This notebook is a great resource to understand (1) how sub-grid parameterizations work in a climate modeling context (2) what is the Lorenz96 model (3) Explore different machine learning techniques applied to the same problem!
3/3 🌊

We then demonstrate machine learning techniques to learn the sub-grid components.
This notebook was a collaborative project between members of the @m2lines.bsky.social
The published article and link to code is here: jose.theoj.org/papers/10.21...
2/3

📌

Organized by: Simon Driscoll ( @simon-driscoll.bsky.social ), Sara Shamekh, Arun Kumar, Laura Mansfield ( @lauramansfield.bsky.social ), Michael G Bosilovich, Will Gregory ( @willjgregory.bsky.social ), and me! 🌊

"3. Refining ocean grid spacing from 0.25 deg to 0.125 deg has systematic improvements across a number of climate relevant features."

Key points Part 2:
"1. We present case studies of selected features of the GFDL-CM4X climate model from CMIP6 piControl, historical, and SSP5-8.5 simulations.
2. Case studies include sea level, eastern boundary upwelling, sea ice, Southern Ocean circulation, and North Atlantic Ocean circulation."

"3. CM4X-p125’s active eddies and negligible spurious mixing render an equilibrated
pre-industrial ocean with 400ZJ less heat than present day."

Key points from Part 1:
"1. CM4X-p25 and CM4X-p125 are designed to study ocean and sea ice physics, focusing on effects from eddies and boundary currents.
2. CM4X-p125 reaches pre-industrial thermal equilibrium in 150 years whereas the coarser CM4X-p25 has yet to equilibrate after 1000 years."

And Part 2:
dx.doi.org/10.22541/ess...
🌊

Thanks for the initiative, please add me!

📌 testing

Nice!