We are pleased to announce an upcoming 4D-BGC webinar on Wednesday, September 30th, at 14:00 UTC / 16:00 CEST. Arianna Olivelli from Flanders Marine Institute (VLIZ) will present (Un)Detectability of ocean deoxygenation trends across machine learning reconstructions differing in climate predictors. More information can be found below:
Wednesday, 30 September, 2026 at 4 pm CEST / 2 pm UTC
Registration Link: https://washington.zoom.us/meeting/register/V768mHtxSD2P4xZeIgY1Mw
Meeting ID: 988 9348 8540 ● Passcode: 953324
Find your local number: https://us06web.zoom.us/u/kdy6pLJAeC
SCOR Working Group 168 Webinar #11
Title: (Un)Detectability of ocean deoxygenation trends across machine learning reconstructions differing in climate predictors
Presenter:
Arianna Olivelli
Flanders Marine Institute (VLIZ)
Machine learning models have become a widely used tool for reconstructing global ocean dissolved oxygen (O₂) concentrations, yet some key methodological choices remain poorly constrained. Most existing ML-derived products include a non-physical measure of temporal changes, such as year of observation or number of months, as a model predictor, and its effect on reconstructed deoxygenation trends has not been systematically assessed. In this talk, I will present the results of a multi-model experiment comprising of eight feed-forward neural networks, which represent four different climate predictor configurations at 0.25° and 1.0° monthly resolution, and their resulting gridded products of O₂ concentrations from the surface to 1500 m depth between 1993 and 2022. I will show that while all experiments perform comparably well against a held-out test set, climate predictor choice is the dominant source of methodological uncertainty in reconstructed deoxygenation trends. Additionally, using inter-experiment consistency as a measure of signal detectability, I will show that robust detection of global ocean deoxygenation over 1993–2022 is currently limited to the surface ocean, while in the ocean interior methodological uncertainty renders robust detection of deoxygenation trends largely unachievable.
You can view and subscribe to the 4D-BGC webinar calendar here:
https://calendar.google.com/calendar/u/2?cid=c2Nvci53Zy4xNjhAZ21haWwuY29t
4D-BGC webinars are recorded, and previous presentations can be found at https://www.youtube.com/@4D-BGC