Bio-optical discrimination of diatoms from other phytoplankton in the surface ocean: Evaluation and refinement of a model for the Northwest Atlantic

Diatoms dominate global silica production and export production in the ocean; they form the base of productive food webs and fisheries. Thus, a remote sensing algorithm to identify diatoms has great potential to describe ecological and biogeochemical trends and fluctuations in the surface ocean. Despite the importance of detecting diatoms from remote sensing and the demand for reliable methods of diatom identification, there has not been a systematic evaluation of algorithms that are being applied to this end. The efficacy of these models remains difficult to constrain in part due to limited datasets for validation. In this study, we test a bio-optical algorithm developed by Sathyendranath et al. (2004) to identify diatom dominance from the relationship between ratios of remote sensing reflectance and chlorophyll concentration. We evaluate and refine the original model with data collected at the Martha's Vineyard Coastal Observatory (MVCO), a near-shore location on the New England shelf. We then validated the refined model with data collected in Harpswell Sound, Maine, a site with greater optical complexity than MVCO. At both sites, despite relatively large changes in diatom fraction (0.8–82% of chlorophyll concentration), the magnitude of variability in optical properties due to the dominance or non-dominance of diatoms is less than the variability induced by other absorbing and scattering constituents of the water. While the original model performance was improved through successive re-parameterizations and re-formulations of the absorption and backscattering coefficients, we show that even a model originally parameterized for the Northwest Atlantic and re-parameterized for sites such as MVCO and Harpswell Sound performs poorly in discriminating diatom-dominance from optical properties.