Data-Driven modelling and control of ungauged Watershed: A Case-Study of Namal Lake

Abstract

This thesis considers the effective management of reservoirs in ungauged watersheds at mon soon margins prone to extreme hydrological events of droughts and flash floods. Prioritizing modeling and management of vulnerable small watersheds, using the principles of Integrated Water Resources Management (IWRM), is essential to protect water resources and rural liveli hoods from climate change impacts. Migrating boundaries of the monsoonal rain belt and data scarcity make accurate streamflow predictions a challenging problem, which is crucial for op timal reservoir operations using Model Predictive Control (MPC) like frameworks. The study proposes an Adaptive Scenario-based (AS) MPC framework that is more robust against unreli able forecasts, extended to an Adaptive Scenario Tree-Based (ASTB) MPC to utilize Ensemble Streamflow Forecasts (ESFs). By integrating the ICON Numerical Weather Predictor (NWP) model with a calibrated SWAT Hydrological Model, a 168-hour (7-day) streamflow forecasting system is developed for the Namal watershed, the study area under consideration. Using the streamflow forecasting system, we test the performance of AS-MPC through hindcast simula tions of extreme historical events. Comparison with Tree-Based MPC and a Perfect Forecast MPC highlight the suitability of AS-MPC for real-time deployment in practical scenarios, bal ancing computational efficiency and performance in uncertain hydrological conditions. No on ground experiments have been conducted yet due to governmental restrictions. Moving forward, the work in this thesis will motivate authorities to implement this solution on-site.