Integrating field marine plankton and satellite seascapes observations: Science products for management in the Southeast Marine Biodiversity Observation Network (MBON)
Enrique Montes, NOAA Atlantic Oceanographic and Meteorological Laboratory, enrique.montes@noaa.gov (Presenter)
Maria T Kavanaugh, Oregon State University, maria.kavanaugh@oregonstate.edu
Tyler Christian, NOAA Atlantic Oceanographic and Meteorological Laboratory, tyler.christian@noaa.gov
Rachel Cohn, NOAA Atlantic Oceanographic and Meteorological Laboratory, rachel.cohn@noaa.gov
Frank E Muller-Karger, University of South Florida, carib@usf.edu
Luke R Thompson, NOAA Atlantic Oceanographic and Meteorological Laboratory, luke.thompson@noaa.gov
Christopher R Kelble, NOAA Atlantic Oceanographic and Meteorological Laboratory, chris.kelble@noaa.gov
Sustained observations of marine plankton are essential to understanding how environmental drivers shape trophic structure, food web dynamics, and ultimately living marine resources. Surveying marine plankton with traditional techniques, however, is time-consuming and requires taxonomic expertise and resources not always readily available. Image-based deep-learning applications are tools that enable effective repeated observations of plankton communities. This study merged image-based plankton observations collected during surveys in south Florida waters with satellite seascapes records to test whether plankton assemblages show distinct affinities with seascape classes. Satellite seascapes are a deep learning classification of surface biomes based on remotely-sensed variables of ocean color and physical properties. Three field campaigns aboard the R/V Walton Smith (U. Miami) were conducted in December 2022 and January and March 2023. A Continuous Particle Imaging and Classification System (CPICS) mounted on a CTD rosette was used to collect plankton image profiles from ~70–90 stations in the Florida Keys, Florida Bay, and the eastern Gulf of Mexico. Collections at 10 frames per second and ~4.5 µm per pixel were processed to identify dominant micro-phytoplankton species (> 20 µm) and zooplankton taxa between ~ 100 µm and 10 mm. Overall, ~300,000 image segments were automatically classified using a convolutional deep neural network Yolo algorithm trained and validated with images of diatoms (553), dinoflagellates (78; i.e. Noctiluca scintillans), copepods (455), radiolarians (71), polychaete (63), and gelatinous (220) species. Each plankton occurrence was matched to a unique seascape class observed during sampling events. Results from this effort will shed light on biogeographic and seasonal patterns of plankton communities in the Gulf of Mexico and provide fundamental information on biodiversity drivers for the Florida Keys National Marine Sanctuary.
Poster: Poster_Montes_3-1_150_35.pdf
Associated Project(s):
Poster Location ID: 3-1
Presentation Type: Poster
Session: Poster Session 3
Session Date: Thu (May 11) 3:00-5:00 PM
CCE Program: MBON