Explore the vast range of titles available.

Karst springs are essential drinking water resources, however, modeling them poses challenges due to complex subsurface flow processes. Deep learning models can capture complex relationships due to their ability to learn non‐linear patterns. This study evaluates the performance of the Transformer in forecasting spring discharges for up to 4 days. We compare it to the Long Short‐Term Memory (LSTM) Neural Network and a common baseline model on a well‐studied Austrian karst spring (LKAS2) with an extensive hourly database. We evaluated the models for two further karst springs with diverse discharge characteristics for comparing the performances based on four metrics. In the discharge‐based scenario, the Transformer performed significantly better than the LSTM for the spring with the longest response times (9% mean difference across metrics), while it performed poorer for the spring with the shortest response time (4% difference). Moreover, the Transformer better predicted the shape of the discharge during snowmelt. Both models performed well across all lead times and springs with 0.64–0.92 for the Nash–Sutcliffe efficiency and 10.8%–28.7% for the symmetric mean absolute percentage error for the LKAS2 spring. The temporal information, rainfall and electrical conductivity were the controlling input variables for the non‐discharge based scenario. The uncertainty analysis revealed that the prediction intervals are smallest in winter and autumn and highest during snowmelt. Our results thus suggest that the Transformer is a promising model to support the drinking water ion management, and can have advantages due to its attention mechanism particularly for longer response times. Key Points The Transformer architecture was applied in karst hydrology for the first time, showing high performance for discharge forecasting Monte Carlo dropout revealed that the prediction intervals are smallest and cover the measured discharges best in winter and autumn The high temporal resolution of the input data sets improved the forecasting performance

Accurate river discharge forecasting is essential for effective water resource management, particularly in regions prone to monsoonal variability and extreme weather events. This study presents an interpretable deep learning framework for daily river discharge forecasting in the Subarnarekha river basin (SRB), integrating Kolmogorov Arnold networks (KAN) with Shapley additive exPlanations (SHAP). Leveraging hydroclimatic inputs from five coupled model intercomparison project phase 6 (CMIP6) general circulation models (GCM) under the high emissions shared socioeconomic pathway (SSP585) scenario, the model was trained and evaluated across four active gauging stations: Muri, Adityapur, Jamshedpur, and Ghatsila covering the period 1980 to 2022, with projections extending to 2100. The main findings of this study are (1) KAN demonstrated high predictive performance with root mean squared error (RMSE) values ranging from 42.7 to 58.3 m 3 /s, Nash–Sutcliffe efficiency (NSE) between 0.80 and 0.87, mean absolute error (MAE) between 28.9 to 52.7 and R 2 values between 0.84 and 0.90 across stations. (2) SHAP based feature contribution analysis identified Relative humidity (hurs) , specific humidity (huss) , and temperature (tas) as key predictors, while ( pr) showed limited contribution due to spatial inherent inconsistencies in GCM precipitation data. (3) The bootstrapped SHAP distributions highlighted substantial variability in feature importance, particularly for humidity variables, revealing station specific uncertainty patterns in model interpretation. (4) The KAN framework results indicate strong temporal alignment and physical realism, confirming KAN’s robustness in capturing seasonal discharge dynamics and extreme flow events under monsoon influence environments. (5) In this study KAN with SHAP (SHapley additive exPlanations) is implemented for hydrological modeling under monsoon-influenced and data-limited regions such as SRB, offering improved accuracy, functional precision and efficiency compared to traditional models. The explainability offered by SHAP confirms informed water resource planning. This novel framework presents a reproducible and climate-resilient decision support tool, particularly suitable for monsoon-influenced, data-limited basins susceptible to extreme hydroclimatic events.

The Himalayan rivers are prone to frequent floods and pose serious risks to human lives and infrastructure. Accurate discharge prediction is crucial for effective flood mitigation and sustainable water resource management. This study focuses on the Sindh River, a vital water source in the Kashmir valley, supporting hydropower and irrigation. Advanced artificial intelligence techniques were applied to analyze 40 years of historical discharge data to address its complex hydrological dynamics. The study evaluated various machine learning models, including K-Nearest Neighbors, Support Vector Regression, Gradient Boosting, Extreme Gradient Boosting, Random Forest (RF), Artificial Neural Network (ANN), and Seasonal Autoregressive Integrated Moving Average (SARIMA). A hybrid RF-SARIMA model was also developed to improve prediction accuracy. The dataset was split into 80% for training and 20% for testing. Model performance was assessed using statistical metrics such as coefficient of determination ( R ²), mean squared error, mean absolute error, and root mean squared error, along with visual tools like box plots, scatter plots, Taylor diagrams, and time series analyses. Results revealed that RF, SARIMA, and ANN performed well among standalone models. However, the hybrid RF-SARIMA model delivered the best results, with an R ² of 0.88 and a correlation coefficient above 0.9 for monthly discharge predictions. This study highlights the hybrid model’s potential to enhance discharge forecasting for the Sindh River, providing valuable insights for flood management and sustainable water planning in the Himalayan regions.

Accurate snowmelt runoff prediction is critical for water resource management in mountainous regions where seasonal snowpack constitutes the dominant water supply. This study demonstrates operational application of the degree-day-based Snowmelt Runoff Model (SRM) integrated with Geographic Information Systems (GIS) and multi-platform remote sensing for discharge forecasting in the Kirkgoze Basin (242.7 km2, 1823–3140 m elevation), Eastern Anatolia, Turkey. Three automatic weather stations spanning 872 m elevation gradient provided meteorological forcing, while MODIS MOD10A2 8-day composite products supplied operational snow cover observations validated against Landsat-5/7 (30 m resolution, 87.3% agreement, Kappa = 0.73) and synthetic aperture radar imagery (RADARSAT-1 C-band, ALOS-PALSAR L-band). Uncalibrated model performance was modest (R2 = 0.384, volumetric difference = 29.78%), demonstrating necessity of site-specific calibration. Systematic adjustment of snowmelt and rainfall runoff coefficients yielded excellent calibrated performance for 2009 melt season: R2 = 0.8606, correlation coefficient R = 0.927, Nash–Sutcliffe efficiency = 0.854, and volumetric difference = 3.35%. Enhanced temperature lapse rate (0.75 °C/100 m vs. standard 0.65 °C/100 m) reflected severe continental climate. Multiple linear regression analysis identified temperature, snow-covered area, snow water equivalent, and calibrated runoff coefficients as significant discharge predictors (R2 = 0.881). Results confirm SRM’s operational feasibility for seasonal forecasting and flood warning in data-scarce snow-dominated basins, with modest requirements (daily temperature, precipitation, and satellite snow cover) aligning with operational monitoring capabilities. The methodology provides a transferable framework for regional water resource management in climatically vulnerable mountain environments where snowmelt supports agriculture, hydropower, and municipal supply.

Accurate short-lead river discharge forecasting is critical for effective flood risk mitigation in small mountainous basins, where rapid hydrological responses pose significant challenges. In this study, we focus on the Fuhu Stream in Emeishan City, China, and utilize high-resolution 5-min time series of upstream precipitation, stage, and discharge to predict downstream flow. We benchmark three data-driven models—SARIMAX, XGBoost, and LSTM—using a dataset spanning from 7 June 2024 to 25 October 2024. The data were split chronologically, with observations from October 2024 reserved exclusively for testing to ensure rigorous out-of-sample evaluation. Lagged correlation analysis was employed to estimate travel times from upstream to the basin outlet and to inform the selection of time-lagged input features for model training. Results during the test period demonstrate that the LSTM model significantly outperformed both XGBoost and SARIMAX across all regression metrics: it achieved the highest coefficient of determination (R2 = 0.994) and the lowest prediction errors (RMSE = 0.016, MAE = 0.011). XGBoost exhibited moderate performance, while SARIMAX showed a tendency toward mean reversion and failed to capture low-flow variability. Accuracy evaluation reveals that LSTM accurately reproduced both baseflow conditions and multiple flood peaks, whereas XGBoost and SARIMAX failed. These results highlight the advantage of sequence-learning architectures in modeling nonlinear hydrological propagation and memory effects in short-term discharge dynamics. Feature importance analysis indicates that the LSTM model was highly effective for real-time forecasting and that the WSQ/LY sites served as critical monitoring nodes for achieving reliable predictions. This research contributes to the operationalization of early warning systems and provides support for decision-making regarding downstream flood disaster prevention.

Discharge forecasting is a key component for early warning systems and extremely useful for decision makers. Forecasting models require accurate rainfall estimations of high spatial resolution and other geomorphological characteristics of the catchment, which are rarely available in remote mountain regions such as the Andean highlands. While radar data is available in some mountain areas, the absence of a well distributed rain gauge network makes it hard to obtain accurate rainfall maps. Thus, this study explored a Random Forest model and its ability to leverage native radar data (i.e., reflectivity) by providing a simplified but efficient discharge forecasting model for a representative mountain catchment in the southern Andes of Ecuador. This model was compared with another that used as input derived radar rainfall (i.e., rainfall depth), obtained after the transformation from reflectivity to rainfall rate by using a local Z-R relation and a rain gauge-based bias adjustment. In addition, the influence of a soil moisture proxy was evaluated. Radar and runoff data from April 2015 to June 2017 were used. Results showed that (i) model performance was similar by using either native or derived radar data as inputs (0.66 < NSE < 0.75; 0.72 < KGE < 0.78). Thus, exhaustive pre-processing for obtaining radar rainfall estimates can be avoided for discharge forecasting. (ii) Soil moisture representation as input of the model did not significantly improve model performance (i.e., NSE increased from 0.66 to 0.68). Finally, this native radar data-based model constitutes a promising alternative for discharge forecasting in remote mountain regions where ground monitoring is scarce and hardly available.

In dam operations, sudden discharges during extreme rainfall events can pose severe flood risks to downstream communities. This study developed a dam discharge-based river water level forecasting model using a data-driven deep learning approach, long short-term memory (LSTM). To enhance predictive performance, physics-based HEC-RAS simulation outputs, including extreme events, were incorporated as additional inputs. The Seomjin River Basin in South Korea, which recently experienced severe flooding, was selected as the study area. Hydrological data from 2010 to 2023 were utilized, with 2023 reserved for model testing. Forecasts were generated for four lead times (3, 6, 12, and 24 h), consistent with the operational flood forecasting system of the Ministry of Environment, South Korea. Using only observed data, the model achieved high accuracy at upstream sites, such as Imsil-gun (Iljung-ri, R2 = 0.92, RMSE = 0.27 m) and Gokseong (Geumgok Bridge, R2 = 0.91, RMSE = 0.35 m), for a 6-h lead time. However, performance was lower at Gurye-gun (Songjeong-ri, R2 = 0.72, RMSE = 1.48 m) due to the complex influence of two dams. Incorporating enhanced inputs significantly improved predictions at Gurye-gun (R2 = 0.91, RMSE = 1.17 m at 3 h). Overall, models using only observed data performed better at upstream sites, while enhanced inputs were more effective in downstream or multi-dam regions. The 6-h lead time yielded the highest overall accuracy, highlighting the potential of this approach to improve real-time dam operations and flood risk management.

The Bekhma Dam was once proposed as a massive solution to Iraq’s growing water challenges. But after decades of political delays, environmental concerns, and engineering complications, it remains incomplete. Instead of waiting for a project of such scale, this study explores more flexible and sustainable options. Using historical discharge data from the Eski Kalak Gauge Station, machine learning (XGBoost) was applied to project future water availability under changing climate conditions. The model showed a clear downward trend in discharge, indicating a dominant dry pattern in the coming decades and reinforcing the urgency of adaptive planning. In response, four upstream dam sites were identified through GIS and DEM analysis and evaluated based on reservoir capacity, topography, and regional fit. Bekhma itself was also reconsidered this time as a scaled-down version supported by those smaller alternatives. The idea is simple: by capturing floods and sediment in the upper basin, we can reduce Bekhma’s dead storage requirement and make those construction more realistic. This integrated approach could deliver the key benefits water security, storage, flood mitigation while reducing cost, social impact, and structural risk. Importantly, the study clarifies that these alternatives are not a complete replacement for Bekhma, but rather a practical and scalable support system to strengthen resilience. The study argues that combining predictive hydrology with decentralized infrastructure offers a smarter path forward for the Greater Zab Basin. Based on these findings, the study recommends adopting the proposed upstream alternatives alongside a reduced Bekhma structure to create a flexible, climate-resilient water management system. لقد طرح سد بيخمة في السابق كحل ضخم لمواجهة التحديات المائية المتزايدة في العراق، لكنه ظل غير مكتمل لعدة عقود بسبب التأخيرات السياسية والمعوقات البيئية والتعقيدات الهندسية. وبدلًا من انتظار مشروع بهذا الحجم، تستكشف هذه الدراسة خيارات أكثر مرونة واستدامة من خلال استخدام بيانات تصريف محطة أسكي كلك وتطبيق تقنية تعلّم الآلة (XGBoost) للتنبؤ بتوفر المياه مستقبلًا في ظل التغيرات المناخية. أظهرت نتائج النموذج اتجاهًا واضحًا نحو انخفاض التصريف، وأكدت هيمنة الظروف الجافة على المدى المتوسط، مما يعزز الحاجة إلى تخطيط تكيفي سريع. واستجابة لذلك، تم تحديد أربعة مواقع للسدود صغيرة في أعالي الحوض باستخدام نظم المعلومات الجغرافية (GIS) ونماذج الارتفاع الرقمية (DEM)، وتقييمها من حيث السعة التخزينية والطبوغرافيا والملاءمة الإقليمية. كما أُعيد النظر في سد بيخمة نفسه، ولكن هذه المرة بنسخة منخفضة تعتمد على بدائل صغيرة. الفكرة ببساطة: من خلال حجز الفيضانات والرسوبيات في أعلى الحوض، يمكن تقليل حجم التخزين الميت في بيخمة، وجعل بنائه أكثر واقعية. هذا النهج المتكامل يمكن أن يوفر الفوائد الرئيسية نفسها لأمن المياه والتقليل من تكاليف التخزين، مع تقليل الكلفة والمخاطر الاجتماعية والإنشائية. وتوضح الدراسة أن هذه البدائل لا تهدف فقط إلى استبدال سد بيخمة، بل تمثل نظامًا داعمًا أكثر مرونة وكفاءة، وتعدل الدراسة بأن الجمع بين التنبؤات الهيدرولوجية والبنية اللامركزية يمثل مسارًا أكثر ذكاءً لمستقبل حوض الزاب الأعلى. وبناءً على ذلك، توصي الدراسة باعتماد استراتيجية سدود صغيرة متعددة كبديل عملي مستدام لمواجهة التحديات المائية في المنطقة.

One of the primary goals of watershed management is to proactively monitor and forecast flood water levels to provide early warning for timely evacuation plans and save lives. One of the most economical ways to accomplish this objective is to use remotely-sensed satellite signals. Previous studies have indicated that an Advanced Microwave Scanning Radiometer (AMSR) sensor can be used for river water level monitoring combined with a few in-situ hydrometric gauges for the ground-truth data collection. However, space-based signals are influnced by many error-inducing natural factors, such as dust and cloud cover. Hence, a hybrid method is proposed, which comprises of a multi-objective particle swarm optimization model, a decision tree classification algorithm, the Hotelling’s \\(T^{2}\\) outlier detection, and a regression model to identify and replace inaccurate space-based signals. This complex hybrid method will be referred to, in this study, with the acronym (OCOR). In the first phase of this hybrid method, the outlier signals are detected and eliminated from the dataset, and in the second phase, the eliminated signals along with signals lost due to satellite technical problems are estimated by ground-truth data calibration using in situ hydrometric stations. The two case studies of the White and Willamette Rivers demonstrate the performance of OCOR in practical situations.

To achieve a robust data-driven flood forecasting model, features such as fast learning, appropriate training using insufficient data and reliable prediction of flood flows are of essential importance. These models also have notable vulnerabilities such as decreased accuracy in forecasting peak discharges, challenging simulation of rainy events, performance deterioration in confronting with inadequate training data and weakness due to reduced number of training epochs. In this paper, the supervised brain emotional learning (SBEL) neural network has been used in daily rainfall–runoff modeling of the Dez Dam watershed in the southwest of Iran, as its first application in the field of hydrology. SBEL is a supervised neurocomputing model inspired by the limbic system in the mammalian brain. To create the right responses, the SBEL models the processing of emotional stimuli and the inhibitory mechanism of incorrect responses to stimuli in the emotional brain. The performance of SBEL was compared to the well-known multilayer perceptron (MLP) with 15–8–1 architecture, through different perspectives. The SBEL outperforms MLP in peak flow prediction, limiting the training epochs, reducing the training samples and predictions of rainy events, while improving the mean relative error by 21%, 59.4%, 74.5% and 14.4%, respectively. By placing reduced training data in dry, normal and wet periods, it has been observed that SBEL has more generalization ability in all flow regimes. Overall, the use of this type of emotional intelligence-based model can be of particular interest in developing reliable early flood forecasting and warning systems.