research

The group works on atmospheric turbulence across scales. The threads below cut across most of our projects.

Atmospheric boundary-layer processes

Structure and scaling of turbulence in the stable and convective boundary layer: similarity theory, the very stable regime and model “run-away” cooling, low-level jets, boundary-layer height, and the dissipation rate of velocity and temperature fluctuations.

Atmospheric optics

Refractive-index and temperature structure parameters ($C_n^2$, $C_T^2$) and their effect on imaging and laser propagation — physically based and machine-learned parameterizations, mesoscale prediction of optical turbulence, ray tracing through refractive structures, and von Kármán vortex streets. Part of NATO SET groups on atmospheric effects on electro-optical systems.

Large-eddy simulation

Scale-dependent dynamic subgrid-scale modeling, grid-resolution sensitivity in stable boundary layers, and LES intercomparison (chair of the 3rd GABLS LES intercomparison). We develop JAX-ALFA, an open-source JAX-based LES framework that runs on CPUs and GPUs.

Mesoscale modeling & numerical weather prediction

WRF and gray-zone modeling for wind energy and optical applications: low-level jets, coastal and offshore flows, island wakes, downscaling of reanalyses, and the vertical resolution needed to represent optical turbulence.

Wind energy

Inflow generation for turbine-load analysis, thunderstorm downbursts, wind-power ramps and Dunkelflaute events, offshore resource assessment, and forecasting. Contributor to the Grand Challenges in wind energy and IEA Wind Task 51.

Machine learning for the Earth sciences

Parameterization, downscaling, nowcasting, and data-driven surrogates — gradient boosting, TabNet, convolutional autoencoders, neural operators, and foundation models — for optical turbulence, wind gusts, wind power, and air quality.

Competitions

I regularly enter applied machine-learning and forecasting challenges, often with students (Kaggle profile):

  • Shell.ai Hackathon — 1st place (solar radiation forecasting, 2021); Level-2 finalist (fuel-blend properties, 2025)
  • NASA Airathon — 3rd (trace gas) and 6th (particulates), 2022
  • IEEE Hybrid Energy Forecasting and Trading Competition — 6th overall, 2024
  • HEROx American-Made Net Load Forecasting Prize — 9th overall; 1st for Amity, OR, 2023
  • NASA & DNV UQ Challenge — 5th of 40, 2025
  • Solafune aerosol optical depth — 25th of 307 (bronze), 2024
  • DrivenData tropical storm intensity — 32nd of 733, 2020
  • European Electricity Market (EEM) forecasting — 3rd, 2020