Classification Of Natural Phytoplankton Populations With Fluorescence Excitation-Based Imaging Multivariate Optical Computing

Phytoplankton account for the majority of the primary productivity in the ocean and contribute significantly to the global carbon cycle through photosynthesis. A quantitative characterization of phytoplankton cell size and taxonomic composition is essential for understanding marine biogeochemical cycles, quantifying carbon export, and for predicting the ocean’s response to future climate change. Our labs have developed a new instrument for this purpose that combines fluorescence excitation spectroscopy with an all-optical approach to multivariate statistics called multivariate optical computing (MOC). The instrument, known as the Shipboard Streak Imaging Multivariate Optical Computing (SSIMOC) photometer, is a simple filter photometer that images the chlorophyll a fluorescence response of individual cells after excitation with a spectrum of light tailored specifically to differentiate species of phytoplankton. The images captured by the SSIMOC photometer carry information about the size and spectral characteristics of phytoplankton cells that are related to their taxonomic identity. Due to a relatively large depth of field, non-linear multivariate calibration was need to provide accurate estimates of size regardless of focus quality. The extracted size and spectral information were then used in multivariate classification to both assess the ability of SSIMOC to differentiate between cultured species of phytoplankton as well as to explore the spectral and pigment characteristics that create the observed distinctions between species. During preliminary field studies, the methodologies for sizing and classifying phytoplankton were applied to samples measured in situ. These studies demonstrate the ability and efficiency of the SSIMOC photometer for detecting phytoplankton, its ability to characterize their size distribution, and its effectiveness at classifying phytoplankton in their natural environment. During the field studies classes of phytoplankton that form chains were also imaged. In these cases, the task of extracting the desired spectral information becomes more difficult since the analysis software has been optimized for single cells. An approach is presented for detecting these special types of phytoplankton and identifying their class based on frequency patterns created by the repetition of cells