Soil infiltration is a critical process shaping water security in Ethiopia’s northwestern highlands, determining whether rainfall nourishes crops, recharges aquifers, or is lost as runoff. In the Debre Markos region, infiltration varies widely with land use, soil texture, and seasonal rainfall, yet many models still assume uniform behaviour. To address this gap, we developed a two-stage hybrid framework that integrates the physically based Green-Ampt model with Support Vector Machine (SVM) learning. The Green-Ampt model provided baseline infiltration estimates using calibrated soil hydraulic parameters (RMSE = 23.1mm/hr, R² = 0.73), while SVM regression corrected residual biases. The (RMSE = 19.5mm/hr, R² = 0.79, representing a 15.6% error reduction) and SVM classification delineated infiltration zones into four categories based on steady-state infiltration rates. These thresholds were selected to align with Ethiopian highland soil and land-use management priorities, where very low rates indicate severe compaction or surface sealing, and high rates correspond to well-vegetated, permeable soils. Results revealed that grasslands on silty loam exhibit high point-scale infiltration capacity (>300mm/hr), while urban and grazed lands on clay loam show suppressed rates (<150mm/hr), indicating their relative potential for local water entry rather than direct groundwater recharge. These findings demonstrate how combining physical models with AI enhances predictive reliability and provides a reproducible framework for identifying recharge-prone areas and erosion-vulnerable zones, supporting site-specific watershed management and climate adaptation in Ethiopia’s highlands.
Reliable daily streamflow prediction is essential for water resources management in California’s Sierra Nevada, where increasing hydroclimatic variability has intensified flood and drought risks. Although regional machine learning models can improve generalization across watersheds, conventional shared architectures may not adequately capture watershed-specific characteristics and distinct flow regimes. To address this limitation, we propose MoE-τ, a mixture-of-experts framework that combines multiple LSTM experts through an attention-based temporal gating network with a learnable temperature mechanism for adaptive expert weighting. We evaluated MoE-τ across 14 heterogeneous Sierra Nevada watersheds against learning-based models and the SAC-SMA process-based model. Under the in-domain temporal prediction benchmark, MoE-τ attained the highest mean NSE (0.873) among the learning-based models, a modest improvement over the strongest recurrent baseline (Bi-LSTM, 0.857) and a larger margin over SAC-SMA (0.762). It also remained robust under an alternative train–test split. Ablation and parameter-scale analyses showed that the performance gains arise from the proposed gating and expert-based design rather than increased model size. Model interpretation analyses revealed expert specialization across flow regimes and a strong relationship between hydrometeorological distribution differences and expert-usage differences (Spearman’s r=0.687). In held-out watershed experiments, MoE-τ outperformed LSTM on nearly all metrics across watersheds with distinct hydrologic conditions. These results demonstrate that adaptive expert selection improves both in-domain prediction and generalization to unseen watersheds, providing a robust and interpretable method for regional streamflow modeling.
Groundwater recharge remains a critical yet poorly quantified component of the hydrological cycle, particularly in data-scarce, heterogeneous basins. This study employs a novel comparative framework that integrates stable isotope mass balance with the spatially distributed WetSpass-M hydrological model to assess groundwater recharge dynamics across three climatically and topographically distinct watersheds (Gilgel Gibe, Gombora, and Neri) within the Omo Gibe River Basin, Ethiopia. We analyzed stable isotopes of Oxygen (δ18O) and Hydrogen (δ2H) in precipitation and groundwater to establish Local Meteoric Water Lines (LMWLs) and quantify seasonal recharge contributions. Concurrently, the WetSpass-M model was calibrated and validated against stream flow data to simulate high-resolution spatiotemporal components of the water balance. Key findings include: (1) groundwater recharge is primarily driven by wet-season precipitation, with isotopes revealing that 88.1% (Gilgel Gibe) and 83.7% (Gombora) of annual recharge occurs during the wet season; (2) in the bimodal Neri watershed, a more balanced seasonal distribution is observed (64.7% wet season), reflecting its distinct rainfall regime; and (3) the WetSpass-M model corroborates these seasonal proportions while providing high-resolution spatial mapping of recharge hotspots. The WetSpass-M model provided detailed spatial mapping, confirming that recharge is highest in vegetated highlands with permeable soils and is significantly reduced in urbanized areas. The strong agreement between the two methods underscores the robustness of our synergistic framework, in which isotope data validate the model's temporal dynamics and the model elucidates the spatial context of the isotopic signatures. This study suggests that sustainable groundwater management in the basin should prioritize wet-season water harvesting, land-use planning to protect infiltration zones, and climate adaptation strategies.
Floods and landslides are significant natural hazards that frequently threaten the Indian Himalayas, an extremely sensitive region both geologically and climatically. Shallow slope failure due to heavy rainfall during rainstorms is common in mountain areas. Existing forecasting models are usually either purely data-driven and thus suffer from the "black-box" issue and poor generalization, or purely physics-based, which are computationally expensive and extremely challenging to calibrate. This research presents a hybrid framework combining a Physics-Informed Neural Network (PINN) for flood forecasting and an Extreme Gradient Boosting (XGBoost) model for landslide susceptibility prediction. It serves as a compromise between purely data-driven and physics-based models. Our method is an integration of observed data consisting of high priority variables - precipitation records, soil moisture data, and Digital Elevation Models (DEMs) with the underlying governing partial differential equations (PDEs). By incorporating physical equations and laws into the neural network, the Hydrological model obtains proficiency such that the results align well with observed data and physical behavior, governed by the physical laws. The proposed method is applied in a highly dynamic Himalayan region (Himachal Pradesh), a region that frequently witnesses the two natural hazards.
Artificial groundwater recharge (AGR) is an effective method to replenish depleted aquifers in arid and semi-arid regions facing over-extraction, droughts, and rising demand. By channeling surface water underground (e.g., via infiltration basins or wells), AGR boosts storage, improves quality through soil filtration, and supports sustainable management in water-stressed areas. This research develops a novel framework integrating Ordered Weighted Averaging (OWA) and Principal Component Analysis (PCA) for optimal AGR site selection in water-scarce regions, using the Yazd-Ardakan watershed as a case study. From 17 geospatial layers, eight critical factors were identified (correlation >0.5): land use, slope gradient, proximity to aquifers, geological formations, pedological characteristics, elevation bands, temperature variations, and microclimate zones. PCA showed the first principal component (PC1) explaining 35.9% of variance, with dominant influence from pedological properties (weight=0.32), geological features (0.29), slope gradients (0.27), and elevation ranges (0.25). OWA and PCA outputs had a strong 64% positive correlation, validating these factors as key determinants for recharge suitability. The integrated approach reduces data processing by 30-40% compared to conventional methods, provides statistical validation of weights via PCA, and adapts well to arid/semi-arid environments. It offers water managers a cost-effective tool for sustainable AGR implementation and water security in climate-vulnerable regions.
Citizen science (CS) empowers communities in data collection and problem-solving, enhancing research inclusivity and local relevance. However, the absence of a comprehensive framework limits their development and potential contributions to sustainable practices. In this study, we developed and tested a CS framework for hydrological data collection and the co-creation of nature-based solutions (NbS). The framework has 3 major stages: establishment, operation and knowledge generation - each involving multiple activities. It was piloted in the Kelekindo watershed of Central Rift Valley Lakes Basin of Ethiopia, engaging the citizen scientists (CSts) in monitoring of rainfall, river water level, shallow groundwater (SGW) and water quality. The CS data allowed to investigate poorly understood hydrological behaviour of the watershed, including the fill-and-spill effect and characterization of an extreme runoff event. Co-interpretation of the data showed that SGW recharge could support thousands of farmers in practicing backyard irrigation. The river and well water in the study area were within safe limits for irrigation in terms of the commonly monitored water quality parameters. CSts were engaged in the selection and co-design of a priority NbS for watersheds. The study showed that a well-designed citizen science framework can generate reliable hydrological data and meaningfully engage communities.
The Pearson type III (P-III) distribution is widely applied in rainfall frequency analysis, yet quantifying parameter uncertainty remains challenging due to reliance on prior assumptions or resampling from point estimates. This study proposes a pivotal-quantity-based method to construct interval estimations for the location and scale parameters, while the shape parameter is estimated separately. The proposed framework avoids prior specification and reduces dependence on point-estimate-based resampling. The method is applied to rainfall frequency analysis using annual maximum rainfall data from 13 gauges in Shanghai, and its performance is evaluated against Bayesian and bootstrap approaches through simulation experiments under varying sample sizes. Results indicate that uncertainty in design rainfall estimates tends to be higher in northern and central Shanghai. In addition, the proposed method reduces root mean square error (RMSE) by approximately 10–30% compared to the bootstrap method across different sample sizes, while showing comparable performance to the Bayesian approach. These results suggest that the proposed method provides an alternative framework for uncertainty analysis of the P-III distribution under the conditions considered in this study, particularly for small to moderate sample sizes.
This study examines the runoff seasonality, focusing on low-flow and high-flow events in five headwater catchments of the Krkonoše (Giant Mts.) National Park, Czechia over the 1964–2023 period. The research aims to quantify the timing, distribution, and trends of streamflow extremes in order to understand the impact of climate change on hydrological dynamics in this protected area. The results show significant shifts in runoff seasonality. Low-flow events are increasingly clustered in late summer and early autumn, occurring several weeks earlier in most catchments. High-flow events have moved earlier in spring, driven by changes in the timing of snowmelt due to rising temperatures and decreasing winter snowpack. Analysis of the seasonality index shows an increase in the regularity of both low and high flows across in all study catchments. Statistical tests identified significant decreases in both minimum and maximum streamflow, particularly for annual 7-day and 30-day maximum flows in most catchments. Strong declines in monthly flows are most evident during summer, while notable increases were observed between January and March. The findings highlight the influence of climate change on streamflow seasonality in European montane regions, emphasizing the risk for protected wetland ecosystems within headwaters, as well as for water management downstream.
River ice plays a central role in winter transportation, safety, flood dynamics, and overall water resource operations in northern regions. This study presents the CrowdICE smartphone application developed for citizen science-based river ice monitoring in Finland. The app was designed to collect current and historical observations and to draw future expectations of river ice conditions. During the pilot period (April–June 2025), citizens submitted 87 observations and 221 photos across Finland, with 68 observations and 184 photos originating from the case study of Kiiminkijoki. It was noticed that engagement was highest in areas where the project was actively promoted. Results revealed a greater number of historical photo submissions compared to current photo additions. Comparison of citizen observations with in-situ camera images revealed good agreement, although the small dataset limited the conclusions that could be drawn. Participants’ perceptions revealed generational differences in views regarding changes in river ice cover and future needs. Overall, CrowdICE was applicable for enhancing river ice monitoring by engaging citizens to use mobile technologies. In the future, system development should focus on improving the app interface, questionnaire, and data verification. The app could be integrated into broader environment monitoring systems and related machine learning applications.
Riverbed sand mining is one of the main anthropogenic drivers of sediment imbalance and riverbank erosion in the Mekong Delta. To investigate its morphological impacts, we used the hydrodynamic–morphological model HYDIST, which was coupled with a sand mining source function (Ssm) that applies prescribed extraction rates to simulate sand mining as a time-dependent process. The model is applied to the Vam Nao River, a confluence connecting the Tien and Hau rivers.Model simulations reveal that sand mining increases riverbed erosion and reduces the natural balance between erosion and deposition. In the Vam Nao reach (Zone 1), the depositional proportion decreased from 27% to 13%, while erosion accounted for 87% of total morphological change, resulting in a shift to an erosion-dominated regime. Total erosion was about 2 times higher than under the no-mining scenario, with approximately 74% of the additional erosion occurring within the mining area and 26% outside it. In the Hau River reach (Zone 2), deposition dominated under non-mining conditions, accounting for 84% of total bed change. Sand mining reversed this pattern and increased total erosion volume by a factor of about 8. Approximately 76% of the additional erosion occurred within the mining area, while the remaining 24% occurred outside it. At the Vam Nao–Hau River confluence (Zone 3), deposition decreased by 30%, and erosion accounted for 70% of total bed change, demonstrating pronounced downstream propagation of mining impacts. Overall, the findings highlight that sand mining increased riverbed erosion, with the strongest impacts occurring at the river confluence.
To evaluate the hydrochemical characteristics and driving mechanisms of shallow groundwater in the Fuyang section of the Ying River Basin, 167 groundwater samples were collected and systematically analyzed using Piper diagrams, Gibbs plots, ion ratio analysis, inverse hydrogeochemical modeling, and multivariate statistical methods (PCA and HCA). The results indicate that the groundwater is weakly alkaline (pH 6.68–8.81), with TDS averaging 698.36 mg/L, predominantly freshwater, and only 8.38% classified as brackish water (TDS>1000 mg/L). High-TDS zones (>1000 mg/L) are distributed on both sides of the Ying River, while low-TDS zones (<500 mg/L) are located along the river channel. The dominant hydrochemical type is HCO3-Ca, primarily controlled by carbonate/silicate weathering and cation exchange, with evaporation-concentration as a secondary factor. Inverse modeling quantitatively reveals the dissolution of Ca-montmorillonite and dolomite, precipitation of calcite, and cation exchange processes along groundwater flow paths. HCA classifies the samples into three groups: Group 1 (natural background, HCO3-Ca type, TDS mean 485.6 mg/L, n = 89), Group 2 (agricultural impact, HCO3-Ca·Na type, TDS mean 762.3 mg/L, n = 52), and Group 3 (industrial/domestic pollution, HCO3·Cl-Ca/HCO3·SO4-Ca types, TDS mean 1123.8 mg/L, n = 26). PCA quantitatively identifies: F1 (42.54%) as anthropogenic inputs (Cl-/SO42-), F2 (22.30%) as rock weathering and cation exchange (Na+/HCO3-), and F3 (15.34%) as geological controls under river influence (pH/Ca2+), indicating that most ions are governed by anthropogenic activities and rock weathering, while pH and Ca2+ are related to the primary geological environment. This study provides a theoretical basis for the protection and utilization of groundwater resources in the Ying River Basin.
Hydraulic structures like dams, weirs, and culverts disrupt river connectivity, threatening aquatic biodiversity by impeding fish migration. This study evaluates the hydraulic performance of one-sided wart-type baffles in a diversion tunnel under supercritical flow, focusing on their impact on velocity, turbulence intensity (TI), and low-velocity refuges for fish passage. Experiments were conducted in a 7.0 m tilting flume at Kyoto University, with a 2% bed slope and 0.20 m baffle spacing. Baffles reduced velocities by 10–20% compared to smooth channels, creating heterogeneous flow fields and resting zones critical for fish. Mean TI increased from 7.59% (smooth) to 10.34% (baffled), indicating enhanced turbulence and energy dissipation. The Hydraulic Passage Index showed higher velocity deficits (9.57% vs. 7.94%), reflecting reduced flow efficiency but improved ecological functionality. These findings highlight baffles’ effectiveness in facilitating fish passage, offering insights for optimizing fishway designs.
Compound flooding in coastal cities arises from nonlinear interactions among rainfall, tide, and upstream stage, yet the hydraulic transition mechanisms governing such interactions remain poorly understood. This study develops a threshold-oriented framework that combines synchronization-focused empirical scenarios with a fully coupled hydrodynamic model to quantify rainfall–tide–stage interactions in Lianyungang, China. Results demonstrate a two-phase amplification mechanism, with synchronization-induced flooding followed by delayed drainage persistence, revealing the temporal asymmetry of compound processes. The phase–stage matrix shows that amplification is strongest under a low upstream stage and progressively weakens as the upstream stage rises, indicating nonlinear attenuation rather than monotonic scaling. Pipe-level surcharge analysis further reveals selective backwater-driven redistribution within coastal and downstream-connected corridors, while more than 90% of pipes remain structurally persistent. More than 90% of inundation depths remain below 0.3 m, forming a robust shallow-water skeleton governed by drainage–topography interactions. The proposed hydraulic stress and synchronization indicators (BTI, SAF, TRC), together with joint design curves, provide phase-resolved sensitivity metrics that can support adaptive pumping operation and blue–green–grey infrastructure planning for urban flood resilience.
Examining hydrological variability is essential for understanding system dynamics and supporting effective water resource management. Recurrence analysis, a nonlinear time series technique, can uncover hidden patterns in hydrometeorological data through recurrence plots (RPs) and recurrence quantification analysis (RQA). Despite its effectiveness, its application remains limited, particularly in the Nordic region. This study applies recurrence plots (RPs) and recurrence quantification analysis (RQA) to daily air temperature, precipitation, and streamflow data (2000–2022) from the regulated Kemijoki and pristine Tornionjoki river basins, explicitly considering frozen and unfrozen seasons. Streamflow exhibited the strongest recurrence characteristics, with Tornionjoki showing consistently higher determinism, laminarity, and recurrence persistence than Kemijoki, indicating stronger hydrological memory and seasonal organization. Seasonal analysis showed that recurrence structures were generally more persistent during the unfrozen period, while frozen-season dynamics differed markedly between the two basins. Temperature displayed moderate to high recurrence characteristics associated with strong seasonal forcing, whereas precipitation remained the least deterministic variable, reflecting its inherently stochastic nature. Surrogate data analysis further confirmed that the observed recurrence characteristics cannot be explained solely by linear stochastic processes. These findings demonstrate that recurrence-based metrics can effectively distinguish natural and regulated hydrological regimes and provide a robust framework for understanding seasonal dynamics and anthropogenic impacts in cold-region river systems.
Predicting thermal stratification in deep reservoirs is challenging: 3D hydrodynamic models demand intensive computation, while pure deep learning often systematically overestimates hypolimnetic temperatures—failing to capture the density-induced suppression of vertical mixing at the thermocline, limiting the reliability of AI models for operational decision-making. To address this, we developed a hydrodynamic-informed surrogate framework for Miyun Reservoir. Utilizing the CE-QUAL-R1 model as a physical data engine (2014–2022), we decomposed the thermal sequences into low-frequency seasonal trends—captured by an LSTM-Transformer—and high-frequency meteorological fluctuations. Crucially, an XGBoost model explicitly fitted the nonlinear residuals induced by the density gradient that reduces vertical heat exchange, correcting the systematic overestimation. This targeted correction reduced unrealistic downward heat propagation, reducing the mean absolute error in the metalimnion by 78% during validation. Under an idealized + 20% air-temperature stress test, the framework revealed an asymmetric thermodynamic response: surface warming (+0.26 °C) sharpened the epilimnetic density gradient, reinforcing the density gradient. Consequently, the metalimnion (−0.34 °C) and hypolimnion (−0.25 °C) exhibited reduced warming (i.e., relative cooling compared with the no-warming baseline scenario), expanding the surface–bottom temperature gradient by 1.1 °C. The proposed surrogate framework alleviates systematic stratification-related prediction biases commonly observed in conventional deep learning models, while offering a computationally efficient pathway for quantifying stratification intensity and thermodynamic divergence under extreme meteorological forcing, thereby strengthening the role of AI in climate-resilient reservoir management.
Soil erosion begins with the impact of raindrops, which create a splashing effect. The rainfall erosivity factor (R-factor) quantifies this impact and is used to estimate soil erosion in the Revised Universal Soil Loss Equation (RUSLE) model. Gridded daily rainfall and temperature data from the Indian Meteorological Department (IMD) were collected. The R-factor was computed for the period 1989–2024. A preliminary study was carried out at a selected grid point to develop a model using various machine learning techniques. Daily rainfall, maximum temperature, and minimum temperature were chosen as input features, while the R-factor was used as the prediction target. A 5-fold cross-validation approach was applied to assess model performance. The evaluation metrics indicated that the Multi-Layer Perceptron (MLP) achieved the best predictive accuracy. Predicted rainfall and temperature data were then obtained from the Bias-Corrected Climate Projections of CMIP6 Global Circulation Models (GCMs) for South Asia. To identify the most suitable GCM, rainfall data from 2015 to 2024 were compared between IMD observations and GCM outputs using Taylor diagram. The selected GCM was then used to provide future rainfall and temperature data, which were applied to predict the R-factor for upcoming years. The projected changes in rainfall erosivity provide important insights for soil conservation planning and sustainable land management under future climate change scenarios.
Reliable drought monitoring based on standardized indices requires long, gap-free hydrological datasets. Reanalysis products such as ERA5-Land are increasingly used for this purpose due to their spatial coverage and temporal continuity; however, their biases in temperature and precipitation may propagate into drought indices such as SPI and SPEI. This study assesses the suitability of ERA5-Land for drought monitoring over Sicily—an ideal testbed thanks to its complex terrain, recurrent droughts, and availability of long-term observations. ERA5-Land temperature and precipitation were compared with a gridded observational dataset (1951–2013) using Nash–Sutcliffe Efficiency (NSE), RMSE, and correlation.ERA5-Land reproduced temperature well (correlation >0.9, NSE >0.8, RMSE <3 °C), whereas precipitation showed weaker skill (correlation 0.6–0.8, NSE often <0.5, RMSE 20–80 mm). Biases propagated into drought indices: long-term SPI and SPEI (24–48 months) were reasonably consistent with observations (correlation 0.75–0.9), while short-term indices exhibited limited skill, including negative NSE values.Results highlight both the potential and limitations of ERA5-Land for drought assessment in Mediterranean environments. They underline the need for spatial and temporal bias correction when using reanalysis-driven drought indices for agricultural planning and water resource management in regions with complex climatic and topographic settings.
This study applied Multifractal Detrended Fluctuation Analysis (MFDFA) to investigate the scaling behaviour of streamflow, precipitation, and potential evapotranspiration (PET) in nine Ugandan catchments during 1990–2013. We estimated, among others, the Hurst exponent (H), mass exponent τ(q), multifractal spectra f(h), and multifractal width (Δh). All hydroclimatic variables exhibited H values greater than 0.5 and were significantly (p<0.05) different from white noise, confirming long-range dependence (LRD). The order of persistence from the lowest to the highest was for PET, precipitation, and streamflow. Strong multifractality was observed in streamflow (1.04≤Δh≤4.81) and precipitation (1.4≤Δh≤3.9), whereas PET exhibited weak multifractality (0.12≤Δh≤0.32). The f(h) were predominantly right-skewed, suggesting a stronger contribution from small fluctuations. The τ(q) further confirmed multifractality driven by both small- and large-magnitude variations. The results indicate that streamflow is the most persistent hydroclimatic variable, while precipitation behaves as a climate-synchronized but weakly integrated process, with PET playing an intermediate role largely associated with short-term moisture modulation. Streamflow, through precipitation–runoff generation, partially inherits its scaling behaviour more from precipitation than PET. These results underscore the multiscale complexity of hydrological processes across Uganda and emphasize the significance of incorporating multifractality and LRD in hydrological modelling and water resources planning.
Precipitation is a primary component of the hydrological cycle. Depending on the climate system, it significantly changes over time and space. Such a spatiotemporal change escalates various natural extremes, primarily floods and droughts, resulting in low-to-high-scale disastrous effects at basin-scale, making it essential to understand how precipitation varies spatially within basins. Particularly, the spatial behavior of the monthly total precipitation needs to be known as a main requirement for most of the hydrologic practices. Spatial maximum and minimums mark the limits of the precipitation changes within basin, therefore provide a realistic concise image for the variation of monthly precipitations in basins. This study suggests global and basin-scale statistical models to identify spatial variations of monthly total precipitations within world’s major basins. Basin-scale models are linear regression equations described by basin-specific periodic functions with 12-month periods. Global models rely on the statistical similarity of those periodic functions of different basins. Both the global and basin-scale models successfully estimate the spatial maximum and spatial minimum statistics of monthly total precipitations over time. The variance explanations of the basin-scale models are 97% and 89% for the spatial maximum and minimum precipitations, while the global models suggest 94% and 82% variance explanations, respectfully.
Identifying the relationship between discharge and water level is challenging in lowland streams because seasonal growth of submerged aquatic vegetation causes complex, time-varying channel roughness. We propose a novel differentiable modelling framework integrating (1) an LSTM model that predicts the dynamics of the channel roughness, with (2) a hydraulic river model providing physics-based routing. The integrated framework is trained to learn time-varying channel roughness based on historical observations of meteorological variables, river water levels, and discharge. Once trained, we demonstrate that the framework can be applied to independent periods for deriving discharge estimates using backpropagation with only observed water levels and meteorological variables as inputs, thereby acting as a time-varying rating curve. Synthetic experiments confirm accurate recovery of (a) time-varying roughness and (b) discharges imposed during hydraulic model simulations. Applied to a real Danish river, the framework achieves an NSE for water levels of 0.76, compared to 0.19 for a constant-roughness benchmark. Incorporating temporal roughness dynamics reduces the Mean Absolute Relative Error (MARE) in discharge estimation from 21.4% to 10.5%. This work takes a significant step towards setting up hydraulic models directly from primary data, bypassing the explicit construction of complex, time-varying rating curves.
