Automated Surveying of Phytoplankton Population Development in a Mesocosm Experiment

Joe Walker, University of California San Diego, La Jolla, United States; Scripps Institution of Oceanography, La Jolla, United States, Jules S Jaffe, Scripps Institution of Oceanography, La Jolla, CA, United States, Eric Coughlin Orenstein, Monterey Bay Aquarium Research Institute, Moss Landing, United States and Sarah Amiri, University of California Santa Barbara, Santa Barbara, CA, United States
Abstract:
Atmospheric aerosols profoundly impact human health and the Earth’s climate, yet the transfer dynamics of ocean-derived aerosols to the atmosphere is not well understood. During the summer of 2019, researchers from the Center for Aerosol Impacts on Climate and the Environment (CAISE) set up a mesocosm experiment to investigate the impact of planktonic bloom events on the composition of sea spray aerosols (SSA). To find a reliable mapping of state variables and biological activity to SSA, it is crucial to create an accurate profile of the constituents. Monitoring the time varying composition of the plankton population during the experiment is critical to this effort. To aid in this analysis we installed the Scripps Plankton Camera, an in situ imaging system, to continuously sample the population. Here we present automated machine learning methods to classify the ~2.5 million images taken during the 24-day sampling period. The computer labeled images are used to estimate the species richness and density. Of particular interest are certain species of phytoplankton that are known to produce the nucleating particles that comprise SSA. This data stream could provide insight to the CAISE team as they unravel the complex physical and biological interactions at the air-sea interface.
Back to: Artificial Intelligence Systems for Advancing the Study of Aquatic Ecosystems II Posters