Spatial prediction of organic carbon concentrations and turnover using machine learning

durch Naveenkumar Parameswaran (GEOMAR Helmholtz Zentrumk Kiel)

Mittwoch 21.10.2026, 13:15 → 14:30 Europe/Berlin

5-1.213 - PAZIFIK / PACIFIC- Rechte Seite - Großer, unterteilbarer Konferenzraum (GEOMAR - Standort Ostufer / GEOMAR - East Shore)

Abstract:

Marine sediments store organic carbon over long timescales, but seabed measurements of sediment accumulation and particulate organic carbon (POC) are sparse and uneven. Using machine learning trained on seafloor observations and oceanographic features, we predict global maps of sediment mass accumulation rate (MAR), organic carbon content, and benthic carbon turnover. Continental shelves receive more than half of the POC rain to the seabed and bury most organic carbon, because burial efficiency is roughly twice that of the deep ocean. These models yields maps that combine low test error with a sediment budget consistent with independent particle-flux estimates, and highlights where new measurements would most improve future predictions.

Meeting-Link:
https://geomar.webex.com/geomar/j.php?MTID=m5aafe889ba4ee9d8d574eab3fbd47e34