Use of Discriminant and Fourth-Derivative Analyses With High-resolution Absorption Spectra for Phytoplankton Research: Limitations at Varied Signal-to-Noise Ratio and Spectral Resolution
Future management efforts aimed at inhibiting harmful algal blooms will require extensive temporal and spatial monitoring of phytoplankton community composition. A cost-effective approach to delineating phytoplankton community composition may be through analysis of absorption spectra, measured in situ with instruments deployed on moorings or by remote sensing. Classification techniques relying on absorption spectra include discriminant and fourth-derivative analysis. We investigated how well these techniques performed theoretically at varied signal-to-noise ratio and spectral resolution representative of a new absorption and attenuation instrument called HiStar. Our findings suggest that discriminant analysis of absorption spectra is a highly useful technique for categorizing green algae, cyanobacteria, noxious bloom-forming dinoflagellates, diatoms, and other chrysophytes. For the purposes of discriminating dinoflagellates from the other algae groups, discriminant analysis worked well with either low- or high-resolution spectral data. The discriminant analysis technique was able to delineate a noxious bloom-forming dinoflagellate species, Prorocentrum minimum, at signal-to-noise ratios as low as ~17. The current noise level in the HiStar, however, is ~28-fold too high to allow correct classification of this dinoflagellate at concentrations where shellfisheries are closed. Improvements to the discriminant analysis (e.g., inclusion of scatter properties) or to the HiStar must be accomplished before this technique becomes useful for harmful algal bloom management applications. Fourth-derivative analysis of absorption spectra, also a useful classification technique and a possible approach to assess physiological state of some algae, required at least 4 nm spectral resolution for assessment of chlorophylls a and b. The spectral resolution of HiStar (3.3 nm) meets this requirement.
Roelke, D., C. Kennedy and A. Weidemann.
Use of Discriminant and Fourth-Derivative Analyses With High-resolution Absorption Spectra for Phytoplankton Research: Limitations at Varied Signal-to-Noise Ratio and Spectral Resolution.
Gulf of Mexico Science
Retrieved from https://aquila.usm.edu/goms/vol17/iss2/2