Assessing phytoplankton community composition from hyperspectral measurements of phytoplankton absorption coefficient and remote-sensing reflectance in open-ocean environments

摘要

This study assesses the ability of hyperspectral optical measurements to discriminate changes in the composition of phytoplankton communities in open-ocean non-bloom environments. A large set of in situ near-surface measurements, comprising phytoplankton pigment determinations and hyperspectral optical data of phytoplankton absorption coefficient, aph(X), and remote-sensing reflectance, R-rs(lambda), are used in the analysis. Measurements were collected in different ecological provinces in the Pacific and Atlantic Oceans with chlorophyll-alpha concentrations ranging from about 0.02 to 1.5 mg m(-3). Hierarchical cluster analysis was applied to measured spectra of a(ph)(lambda) and R-rs(lambda) and the second-derivative spectra of these optical variables. The resulting optical-based classifications of the examined stations compared favorably (similarity index >= 0.73) with a classification of phytoplankton community composition calculated from pigment measurements. Similarities between pigmentbased and optically-based classifications were better for the optical data of a(ph)(lambda) than R-rs(lambda), with only slight improvements resulting from the use of the second derivative spectra as opposed to the non-differentiated spectra. An Empirical Orthogonal Function (EOF) analysis was applied to the optical spectra to examine the correlation of dominant modes of variability with several bio-optical and biogeochemical properties. This analysis supports the notion that the performance of the optical approach is strongly associated with the effects of differences in pigment composition, cell size, and intracellular pigment concentration among different phytoplanIcton communities on the optical properties of the ocean. (C) 2015 Elsevier Inc. All rights reserved.