Machine Learning with COSIMA data

COSIMA Ocean
,
1

Hi Folks,
I am looking for examples where AI/ML was applied to COSIMA data. A quick “machine” word search on paper titles that cite Kiss et al. 2020 comes up empty. If you know of any working examples please let me know :slight_smile:
Thank you,
Paul

2

@taimoorsohail @pearseb @nmalan all have work that fits this category.

2 Likes
3

Taimoors paper is here https://www.authorea.com/doi/full/10.22541/au.173264116.62054531
Pearse’s paper is here: EGUsphere - Optimisation of the World Ocean Model of Biogeochemistry and Trophic-dynamics (WOMBAT) using surrogate machine learning methods

4

Hi Paul,

I used ACCESS-OM2-01 to estimate the distribution of crabeater seals. I used two machine learning algorithms: boosted regression trees and Random Forests together with two regression based methods.

This is under review at the moment.

Not the usual application but may be of interest.

1 Like
5

I believe @taimoorsohail, @janzika and collaborators of theirs have more than one relevant papers on this.

@PSpence, I’m intrigued tho: why you want to collate such a list? You didn’t provide us any context.

6

ACCESS-OM2

  1. Sohail, T. & Zika, J. D. Unsupervised classification identifies warm, fresh and dense regimes of the Antarctic margins. Journal of Physical Oceanography(2024), 54, 1229–1242. doi:10.1175/JPO-D-23-0153.1
  2. Sohail, T., Zika, J. D., & Ehmen, T. How accurate are salinity measurements around Antarctica? A machine learning based approach. Submitted to Machine Learning: Earth (2025). Preprint

ACCESS-CM2 (which has an ACCESS-OM2 ocean?)

  1. Piedagnel, E., Sohail, T., & Zika, J. D. Mapping ocean salinity data using Gaussian Mixture Modelling. Submitted to Artificial Intelligence for the Earth Systems (2025). Preprint
7

@janzika also!

8

there is no pre-decided notion of what is considered “usual applications” for this :slight_smile:

9

Thanks @navidcy! It’s a bit unclear what constitutes ML or AI. There are the typical neural networks, random forest, etc, but there’s also unsupervised clustering methods which could be considered more stats/data science.

Anyway, depending on the actual interpretation, there’s an additional paper myself and co-authors wrote which uses an image compression algorithm to subset the ACCESS-CM2 ocean:

Sohail, T., Holmes, R. M., Zika, J. D. Watermass co-ordinates isolate the historical ocean warming signal. Journal of Climate (2023), 36, 3063–3081, doi:10.1175/JCLI-D-22-0363.1