Human threats to freshwater ecosystems overall compromise water security, and identifying these threats is essential to set up the required strategies in managing water resources. The article focuses on Spain as the region of study since this country maintains a long-lasting conflict between the use of water resources and the impacts on freshwater ecosystems. Spain has a double-sided water reality, both in spatial (dry Mediterranean versus humid Atlantic ecoregions) and in temporal terms (interannual climate variability). Up to 30 threats for freshwater ecosystems have been identified and prioritized after a deliberative process composed by three main methodological stages aimed at (I) documenting, (II) screening, and (III) assessing based on academics and water managers perspectives. Climate and land use change are the highest-ranked threats, while pollution is ranked lower. Overall, quantity issues lead to concern among both the experts and Júcar River Basin managers, whereas quality-centred threats displayed discrepancies between the two perspectives. These observed discrepancies between academic and managerial priorities underscore the challenges of operationalizing Integrated Water Resources Management (IWRM), highlighting the need for enhanced knowledge co-creation in water governance. Thus, this twofold assessment, as a co-creation and prioritization exercise, establishes a new knowledge co-creation approach and represents a potential framework in designing strategies for assuring freshwater ecosystem preservation as a foundation for water security.
IntroductionMountain springs are the primary source of drinking water and support the livelihoods of many people in the Indian Himalayan Region. These springs face pressure from human activities, land-use changes, and climate change. Detailed records of springs in the Kashmir Himalaya are scarce, making planning and conservation difficult.MethodsThis study develops a baseline geospatial inventory and assessment of spring resources in South Kashmir, focusing on the Anantnag and Kulgam districts. An extensive field survey using GPS identified 649 springs in 2021 and 2022. For each spring, the study recorded its location, elevation, water flow, seasonality, ownership, use, dependency, and conservation status. The study used a GIS environment to assess spring distribution across physiographic, land-use, and elevation units.ResultsAmong the springs surveyed, 79.8% were perennial, 8.5% were seasonal, and 11.7% were dry at the time of the study. Most springs were found in low-elevation areas below 1700 meters, on flat landscapes, and in horticultural regions. The majority of the springs had discharges corresponding to the fifth and sixth magnitudes in the Meinzer Classification. The local community reported that spring discharge and reliability have decreased over the past 10 years, but these reports could not be confirmed due to the lack of long-term hydrological records.Discussion and conclusionThe spatial inventory highlights areas where spring systems may be at greater risk and provides evidence to help prioritize monitoring, conservation, and restoration efforts. While this study does not show direct links between spring conditions and potential causes such as climate change, land use, or groundwater use, it does provide a solid starting point for future research, long-term monitoring, and climate adaptation planning in the Himalayan region, where data are limited.
Accurate flood forecasting in arid inland basins remains difficult because observed events are few, runoff generation is strongly nonlinear, and snowmelt and rainfall may interact within short forecasting windows. This study evaluates an improved Long Short-Term Memory (LSTM) framework for the Chinese Ili River Basin in northwestern China, where flood processes are dominated by snowmelt and can be modified by rainfall recharge. Hourly temperature, precipitation, and runoff observations from nine 24-h flood events were used for model development and testing. The framework combines Pearson-based feature screening, sliding-window sample reconstruction with training-only low-level noise injection, Dropout–L2 regularization, and grid-search tuning within nested leave-one-event-out validation; temporal interpolation is not used. Deterministic hyperparameter sensitivity was further examined using Latin hypercube sampling and Monte Carlo analysis, separately from predictive uncertainty quantification. Across the nine events, the improved LSTM yielded mean NSE and R2 values of 0.9633 and 0.9015, respectively, with a mean peak-discharge error of 3.90%. The temperature-runoff lag averaged 11.67 h, supporting the dominant role of cumulative thermal forcing in the snowmelt-driven events, whereas rainfall mainly modified peak behavior in the mixed events. The hyperparameter-sensitivity analysis indicates that hidden neurons, learning rate, and time step are the most influential parameters. These results suggest that the pro-posed framework can provide a useful event-scale forecasting basis for snow-melt-dominated floods in data-scarce arid basins.
IntroductionPrediction of an oil spill event is fundamentally a spatiotemporal problem, impacted by several environmental variables such as wind speed, ocean currents, and volume of the spill. Current solutions largely depend on either physics-based models or data-driven models, which makes it difficult for them to simultaneously account for temporal and spatial aspects. The current research work suggests a novel approach combining Gated Recurrent Unit (GRU) and Graph Neural Netowrks (GNN) to predict multi-horizon marine oil spills.MethodologyGRU model relies on a log-space formulation in order to achieve stable and monotonic predictions regarding temporal spill growth. The GNN takes into account spatial spill propagation via the graph-based representation of the marine environment and is fed with wind and ocean current data. Physics-informed scaling, normalization, and monotonicity constraints are applied to enhance stability and keep consistency with physics.ResultsExperimental results confirm stable training and good-quality multi-step predictions yielding log-space RMSEs equal to 0.1425, 0.2537, 0.3806, and 0.5934 for 1 h, 3 h, 6 h, and 12 h of prediction time horizon, correspondingly. All cases tested have 100% of monotonicity with no unphysical decrease of forecasted values. High correlation between physical processes and model predictions is achieved, yielding correlation coefficients of 0.730 for spill volume, 0.405 for wind speed, and 0.413 for ocean current speed.ConclusionThe suggested GRU-GNN model can offer a physically sound method for forecasting the oil spill in space–time and thereby provide useful results for marine environmental monitoring in the short-to-medium term.
Wastewater treatment plants (WWTPs) are potential reservoirs and dissemination pathways for antibiotic-resistant bacteria (ARB) and antibiotic resistance genes (ARGs) into aquatic environments. Most South African WWTPs use activated sludge processes that may incompletely remove ARB and ARGs, potentially contributing to environmental antimicrobial resistance (AMR). This study investigated physicochemical parameters, bacterial occurrence, antimicrobial susceptibility, and genomic characteristics of presumptive Enterococcus and Staphylococcus spp. from two WWTPs and receiving rivers in Gauteng, South Africa. Effluent physicochemical parameters remained within regulatory pH and temperature ranges (pH 7.25–7.51; temperature 16.73–22.33 °C), with DO increasing (26.77–125%) and TDS ranging from 170–326 ppm. Seventy-two water samples were collected from influent, effluent, upstream, and downstream sites during three sampling periods in June, August, and late October 2023. Target bacteria were enumerated by membrane filtration and selective culture, followed by biochemical identification and antimicrobial susceptibility testing (AST). Presumptive Enterococcus and Staphylococcus concentrations ranged from 6.074–7.423 and 6.576–7.255 log CFU/mL, respectively. WWTP1 showed limited bacterial removal (0.68–10.76%), whereas WWTP2 achieved 89.5–90.5% reduction; downstream concentrations generally exceeded upstream levels. From 54 isolates subjected to AST, six met the predefined MDR criterion. Enterococcus isolates exhibited resistance to teicoplanin, azithromycin, and tetracycline. Whole-genome sequencing identified four E. faecium sequence types (ST1052, ST1058, ST94, and ST32) and one MDR S. aureus sequence type (ST243). E. faecium harboured msr(C), erm(B), and lsa(A), while S. aureus carried norA and norR. Overall, detection of MDR bacteria and clinically relevant resistance determinants in treated wastewater and receiving rivers indicates potential WWTP-associated environmental dissemination of resistant bacteria in aquatic environments.
