ABSTRACT Reliable river water level forecasting is crucial for flood management and sustainable water resource planning in climate-vulnerable deltaic regions like Bangladesh, where conventional hydrological models face challenges due to data scarcity and complex monsoon-driven dynamics. This study evaluates five machine learning and deep learning models – linear regression, random forest (RF), XGBoost, light gradient boosting machine (LGBM), and long short-term memory (LSTM) – to predict daily water levels in the Old Brahmaputra River under four scenarios. Using 26 years (1999–2024) of hydrological and meteorological data, we optimized hyperparameters via grid search and 5-fold cross-validation, incorporating lagged variables (1–5 days) for temporal dependencies. Performance was assessed using six metrics, with principal component analysis for model ranking. Results revealed exceptional accuracy in the water level-only scenario (RF: MAE 0.1445 m, R2 0.9916, NSE 0.9916). Climate-driven scenarios demonstrated LSTM's superiority, achieving R2 of 0.8145 (rainfall-temperature), 0.8064 (rainfall-only), while temperature-only scenarios showed limited predictive capability (R2 0.5346). Spatial transferability assessment at Sarishabari station validated robust cross-station performance without recalibration (LGBM: MAE 0.1205 m, R2 0.9917 for water level scenario). The study provides operational frameworks for climate-driven forecasting in data-scarce deltaic environments, highlighting LSTM's exceptional capability for ungauged catchments.

ABSTRACT Anthropogenic activities, particularly in regions with high population growth, have profoundly impacted land use and land cover (LULC), driving significant changes that affect global hydrological processes, watershed soil and water management, and streamflow dynamics. This study assessed LULC dynamics and predicted future trends in the Suha watershed, Upper Blue Nile, Ethiopia. LULC maps for 1992, 2007, and 2022 were developed using Landsat data from USGS Earth Explorer and maximum likelihood supervised classification. Future LULC changes for 2037 and 2052 were predicted using the CA-Markov model. Results revealed substantial changes, with agricultural land expanding from 77.92% in 1992 to 87.42% in 2022, while shrubland, forestland, and grassland declined by 7.3, 4.76, and 0.37%, respectively. Projections indicate further reductions in forest cover, from 2.48% in 2022 to 0.41% by 2052, alongside an increase in built-up areas to 8.67% by 2052. These findings highlight rapid LULC transformations, with significant conversions of forest, shrubland, and grassland into agricultural land expected. This study provides critical insights to support sustainable land-use planning and management, guiding decision-makers in mitigating ecological and environmental impacts.

ABSTRACT This study quantifies hydropower generation under historical and future climatic conditions, to examine the impact of climate change on hydropower projects, focusing on Kankai River basin. We used projected temperature and precipitation data from the six GCMs under CMIP6 scenarios after removing the biases through linear scaling which were forced into the well calibrated and validated SWAT model to obtain the streamflow projections based on which future hydropower generation was analyzed. Future projections reveal an increase in annual precipitation up to 42.11% in SSP 585 and both maximum and minimum temperatures rising up to 29.78% in SSP 585 by the end of the century compared to baseline. The streamflow illustrates the increasing trend marking the typical South Asian climate of monsoon peak flow with a substantial rise in the far-future of SSP 585. The analysis of power generation of the three distinct hydropower projects in the basin showed a decrease in average annual energy generation in the near future while an expected increase in the mid-future and far-future with a more pronounced increase in the far-future of SSP 585. The inter-annual fluctuations raise challenges in the operations and energy supply-demand balance from run-of-the-river projects necessitating strategic planning sustainable energy management.

ABSTRACT Flooding in Nepal's Terai region poses significant threats to public infrastructure, yet the exposure of facilities to flood hazards remains underexplored. Using a two-dimensional hydrodynamic model, this study simulated flood extents for 20-, 50-, 100-year return periods (YRP) and assessed the exposure of buildings and transportation networks in Mohana-Khutiya River Basin, Nepal. Chure Rural Municipality and Godawari Municipality experienced the highest flood depths of 17.6 and 17.5 m, respectively, for all YRPs. Building exposure increased by 40% in 50 YRP and by 43% in 100 YRP, compared to 20 YRP. Dhangadhi Sub-metropolitan City was identified as the most exposed in terms of building exposure. Health facilities' exposure was limited, with two buildings affected in 20 YRP, and three affected in 50 and 100 YRPs. Dhangadhi's exposure to school increased across all return periods. Godawari Municipality had the highest bridge exposure. 223 km of roads were exposed in 20 YRP, rising to 296 km in 100 YRP. Critical infrastructure in Dhangadhi Sub-metropolitan City, Godawari, and Krishnapur Municipalities was increasingly exposed to higher return periods. The findings underscore the importance of integrating flood hazard assessments into urban planning for effective disaster risk management.

ABSTRACT This study evaluates various water extraction indices to identify surface features using Sentinel-2 satellite images. Eight study sites were selected with three indices employed in this study, such as the water ratio index (WRI), the modified normalized difference water index (MNDWI), and the normalized difference water index (NDWI). The results of the accuracy assessment revealed that the producer accuracy (PA) for the NDWI ranged from 81.11 to 94.56%, that for the MNDWI ranged from 85.3 to 96.7%, and that for the WRI ranged from 96.98 to 97.98%. The user accuracies for the NDWI, MNDWI, and WRI results are 68.1–81.6%, 74.81–86.14%, and 80.33–88.78%, respectively. This study reveals that the WRI has greater PA (97.98%), user accuracy (88.78%), and kappa statistic (0.88) than the MNDWI and NDWI. The WRI is particularly suitable for extracting surface bodies surrounded by bare lands with minimal vegetation cover. Conversely, the NDWI is more applicable in areas with moderate vegetation and shadows, such as agricultural regions or wetlands. The MNDWI is suitable to avoid confusion water bodies with built-up areas and used to extract water bodies with various land cover. Therefore, establishing optimal thresholds is essential for policymakers and planners to effectively monitor water surface features and changes in water sources.