Phytoplankton communities quantified from hyperspectral ocean reflectance correspond to pigment-based communities

Community detection analysis is a powerful tool to separate groups of samples that are similar based on their composition. Here, we use a paired global dataset of in-water hyperspectral remote sensing reflectance (Rrs) spectra and high performance liquid chromatography (HPLC) pigment concentrations to investigate the similarity in phytoplankton composition of the communities detected from each method. Samples were separated into optical communities using network-based community detection analysis applied to the Rrs residual (dRrs), which is calculated by subtracting a modeled hyperspectral Rrs spectrum from a measured hyperspectral Rrs spectrum. The dRrs spectrum accentuates short spectral scale features (<=10nm) that are correlated with phytoplankton pigment composition and therefore relate to phytoplankton community composition. To test the similarity of these optical communities to communities assessed from phytoplankton pigments, we also used network-based community detection analysis to separate HPLC pigment-based communities from twelve accessory pigment ratios to total chlorophyll-a. The results of these two community detection analyses were directly compared to identify similarities and differences in reflectance-based vs. pigment-based phytoplankton communities. A majority of samples (74%) were assigned to the same phytoplankton community between optical and pigment-based methods. Our results demonstrate that three distinct phytoplankton communities can be separated from hyperspectral Rrs data even after removing the dominant optical signals associated with total chlorophyll-a, non-algal particles, and CDOM. Notably, the optical communities assigned here offer a new tool for assessing phytoplankton community composition on global scales using the Ocean Color Instrument (OCI) on NASA’s new Plankton, Aerosol, Cloud, ocean Ecosystem (PACE) satellite.