Short description and objective of the project:
Tidal sandbanks are large-scale bedforms found in sandy coastal shelf seas of tens of meters in height, tens of kilometres long and crest-to-crest spacing of around 5 kilometres. They are dynamic, meaning they slowly move over time, interacting with the tides that shape them. They provide ecologically important habitats, and understanding their formation mechanisms, dynamics and shape is important for sand extraction, windfarms and pipelines.
This MSc project will be part of the BANX project, where we study tidal sandbank dynamics and the impact of sand extraction in sediment-scarce environments, like the Belgian Continental Shelf. As part of this project, a process-based numerical model has been developed, which describes the cross-sectional dynamics of tidal sandbanks over long timescales. This shows a highly complex interaction between the input space (the environmental conditions: location, water depth, tide and sediment characteristics) and the equilibrium sandbank shape. Usually, a sensitivity analysis is performed to understand the relationship between input and output. However, there is a need for a quick toolbox that models the equilibrium profiles of sandbanks for certain environmental conditions, and their response to sand extraction. A single run takes around an hour. Therefore, surrogate modelling, i.e. the use of a neural network between input and output space, can help to quickly generate insight into these systems.
In this MSc project, we aim to develop a machine learning model / neural network that helps us understand how tidal sandbanks are related to environmental conditions. In the first part of the project, you will get familiar with the relevant literature on the topic and the process-based numerical model that is developed. The main research will focus on the development of a useful surrogate model, the lessons we can learn from it, and how it can be used by practice. Furthermore, the options are wide, and can include the development of a user interface, expanding the underlying numerical model, using the neural network to obtain mechanistic insight, and more. |
References:
Geerts, S. J., Roos, P. C., & Hulscher, S. J. (2026). Semi-implicit time integration of the Exner equation for bedload transport over a non-erodible layer: Application to a tidal sandbank model. Environmental Modelling & Software, 107022.https://doi.org/10.1016/j.envsoft.2026.107022
Other relevant information:
This project is part of the BANX project, which is a collaboration between Belgian and Dutch companies, knowledge institutes and universities. You will be part of this network and can be involved in collaborations with these partners.
