Dr Tom Dunstan

Areas of expertise

• Numerical methods for flow simulations
• Numerical methods for flow simulations
• Orographic form drag and boundary layer turbulence
• Wind energy modelling

Current Activities

Tom is a scientist working on flows over complex terrain, and on the application of machine learning to improve the accuracy and speed of weather and climate models.

Tom's current work centres on understanding how surface features such as hills, buildings, or wind turbines interact with the larger scale flow. Correctly representing the drag exerted by surface obstacles on the atmosphere is important for improving the accuracy of weather and climate models. Estimating the energy available from wind farms, or correctly modelling the surface exchange of heat and moisture also depends on a clear understanding how near surface turbulence modulates the exchange of momentum between the surface and the boundary layer.

To investigate these issues, Tom uses high resolution modelling and digital elevation data sets to simulate flow over complex terrain. He has extended the Met Office – NERC Large Eddy Model (MONC) to enable it to represent complex surfaces such as steep hills, buildings, and wind turbine arrays using the immersed boundary method.

Working closely with colleagues at the Environmental Fluid Mechanics Group at Oxford University, Tom also collaborates on the modelling of large wind turbine arrays. Simple analytical models for wind turbine arrays can be valuable tools for understanding their effects on the atmospheric boundary layer and to estimate their potential performance. Validating these models against data from high-resolution configurations of the Unified Model and large-eddy simulations is an essential part of their development.

Tom also works on the application of machine learning methods in weather and climate models, with a particular focus on how parametrizations can be emulated using deep learning. He completed a pilot project to emulate the radiative transfer scheme SOCRATES using a deep convolutional neural net which demonstrated that significant improvements in both accuracy and speed are possible.

Previously, Tom has also worked with the Renewables Applications team in using machine learning to identify the sources of uncertainty in currently used downscaling techniques such as the "Intelligent" Virtual Met Mast™ and Virtual Met Mast Plus™.

 

Career Background

Tom has been a member of the The Atmospheric Boundary Layer group within Atmospheric Processes and Parametrizations in 2012. In 2019 he moved to the Orography group. Before joining the Met Office Tom completed a PhD at Cranfield University on turbulent combustion modelling using direct numerical simulation, and continued working in this field for four years as a Research Associate (postdoc) at Cambridge University Engineering Department.