Estimating relative phytoplankton abundances in the Santa Barbara Channel
Remote sensing enables monitoring of large ocean regions over extensive spatial and temporal scales that would not otherwise be possible. However, while remote sensing can monitor primary productivity (chlorophyll-a) it is limited in its ability to detect relative abundances of phytoplankton taxa, especially in coastal waters. Algorithms designed to estimate phytoplankton relative abundance based on remote sensing reflectance (Rrs) spectra of ocean pixels are often susceptible to poor atmospheric corrections, as well as the low Rrs values, which are often associated with ocean pixels. This study sought to improve our ability to detect relative abundances of phytoplankton taxa in the Santa Barbara channel by comparing the performance of PHYDOTax, a local algorithm that relies on pre-determined Rrs spectra, and spectral unmixing using user-selected spectral end-member libraries, in measuring relative phytoplankton abundances of poorly atmospherically-corrected AVIRIS imagery. Results indicate that spectral unmixing using end-member spectral libraries from images out performs PHYDOTax for AVIRIS imagery subject to poor atmospheric correction. Moving forward, this suggests that spectral unmixing using end-member libraries from images may be the best technique for measuring relative phytoplankton abundance of ocean pixels, as it can be used on poorly atmospherically corrected imagery.