Postdoc: Machine learning for wind flow prediction in coastal dunes
Posted Updated
Description
Coastal dunes are dynamic landforms that provide flood protection and crucial habitat in the Netherlands and across the world. Coastal dunes are formed by sand transport, which strongly relies on the interaction between wind and dune shape. Recent improvements in airflow modelling and coastal dune modelling are promising. However, getting accurate wind field predictions remains challenging, especially in dunes with complex topography.
Your job
In this 2-year postdoc position, you will use numerical modelling and machine learning techniques to increase the accuracy of coastal dune models. As a result, your project will inform and enhance decision-making in coastal dune management.
The ultimate goal of this project is to build a surrogate wind field model that can feed accurate wind field predictions into numerical coastal dune models. You will start by evaluating current wind field predictions from the coastal dune model AeoLiS by comparing model output with existing field measurements and/or Computational Fluid Dynamics (CFD) simulations. Specifically, you will assess where simplified assumptions in the model fall short. You will then train a machine learning model (such as PySR or neural networks) on CFD data to develop a fast, data-driven wind field predictor. You will combine this surrogate model with AeoLiS and evaluate the accuracy of the new model setup by applying it to existing case studies. Where needed, you will contribute to field data collection to support model validation. You will share your results in stakeholder meetings, scientific conferences, and academic journals.
Your job
In this 2-year postdoc position, you will use numerical modelling and machine learning techniques to increase the accuracy of coastal dune models. As a result, your project will inform and enhance decision-making in coastal dune management.
The ultimate goal of this project is to build a surrogate wind field model that can feed accurate wind field predictions into numerical coastal dune models. You will start by evaluating current wind field predictions from the coastal dune model AeoLiS by comparing model output with existing field measurements and/or Computational Fluid Dynamics (CFD) simulations. Specifically, you will assess where simplified assumptions in the model fall short. You will then train a machine learning model (such as PySR or neural networks) on CFD data to develop a fast, data-driven wind field predictor. You will combine this surrogate model with AeoLiS and evaluate the accuracy of the new model setup by applying it to existing case studies. Where needed, you will contribute to field data collection to support model validation. You will share your results in stakeholder meetings, scientific conferences, and academic journals.