ProfessorMaarten van Reeuwijk

My group works on transport processes in fluids — from city streets to cloud edges — generally involving wall-bounded turbulence and buoyancy effects. Application areas span urban environments (urban heat island, pollutant dispersion, microclimate), atmospheric convection, building ventilation (stratified environments, plumes/jets, exchange flows), oceanographic and coastal flows, and the dynamics of inertial particles (droplets, aerosols, microplastics) in turbulent and stratified carrier flows.

Urban fluid mechanics is our flagship area. Using the open-source LES framework uDALES that we develop, we resolve airflow around individual buildings and trees and chase the consequences from a single street up to a whole neighbourhood — heat, pollutants, ventilation, exposure. A coupled urban surface energy balance (multi-reflection radiation, conduction, vegetation drag and transpiration, blue-green infrastructure) ties the radiative, thermal and dynamical sides together. Sub-kilometre weather prediction needs the same physics in a different form, so we develop morphology-aware drag closures and surface-energy parameterisations that operational NWP models can use without resolving every wall.

Underpinning this is a sustained programme on turbulence and mixing: jets, plumes, fountains, gravity currents, and the structure of turbulent / non-turbulent interfaces. Energy-consistent entrainment relations, generalised plume theory, and DNS-resolved interface dynamics provide the closures that the urban and multiphase work depends on.

Multiphase flows extend the same machinery into cloud microphysics — DNS with Lagrangian particles at cloud edges, droplet evolution under intermittent turbulence, aerosol-cloud interactions through ship tracks — and into particle-laden environmental flows, including microplastics and turbidity currents.

Machine learning cuts across all of it. We build emulators that compress expensive urban land-surface physics into something operational weather models can run at every grid cell, Gaussian-process surrogates that turn sparse sensor networks into street-level pollutant fields with calibrated uncertainty, and Ensemble Kalman Filter data assimilation that fuses in-situ, UAV and satellite observations into hydrodynamic models.

Tools

Methodologically we span Direct Numerical Simulation (DNS), Large-Eddy Simulation (LES), and reduced-order theory. The largest DNS calculations run at resolutions exceeding 10 billion points across more than 250,000 cores on Europe's largest supercomputers. Insights from these high-fidelity simulations feed directly into the parameterisations and surrogates we deliver to operational models, and into very fast reduced-order tools that city planners and architects can use during early-stage design, where rapid scenario exploration matters more than full-fidelity precision.

Join us

There are regular opportunities for PhD and postdoctoral positions in my group — please get in touch if you are interested.