Towards future flood hazard assessment under climate and land use change in the Karnali River Basin, Nepal : an integrated LSTM neural network and HEC-RAS modeling approach | |
| Author | Lakhe, Hanik |
| Call Number | AIT Thesis no.WM-26-11 |
| Subject(s) | Flood forecasting--Karnali River Basin--Nepal Climatic changes--Karnali River Basin--Nepal Long-term memory Short-term memory Hydraulic models |
| Note | A thesis submitted in partial fulfillment of the requirements for the degree of Master of Science in Water Engineering and Management |
| Publisher | Asian Institute of Technology |
| Abstract | Flood hazards in mountainous river basins are increasingly exacerbated by the combined impacts of climate change and land-use change, particularly in data-scarce regions such as the Karnali River Basin (KRB), Nepal. This study developed an integrated modelling framework combining Long Short-Term Memory (LSTM) neural networks, the Dyna-CLUE land-use change model, and two-dimensional HEC-RAS hydraulic simulations to evaluate present and future flood hazards under climate and land use scenarios. Climate projections indicated a pronounced intensification of climatic conditions across the basin. Ensemble median precipitation is projected to increase by approximately 16.1% under SSP2-4.5 and 38.9% under SSP5-8.5 by the late century, accompanied by increases of up to 4.17°C in maximum temperature and 4.87°C in minimum temperature. Land-use simulations demonstrated satisfactory model performance, with overall accuracies of approximately 76– 77% and a Kappa coefficient of about 0.69. Scenario projections suggested continued forest expansion under the Low Urbanization pathway, reaching 48.6% of the basin area, whereas built-up areas are projected to increase to 9.2% under the Medium Urbanization pathway by 2100. The LSTM model successfully reproduced daily discharge dynamics, achieving strong predictive performance (training NSE = 0.952, validation NSE = 0.854, and testing NSE = 0.886). Future simulations revealed substantial intensification of extreme streamflow, with far future peak discharge projected to increase by up to 112% under SSP5-8.5 with Medium Urbanization, while the 99th percentile discharge may increase by approximately 110%. Flood frequency analysis further indicated significant amplification of design floods, with projected 50-year discharges exceeding 20,000 m³/s and 100-year discharges surpassing 23,000 m³/s. Hydraulic simulations demonstrated a considerable expansion of flood hazard zones, with very high hazard areas projected to increase from approximately 53 km² during the historical period to nearly 139 km² under the 100-year return-period scenario in the late century. Overall, climate change emerged as the dominant driver of increasing flood hazard in the Karnali River Basin, while urbanization further amplified flood magnitude and spatial extent. The integrated AI– hydrodynamic modelling framework provides a robust and scalable approach for assessing climate-driven flood risks in data-constrained mountainous basins and offers valuable insights for flood management and disaster risk reduction. |
| Year | 2026 |
| Type | Thesis |
| School | Faculty of Civil and Environmental Engineering (2026) |
| Department | Other Field of Studies (No Department) |
| Academic Program/FoS | Water Engineering and Management (WEM) |
| Chairperson(s) | Shrestha, Sangam; |
| Examination Committee(s) | Sarawut Ninsawat;Sundaram, S. Mohana;Natthachet Tangdamrongsub;Khadka, Dibesh; |
| Scholarship Donor(s) | Global Water & Sanitation Center (GWSC);AIT Scholarship; |
| Degree | Thesis (M. Sc.) - Asian Institute of Technology, 2026 |
 | |