These aspects draw paramount significance in catchments with large heterogeneity and spatial complexities such as mountainous and urban catchments, data scare regions, and low-income countries where investment in hydrological and water quality monitoring networks and installation of IoT sensors is very limited. It is therefore warranted to leverage geospatial, machine learning, decision science, statistical and modelling techniques to improve the understanding of catchment hydrology and consequences of climate change and anthropogenic drivers on surface and groundwater resources at various scales. The worldwide readily available satellite remote sensing data and global data products enable us to leverage these techniques in addressing water and environmental challenges.
We solicit novel contributions from researchers in catchment hydrology by utilizing Remote Sensing, GIS, Artificial Intelligence (AI), Machine Learning (ML), Decision Science, and advanced statistical techniques for addressing pressing challenges of water sustainability in mountainous and urban catchments and data scarce regions. The combined use of these technologies will revolutionize understanding of complicated hydrological, hydrodynamic and biogeochemical processes, and will be useful in evolving effective water resource management and ecosystem-based adaptation strategies to foster sustainable development.
Orals: Mon, 4 May, 08:30–12:30 | Room 2.15
Accurate representation of transport processes is essential for understanding water quality dynamics in surface flow systems, particularly under turbulent conditions where observations are limited in space and time. In such environments, sediment and sediment-associated constituent transport is strongly influenced by multiscale turbulence, intermittency, and correlated particle dynamics, processes that are not adequately captured by conventional deterministic modeling approaches.
This study presents a Lagrangian stochastic framework for modeling particle transport in turbulent flows, with particular emphasis on addressing unresolved variability and the limited availability of Eulerian observations. Particle motion, entrainment, and dispersion are formulated using multivariate and multi-layer stochastic differential equations that explicitly incorporate turbulence-induced intermittency, particle memory, and scale-dependent correlations. Near-threshold sediment entrainment is represented through physically based probabilistic criteria, enabling the modeling of intermittent transport events that dominate sediment flux in regimes close to the threshold of sediment motion.
To capture relative dispersion and correlated motion driven by multiscale turbulent structures, the framework extends beyond single-particle formulations to include two-particle stochastic dynamics. Model development and validation are informed by Direct Numerical Simulation (DNS) data, which provide flow statistics for quantifying particle position, velocity, and correlation structures. This integration allows critical transport characteristics to be inferred even when field-scale monitoring data are limited in space or time.
The proposed stochastic framework provides a physical framework for modeling the transport of particle-associated constituents in surface flows. By emphasizing process-based stochastic representations rather than data-intensive deterministic closures, the approach offers a robust pathway for advancing transport modeling in turbulent flows under data-limited conditions.
How to cite: Tsai, C.: Physically Based Lagrangian Stochastic Modeling of Particle Transport in Data-Limited Turbulent Flows , EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-17157, https://doi.org/10.5194/egusphere-egu26-17157, 2026.
Anomalous sediment transport is often observed in turbulent flows. Under these conditions, particle motion frequently deviates from the classical Fickian diffusion assumption due to long-term correlations and complex interactions between flow and sediment. Although many models have been developed to describe this behavior, it remains challenging to link particle-scale dynamics, field-scale transport processes, and statistical descriptions of concentration distributions within a single physical framework. As a result, parameters used in statistical or fractional-order models are often obtained through empirical fitting, and their physical interpretations remain unclear.
This study presents a multiscale framework for interpreting memory-driven anomalous sediment transport by linking particle dynamics, continuum transport behavior, and statistical descriptions. At the particle scale, a Fractional Sediment Diffusion Particle Tracking Model (FSDPTM) is employed to simulate sediment motion with temporal memory. Under this setting, anomalous diffusion emerges from non-Markovian particle dynamics. The mean-square displacement (MSD) is then analyzed to quantify anomalous transport behavior at the particle scale and to describe the strength of temporal correlations.
At the macroscopic scale, transient concentration fields obtained from particle trajectories are used to guide the fractional advection–diffusion equation (FADE). This step connects the particle-scale memory effect with the field-scale Eulerian description. Since experimental observations of transient concentration evolution are often difficult to obtain, the proposed method focuses on cross-scale internal consistency rather than direct data fitting. The steady-state concentration profiles produced by the particle model are then compared with laboratory measurements to assess whether the long-term transport behavior is physically reasonable.
Building on the validated steady-state profiles, a fractional entropy formulation is used to describe the statistical structure of sediment concentration distributions. The entropy parameter is not an empirical fitting coefficient, rather, it is interpreted as a potential indicator reflecting the cumulative effects of memory-driven transport processes. By comparing the mean-square displacement (MSD) at the particle scale, the FADE parameters at the field scale, and the entropy-based description, this study demonstrates that entropy parameter may be related to anomalous transport characteristics associated with long-term particle memory.
Overall, this study presents a multiscale interpretation of anomalous sediment transport in which particle dynamics, continuum transport equations, and statistical descriptions are treated in a mutually consistent manner. The results suggest that entropy-based parameters may have the potential to serve as compact and physically interpretable indicators of anomalous transport intensity. This framework provides a structured approach for connecting transport dynamics across scales and for extracting physical insights from limited observable information.
Keywords:Anomalous diffusion;Memory-driven transport; Multiscale processes; Fractional dynamics; Particle-based modeling; Statistical characterization
How to cite: Wu, H. H. and Tsai, C. W.: A Multiscale Interpretation of Memory-Driven Anomalous Sediment Transport, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-10106, https://doi.org/10.5194/egusphere-egu26-10106, 2026.
Over the past two decades, microplastics (MPs) pollution has been recognized as a significant risk to public health and to a wide range of environments, particularly riverine, estuarine, and oceanic systems. However, much of the existing research on MPs has focused primarily on large-scale transport behavior in ocean zones using deterministic approaches. Consequently, many of the underlying fundamental principles governing the transport mechanisms of MPs and their fate in open channel flows remain poorly understood. Unlike sediments, which generally settle downward, MPs exhibit far greater variability in physical properties, including material composition, shape, size, drag, and density. Some MPs are even lighter than water, leading to upward or buoyant motion during transport and introducing additional complexity to the governing hydrodynamics.
To account for the geometric irregularity of particles, this study employs a stochastic diffusion particle tracking model (SD-PTM) that incorporates a modified vertical velocity formula to better represent the effects of inertial and viscous drag forces on MPs. In this model, the movement of suspended MPs is modeled as a stochastic process composed of a drift term and a random term, to represent particle transport in open channel flow. In addition, the genetic algorithm (GA) is applied to optimize the drag coefficients, thereby enhancing model robustness under data-limited conditions.
Compared with traditional models without consideration of MPs’ physical properties, the proposed modified stochastic model investigates not only the settling motion of MPs, but also extends, for the first time, stochastic modeling approaches to buoyant particles. The model results are compared with the experimental data provided by Born et al. (2023) across a range of flow conditions to calibrate the model coefficients. This study offers a new perspective on both rising and settling MP motion, thereby advancing the understanding of microplastic fate and transport in open channel flows.
How to cite: Chen, M. T. and Tsai, C. W.: Modified Stochastic Model for Settling and Rising Microplastic Transport in Open Channel Flows, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-10114, https://doi.org/10.5194/egusphere-egu26-10114, 2026.
Glacial Lake Outburst Floods (GLOFs) are emerging as a significant hazard in high-mountain regions due to accelerated glacier retreat and lake expansion resulting from climate warming. The present study employs a one-dimensional hydrodynamic modelling framework to simulate the propagation of GLOF and downstream flood characteristics for the Gepan Gath Lake–Chandra Basin in the Western Indian Himalayas. The selected study area represents one of the most rapidly evolving and hazard-prone glacial lake settings in the region. Unsteady flow simulations are performed using the HEC-RAS hydraulic model to route scenario-based GLOF hydrographs along the downstream river corridor. Breach outflow hydrographs have been generated using plausible combinations of lake volume and dam failure mechanisms, and are applied as upstream boundary conditions. River geometry is represented through cross-sections extracted from the ALOS PALSAR digital elevation model and supporting geospatial datasets. The simulations capture the temporal and spatial evolution of discharge and water surface elevation along the river network under multiple GLOF scenarios. Results indicate rapid flood wave propagation in steep upstream reaches, followed by attenuation and lateral spreading in wider downstream valleys. Peak discharge, inundation depth, and flood arrival time exhibit strong spatial variability, primarily governed by valley morphology and hydraulic connectivity. The modelling outputs enable identification of critical downstream impact zones and provide first-order estimates of exposure to GLOF hazards. This study demonstrates that 1D hydrodynamic modeling using HEC-RAS, combined with remotely sensed terrain data, provides an efficient and robust approach for regional-scale GLOF hazard assessment, supporting the design of early warning systems and disaster risk reduction planning in data-scarce Himalayan environments.
How to cite: Mishra, N., Keshari, A. K., and Chahar, B. R.: Scenario-based 1D hydrodynamic modelling of glacial lake outburst floods in the Western Indian Himalaya, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-21229, https://doi.org/10.5194/egusphere-egu26-21229, 2026.
This study develops a three-layer embedded Lagrangian stochastic (LS) model for simulating suspended sediment transport in open-channel flows. The model describes particle motion at three levels: position, velocity, and acceleration, using multiple Ornstein–Uhlenbeck (OU) processes within a coupled stochastic system. This construction preserves intrinsic stochasticity while allowing the velocity process to be differentiated in time to obtain particle acceleration, enabling a consistent description of particle motion at small time scales.
In conventional LS models, random forcing is typically represented by a Wiener process. Since this process is nowhere differentiable, it limits the interpretation of higher-order kinematic quantities. In this study, an embedded Ornstein–Uhlenbeck formulation is employed, where the random forcing is described by a finite-order system of coupled stochastic ordinary differential equations. Compared with conventional two-layer LS models, the three-layer formulation produces smoother Lagrangian velocity trajectories by improving the differentiability of the velocity process. This formulation reduces abrupt fluctuations in the simulated velocity signal and allows acceleration to remain finite and well-behaved.
As a result, the model provides a clearer basis for describing short-time-scale particle motion and for exploring rapid turbulent effects near the bed. Model parameters are determined based on laboratory experimental data and commonly used turbulence scaling relations reported in the literature.
Overall, the proposed framework provides a stochastic description of particle motion that allows velocity and acceleration to be consistently represented at small time scales and offers a basis for further investigation of near-bed particle behavior and suspended sediment transport processes.
How to cite: Chen, C. Y. and Tsai, C. W.: Three-Layer Ornstein–Uhlenbeck Model for Turbulent Flow Simulation, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-10246, https://doi.org/10.5194/egusphere-egu26-10246, 2026.
Water quality monitoring in subsurface environments is often limited by sparse, irregular, and uncertain measurements, complicating the calibration process and reliability of transport models. In this study, we propose a Finite Volume (FV) residual Physics Informed Neural Network (PINN) framework for contaminant transport through subsurface media governed by the advection-dispersion equation (ADE), with a focus on generating predictions considering parameteric uncertainty for data-scarce environments. The core idea is to replace the strong-form PDE residual typically used in PINNs with a control-volume conservation imbalance derived from a discrete FV balance. Neural network predictions are used to evaluate advective and dispersive numerical fluxes at cell faces, and training minimizes the resulting cell-wise flux imbalance while enforcing initial and boundary conditions. This conservative formulation enables transport-specific numerical flux treatments (e.g., upwind/TVD advection and consistent boundary fluxes), and we assess performance for advection-dominated systems with sharp concentration fronts.
To represent heterogeneity and uncertainty in dispersion, we parameterize the dispersion coefficient as a strictly positive random field using a low-dimensional basis. Uncertainty is propagated through the learned surrogate using Monte Carlo sampling to obtain prediction intervals and monitoring-relevant risk metrics such as threshold exceedance probabilities at selected locations. We outline two uncertainty workflows: (i) an ensemble strategy that trains FV-PINN models across sampled dispersion realizations, and (ii) a prospective conditional FV-PINN that takes random-field coefficients as additional inputs, enabling efficient Monte Carlo evaluation after a single training stage. The application of the methodology is demonstrated on simple benchmark examples designed to represent sparse monitoring data, showing how conservative learning and random-field uncertainty propagation can support reliable transport predictions when observations are limited.
How to cite: Jain, S., Dey, S., and Chahar, B. R.: A Conservative FV-Residual PINN Framework for Solute Transport through Subsurface Media with Dispersion Uncertainty for Data-Scarce Environments, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-6256, https://doi.org/10.5194/egusphere-egu26-6256, 2026.
Water quality monitoring is essential for addressing water contamination and ensuring public safety. Particularly, managing nitrate levels has become a major concern due to their direct impact on eutrophication. Despite the high accuracy of conventional analysis methods, their practical application is often limited by high costs, labor-intensive processes, and a lack of real-time monitoring capabilities. This study presents a novel framework for real-time water quality monitoring by integrating hyperspectral and multi-sensor data through deep learning-based data fusion. The multi-sensor data includes pH, electrical conductivity (EC), dissolved oxygen (DO), and oxidation-reduction potential (ORP). A transformer-based deep learning model was applied to predict water quality concentrations by capturing correlations within time-series hyperspectral absorbance and multi-sensor data. Furthermore, transfer learning was employed to improve the performance in target domains by transferring the information contained in a pre-trained model. The data-fusion transformer model predicted water quality concentrations with high accuracy, achieving a coefficient of determination (R2) exceeding 0.99 in both deionized and tap water conditions. Specifically, the integration of multi-sensor data improved model robustness and performance compared to applying spectral data alone. This research also demonstrated that transfer learning effectively supported the model in adapting to varying flow conditions. The proposed deep learning-based data-fusion framework provides a reliable solution for real-time water quality monitoring, with aims to extend the model application to predict multiple water parameters simultaneously.
How to cite: Yun, D., Gwon, N.-H., Jung, J., and Baek, S.-S.: Deep Learning-Driven Hyperspectral Data Fusion for Real-Time Water Quality Monitoring, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-16044, https://doi.org/10.5194/egusphere-egu26-16044, 2026.
Karstic terrains are formed by the dissolution of carbonate rocks and are essential zones for groundwater reservoirs but are susceptible to geological and climatic conditions. Thus, delineation and characterization of potential karst development sites are necessary, especially in areas with limited data on karst development, which hinders accurate groundwater assessment, hazard mitigation, and sustainable land-use planning, especially in remote areas such as Siargao Island, Philippines. By applying a data-driven geospatial framework that combines statistical analysis with Geographic Information System (GIS) techniques, it is possible to evaluate the island’s karstification potential as support for future water resource management strategies.
Principal Component Analysis (PCA) was applied to eight initially selected variables, which were then reduced to four key components: geology, slope, precipitation, and vegetation. These components were used for GIS-based multi-criteria evaluation to generate a karst potential map of Siargao Island. Results show strong spatial variability in karst development wherein high to very high potential zones are in the southern and southeastern regions, characterized by mature cockpit karsts, caves, and sinkholes. The eastern and western parts of the island, where transitional stages of karst development are present, exhibit moderate karstification potential. Non-carbonate areas with minimal karst expression in the central and northern regions showed low to very low potential zones. Field observations, existing geomorphological maps, and sinkhole inventory data were utilized for model validation, resulting in an overall accuracy of 80.6% and a Kappa coefficient of 0.44, indicating moderate agreement between the predicted and observed karst features.
Through this approach, a cost-effective monitoring strategy for assessing groundwater resources and geohazards in data-scarce, remote areas with karstic terrains, such as Siargao Island, can be developed. The generated karst potential map provides a baseline for sustainable water resource management, groundwater protection, and land-use planning. Furthermore, this study demonstrates the use of geospatial and decision-support methods to strengthen hydrological management in remote environments.
How to cite: Varela, R. K., Barrios, E. D., Bondad, F., and Cortejos, Y. A.: GIS-Based Assessment of Karstification Potential in Siargao Island, Philippines, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-19227, https://doi.org/10.5194/egusphere-egu26-19227, 2026.
Ecosystem services (ESs) represent the essential ecological contributions that support human well-being and socioeconomic subsistence. This study employs multi-temporal remote sensing (RS) datasets from 1995 - 2022 to quantify the Ecosystem Service Value (ESV) of key ecosystem functions within a representative Tier-2 Indian city. Land Use/Land Cover (LULC) classification is performed using a Random Forest (RF) supervised machine learning algorithm to delineate ecosystem units, producing high-precision classification results with strong overall accuracy and optimized Kappa coefficients. Valuation is conducted using benefit transfer methods, with values expressed in million US dollars per year. The results indicate that, after vegetative cover, built-up areas, croplands, waterbodies, and barren land are the next major contributors to the total ESV. The key findings of the study are that Vishakapatnam, Tier-2 city in India is highly sensitive to LULC transitions, where rapid urbanization significantly alters the trajectory of provisioning, supporting, regulatory, and cultural ecosystem services. In addition, the study examines spatio-temporal relationships between ecosystem service trade-offs and synergies, demonstrating that high-resolution ESV mapping serves as a reliable diagnostic tool for assessing the impacts of human overexploitation and poor resource management. Overall, the study provides a robust quantitative framework for ecological valuation, offering a critical foundation for evidence-based policy interventions and sustainable urban planning in rapidly transforming urban environments.
How to cite: Agrahari, S., Swetha , D., and Pal, M.: Spatiotemporal Assessment of Ecosystem Services in a Tier-II Indian City: A Case Study of Visakhapatnam (1995–2022), EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-21599, https://doi.org/10.5194/egusphere-egu26-21599, 2026.
Intensive groundwater abstraction, land-use changes, and climate variability have significantly altered natural discharge and flow patterns within groundwater systems, threatening long-term groundwater sustainability. These disruptions increase the risk of degradation in ecosystems that rely directly or indirectly on groundwater discharge, i. e. groundwater-dependent ecosystems (GDEs).
Mexico is particularly vulnerable to declining water table levels, a situation accelerated by gaps in groundwater management that fail to incorporate GDEs into decision-making processes. This issue is especially critical in northeastern Michoacán, home to two of the country’s largest lakes: Pátzcuaro and Cuitzeo lakes, which represent a key study area for studying growing threats to GDEs caused by pollution, climate change, and intensive groundwater abstraction. In order to preserve GDEs, along with their associated biodiversity and ecosystem services, accurate mapping is essential to secure their future integration into groundwater sustainability policies and conservation initiatives.
To address this issue, we compared four methods usually used in geospatial mapping: the Analytical Hierarchy Process (AHP), Weights of Evidence (WoE), and two machine learning models: Logistic Regression (LR) and Random Forest (RF), using environmental variables associated with GDE presence obtained from geospatial data and remote sensing products.
Model performance was evaluated using a validation dataset derived from local inventories and fieldwork conducted in 2024, applying Receiver Operating Characteristic (ROC) curves and the Area Under the Curve (AUC) metric. Results showed that RF (AUC = 0.82) and LR (AUC = 0.70) outperformed WoE (AUC = 0.61) and AHP (AUC = 0.59), with RF demonstrating the highest predictive accuracy and best performance in cross-validation folds.
The GDEs prediction map derived from RF highlights areas primarily along the shores of both lakes, where volcanic lithology contacts with lacustrine deposits, inducing groundwater discharge through springs that sustain wetlands. Additional GDEs areas occur along fault zones that enhance discharge within volcanic lithology near Morelia City and in perennial streams located at intermediate elevations.
The study faces limitations related to varying spatial resolutions, independent errors in geospatial datasets, and uneven data quality across local zones within the study area. Furthermore, the absence of direct field verification for areas with the highest predicted GDE potential constrains the overall impact of the study. Nevertheless, this research provides significant evidence of the advantages of using machine learning approaches in regions lacking detailed hydrogeological information, supporting the integration of GDEs into groundwater sustainability management.
How to cite: Salgado-Albiter, A. C., Olea-Olea, S., Ramírez-Serrato, N. L., Morales-Casique, E., Ramírez-González, L., and Llanos-Solis, A. G.: Comparison of expert-knowledge and machine learning approaches for mapping groundwater-dependent ecosystems in a regional setting in Central Mexico, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-4669, https://doi.org/10.5194/egusphere-egu26-4669, 2026.
Mountain aquifers are receiving increasing attention as a key component of the so-called water towers. They sustain important freshwater ecosystems, river flow during droughts, and are a key water resource for populations living in mountain valleys and the nearby floodplains. These aquifers are exposed to emerging pollutants, such as pharmaceuticals, PFAS, and microplastics, whose adverse effects on ecosystems and human health are exacerbated by overexploitation. The interaction between surface and subsurface waters increases the risk of groundwater contamination by untreated sewage waters, and in several cases also by treated waters, because in most countries sewage treatment systems are not yet designed to remove pharmaceutical and emerging contaminants. A significant challenge that modelers face when dealing with these systems is the endemic lack of data to constrain the models, which limits their reliability in risk analysis and in the comparison of the effectiveness of alternative remediation actions. An example of application in a mountain valley aquifer of northeastern Italy is used to discuss how to make a convenient use of available data to reduce the uncertainty affecting groundwater modeling in such environments, where lateral fluxes stemming from hillslopes and the surface/subsurface water exchange fluxes are difficult to constraint and a source of large uncertainties in modeling both groundwater availability and groundwater contaminant transport. In particular, we explored the gain in model consistency that can be obtained by supplementing groundwater head data with geochemical and groundwater concentration data of a target contaminant at a few controlling groundwater wells. The geochemical data refer to river water and to springs emerging from the lateral hillslopes. Electrical conductivity and other geochemical data typically collected as part of the standard water quality monitoring performed by Environmental Protection Agencies may help in reducing the uncertainty in the lateral and surface/subsurface exchange fluxes and in improving the reliability of the transport model, when used in combination with contaminant concentration data at the available groundwater monitoring wells. The analysis suggests that considering the valley aquifer as part of a more complex system, including the contribution of the lateral mountain aquifers, and the exchange with surface water, is an opportunity for producing realistic models rather than an unnecessary complication.
How to cite: Bellin, A., Betterle, A., and Di Dato, M.: The value of data in reducing uncertainty in mountain groundwater modeling, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-11919, https://doi.org/10.5194/egusphere-egu26-11919, 2026.
Accurate streamflow monitoring is essential for water resources management, yet many Brazilian watersheds lack sufficiently long historical records to support effective decision-making. This challenge is particularly critical in the Metropolitan Region of Belo Horizonte (RMBH), which depends on major reservoirs located within its territory – such as Rio Manso, Serra Azul, Vargem das Flores, and the Ibirité (REGAP) reservoir – for industrial and domestic water supply. Several of these strategic systems suffer from limited or inconsistent hydrological monitoring, complicating operational planning, increasing the risk of water shortages and of compromising reservoirs flow outcome capacity. Transfer Learning (TL) with Long Short-Term Memory (LSTM) networks emerges as a promising strategy to overcome this limitation, enabling the development of hydrological models in watersheds with little or no historical data. This study investigates the application of TL to enhance daily streamflow prediction in data-scarce basins of the Metropolitan Region of Belo Horizonte (RMBH), while assessing the optimal length of local streamflow records required to improve hydrological modelling through fine-tuning of a regional TL model. For this, 23 watersheds with similar hydrological behaviour and geomorphological characteristics were previously selected in the RMBH to evaluate the feasibility of reconstructing streamflow time series in data-scarce regions. Satellite-derived products and reanalysis datasets were employed as inputs to overcome limitations in hydrometeorological data availability. Furthermore, eXplainable Artificial Intelligence (XAI) methods are employed to explore the physical feasibility of knowledge transfer, with the potential to identify which watershed attributes – such as drainage area, elevation, soil-moisture dynamics, land-use composition, and climatic seasonality – most strongly influence whether hydrological behaviour learned in source basins can be meaningfully transferred to target basins. Significant performance gains can be achieved with only one to two years of local data, allowing accurate models to be developed rapidly even in newly monitored watersheds. This improves considerably the decision-making in data scarce regions, primarily those ones with some water conflicts. XAI analyses confirmed the physical soundness of the predictions, supporting more reliable streamflow reconstruction. However, further methodological improvements are required, as some watersheds were unable to benefit from transfer learning. Overall, TL represents a powerful direction for streamflow modelling in regions with limited monitoring, while XAI provides a framework to understand the physical consistency of the transferred knowledge and to determine the minimum monitoring effort required to build reliable local models.
How to cite: Rodrigues, A., Maia, T., Macedo, M., Perdigão, R., Eleutério, J., and Brentan, B.: Transfer Learning for Hydrological Modelling and XAI-Based Physical Consistency Assessment in Reconstructing Streamflow Time Series in Data-Scarce Regions, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-841, https://doi.org/10.5194/egusphere-egu26-841, 2026.
Lakes are the essential asset for the inhabitants of our planet since these are vital sources of water. It is understood that these lakes become more crucial in the regions where water is not easily available such as in Himalayas, drought-prone, and arid regions. However, it has been noticed that the dual problems have arisen at the same time due to climate change, i.e., water scarcity in the arid or drought-prone regions due to rapid extinction of some of the lakes and flood devastations in Himalayan due to overtopping of water from the vulnerable lakes. Climate Change extremes cannot be blamed alone for the extinction of these lakes while overexploitation, improper maintenance and non-civic senses have also exaggerated the process. While the catastrophic events due to these lakes called Glacier Lake Outburst Floods (GLOFs) are mostly occurring due to extremes rainfall events causing regular expansion and contraction of the lakes. However, these extreme events are more intense and frequent due to climate change and tends to increase in the future, making these lakes more vulnerable and responsible for such events. It is essential to monitor the lake water dynamics not only for sustainable water resources management but also for mitigating future catastrophic event risk arising due to these lakes. While the monitoring of lakes is not always easy either due to data-scarcity in the catchments or impossible in-situ measurements due to inaccessible catchment terrain like in Himalayas. The availability and accessibility of advanced remote satellite sensing data such as altimeter, and space-borne Light Detection and Ranging (LiDAR) have been enabled us lake monitoring, however, their processing demands modern approaches. Hence, the present study aims to develop a machine learning model integrated with geospatial approach to process these advance remote sensing data for the spatio and temporal monitoring of water dynamics of lakes. The present study utilizes Icesat-2 as space-borne LiDAR and Surface Water and Ocean Topography (SWOT) as wide swath altimeter data. The study provides a reliable and precise remote sensing derived Water Surface Elevation (WSE) for the lakes at spatial and temporal scales. The derived WSE for lakes would help us to identify the vulnerable lakes and to evolve robust policies to solve dual lake problems at greater extent, i.e., water scarcity in drought or arid-prone regions as well as in the regions like Himalayas for mitigating catastrophic events due to glacier lakes. Further, the developed model would be easily applicable to any lake while the finer adjustment may be required due to different topographic conditions.
Keywords: Lake water dynamics, Space-borne LiDAR, Altimeter, Machine Learning, and Geospatial.
How to cite: Ranjan, R., Rai, A. K., Dhote, P. R., and Keshari, A. K.: Leveraging Advanced Remote Sensing with Machine Learning and Geospatial Techniques for Spatio-Temporal Monitoring of Lake Water Dynamics in Inaccessible and Data-Scarce Catchments , EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-6020, https://doi.org/10.5194/egusphere-egu26-6020, 2026.
Within the APRIORA project, an open-source, geospatial QGIS plugin was developed to support the implementation of the EU Urban Wastewater Treatment Directive in 2025 by assessing environmental risks from human pharmaceuticals. This multidisciplinary deterministic model estimates annual loads from wastewater treatment plants, distributes them spatially through river networks and calculates the Predicted Environmental Concentration (PEC) for each reach.
The practical application of the tool encountered a key limitation in data-scarce regions, where missing catchment-scale flow data and API consumption data prevented the calculation of PECs. Existing hydrological models often present barriers due to high computational demands, intensive calibration needs and strict data requirements. To bridge this gap, a new, integrated hydrological module for the QGIS plugin was developed, offering a flexible, efficient solution that operates with minimal and easily accessible geospatial inputs. In that way, the tool became applicable in data scarce catchments of the project with limited monitoring networks as Poland and Latvia.
The module consists of four tools designed to operate sequentially. The first, “Fix river network”, establishes topological contributing relationships between river sections. The second, “Contributing area of gauging station”, delineates subcatchments contributing to any available stream gauges, defining the areas for model calibration and validation. This step can be omitted in fully ungauged catchments. The third, “Calculate geofactors”, computes physiographic and climatic predictors (e.g., mean elevation, slope, share of forest and settlement area, mean annual precipitation) for each subcatchment. It is important to note that the model makes use of freely available continental-scale datasets (e.g., Copernicus DEM (30m resolution), Corine Land Use Land Cover (100m resolution) and ERA5 monthly total precipitation) thereby ensuring its applicability in regions where data is scarce. The fourth tool, “Flow estimation”, employs a machine learning approach (specifically a Random Forest Regressor) where the previously calculated geofactors act as independent variables to predict the flow measured in gauged subcatchments.
In order to guarantee its applicability in regions without local gauges, the tool allows the use of pre-calibrated, averaged model parameters derived from the project’s partner countries. This provides a transferable solution despite underlying regional hydrological uncertainties. The model estimates annual mean flow and annual mean low flow for regional river sections. This temporal resolution aligns with annual API consumption statistics and also represents the worst-case condition for pollution dilution and environmental risks.
In this presentation, we will present the tool itself as well as results from three different Baltic Sea catchments.
Acknowledgement - The authors thank the Interreg Baltic Sea region funding programme – co-founded by the European Union (ERDF) – and all the APRIORA project partners contributing to this work.
How to cite: Guidi, C., Seidenfaden, A., Marzahn, P., and Tränckner, J.: Regional Annual Flow Estimation by Machine Learning Tool in QGIS for Data-Scarce Catchments, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-11882, https://doi.org/10.5194/egusphere-egu26-11882, 2026.
High-precision and accurate runoff simulation is crucial for the management and allocation of water resources, the operation of hydraulic engineering, and the prevention of flood and drought disasters. However, there is currently no consensus on how to effectively filter and reshape the impact of numerous external factors influencing runoff, and also there is a lack of sufficient theoretical support. To maximize the metrics accuracy of the result of runoff simulation and better capture the internal hydrological characteristics of runoff, the concept of granular computing from the field of artificial intelligence was drawn on, terrain factors were extracted and their attribute features were optimal-selected based on granulation rules, and a Long Short-Term Memory (LSTM) model incorporating the climate characteristic index (LSTM-new) was developed based on delineated sub-region areas in this study. Finally, a unidirectional feedback framework was proposed, combining process-driven method based on the Variable Infiltration Capacity (VIC) model with a data-driven method using the established LSTM (CopulingVIC-new), to enhance the hydrological process characteristics of the simulated runoff and improve simulation accuracy. The results showed that the average NSE, R2, KGE, and RMSE of CopulingVIC-new during training, validation, and testing periods achieved 0.93, 0.92, 0.91, and 334.86 m3/s, respectively, which increased by 7.29%、2.97%、9.73%、-19.41% and 13.41%, 12.19%, 19.73%, -46.95% compared to uncoupled LSTM and VIC. Additionally, the proposed framework effectively captured the interannual variation trend of runoff in all seasons except late spring and summer, though it also overestimated the risk of the occurrence of annual maximum daily peak flow (AMDPF) and total flood volume of annual continuous maximum 5-day (TFAM5D) and thier joint variables. The overall results indicated that the scheme of introducing climate characteristic index, based on sub-region division, can more accurately capture extreme runoff in the study area, as well as the variation of seasonal runoff on both intra-annual and interannual scales. Although CouplingVIC-new still had limited ability to capture extreme flow, the structure of extreme value of the output runoff became more robust after unidirectional coupling. This research can help to expand the application of machine learning in hydrological modelling and provide a useful reference for related studies.
How to cite: Zhao, Y.: Runoff simulation based on granular computing by introducing terrain factors to construct climate characteristic index, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-18485, https://doi.org/10.5194/egusphere-egu26-18485, 2026.
Physics-Informed Neural Networks (PINNs) offer a promising framework for groundwater modeling in regions where hydrogeological data are limited. However, their performance significantly depends on the choice of constraint weights associated with governing equations and derivative-based regularizations. In this study, we develop a constraint-weight selection strategy for PINNs to simulate groundwater head dynamics in data-sparse environments where aquifer properties such as hydraulic conductivity (K) and specific yield/storativity (S) are unavailable. The proposed formulation incorporates first-, second-, and third-order spatial and temporal derivatives of hydraulic head and aquifer properties into the PINN loss function, enabling the model to capture fine-scale spatiotemporal variations without explicit knowledge of subsurface parameters. The approach is applied to a small section of the Varuna River Basin, using groundwater-level observations collected from 37 monitoring stations between 2022 and 2024. The dataset contains several missing values that the PINN framework handles seamlessly, unlike conventional simulation models such as MODFLOW, which require complete and continuous input fields for stable execution. An iterative optimization scheme is employed to balance data fidelity, physical constraints, and derivative-based regularization during training. The proposed method achieves a training R² of 0.986 and a testing R² of 0.947, with corresponding RMSE values of 0.721 and 1.416 meters, respectively. These results demonstrate that adaptive constraint weighting significantly improves prediction accuracy, robustness, and convergence compared to fixed-weight PINN formulations. Overall, the study highlights the potential of derivative-enhanced PINNs for groundwater modeling in data-sparse aquifers and provides a generalized framework for physics-guided learning under missing or incomplete observations.e data scarcity.
How to cite: Bajpai, M., Gaur, S., and Singh, K.: Derivative-Enhanced Constraint Weights for PINNs in Groundwater Flow Modeling Under Unknown Aquifer Properties, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-896, https://doi.org/10.5194/egusphere-egu26-896, 2026.
In addition to supplying the water that people need daily, groundwater also affects agricultural methods, preserves natural balance, and promotes industrial development. The 108,300 km2 Shabelle River Basin served as the site of the current study. Monitoring, evaluating, and conserving groundwater supplies for water resource management and development is made possible by the effective integration of remote sensing data and GIS in hydro-geological research. The Shabelle basin area's Ground Water Potential Zones were defined by combining seven thematic layers—geology, land use/land cover, drainage density, slope, lineament density, rainfall distribution map, and soil map—into a GIS platform using the spatial analyst tool in Arc GIS 10.8. The analytical hierarchy process (AHP) technique is used to find the weighted values for each parameter and its sub-parameters based on the relative importance of the influencing elements for groundwater recharge. Four groups were identified on the final groundwater potential zonation map of the study area: low potential zones of 1,548.7 km2 (1.43%), moderate potential zones of 25,786.23 km2 (23.81%), high potential zones of 22,353.12 km2 (20.64%), and very high potential zones of 55,341.3 km2 (54.10%). According to this study, high and very high groundwater potential zones dominate in the basin in 75% of the entire studied region. These zones are found in the basin's northern and central regions, where low slopes, fractured geological formations, and porous soil are present. However, because to their steep slopes, strong geological formations, and low rainfall zones, the south and southwest regions of the basin have poor potential zones. When well data was utilized to validate the accuracy of this data, there was a high degree of agreement between the expected and observed well performance. The Shabelle river basin's water management policies, effective use of natural resources, physical design, and sustainable groundwater development should all benefit greatly from the findings, particularly as the adverse effects of climate change on human life become closer. Anywhere else in the world, the study's methodologies can be used. The findings of this study can be applied to future research on agriculture, basin management, sustainable groundwater, and the interaction between groundwater and climate change.
How to cite: Alasow, I. M. A., Hussein, M. A., Tiwari, S. K., and Bhatla, R.: Identification of Groundwater Potential Zones Using GIS and Multi-Criteria Decision-Making Techniques: A Case Study of the Shabelle River Basin (Somalia), EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-95, https://doi.org/10.5194/egusphere-egu26-95, 2026.
Pharmaceuticals, ubiquitous in human, veterinary and agricultural use, are prevalent emerging contaminants in Chinese surface waters. Although not highly persistent, their low removal in conventional wastewater treatment leads to continuous discharge, creating "pseudo-persistence." This chronic exposure poses significant ecological and human health risks, including hormonal disruption of female reproduction and antibiotic-induced gut microbiota alterations and antimicrobial resistance in aquatic biota.
Numerous pharmaceuticals (>100) have been detected in China's surface waters. However, clear regulatory priorities are lacking, and nationwide monitoring is insufficient, leaving many regions without concentration or risk data. This study aims to: (1) identify pharmaceuticals posing the highest human and environmental hazards; (2) develop nationwide predictive concentration models using machine learning; and (3) generate a health risk map for pharmaceuticals in China's surface waters.
Through systematic keyword searches in Web of Science and CNKI, we compiled data from 227 peer-reviewed articles (2010-2023), covering approximately 13,000 sampling sites across China's nine major river basins. Pharmaceutical concentrations, detection frequencies, and sampling metadata were extracted. To assess environmental behavior and risks, four key indicators were selected: octanol-water distribution coefficient (LogDow) for bioaccumulation potential, degradation half-life (T1/2) for persistence, predicted no-effect concentration for aquatic ecosystems (PNECeco) for ecotoxicity, and predicted no-effect concentration for human exposure (PNEChum) through drinking water and fish consumption.
Principal component analysis (PCA) integrated four indicators into a composite hazard score (HP) and to combine concentration and detection frequency into an exposure potential score (EP). Pharmaceuticals were preliminarily screened based on reference thresholds for HP and EP values, and then ranked by the product of HP and EP to establish priority control lists for each river basin. Roxithromycin and erythromycin, exhibiting high toxicity and extensive data, ranked highest across all basins. Antibiotics were consistently high-priority in all nine basins. In densely populated basins (Haihe, Yangtze, Pearl), bezafibrate, indomethacin, and ibuprofen require additional attention. Hormones (estrone, estriol, ethinylestradiol) showed elevated concentrations and risks in Songhua/Liao basins. Increased monitoring is strongly recommended for data-scarce inland basins.
Four representative pharmaceuticals (erythromycin, ciprofloxacin, norfloxacin, carbamazepine), selected based on high toxicity or exposure potential, were modeled nationally. Predictors included 27 variables across five categories: Socioeconomic, Healthcare, Agricultural and aquacultural, Natural environmental, and Water quality indicators. Seven machine learning algorithms were evaluated (DT, ExtraTrees, GB, KNN, RF, SVM, XGBoost). RF demonstrated superior performance and was selected for feature selection (via weighted backward stepwise regression) and hyperparameter tuning (grid search with 10-fold CV). The optimal model was chosen based on R² and RMSE.
Predicted concentrations were then input into the USEPA-recommended human health risk assessment model. Carbamazepine, ciprofloxacin, and norfloxacin exhibited low risks nationwide (HQ < 1). Erythromycin exceeded safe levels (HQ > 1) in eastern regions (Yangtze River Delta, Bohai Rim, Pearl River Delta). Spatially, erythromycin and norfloxacin risks displayed a distinct east-west gradient (higher east), while carbamazepine and ciprofloxacin showed minimal spatial variation.
How to cite: Li, J.: Nationwide Prioritization and Machine Learning-Based Risk Prediction of Pharmaceuticals in China's Surface Waters, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-6360, https://doi.org/10.5194/egusphere-egu26-6360, 2026.
The overexploitation of groundwater has emerged as a critical environmental issue due to the increasing pressure placed on this vital freshwater resource by rapid urbanization and population growth. Understanding future groundwater availability near urban expansion is essential for sustainable urban planning and water-resource management. This study investigates the influence of land-cover change on groundwater depletion while also examining the spatial patterns of urban growth and their effects on surface thermal conditions using Land Surface Temperature (LST) and the Normalized Difference Vegetation Index (NDVI). Groundwater storage variations were monitored using data from the Gravity Recovery and Climate Experiment (GRACE), while Landsat imagery was used to derive land-cover maps, NDVI, and LST. To assess the relationship between climate variability and groundwater recharge, GRACE-derived groundwater storage anomalies were correlated with precipitation data obtained from the Global Precipitation Measurement (GPM) mission. Time-series analyses of groundwater storage and land-cover changes were conducted at five-year intervals from 1990 to 2025 to quantify the impacts of urbanization on groundwater dynamics. The results reveal a significant acceleration in groundwater depletion and urban expansion over the past decade. Concurrently, LST exhibits an increasing spatial trend that closely corresponds with declining vegetation cover and expanding built-up areas, indicating that urbanization has contributed substantially to rising surface temperatures. These findings underscore the urgent need for effective groundwater management policies and integrated urban planning strategies to ensure the long-term sustainability of freshwater resources.
How to cite: Ali, M. Z. and Benaafi, M.: Impact of Urbanization on Groundwater Storage and Surface Temperature Changes: A Case Study of Riyadh, Saudi Arabia, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-10496, https://doi.org/10.5194/egusphere-egu26-10496, 2026.
In West Africa, limited access to hydrometric data remains a major challenge for advancing surface water research and improving water management. Since the early 1980s, many gauging stations have been decommissioned, leaving gaps in reliable streamflow records across numerous catchments. Parameter regionalization of hydrological models is commonly employed to enable runoff prediction in ungauged catchments. This study represents an assessment of rainfall-runoff model regionalization across West Africa. We used an unprecedented dataset of 189 near-natural catchments to compare two contrasting approaches: (i) a benchmark conceptual modeling framework using the GR4J model, regionalized with three parameter-transfer techniques (spatial proximity, physiographic similarity, and Random Forest), and (ii) a data-driven framework based on Long Short-Term Memory (LSTM) neural networks. Using a leave-one-out resampling approach, regionalization approaches were evaluated using different performance metrics: (i) the Kling-Gupta Efficiency (KGE), calculated between simulated and observed streamflows, (ii) the relative bias (rBias) on several hydrological signatures computed with observed or simulated discharge and (iii) the difference between observed and simulated flood quantiles. Results show that the conceptual modeling approach with traditional parameter-transfer techniques consistently underperforms compared to the LSTM, failing to reproduce key hydrological signatures. In contrast, the LSTM model showed better generalization performance, accurately simulating streamflow with a median KGE of 0.67 and reliably capturing hydrological signatures and flood quantiles across West Africa’s diverse climates and landscapes with lower biases. These findings highlight the potential of data-driven approaches to enhance hydrological prediction in data-scarce regions, supporting more effective flood risk management and water resource planning.
How to cite: Tramblay, Y., Diop, S. B., Kate, F., Souassi, I., Dieppois, B., Bodian, A., Guerin, J., Hostache, R., Johannet, A., Kratzert, F., Oudin, L., Sivelle, V., and Traoré, K.: Large-scale streamflow regionalization in ungauged West African catchments: How do classical and deep learning approaches compare?, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-1536, https://doi.org/10.5194/egusphere-egu26-1536, 2026.
Regional frequency analysis relies heavily on robust goodness-of-fit (GOF) testing for selecting an appropriate probability distribution, which directly influences the accuracy of estimated quantiles. However, existing statistical approaches often involve strong assumptions and computational overheads that limit their effectiveness, particularly for large regional datasets. The widely used L-moment-based approach requires scaling each site’s data by its own mean, which raises concerns about potential distortion of the original distributional characteristics. To overcome this limitation, the present study proposes a novel Deep Learning (DL)-based GOF test that identifies the regional distribution without performing mean-based scaling. The proposed methodology employs a Deep Neural Network (DNN) trained to classify regional distributions based on the distinctive behavior of Generalized Extreme Value, Generalized Pareto, Generalized Logistic, Generalized Normal, and Pearson Type III distributions under specific mathematical transformations. These transformations yield distribution-specific signatures that form the basis of the DNN training process. For a given dataset, the transformations are applied, and kernel density estimates derived from the transformed data are used as inputs to a pre-trained DNN model to identify the most suitable regional distribution. The DNN classifier achieved an accuracy of 95.09% on the training dataset and 94.86% on the test dataset. A comprehensive simulation study was conducted for multiple regional configurations to assess the performance of the proposed DL-based GOF test. The results were compared against the conventional L-moment-based GOF approach. The proposed method demonstrated comparable classification accuracy for smaller region sizes and marginally improved accuracy for larger datasets. The proposed DL-based GOF framework shows significant promise, particularly due to its substantially lower computational cost compared to the conventional L-moment methodology. The findings suggest that this approach can facilitate accurate and efficient estimation of quantiles, thereby supporting informed decision-making planning, management and risk assessment.
How to cite: Kaur, S. and Chavan, S. R.: Proposing a Deep Learning based Regional Goodness-of-Fit test for identification of regional distribution , EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-5433, https://doi.org/10.5194/egusphere-egu26-5433, 2026.
Access to safe drinking water remains a daily challenge for millions of urban residents around the world, particularly in sub-Saharan Africa. This study provides a detailed assessment of inequalities in the realization of the human right to water in urban neighborhoods in Malabo, Equatorial Guinea. Clustering techniques combined with GIS analysis were used to map and assess access to water throughout the study area. The clustering results were compiled into a matrix assessing six key indicators: the physical availability of improved water sources; transport time; water quality; water quantity; reliability; and affordability. More than 500 household surveys were conducted and over 200 water points were sampled for this work. The results indicate that access to water is severely limited by poor quality, insufficient quantity and an unreliable supply. Below 3% of households meet the standard for safely managed drinking water, and less than 22% have at least basic access, which contrasts sharply with official statistics. Considering these results in the context of current literature highlights the importance of taking all relevant factors into account when making reliable estimates of water access. Current rates of access to this resource tend to be significantly lower than reported, and despite global progress, humanity is still far from fulfilling the fundamental human right to water. These findings emphasise the urgent need for targeted interventions to address inequalities and enhance the water supply in urban areas.
How to cite: Rodríguez del Rosario, M., Nsue-Mikue, S. M., Gómez-Escalonilla, V., Montero-González, E., Díaz-Alcaide, S., and Martínez-Santos, P.: Assessing urban water access in African cities: a GIS clustering approach in Malabo, Equatorial Guinea, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-19960, https://doi.org/10.5194/egusphere-egu26-19960, 2026.
Water quality in ice-covered lakes is strongly affected by winter physical conditions, particularly in shallow systems where ice cover influences mixing, oxygen availability, light conditions, and biogeochemical processes. Changes in ice thickness and duration can therefore have substantial impacts on key water quality parameters, including dissolved oxygen and nutrient dynamics. However, long-term observations of both water quality and ice conditions are sparse and unevenly distributed across Finnish lakes, limiting comprehensive assessments. In this study, we apply a machine-learning approach based on the gradient boosting algorithm to model water quality and ice conditions on shallow lakes in Finland over the period 1965–2024. The model demonstrates strong predictive performance, evaluated using the root mean square error (RMSE), enabling the reconstruction of water quality dynamics under data-scarce conditions.
How to cite: Nourinezhad, S., Fazel, N., Postila, H., and Torabi Haghighi, A.: Ice-Regulated Water Quality Dynamics in Finnish Shallow Lakes: A Machine-Learning Reconstruction, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-21064, https://doi.org/10.5194/egusphere-egu26-21064, 2026.
Accurate prediction of river water quality is essential for effective watershed management, yet it is often hindered by practical monitoring constraints, including infrequent grab sampling (e.g., monthly observations) and the lack of reliable streamflow data. These limitations restrict the applicability of conventional process-based water-quality models and necessitate alternative analytical tools. In this study, we propose a graph-based machine learning framework that integrates prediction and diagnostic analyses of river water quality, with chromaticity prediction in the Hantan River Basin, Republic of Korea, as a case study. Graph-based models outperformed purely temporal baselines, with the Graph Sample-and-Aggregate (GraphSAGE) model achieving a test R² of 0.82. Its sampling-based spatial aggregation integrates localized and distributed upstream information across the river network, allowing the model to capture nonlinear relationships mediated by implicit flow connectivity. Graph explanation analyses using PGExplainer identify the SC sub-watershed as the dominant pollution source and primary intervention area. In addition, feature attribution analyses distinguish persistent long-term drivers (e.g., TOC associated with major wastewater treatment plant discharges) from short-term episodic influences linked to facility-specific effluent spikes. Overall, these results demonstrate that graph-based machine learning can serve as a useful framework for both prediction and diagnostic interpretation of key water-quality drivers in data-limited river systems.
How to cite: Choi, S., Jung, E., Shin, H.-S., Song, J.-H., You, H., Son, H., Choi, I., Yang, J., and Moon, H.-C.: Graph-based machine learning approach for river water quality prediction under data limitations, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-4811, https://doi.org/10.5194/egusphere-egu26-4811, 2026.
Posters virtual: Wed, 6 May, 14:00–18:00 | vPoster spot A
EGU26-9498 | ECS | Posters virtual | VPS9
Thermal Remote Sensing for Qualitative Analysis of Surface Water and Groundwater Interaction: A Case Study of the Kharun RiverWed, 06 May, 14:00–14:03 (CEST) vPoster spot A
Surface-groundwater interactions (SGI) plays a crucial role in maintaining stream thermal regime and ecological balance, keeping a check on this is logistically and financially challenging. This study utilises multi-temporal Landsat-8 Thermal Infrared Sensor (TIRS) data to compute Stream Surface Temperature (SST), its anomaly (SSTA), Robust Thermal Deviation Index (R-TDI) and further classifies groundwater influence on the Kharun River, a semi-arid urban catchment in India (approximately 4109 km²).
Due to changes in weather and season, surface water is subject to heating and cooling, but the water system beneath the land surface will be at a constant temperature. The stream reach, influenced by groundwater, will show a relatively stable thermal signature across all seasons. Stream Surface Temperature (SST) derived through radiometric calibration and emissivity-adjusted retrieval across pre-monsoon, monsoon, and post-monsoon periods. To isolate localized hydrological processes from regional climatic forcing, we computed Stream Surface Temperature Anomalies (SSTA) by subtracting reach-wise median SST from pixel-scale values. To account for the non-normal nature of SST, a Robust Thermal Deviation Index (R-TDI) framework was utilised which minimizes atmospheric noise and mixed-pixel interference, allowing for the isolation of persistent thermal signals.
Using statistically defined TDI thresholds, a classification approach was finalised putting stream stretches into high, moderate, and low groundwater influence zones. Results identify spatially consistent cold-water anomalies indicative of groundwater discharge primarily during pre-monsoon and warmer-water anomalies during post-monsoon seasons when thermal contrasts are most pronounced. These zones coincide with structurally controlled segments and urbanized stretches, suggesting a complex interplay between hydrogeology and anthropogenic modifications. By leveraging open-access satellite data, this research provides a scalable tool for evidence-based river restoration and climate-resilient water management in rapidly urbanizing regions.
Key Words: Thermal remote sensing; Landsat-8 TIRS; Stream surface temperature; Thermal anomaly; Surface–groundwater interaction; Data-scarce catchments
How to cite: Chandel, R. and K. Singh, C.: Thermal Remote Sensing for Qualitative Analysis of Surface Water and Groundwater Interaction: A Case Study of the Kharun River, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-9498, https://doi.org/10.5194/egusphere-egu26-9498, 2026.
EGU26-19536 | ECS | Posters virtual | VPS9
Estimating Surface and Subsurface Water in Ondiri Swamp, Kenya, Using Multi-Sensor Embedded Data and Preliminary Water BalanceWed, 06 May, 14:03–14:06 (CEST) vPoster spot A
Wetlands play a vital role in the hydrologic cycle since they impact stream flow, add to water storage capacity, provide habitat for many species, and provide resiliency in ecosystems. Ondiri Swamp, which is a peatland located in Kenya with approximately an area of 33 hectares and is the headwaters of Nairobi River has very little hydrological understanding, especially regarding subsurface and groundwater contributions since there have not been any continuous in-situ data measurements. This study aims to quantify water storage and investigate potential groundwater presence using an embedded, multi-sensor dataset.
To overcome the limitation of only using traditional optical index methods for surface water detection under dense vegetation, water occurrence data (Global Surface Water), Sentinel-1 SAR and Sentinel-2 multitemporal optical images, DEM images (Copernicus DEM), and NDVI derived vegetation index data will be combined. Measurements of swamp depth and peat thickness will be collected from short-term field campaigns for calibration of volume estimates and provide preliminary data for a preliminary water balance. The precipitation data (CHIRPS) and ET data (FAO WaPOR) will be combined with inflow and outflow estimates to create a preliminary water balance. Surface storage will be estimated, and potential groundwater contributions will be inferred without long-term observatory data sources. The methods used for the quantitative and qualitative assessment of wetland water resources will generate probabilistic wetland water maps using a multi-temporal remote sensing-based classification of existing datasets, as well as using terrestrial calibrations from field data.
The study will be able to quantify total wetland water storage, determine the degree to which groundwater may influence wetlands, and identify the seasonal dynamics of wetland hydrology. Through a combination of remote sensing, existing datasets, and terrestrial calibrations from field studies, the study provides a strong, scalable framework for conducting wetland hydrology research, managing wetland ecosystems and planning wetland water resources in areas where very few, if any, hydrological observations are available.
How to cite: Ouedraogo, A.: Estimating Surface and Subsurface Water in Ondiri Swamp, Kenya, Using Multi-Sensor Embedded Data and Preliminary Water Balance, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-19536, https://doi.org/10.5194/egusphere-egu26-19536, 2026.
EGU26-5160 | ECS | Posters virtual | VPS9
Solute dispersion from continuous point sources in ice-covered turbulent flows with bed absorptionWed, 06 May, 14:06–14:09 (CEST) vPoster spot A
A numerical investigation is conducted to study the steady-state concentration field
when a solute is released from multiple continuous line sources in an ice-covered
channel with an absorbing bed under turbulent flow conditions. The governing
equations are solved using the Crank-Nicolson scheme by adopting a two-power law
velocity and a quartic eddy diffusivity profile, which is influenced by the roughness of
the bed layer and the ice cover. Validation against earlier numerical results for a
specific case reveals strong consistency in the concentration profiles. The findings
highlight how the roughness of the boundaries affects the solute concentration. It
further demonstrates the effect of the bed absorption parameter in the early mixing
stages when solute is released near the bed. For zero bed absorption, the solute
concentration asymptotically attains a uniform far-field value of unity, while any non-
zero bed absorption leads to complete depletion of solute downstream.
How to cite: Paul, S. and Ghoshal, K.: Solute dispersion from continuous point sources in ice-covered turbulent flows with bed absorption, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-5160, https://doi.org/10.5194/egusphere-egu26-5160, 2026.
