POSTER SESSION 3
15:30–17:30, Wednesday, September 16
Poster Session | 1 | 2 | 3 | 4 | Schedule at a Glance
ABSTRACT 1305 | POSTER 223
PSEUDO-NITZSCHIA BLOOM DETECTION FROM SATELLITE RADIOMETRY IN THE WEST MEDITERRANEAN COAST USING CONVOLUTIONAL NEURAL NETWORKS
Blooms of the Pseudo-nitzschia can lead to domoic acid production and downstream risks for shellfish, marine wildlife, and coastal management. We present a machine learning framework that links in situ biological observations, meta-omic indicators, and satellite-derived oceanographic variables to classify Pseudo-nitzschia bloom conditions in the Balearic Sea and adjacent northwestern Mediterranean coastal waters. The training data combine more than 2,300 weekly cell abundance observations from 15 monitoring sites with 602 meta-transcriptomic observations from three oceanographic stations. These biological datasets are matched with Copernicus Marine variables describing ocean optics, nutrients, carbon system properties, and physical conditions, including remote sensing reflectance, chlorophyll a, diatom concentration, nutrient ratios, pH, sea surface height, temperature, salinity, and mixed layer depth. Separate 3D convolutional neural networks are trained as feature extractors for grouped environmental variables and biological targets, and their embeddings are combined with metadata in a final classifier. Bloom intensity is represented using three ecological classes: low, medium, and high. Preliminary evaluation on an independent test set indicates promising predictive skills, with an overall F1 score of 0.60 and correct classification rates of 63.2, 65.4, and 58.2 percent for low, medium, and high bloom conditions, respectively, supporting the feasibility of using ocean color and multi-source environmental data for early detection of harmful Pseudo-nitzschia bloom risk.
Marco Talone, Institut de Ciències del Mar ICM-CSIC, [email protected], https://orcid.org/0000-0002-9723-2080
Borja Sánchez-López, Institut de Ciències del Mar ICM-CSIC, [email protected], https://orcid.org/0000-0002-8768-5422
Margarita Fernández-Tejedor, Institut de Recerca i Tecnologia Agroalimentàries IRTA, [email protected], https://orcid.org/0000-0002-2875-1135
Rosa Trobajo, Institut de Recerca i Tecnologia Agroalimentàries IRTA, [email protected], https://orcid.org/0000-0001-9498-3797
David Mann, Institut de Recerca i Tecnologia Agroalimentàries IRTA, [email protected], https://orcid.org/0000-0003-0522-6802
Pierre Ramond, Laboratoire d’Ecogéochimie des Environnements Benthiques LECOB, [email protected], https://orcid.org/0000-0002-2314-0116
Gonzalo Martínez-Fornos, Institut de Ciències del Mar ICM-CSIC, [email protected], https://orcid.org/0009-0000-8860-8667
Poster Session | 1 | 2 | 3 | 4 | Schedule at a Glance
