ABSTRACT
Frequent Aureoumbra lagunensis blooms in the Indian River Lagoon (IRL), Florida, have devastated populations of seagrass and marine life and threaten public health. To substantiate a more reliable remote sensing early-warning system for harmful algal blooms, we apply varimax-rotated principal component analysis (VPCA) to 12 images spanning ~1.5 years. The method partitions visible-NIR spectra into independent components related to algae, cyanobacteria, suspended minerals, and pigment degradation products. The components extracted by VPCA are diagnostic for identifiable optical constituents, providing greater specificity in the resulting data products. We show that VPCA components retrieved from Sentinel-3A Ocean and Land Colour Instrument (OLCI) and a field-based spectroradiometer are consistent despite vast differences in spatial resolution (~50 cm vs. 300 m). Furthermore, the VPCA components associated with A. lagunensis in both spectral datasets indicate high correlations to Ochrophyta cell counts (R2 ≥ 0.92, p < 0.001). Recombining components exhibiting a red-edge response produces a Chl a algorithm that outperforms empirical band ratio algorithms and preforms as well or better than a variety of semianalytical algorithms. The results from the VPCA spectral decomposition method are more efficient than traditional Empirical Orthogonal Function or PCA, requiring fewer components to explain as much or more variance. Overall, our observations provide excellent validation for Sentinel-3A OLCI-based VPCA spectral identification and indicate A. lagunensis was highly concentrated within the Banana River region of the IRL during the study. These results enable improved brown tide monitoring to identify blooms at an early stage, allowing more time for stakeholder response to this public health problem.
