A review of current trends, challenges, and future perspectives in machine learning applications to water resources in Nepal

Abstract

Nepal faces frequent hazards like floods, droughts, sedimentation, glacial melting, and landslides that threaten lives and infrastructure, making effective risk assessment and sustainable management essential. Recently machine learning (ML) approaches have gained popularity worldwide in the water resources sector as it can capture complex relationship between predictors and predictand variables. We reviewed 34 papers published from 2010 to 2025 to identify the current status, challenges and future perspectives in the applications of ML in flood forecasting, streamflow prediction, water quality assessment, groundwater mapping, and landslide susceptibility. The review indicates that the application of ML in water resources in Nepal is on increasing trends. Our study shows that both classical ML and deep learning models consistently achieve higher accuracy than traditional empirical and processed based models. In addition, deep learning and hybrid models outperformed classical ML models in solving water resources problems. Key challenges include sparse observed data, limited computational resources, and insufficient local expertise to train and validate the ML models. Future research should focus on hybrid models and transfer learning to further enhance prediction accuracy and support sustainable water management in Nepal. By providing insights into the current status and future opportunities of ML applications in water resources, this study serves as a valuable resource for novice researchers and practitioners aiming to contribute to this evolving field.