POSTER Session 4

DERIVING MAJOR PHYTOPLANKTON TAXONOMIC GROUPS ON THE WEST COAST OF CANADA USING HIGHLY RESOLVED HYPERSPECTRAL RADIOMETRY DATA

The seasonal dynamics of the major phytoplankton groups (PGs) along a regularly traveled ship of opportunity ferry track in optically Case-2 waters on the west coast of Canada were derived from hyperspectral radiometry data. Here we mainly utilized CHEMTAX-derived phytoplankton community composition data and hyperspectral remote sensing reflectance (Rrs) obtained from above-water radiometers with autonomous solar tracking capability. To achieve this goal, we adapted a previously proven Empirical Orthogonal Function (EOF) algorithm developed using matchups between CHEMTAX data and EOF scores derived from hyperspectral Rrs data (N=103). Our model evaluation using full-fit statistics showed best performance for TChla, followed by diatoms, raphidophytes, green algae, and cryptophytes (R2 =0.81, 0.61, 0.57, 0.57, 0.53, respectively), and moderate retrievals were observed for haptophytes (R2 = 0.47). However, poor correlations were observed for dinoflagellates, dictyochophytes, and cyanobacteria (R2 = 0.24, 0.36, 0.38, respectively). Furthermore, our cross-validation results were comparable to full-fit statistics which demonstrated the robustness of the model. We then applied this algorithm to the hyperspectral Rrs data across the ferry track and derived their dynamics. Phytoplankton groups derived from hyperspectral Rrs data showed remarkable seasonal and spatial variability across this transect, with diatom dominating spring bloom located within the western portion of this transect and Heterosigma akashiwo bloom found close to the Fraser River plume waters. To our knowledge, this is the first study that utilized the potential of continuously obtained hyperspectral data to elucidate major PGs on the West Coast of Canada.

Vishnu Perumthuruthil Suseelan, Norwegian University of Science and Technology, Norway, [email protected], https://orcid.org/0000-0003-1439-4939

Code of Conduct
Privacy Policy