ABSTRACT

Summary Two and three dimensional physically-based models, data-driven models and a hybrid model combining the two categories of model (physically based and data driven) are applied separately in the case study which forms the focus of this Ph.D. research. The present study proposes a framework for hybrid modelling. The ecological water quality variables for deep and long reservoirs, like the Yongdam reservoir, can be successfully simulated using a two-dimensional model (in the vertical), but the modelling of eutrophication still has room for improvement. A three-dimensional model shows better simulation results for reservoir hydrodynamics than the two dimensional model despite it being developed for the ocean. The three-dimensional model as a management tool has a couple of short-comings in the amount of data required for water quality modelling and computing time. Data-driven models give simulation results based on instances; that is, measurements made at specific times and places. If the instance is not in the form of a regular time series, the models seek only a similar case from the historical data set. Actually, the time series database, representing the environmental processes in the reservoir, is difficult to develop. We can acquire only biweekly and monthly data because of budget and analysis limitations. As an alternative, we can use a physically-based model to provide time series simulation results at the time interval specified by modeler for time series data-driven modelling. As the eutrophication process has many unknown mechanisms, the Chl-a concentration may be predicted better with a data-driven model. The hybrid model and the corresponding prediction results of Chl-a concentrations show a better fit with observed data than the physically-based model alone. The results of the hybrid model can be used for a short-term forecasting of water quality in the reservoir and for long term prediction of the reservoir response to remedial activities affecting reservoir water quality. The hybrid model is subsequently implemented in a decision support system for water quality management in the Yongdam reservoir.