Variability in the hyperspectral remote sensing reflectance and performance of OC4v4 algorithm at the coast off Patagonia

FERREIRA, A.; GARCIA, C.A.E.; GARCIA, V.M.T; DOGLIOTTI, A.I.

Conferencia; Ocean Optics XX; 2010

Ocean color images from the Southwestern Atlantic Ocean usually show a band with high phytoplankton biomass during spring and summer periods along the Patagonian Shelf Break. Six cruises were conducted in October 2006, March and October 2007, January and October 2008 and January 2009, when more than 170 oceanographic stations were occupied in the region. Here we present the Remote Sensing Reflectance (Rrs) variability and an evaluation of the SeaWiFS’s empirical algorithm OC4v4 used to estimate chlorophyll a concentration ([Chl a]) based on Rrs ratios. Surface [Chl a] varied from 0.10 to 18.87 mg m-3 (mean of 2.82 ± 3.35 mg m-3). Spectral Rrs (measured using a hyperspectral radiometer profiler) presented a great variability in both magnitude and spectral shape. A hierarchical cluster analysis was applied to Rrs spectra in order to classify the whole set of Rrs spectra into coherent groups. Three spectrally distinct classes were defined, differing significantly from one another by their [Chl a] range. Evaluation of the OC4v4 algorithm showed a relatively good relationship between measured [Chl a] and algorithm-derived concentrations, combining all cruises (r2=0.78, slope of 0.86 and intercept of 0.03), with a positive bias (Mean Relative Percentage Difference, RPD=11.53%). The model was also applied to each class determined by the cluster analysis. Large errors were found for the class related to the intermediate (RPD=31.80%; APD=51.68%) [Chl a] values, but results for the class associated to the lowest [Chl a] are in good agreement with the analyzed algorithm (RPD=-2.63% APD=30.83%). Moreover, positive and negative errors were related to higher and lower phytoplankton and attenuation specific absorption coefficients (i.e. normalized by [Chl a]), respectively, reflecting the impact of the variability in the efficiency of these optical properties on the performance of empirical approaches in the region.