Karst springs are characterized by heterogeneous hydrological responses and complex dynamics of suspended material transport during rainfall events. This study aimed to improve understanding of the processes that determine the hydrological response of a karst aquifer by integrating analyses of the hysteretic relationships of turbidity–discharge and electrical conductivity–discharge. The analysis was based on measurements of rainfall, discharge, turbidity, and electrical conductivity during individual events at the Žrnovnica Spring, focusing on the types of turbidity–discharge and electrical conductivity–discharge hysteresis loops during periods when turbidity exceeded 4 NTU, and on the occurrence of isolated exceeding turbidity episodes. A total of 15 events with turbidity above 4 NTU were identified, dominated by complex hysteresis loops. The results show pronounced temporal variability in the system response and indicate the complexity of the flow, heterogeneous hydrological response, and time-delayed mobilization of previously accumulated sediment within the aquifer. The occurrence of isolated exceeding turbidity episodes during recessionary phases confirms the importance of internal redistribution of suspended material within the aquifer. The results show that an integrated interpretation of these hysteretic relationships can provide a useful approach for interpreting the hydrological processes and functioning of karst aquifers and has potential application to other karst sources.
Groundwater drought is characterized by delayed responses, long duration, and slow recovery, and its evolution is influenced by both local hydroclimatic processes and large-scale ocean–atmosphere variability. To investigate the spatiotemporal evolution and multiscale responses of groundwater drought, a Groundwater Drought Index (GDI) was constructed for the Huaihe River Basin (HRB) using Gravity Recovery and Climate Experiment (GRACE)/GRACE Follow-On (GRACE-FO) and Global Land Data Assimilation System (GLDAS) data from 2003 to 2025. The Bayesian Estimator of Abrupt change, Seasonality, and Trend (BEAST), Modified Mann–Kendall (MMK)–Hurst analysis, wavelet coherence, and Extreme Gradient Boosting–Shapley Additive Explanations (XGBoost-SHAP) were integrated to characterize trend shifts, persistence, multiscale hydroclimatic relationships, and teleconnection-related contributions. The results showed clear temporal variability and spatial heterogeneity in the GDI, with a basin-wide groundwater drought occurring around 2020. The basin-averaged GDI reached a minimum of −2.23 in June 2020, while the lowest sub-basin value was −3.25. The change points in the Shandong Peninsula and Coastal River systems (SPCR) and Yishu-Si River system (YSR) occurred in April 2018 and August 2019, respectively. The MMK–Hurst results indicated predominantly downward GDI trends across the basin, with persistent drying characteristics in parts of the central and northern HRB. Wavelet analysis revealed clear multiscale relationships between GDI and local hydroclimatic variables. Soil moisture (SM) had the largest proportion of significant coherence area (POSP = 19.47%). Shapley Additive Explanations (SHAP) analysis further identified the Interdecadal Pacific Oscillation (IPO) and Tropical South Atlantic Index (TSA) as the two most important teleconnection predictors, with Mean |SHAP| values of 0.1621 and 0.1314. The proposed framework provides a useful basis for groundwater drought monitoring, risk assessment, and basin-scale water-resource management.
This study was conducted in the Upper Olifants River Catchment, located in the Mpumalanga Province of South Africa. The area faces groundwater pollution associated with the current land use and land cover. The primary aim of the present study is to apply multivariate statistical methods to investigate hydrochemical variability and identify the dominant natural and anthropogenic processes associated with water-quality deterioration. In this study, Principal Component Analysis (PCA) and Hierarchical Clustering Analysis (HCA) were applied to hydrochemical data to explore the key factors driving variability in the physical and chemical characteristics of water resources in the Upper Olifants River Catchment. Factor analysis identified three underlying common factors with Eigenvalues > 1, accounting for 80.132% of the total variance in the hydrochemical data. Component 1 explains 52.567% and shows high loadings for TDS, Ca, Mg, Mn, K, and SO4. Component 2 accounts for 15.978% and shows high loadings for Na and Cl. Component 3 accounts for 11.587% of the variance and shows a loading on Fe. Hierarchical clustering yields two clusters: cluster 1 with TDS, Mg, SO4, and Ca; cluster 2 with Cl, Fe, Na, K, and Mn. The presence of TDS, Mg, SO4, and Ca in component 1 and cluster 1 indicates that chemical interactions between water and rocks dominate water chemistry. Higher Na and Cl in component 2 and cluster 2 result from ion exchange and groundwater residence time effects on water quality. Additionally, inverse modelling indicates that sulphate mainly exists as a free ion in water and strongly binds to divalent metals such as Ca, Mg, and Mn. The saturation index indicates that the water is supersaturated with sulphur-rich minerals such as gypsum, while alkaline water is supersaturated with carbonate minerals. Sulphur-rich minerals can cause heavy metals to dissolve into the water, whereas carbonate minerals serve as a buffer, decreasing the water’s acidity. The integrated application of PCA, HCA, and inverse geochemical modelling helps identify spatial patterns in groundwater chemistry and the main processes affecting water quality, thereby offering a solid basis for understanding the causes of groundwater deterioration in the study area.
Machine learning (ML) models utilizing high-frequency Internet of Things (IoT) monitoring data offer significant potential for stream-level nowcasting. The assessment of model performance is often a challenging task because conventional approaches are typically influenced by prevailing low-variability baseline conditions and are unable to assess a model’s ability, or inability, to reproduce rare but critical stream-level surges. To address this limitation, this study proposes a framework combining standard and event-based performance metrics. Using two years of multi-source observations (i.e., stream level, volumetric water content, suction, and rainfall) collected in a small catchment in southern Italy, a set of Random Forest (RF) models was developed to systematically investigate the effect of adopting alternative feature sets, calibration strategies, and target formulations. The assessment of the calibrated models was performed considering 29 stream-level surge events. The results of the analyses revealed that better statistical accuracy over the entire validation period does not necessarily translate into better capability in reproducing critical event dynamics. Three main findings emerged: (i) models predicting short-term stream-level variations consistently outperformed those directly predicting stream level; (ii) calibration using the complete continuous dataset produced more robust models than calibration based exclusively on surge events; and (iii) a reduced set of features provided the most favorable performance balance between predictive performance and model robustness among the tested configurations. Overall, these results suggest that combining conventional and event-based evaluation metrics can provide a complementary assessment of ML forecasting models, highlighting the importance of considering both overall predictive accuracy and event-scale dynamics.
Flood early warning systems (FEWSs) in data-scarce and rapidly responding catchments require monitoring networks that provide sufficient forecasting information with minimal operational cost. This study evaluates the applicability of a data-driven sensor-placement framework to the semi-arid Torghabeh River Basin in northeastern Iran, where flood-producing tributaries affect the city of Torghabeh and downstream developed areas. A 2000-year stochastic rainfall series was generated to represent a wide range of hydrological conditions and was used to drive a basin-scale HEC-HMS model parameterized from catchment physiographic characteristics and available hydrological information. After applying the adopted flood-event extraction and filtering criteria, 466 relevant synthetic events were retained for the sensor-placement analysis. Sub-basin hydrographs were then used as predictors in a Random Forest forecasting model, while permutation feature importance, SHAP, and derivative-based importance were combined to identify the most informative monitoring locations. For a 1 h lead time, the progressive predictor-selection analysis identified five of 24 candidate sub-basins as a parsimonious configuration associated with the performance plateau, corresponding to an approximately 80% reduction in monitoring locations. A different but similarly parsimonious configuration emerged for a 2 h lead time, demonstrating that optimal sensor placement is lead-time-dependent. These results demonstrate the potential of explainable machine learning tools to support parsimonious FEWS monitoring-network design under the evaluated synthetic modeling conditions, while operational application requires further validation using independent observed hydrological data.
