Developing Efficient Model Surrogates For Water Resources And Subsurface Contaminant Management
Supervisor: Dr Domenico Bau
Numerical models are important tools for understanding the movement of fluids and the fate of contaminants in the subsurface. However, lack of data often limit our ability to construct models that are reliable for planning and management of groundwater resources. To address these limitations, these models can be applied under conditions of uncertainty with respect to geological settings and heterogeneous parameters spatial distributions. These application typically require repeated model runs, as in Monte Carlo simulations, that are often computationally expensive if not prohibitive. The goal of this project is to investigate the use of model emulators (surrogates), that is, reduced versions of the full numerical model that require a fraction of its computational cost. Once developed, surrogates can be used to narrow down uncertainties affecting the full model response and derive better calibrated models in relation to the limited knowledge available.
This project is NOT FUNDED, although Departmental/University scholarships are available for applicants who can demonstrate strong evidence of research potential.
Data-driven and Model-driven surrogate models Machine learning techniques and deep learning frameworks Subsurface Flow and Transport Modelling