- 1Univ Angers, Nantes Université, Le Mans Université, CNRS, Laboratoire de Planétologie et Géosciences, LPG UMR 6112 – Université d’Angers, 49000 Angers, France (sikandar.hayat@univ-angers.fr)
- 2CIDIS-Laboratorio de Biologia Marina, Facultad de Ciencias e Ingenieria, Universidad Peruana Cayetano Heredia, Lima, Peru
- 3Nantes Université, Univ Angers, Le Mans Université, CNRS, Laboratoire de Planétologie et Géosciences, LPG UMR 6112 – Nantes Université, 44400 Nantes, France
- 4Aix-Marseille Université, CNRS, IRD, INRAE, CEREGE, Aix-en-Provence 13545, France
- 5IPSL/Laboratoire d’Océanographie et du Climat: Expérimentations et Approches Numériques – CNRS-IRD-MNHN-Sorbonne Universités, Paris, France
- 6Dirección General de Investigaciones Oceanográficas y Cambio Climático (DGIOCC), Instituto del Mar del Perú (IMARPE), Callao, Perú
- 7Kiel University, Center for Ocean and Society (CeOS), Institute for Agricultural Economics, Kiel, Germany
The Peruvian upwelling system (PUS) sits within the intense, shallow oxygen minimum zone (OMZ) of the eastern tropical South Pacific and is strongly influenced by the El Niño Southern Oscillation (ENSO). El Niño suppresses the upwelling off Peru by weakening the trade winds and allowing warm surface waters to shift eastward, which deepens the thermocline. In contrast, La Niña generally strengthens the trade winds, shoals the thermocline, and enhances upwelling. This study reconstructs Peruvian upwelling variability by using fossil benthic foraminifera assemblages to infer past fluctuations in bottom-water oxygen and productivity. We analysed 168 samples from two sediment cores collected offshore Peru from the center of current OMZ; G10 (14.23° S, 76.40° W; 312 m water depth) and G14 (14.38° S, 76.42° W; 390 m water depth) spanning the last 25,000 years, with average resolution of 113 years. Detecting subtle faunal changes typically requires counting at least 300 specimens per sample, and the identification accuracy and speed depends on the experience of the taxonomist. We trained a CNN to identify and count benthic foraminifera, achieving 92.0% classification accuracy, 93.4% precision, and 92.4% recall. Automated results closely matched manual counts across 31 samples (from both cores at multiple depths), including species abundances, diversity metrics, multivariate assemblage patterns, and bottom-water oxygen estimates, demonstrating the model’s suitability for palaeoecological applications.
We next applied the CNN model to the remaining samples to reconstruct downcore changes in assemblage composition and bottom water oxygenation using the extended Benthic Foraminifera Assemblage index (BFAex). Low- diversity, and high-density assemblages dominated by thin, elongated tests persisted throughout much of the record, consistent with typical OMZ communities. Bolivina humilis was the dominant species across most of the record, whereas Fursenkoina spp. dominated in several intervals in the Heinrich Stadial 1 (H1S), coinciding with high denitrification and a modest increase in organic-matter input. Additionally, Suggrunda porosa, B. costata, B. plicata, Epistominella obesa, and Cassidulina limbata were among the major species. Reconstructed bottom-water oxygen was generally below 0.1 mL/L, however, H1S exhibits several peaks, some exceeding 1 mL/L. Moreover, H1S also shows the largest oxygen variability, potentially reflecting a stronger transmission of ENSO-related perturbations to the seafloor when sea level was 100 m lower than today. During the early and late Holocene, oxygen levels remained at or below the modern value (0.1 mL/L), implying a persistently developed OMZ. Several stratigraphic intervals, including early Glacial, the last Glacial Maximum, middle Holocene, Bølling-Allerød (BA), and many samples from late Holocene show a complete absence of benthic foraminifera. CT scans of two 10 cm-long sections (from BA and late Holocene) reveal “ghost” foraminifera outlines and the presence of gypsum crystals. These observations suggest post-depositional removal of carbonate tests, either during core storage or via early diagenetic dissolution. This latter interpretation is more likely and supported by the coincidence of barren intervals with low enrichment of redox-sensitive metals, reduced denitrification, and low sedimentation rates, conditions generally associated with more oxygenated periods on the Peruvian margin.
How to cite: Hayat, S., Mojtahid, M., Cardich, J., Elliot, M., Garidel-Thoron, T. D., Carré, M., Gutiérrez, D., Salvatteci, R., Barras, C., and Geslin, E.: High-resolution reconstruction of Peruvian OMZ bottom-water oxygen since the last deglaciation using automated benthic foraminifera identification, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-20159, https://doi.org/10.5194/egusphere-egu26-20159, 2026.