctness of physical simulations to circumvent absolute forcing errors. Furthermore, a Physiographically Constrained Negative Sampling (PCNS) strategy and a PSO-optimized “Shallow Tree” configuration are introduced to enforce structural regularization against stochastic noise. Empirical results demonstrate that P-PDRF achieves superior generalization (AUC = 0.942), significantly outperforming standard Random Forest, Support Vector Machine, and Analytic Hierarchy Process models. Ablation studies confirm that the dynamic index outweighs the static Topographic Wetness Index in feature importance, effectively correcting topographic artifacts where static models misclassify arid depressions as high-risk zones. This study offers a scalable Physics-Informed Machine Learning solution for the global “Prediction in Ungauged Basins” initiative. Full article

The Turpan–Hami Basin, a critical energy hub in northwestern China, is plagued by frequent ground collapses induced by extensive mining over karst geology, threatening ecology and safety. Current hazard assessment methods, mainly single linear or traditional machine learning models, fail to capture the complex nonlinear interactions inherent to this coupled geo-mining environment. This study addresses this gap by establishing a multi-dimensional “Geology-Mining-Hydrology-Environment” index system comprising 14 critical factors—including lithology, goaf distribution, mining intensity, and their interaction terms. A coupled gradient boosting decision tree and logistic regression (GBDT-LR) model, optimized for the multi-factor coupling characteristics of ground collapse in arid mining basins, was applied for the hazard assessment. The results reveal a distinct spatial pattern of “core agglomeration with multi-level gradient differentiation.” Extremely high-hazard areas, covering 9.21% of the area, are concentrated in the core mining areas northwest of Turpan and southwest of Hami, while high-hazard areas (4.63%) form surrounding belts. The GBDT-LR model (AUC = 0.871) demonstrated significantly superior performance over a single logistic regression model (AUC = 0.813), proving its enhanced capability to identify high-hazard areas by modeling complex factor interactions. This work provides an essential scientific foundation for implementing zonal hazard management and prioritizing disaster prevention projects in key areas of the basin. Full article

The integration of Geographic Information Systems (GISs) with hydrologic science has evolved over seven decades from manual catchment delineation and output visualization to AI-native spatial water intelligence, reshaping how the water cycle is observed, modeled, and managed. This review explores that evolution, from the progressively tightening coupling between GIS software and hydrologic models to an AI-assisted future in which the line between these two fields blurs and eventually dissolves completely. The evolution of GISs in hydrology is traced through four eras, stratified as: (1) the formalization of governing equations and digital terrain representations (1950–1985); (2) the initial GIS–model coupling era and the rise in watershed simulation (1985–2000); (3) open source and the start of the open data deluge (2000–2015); and (4) machine learning and cloud-native computing (2015–present). A four-level vision for the role of artificial intelligence in the next generation of spatial hydrology is then articulated, from AI-assisted GIS operation to spatially aware AI water intelligence that reasons directly over geospatial data without requiring a traditional GIS or simulation software as an intermediary. Broader limitations and challenges are also discussed. Full article

Urban road collapses are characterized by sudden occurrence and strong spatial heterogeneity, posing substantial challenges for proactive infrastructure management. Susceptibility mapping can provide spatially explicit evidence to support targeted inspection and early-warning strategies. Using Futian District, Shenzhen (China) as a case study, a total of 315 road collapse events recorded during 2019–2023 were compiled to develop an integrated framework for urban road collapse relative susceptibility mapping based on multi-source remote sensing and urban spatial data. First, an indicator-based susceptibility index (SI) was constructed using eight conditioning factors, including PS-InSAR-derived deformation, topographic–hydrological conditions, and distance-based infrastructure variables (distance to underground utilities, metro lines, and roads). Factor weights were determined by coupling the Analytic Hierarchy Process (AHP) with the Entropy Weight Method (EWM), producing a comprehensive SI for historical collapse locations. Subsequently, a set of 17 remote-sensing predictors, including Sentinel-2 spectral bands, Sentinel-2 GLCM texture features, and Sentinel-1 SAR backscatter variables, was used to train a Random Forest model to predict SI and generate continuous susceptibility maps at the urban road-network scale. The influence of neighborhood window size on predictive performance was systematically evaluated. Results show that the Random Forest model performed best at the 5 × 5 window scale (R² = 0.70, RMSE = 0.0172, MAE = 0.0122), outperforming both pixel-based inputs (1 × 1) and larger windows. Uncertainty analysis further indicated that the 5 × 5 RF configuration yielded the most stable and spatially coherent predictions, whereas overly small windows and less robust learners produced more fragmented or higher-uncertainty susceptibility patterns. Spatiotemporal analysis indicates that susceptibility patterns remained broadly stable from 2019 to 2023, with moderate susceptibility accounting for 50.82–57.89% and high susceptibility for 21.94–23.30%, while very high susceptibility consistently remained below 1%. Overall, this study demonstrates that integrating multi-source remote sensing with scale-optimized machine learning provides an effective approach for fine-scale susceptibility mapping of urban road collapses, offering practical guidance for differentiated monitoring and risk prevention along critical road corridors. Full article

To address the limitations of traditional runoff prediction methods—namely, the oversimplification of meteorological factor selection, ambiguous interactions among core variables, and the disruptive influence of redundant inputs—this study focuses on the Zijiang River Basin as a representative case. A suite of machine learning models, including Long Short-Term Memory Neural Network (LSTM), Convolutional Neural Network (CNN)-LSTM, Temporal Convolutional Network (TCN), and Gradient Boosting Regression Tree (GBRT), was constructed and trained using 13 distinct combinations of meteorological variables. These configurations were systematically evaluated to assess their compatibility with each model in simulating daily runoff patterns. Additionally, the Shapley Additive Explanations (SHAP) algorithm was employed to quantitatively assess the contribution of each factor to predictive accuracy. Among the models tested, the TCN model consistently demonstrated superior performance, particularly in mitigating the effects of irrelevant or redundant features. The GBRT model showed distinctive strengths in accurately predicting peak flow timings. Of all input configurations, the combination of “runoff + precipitation + evaporation + temperature” emerged as the most effective. Findings indicate that the predictive value of individual meteorological variables hinges primarily on their direct correlation with runoff, while the effectiveness of multi-factor schemes depends on the degree of functional integration—specifically, the coupling of hydrological recharge, consumption, and regulatory processes. The presence of redundant variables was found to impair model performance unless they contributed to a meaningful synergistic relationship with core inputs. The SHAP analysis further reinforced these insights: precipitation-related variables proved to be the most critical to prediction accuracy, whereas temperature and evaporation served more complementary roles. Notably, the inclusion of relative humidity tended to suppress runoff responses and increased deviation in peak timing estimates. These findings shed light on the nuanced interplay between meteorological input design and model selection, offering a robust foundation for optimizing data-driven runoff prediction frameworks. Full article

Trace metal contamination in lotic freshwater systems exhibits pronounced heterogeneity arising from coupled hydrological connectivity, geochemical partitioning, and anthropogenic forcing, complicating exposure characterization in urban and peri-urban catchments. Addressing this complexity requires integrative analytical approaches capable of deciphering system-level controls, prompting an investigation of the environmental structuring and governing controls of dissolved trace metal signatures in a human-impacted stream using a system-oriented computational framework. To capture temporal variability associated with seasonal hydrological contrasts and heterogeneous pollution inputs, a station-based, season-resolved sampling strategy was implemented during the wet and dry seasons. Physicochemical gradients (pH, temperature, dissolved oxygen, and electrical conductivity), inorganic nitrogen species (NH₃, NO₂⁻, and NO₃⁻), and phosphorus fractions (total phosphorus, TP; total orthophosphate, TOP; soluble reactive P, SRP) were jointly analyzed with dissolved concentrations of chromium (Cr), copper (Cu), nickel (Ni), lead (Pb), cadmium (Cd), mercury (Hg), and arsenic (As). Regression-based machine learning models were used to quantify element-specific sensitivities to hydrochemical drivers under wet–dry periods and to identify optimal predictive configurations. Predictive performance was consistently high for trace metals (R² generally >0.95), with Random Forest providing the best accuracy for Cr, Ni, Pb, Cd, As, and Hg, whereas Cu was most reliably captured by an XGBoost tree ensemble (R² = 0.994). Explainability analyses revealed heterogeneous, metal-specific control regimes: Cr was primarily driven by temperature, Ni by NO₂⁻ and redox-sensitive conditions, Cd by NH₃ and temperature, and As by Hg in combination with phosphorus-related and redox-linked proxies, while Pb showed comparatively lower predictability relative to other metals. Trace metal distributions are therefore structured primarily by differential environmental sensitivity rather than uniform source-driven inputs, reinforcing the need for integrative computational frameworks when interpreting freshwater contamination under intensifying anthropogenic and climatic pressures. Full article

Cultivated land fragmentation (CLF) has evolved from a physical landscape phenomenon into a systemic constraint on agricultural sustainability, especially in rapidly urbanizing regions such as the Yangtze River Delta (YRD). Existing studies are limited by static “snapshot” comparisons that obscure continuous trajectories and by linear models that fail to capture nonlinear interactions and threshold effects. This study integrates the Space–Time Cube (STC) model with an interpretable machine learning framework (Extreme Gradient Boosting–Shapley Additive Explanations, XGBoost–SHAP) to explore the spatiotemporal dynamics and driving mechanisms of CLF in the YRD (1990–2020) at a 1 km² resolution. The STC identifies a distinct north–south gradient, with persistent hotspots in low-lying plains and intensifying fragmentation at peri-urban interfaces. SHAP interpretation suggests a “Base–Stabilizer–Amplifier” structure in the modeled relationships: hydrological accessibility and soil fertility form the dominant background linked to higher CLF, whereas topography correlates with lower CLF, and socioeconomic variables exhibit nonlinear, threshold-like increases in fragmentation beyond higher development levels. Overall, CLF reflects coupled natural–anthropogenic interactions with pronounced nonlinear responses. This mechanism-oriented framework provides actionable guidance for adaptive farmland governance. It also offers a transferable methodology for analyzing land system changes in other deltaic agricultural regions worldwide. Full article

Introduction: Traditional approaches to discover teleconnections and quantify uncertainty, such as global sensitivity analysis, Monte Carlo experiments, decomposition analysis, etc., are computationally intractable for large-scale process-based Coupled Human and Natural Systems (CHANS) models. This study hypothesizes that machine-learned emulator models provide “computationally efficient” algorithms for discovering teleconnections and quantifying uncertainty within and across dynamically evolving human and natural systems. Objectives: This study aims to harness machine-learned emulator models to discover the relative contributions of internal- versus external-to-the-lake teleconnected processes driving the emergence of Harmful Algal Blooms (HABs) and trophic regime shifts. Three objectives are pursued: (1) build emulators; (2); quantify uncertainty and (3) identify teleconnections. Methods: Six machine-learned emulator models are trained on ~3.8 million observations for ~52 features derived from 332 scenarios simulated in an integrated process-based CHANS model that predicts water quality in Missisquoi Bay of Lake Champlain under alternate hydro-climatic and nutrient management scenarios for the 2001–2047 timeframe. The regression random forest (RRF), regression LightGBM (RLGBM) and regression XGBoost (RXGB) models predict the average surface mean of ChlA. Further, the classifier random forest (CRF), classifier LightGBM (CLGBM) and classifier XGBoost (CXGB) predict four trophic states of Missisquoi Bay. Relative importance and partial dependence plots are derived from all six emulator models to quantify relative uncertainty and importance of external-to-the-lake (climatic, hydrological, nutrient management) and internal-to-the-lake (P and N sediment release) drivers of HABs. Results: RXGB (R² = 96%, 48 features) outperforms RLGBM (R² = 95%, 37 features) and RRF (R² = 93%, 20 features) in predicting the average surface mean of ChlA. CLGBM (F1 = 96.15, 4 features) outperforms CXGB (F1 = 95.66, 48 features) and CRF (F1 = 93.06, 23 features) in predicting four trophic states. We discovered that predictor variables representing snow, evaporation and transpiration dynamics teleconnect hydro-climatic processes occurring in terrestrial watersheds with the biogeochemical processes occurring in the freshwater lakes. Conclusions: The proposed approach to discover teleconnections and quantify uncertainty through machine-learned emulator models can be scaled up in different watersheds and lakes for informing integrated water governance processes. Full article

A storage-aware water-scarcity risk assessment framework coupling satellite remote sensing, geospatial datasets with the IPCC exposure-hazard-vulnerability (EHV) paradigm was designed to evaluate the spatiotemporal dynamics of global water scarcity risk over the past two decades. To achieve this, a performance-weighted ensemble machine learning approach was employed to reconstruct long-term terrestrial water storage (TWS) from satellite observations, augmented with glacier-mass calibration to improve reliability in cryosphere-affected regions. Global water withdrawal dataset was generated by integrating remote sensing, geospatial dataset, and machine learning to mitigate the dependency of parameterized land surface hydrological models and enable consistent risk mapping. Satellite-derived results reveal obvious TWS declines in Asia, Northern Africa, and North America, particularly in irrigated drylands and glacier-dominated regions. EHV paradigm and big datasets further identified high-water scarcity risk in Asia and Africa, especially in agricultural regions. Water stress has intensified in Africa over the past two decades, while a decreasing trend is observed in parts of Asia. Vulnerability levels in Asia and Africa are approximately eight times higher than those in other global regions. Results reveal a strong connection between water stress and socioeconomic factors in Asia and Africa, reflecting global disparities in water resource availability. Full article

Heavy metal contamination of aquatic systems represents a critical environmental and public health concern due to the persistence, toxicity, and bioaccumulative potential of these elements. Geographic information systems (GISs) have emerged as indispensable tools for the spatial assessment and management of heavy metals (HMs) in water resources. This review systematically synthesizes current research on GIS applications in detecting, monitoring, and modeling heavy metal pollution in surface and groundwater. A bibliometric analysis highlights five principal research directions: (i) global research trends on GISs and heavy metals in water, (ii) occurrence of HMs in relation to World Health Organization (WHO) permissible limits, (iii) GIS-based modeling frameworks for contamination assessment, (iv) identification of pollution sources, and (v) health risk evaluations through geospatial analyses. Case studies demonstrate the adaptability of GISs across multiple spatial scales, ranging from localized aquifers and river basins to regional hydrological systems, with frequent integration of advanced statistical techniques, remote sensing data, and machine learning approaches. Evidence indicates that concentrations of some HMs often surpass WHO thresholds, posing substantial risks to human health and aquatic ecosystems. Furthermore, GIS-supported analyses increasingly function as decision support systems, providing actionable insights for policymakers, environmental managers, and public health authorities. The synthesis presented herein confirms that the GIS is evolving beyond a descriptive mapping tool into a predictive, integrative framework for environmental governance. Future research directions should focus on coupling GISs with real-time monitoring networks, artificial intelligence, and transdisciplinary collaborations to enhance the precision, accessibility, and policy relevance of heavy metal risk assessments in water resources. Full article

The integrated assessment of watershed ecosystems is increasingly critical for sustainable water resource management amid global environmental change. Multi-source data integration—encompassing in situ monitoring, remote sensing, and model-based observations—has significantly expanded the spatial and temporal scales at which watershed processes can be analyzed. Concurrently, advances in model coupling strategies, ranging from loose to embedded architectures, have enabled more dynamic and holistic representations of interactions among hydrology, water quality, and ecological systems. However, a unifying operational framework that links multi-source data, cross-scale coupling, and rigorous uncertainty propagation to actionable, real-time decision support is still missing, largely due to gaps in interoperability and stakeholder engagement. Addressing these limitations demands the development of intelligent, adaptive modeling frameworks that leverage hybrid physics-informed machine learning, cross-scale process integration, and continuous real-time data assimilation. Open science practices and transparent model governance are essential for ensuring reproducibility, stakeholder trust, and policy relevance. The recent literature indicates that loose coupling predominates, physics-informed ML tends to generalize better in data-sparse settings, and uncertainty communication remains uneven. Building on these insights, this review synthesizes methods for data harmonization and cross-scale integration, compares coupling architectures and data assimilation schemes, evaluates uncertainty and interoperability practices, and introduces the Smart Integrated Watershed Eco-Assessment Framework (SIWEAF) to support adaptive, real-time, stakeholder-centered decision-making. Full article

In this study, the geological disasters in Guizhou Province serve as the research object, and a systematic susceptibility evaluation is conducted in light of the province’s prominent problems with frequent geological disasters. The current research primarily focuses on the application of a single model, often with deficiencies in factor interpretation. It has not yet systematically integrated the advantages of the traditional information model and multiple machine learning algorithms, nor introduced interpretable methods to analyze the disaster mechanism deeply. In this study, the information value (IV) model is combined with machine learning algorithms—logistic regression (LR), decision tree (DT), support vector machine (SVM), random forest (RF), and extreme gradient boosting (XGBoost)—to construct a coupling model to evaluate the susceptibility to geological disasters. Combined with the Bayesian optimization algorithm, the geological disaster susceptibility evaluation model is built. The confusion matrix and receiver operating characteristic (ROC) curve were used to evaluate the model’s accuracy. The Shapley Additive exPlanations (SHAP) method is used to quantify the contribution of each influencing factor, thereby improving the transparency and credibility of the model. The results show that the coupling models, especially the IV-XGB model, achieved the best performance (AUC = 0.9448), which significantly identifies the northern Wujiang River Basin and the central karst core area as high-risk areas and clarifies the disaster-causing mechanism of “terrain–hydrology–human activities” coupling. The SHAP method further identified that NDVI, land use type, and elevation were the predominant controlling factors. This study presents a high-precision and interpretable modeling method for assessing susceptibility to geological disasters, providing a scientific basis for disaster prevention and control in Guizhou Province and similar geological conditions. Full article