17:30–17:45 | ABSTRACT 1173
OMICS-BASED PROKARYOTIC DIVERSITY ACROSS THE MEDITERRANEAN SEA USING MACHINE LEARNING, OCEAN COLOR, AND ENVIRONMENTAL DATA
Marine prokaryotic communities are fundamental components of oceans and play key roles in global biogeochemical cycles. Despite their importance, large-scale diversity patterns and the environmental drivers shaping them remain poorly understood. We developed a machine-learning approach to model the diversity of marine prokaryotic communities across the Mediterranean Sea. We used a long-term in situ 16S rRNA gene dataset collected year-round from 2001 to 2023 at coastal and open-water sites, providing extensive temporal and multisite spatial coverage. Community diversity was estimated using the Shannon Diversity Index and linked to satellite-derived and modeled oceanographic variables, including remote-sensing reflectance, chlorophyll-a, nutrients, and sea-surface temperature. The model demonstrated good predictive skill (R² = 0.78 training, 0.70 testing; RMSE = 0.31 and MAPE = 5% for both) and captured broad spatial and seasonal patterns, with greater uncertainty in less-represented areas. Photoperiod emerged as the most influential predictor, highlighting the central role of seasonal light cycles in shaping microbial communities. Additional predictors, such as the Rrs412:Rrs443 optical index of relative colored dissolved organic matter (CDOM) influence, also contributed significantly, with effects varying seasonally and regionally. Prokaryotic diversity was highest in nutrient-rich coastal zones and during winter mixing, and lowest under summer-stratified or nutrient-poor conditions. Basin-wide diversity maps revealed consistent patterns, including a west-to-east gradient and persistent coastal hotspots. These findings show that machine learning can extend discrete field observations to broader spatial scales while identifying key environmental controls on microbial diversity, offering a transferable tool for ecosystem monitoring within initiatives such as Biodiversa+ PETRI-MED.
Christian Marchese, CNR-ISMAR, [email protected]
Maria Laura Zoffoli, CNR ISMAR, [email protected]
Pierre Ramond, ICM CSIC, [email protected]
Ramiro Logares, ICM CSIC, [email protected]
François-Yves Bouget, Laboratoire d’Océanographie Microbienne (LOMIC/CNRS-Sorbonne Université), franç[email protected]
Pierre E. Galand, Laboratoire d’écogéochimie des environnements benthiques (LECOB/CNRS-Sorbonne Université), [email protected]
Tinkara Tinta, National Institute of Biology (NIB), [email protected]
Neža Orel, National Institute of Biology (NIB), [email protected]
Gianluca Volpe, CNR ISMAR, [email protected]
Angela Landolfi, CNR ISMAR, [email protected]
Emanuele Organelli, CNR ISMAR, [email protected]
