- Marchese, Christian
- Zoffoli, Maria Laura
- Ramond, Pierre
- Turk Dermastia, Timotej
- Tinta, Tinkara
- Logares, Ramiro
- Galand, Pierre E.
- Organelli, Emanuele
11 pages, 4 figures, supplementary material https://www.frontiersin.org/articles/10.3389/fmars.2026.1875929/full#supplementary-material.-- Data availability statement: All environmental data used in this study are publicly available from the Copernicus Marine Service (CMEMS; https://marine.copernicus.eu). The analysis code and the modeling dataset (Shannon Diversity Index values and their associated environmental matchups) are available via Zenodo (https://doi.org/10.5281/zenodo.20718853). Sample sequencing runs from the VIDA dataset are publicly available through the European Nucleotide Archive (ENA) under accession number PRJEB111742. The harmonized 18S rRNA gene dataset used to compute the diversity index is not yet publicly archived; it may be made available from the authors upon reasonable request, Machine learning models provide a scalable approach for predicting the diversity of eukaryotic microbial plankton from environmental predictors. However, the extent to which these models generalize to data outside the training set remains poorly quantified. In this study, XGBoost was used to predict the 18S rRNA gene Shannon Diversity Index (SDI) from seven environmental predictors derived from satellite and model data. Surface samples were collected between 2001 and 2025 at two fixed stations in the northwestern Mediterranean (BBMO and SOLA), one fixed station in the northern Adriatic Sea (VIDA), and during the HOTMIX expedition, which sampled an east-west open-sea transect across the Mediterranean Sea. Model performance was assessed using standard repeated K-fold cross-validation (CV), Leave-One-Dataset-Out CV (LODO-CV), and a blocked spatiotemporal CV that combined LODO with temporal forward chaining. Under standard K-fold CV, the model showed moderate performance (R² = 0.44, RMSE = 0.59). In contrast, performance declined substantially under LODO-CV (R² = 0.09, RMSE = 0.73), with uniformly low per-dataset generalization, a pattern also observed with blocked spatiotemporal CV. VIDA and HOTMIX sampled environmental regimes distinct from those at BBMO and SOLA, which may partly explain their poor transferability. Additionally, BBMO and SOLA, despite similar environmental conditions, exhibited poor transferability, indicating that technical differences among independently collected 18S rRNA datasets likely constrain transferability, although their effects cannot be disentangled from environmental variation. Overall, these results highlight the limitations of imbalanced training data and underscore the importance of spatially explicit evaluation, protocol standardization, and environmentally representative coverage, This work was supported by the PETRI-MED project (Plankton Biodiversity Through Remote Sensing and Omics in the Mediterranean Sea; CUP B53C23003650001), funded by Biodiversa+, the European Biodiversity Partnership under the 2021-2022 BiodivProtect joint call for research proposals. It was co-funded by the European Commission (GA N°101052342) and funding organizations: the Italian Ministry of University and Research (MUR), the Spanish Ministry for the Ecological Transition and the Demographic Challenge (MITECO), the Spanish Fundación Biodiversidad, the Slovenian Ministry of Education, Science, and Sport, and the French National Research Agency (ANR). This research also benefited from the data and support provided by the VIDA program, which was funded through the EU Research Infrastructure projects eLTER ERIC and LifeWatch ERIC and the Slovenian Research Agency (Research Core Funding No. P1-0237). Sequencing of the LTER-BBMO datasets was supported by project MAORI (PID2022-136281NB-I00, MICINN, Spain) to R.L. The SOLA work was supported by funding from the French Agence Nationale de la Recherche (ANR) Seasoning project (ANR-24-CE02-7681) to P.E.G., This work contributes to the Institut de Ciències del Mar "Severo Ochoa Centre of Excellence" accreditation CEX2024-001494-S funded by AEI 10.13039/501100011033 of the Spanish Ministry of Science and Innovation, Peer reviewed