Estimation of inherent optical properties and phytoplankton community structure from hyperspectral in-water radiometry

Abstract

Inverse algorithms are developed to retrieve hyperspectral absorption and backscattering coefficients from measurements of hyperspectral upwelling radiance and downwelling irradiance in vertically homogeneous waters. The first inversion algorithm solves the radiative transfer equation using a simplified phase function to produce estimates of the ratio of the backscattering to absorption coefficients at depths where the light field is in the asymptotic regime. These estimates can be used as a starting point in the second implicit inversion algorithm, where the azimuthally-averaged radiative transfer equation is repeatedly evaluated using the Ecolight radiative transfer model, varying absorption and backscattering coefficients until modeled radiance and irradiance spectra match measurements within a specified criterion. Although this inversion problem is ambiguous for the retrieval of total scattering coefficients, unique and stable solutions can be found for absorption and backscattering coefficients when the inversion is constrained using the attenuation coefficient at one wavelength. A comprehensive error budget for absorption and backscattering estimates details the contributions from forward model parameter error, random and systematic radiometric error, and inversion noise. Both algorithms are tested using simulated light fields from a chlorophyll-based case I bio-optical model and radiometric field data. The second algorithm is then applied to a 51 day record of hyperspectral radiometric measurements to measure the evolution of phytoplankton community structure during the North Atlantic spring bloom continuously over 51 days from a Lagrangian float. The retrieved absorption spectra are of sufficient accuracy, bandwidth, and resolution to estimate bio-optical signatures of phytoplankton community structure including overall biomass, the fraction of small phytoplankton, and the absorption due to colored detrital material. Size fraction estimates based on the effect of pigment packaging on the phytoplankton absorption spectrum are found to be ambiguous in the presence of vertical mixing, also a consideration for ocean satellite retrievals of this parameter. Derivative spectroscopy is employed to detect and estimate the relative contribution to absorption of individual phytoplankton pigments, specifically chlorophyll <italic>a</italic>, fucoxanthin and mycosporine-like amino acids. The time series of these two estimates of community composition show the growth of a diatom bloom which then rapidly disappears and is replaced by a community of smaller cells. This description is consistent with nearby ship-based measurements that rely on water samples and laboratory measurements. We conclude that observations of phytoplankton functional types are possible from low-power, autonomous platforms.