Quantitative Imaging Flow Cytometry Reveals Pigment-Based Overestimation of Diatoms and Microplankton in the North Atlantic

Decades of studies on the highly productive and dynamic North Atlantic Ocean still leave many ecosystem questions concerning phytoplankton communities unanswered due to the challenges and limitations of in situ data collection. Several pigment-based algorithms are commonly used in this region to estimate phytoplankton size classes and taxonomic groups, often for use in development and validation of satellite ocean color-based approaches to describing phytoplankton community composition. Here we address the need for further evaluation of pigment-based phytoplankton community composition by taking advantage of recent advances in automated quantitative cell imagery. An extensive imaging flow cytometry dataset was collected in the North Atlantic during 2015–2018 across all seasons. Images were classified taxonomically using both random forest machine learning complemented by manual validation, as well as machine learning methods based on convolutional neural networks accurate to within ~90% for the phytoplankton categories of interest. We evaluate published pigment-based algorithms (namely variations of the Diagnostic Pigment Analysis) that estimate phytoplankton size classes and taxonomic groups, specifically diatoms, by comparing the results of these algorithms to imagery data. Results show that pigment-based algorithms estimate microplankton and diatom percent contributions to total chlorophyll of 22 and 21%, respectively. In contrast, quantitative imagery shows contributions of microplankton and diatoms to total plankton biovolume of 5 and 3%, respectively. This work highlights the application of imaging flow cytometry to both acquire broad spatial data on phytoplankton communities via continuous at-sea deployment of imaging cytometers, as well as independently evaluate algorithms used to derive phytoplankton composition from remote sensing data, which often rely on in situ pigment databases.