Large-Eddy Simulation models or LES are high-resolution atmospheric models that resolve the largest turbulent circulations in the atmospheric boundary layer. They are used to run experiments and get data that is otherwise very difficult to observe directly in the atmosphere. These experiments are used to test and help develop parameterizations in coarser models like general circulation models (GCMs). It is important to test LES against observations from field campaigns to identify biases and gain confidence in the models.
GCMs are fairly coarse models, run with grid boxes that are tens or hundreds of kilometers across. To represent features like clouds and updrafts that are smaller than this, they use parameterizations that predict these small-scale features based on larger, resolved features. These parameterizations are tested against output from high-resolution LES models.
Machine learning is applied in statistical downscaling of coarse model output, and neural networks have been used to emulate existing atmospheric parameterizations with 50-80x speedups. The problem of diagnostic parameterization in atmospheric models can be seen as a supervised learning problem. I am interested in potential applications of machine learning as a driver of GCM parameterizations, with the use of high-resolution models such as LES to create training data.
Model code is often unclear, tightly coupled instead of modular, poorly documented, and without unit tests, causing long delays in research. A model framework called Sympl" is being developed which addresses these issues. It uses Python for flexible, readable, self-documenting overhead code while allowing compiled languages for computation, and can also be used for developing pure Fortran models. You can find the documentation here and the code here.
Jeremy is a research assistant and graduate student in the Department of Atmospheric Sciences at the University of Washington. He received his B. Sc. in Physics from the University of Toronto in 2014.
He is currently researching how machine learning can be applied in parameterizations for global climate models.
Through his activities as co-founder of the Atmospheric Sciences Python group, Jeremy is an advocate for programming best practices in science.