Floods are among the most damaging hazards in Sri Lanka's wet-zone river basins, where short warning lead times constrain preparedness and physically based hydrodynamic models are often impractical under limited data. This study develops a lag-informed deep-learning framework for 10-h-ahead downstream water-level forecasting in the flood-prone Kalu Ganga Basin, Sri Lanka. Flood-wave travel times between Rathnapura, Ellagawa, Millakanda, and the downstream target station Putupaula were estimated from hourly records covering 2021–2024 using event-based peak matching, rising-limb timing, and cross-correlation. Selected travel times were 18 h for Rathnapura–Ellagawa, 23 h for Ellagawa–Putupaula, and 24 h for Millakanda–Putupaula. These lags were used to align upstream water-level inputs, while rainfall remained in its original chronological order. No-lag and lag-informed Long Short-Term Memory (LSTM), Gated Recurrent Unit (GRU), and Bi-LSTM models were evaluated using chronological training, validation, and independent testing. Lag-informed inputs reduced Root Mean Square Error (RMSE) from 0.317 to 0.288 m for LSTM and from 0.339 to 0.286 m for GRU, while Bi-LSTM performance decreased. The lag-informed GRU also outperformed persistence and ARIMA baselines. Stage-aware training further improved high-flow prediction, with the GRU achieving an RMSE of 0.262 m, Mean Absolute Error (MAE) of 0.188 m, and Nash–Sutcliffe Efficiency (NSE) of 0.934, while the LSTM predicted the 5.57 m June 2024 flood peak at 5.44 m. The framework was deployed as a real-time decision-support dashboard named KaluHydroNet, demonstrating that lag-informed and stage-aware deep learning coupled with live data integration offers a practical pathway for short-term flood early warning in the Kalu Ganga Basin.
IntroductionUnderground metro structures can obstruct groundwater flow, causing upstream water-level rise and pore-pressure redistribution. This study investigates the groundwater-blocking effect associated with Jinan Metro and evaluates measures for reducing the resulting engineering risk.MethodsA physical model system was developed to reproduce groundwater seepage during metro station construction and operation under different hydraulic gradients and diversion conditions. Pore-water pressure responses were monitored, and repeated tests were conducted to assess measurement repeatability. Six indicators were selected for risk assessment, with subjective and objective weights determined using the analytic hierarchy process and principal component analysis, respectively. TOPSIS and a normal cloud model were applied to 12 validation cases.ResultsWithout diversion, the maximum upstream water-level rise reached 2.10 m at a hydraulic gradient of 0.07. Diversion reduced water-level backwater by 74.07%–98.09% across the tested conditions. The 3,000 mm2 gravel channel provided the greatest reduction, while the 700 mm2 gravel channel and the seven connecting pipes of 5 mm diameter also performed effectively. The normal cloud model agreed with the reference results in 75.00% of cases, compared with 58.33% for TOPSIS.DiscussionThe experiments clarify the mechanisms of local groundwater blocking and support the selection of diversion measures and risk-based engineering controls. The normal cloud model showed better agreement within the tested cases; however, the limited validation sample and simplified aquifer representation require further verification using site-specific monitoring and hydrogeological data.
IntroductionRiverbank filtration (RBF) is a widely used natural water treatment process; however, the fate and attenuation of microplastics (MPs), within river–aquifer systems is not well understood. This study investigated the occurrence and characteristics of MPs within an RBF system to comprehend the MPs transport behaviour.MethodsFive sampling campaigns were conducted over the course of a year. These covered rivers, groundwater wells and an abstraction well across six locations within the RBF site, which extracts water from the Leine River in Germany. MPs were collected and characterized through stereomicroscopy and chemically identified using ATR-FTIR. In this study, MPs were detected throughout the RBF system but also in groundwater without river influence during all sampling campaigns, demonstrating their continuous presence and persistence within the hydrological continuum.Results and discussionRiver water consistently exhibited the highest MP concentration (74.4 ± 23.5 MPs/100 L), whereas the abstraction well showed low concentration (9.4 ± 7.8 MPs/100 L), significant differences in MP concentrations were observed (p = 0.002). MP transport and retention were strongly influenced by MP characteristics such as shape and type. Overall, fiber shapes (88.5% of all particles) and polyethylene terephthalate (PET) (73.8% of total identified polymers) dominated. The prevalence of fibers suggests that their elongated morphology may enhance mobility through interconnected pore spaces. Within the collected MP size range of 20 to 5,000 μm, the most abundant class was 100–200 μm, indicating that particle size may regulate pore-scale straining. These results indicate that, despite their high density, PET fibers can pass through pore spaces and potentially bypass natural filtration barriers due to their small size. The study emphasizes that RBF is an effective spatial concentration attenuation system for reducing MP contamination in drinking water sources and highlight the importance of natural attenuation process.
IntroductionMonitoring oil spills in the port area is very difficult and can spread rapidly, resulting in considerable damage to the environment. This study proposes HyEff-UNet, a deep learning-based semantic segmentation model for the detection of oil spills using drone-captured RGB images of port environments.MethodsHyEff-UNet employs a modified version of the U-Net model with an EfficientNet-B4 model as an encoder is used as the model architecture to classify the pixels of the input image as oil, water, or other surface regions. A combined categorical cross-entropy and Dice loss function is used to jointly optimize pixel-wise classification and region overlap.ResultsThe results show that the complete architecture achieves the highest performance among the internally evaluated models, while the sequential ablation indicates that the largest performance improvement is observed after the hybrid categorical cross-entropy and Dice loss is introduced following the EfficientNet-B4 and attention-based architecture.ConclusionThese findings demonstrate the effectiveness of the complete architecture within the evaluated configurations. However, the sequential ablation study does not isolate the independent contribution of the hybrid loss. Furthermore, the measured inference latency indicates that further optimization and unmanned aerial vehicle (UAV)-specific benchmarking are required before real-time onboard deployment can be considered.
IntroductionAntibiotic resistance is spreading throughout aquatic ecosystems, especially wetlands that receive urban and industrial effluents. These environments are large reservoirs under anthropogenic selection pressure, favouring the persistence and transmission of multidrug-resistant (MDR) diseases. This study investigated the biofilm-forming capacity and antimicrobial resistance profiles of 32 environmental bacterial isolates recovered in 2024, assessing the microbiological water quality of a sewage-fed aquaculture unit within the East Kolkata Wetlands (EKW), an Indian Ramsar site.MethodsTo characterise persistence-associated biofilm formation, phenotypic antimicrobial resistance, and hemolytic activity, isolates were evaluated using quantitative biofilm assays, antimicrobial susceptibility testing, and blood-agar hemolysis profiling, respectively. We constructed a composite phenotypic risk score using a risk-quadrant approach to classify isolates based on the convergence of multidrug resistance, strong biofilm formation, and hemolytic activity.ResultsThe findings showed a positive correlation between biofilm development and the multiple-antibiotic-resistance (MAR) index, indicating a close relationship between sessile microbial lifestyles and enhanced survival under antibiotic stress. Pseudomonas and Citrobacter, two genera known to flourish in moist environments, were well represented in the taxonomic profile, which was dominated by the genus Enterobacter. Phylogenetic characterisation using 16S rDNA sequencing revealed a diverse community, with distinct groups of Enterobacter, Pseudomonas, and Citrobacter. BLASTn identity scores ranged from 99.43 to 100%, supporting robust taxonomic classification. Alpha hemolysis predominated, suggesting a significant presence of potential opportunistic pathogens, as supported by the identification of high-risk isolates.DiscussionThese evolutionary linkages underscore the wetland’s significance in conserving therapeutically relevant bacterial lineages amid sustained anthropogenic assault. The clustering of these variables suggests the potential for environmental transmission of high-MAR strains via aquatic routes. These findings emphasise the need to integrate ecological monitoring into broader biosafety frameworks to mitigate the risks posed by persistent antimicrobial resistance in wetland reservoirs.
