Parameterisation design for Climate Models

Simulation from the NCAR CCSM Model by Gary Strand Used under CC BY-NC 4.0.

Throughout my career I have spent a lot of time working with parameterisations, especially whilst at ICCS.

What is a parameterisation?

Coupling machine learning to climate models.

I am currently the main developer of FTorch , a library to facilitate easy coupling of Fortran and PyTorch models.

Related to this I convened two conference sessions at EGU and PASC in 2024 on the subject of Coupling physics-based numerical models and machine learning.

Development of an interfacing structure and guidelines for ML emulation of physical parameterisations.

Parameterisation design

More generally, my exposure working with several atmospheric and climate models has allowed me to see what doesn’t work and think about how software could be designed better.

This interest began whilst writing a microphysics scheme as part of my volcanic plumes work but became more pertinent as This became particularly pertinent.

EGU Presentation