Global ocean colour algorithms, used to extract chlorophyll concentration in the ocean surface, normally overestimate pigment values in coastal regions, due to optical interference of water components. The objective of the present investigation was to test the performance of both empirical (SeaWiFS OC4v4) and semi‐analytical (GSM01 and CARDER) algorithms in the south‐western Atlantic. In situ pigment and optical data have been collected in waters influenced by continental discharge from La Plata River and Patos Lagoon. The data was used to develop a regional version of the empirical SeaWiFS OC2v4 algorithm (termed OC2‐LP). The relative percentage difference (RPD) between in situ and algorithm‐derived chlorophyll was 11% in the regional version as compared to the global OC4v4 (RPD = 27%). The GSM01 and CARDER showed RPD of 14% and 31%, respectively. We have also tested the accuracy of the four algorithms (OC4v4, OC2‐LP, GSM01 and CARDER) on SeaWiFS images taken over two cruise periods in the study region (winter of 2003 and summer of 2004). A seasonal difference was observed, where both OC4v4 and OC2‐LP overestimate chlorophyll in summer at a higher magnitude than in winter, and the GSM01 algorithm showed a marked underestimation of chlorophyll in winter. The CARDER model showed a good performance both in winter and summer, when applied to satellite‐retrieved radiances. Our results show that the use of semi‐analytical models does not improve significantly the accuracy of chlorophyll retrievals in coastal areas when not properly tuned with regional inherent optical properties measurements.
No Reference information available - sign in for access.
No Citation information available - sign in for access.
No Supplementary Data.
Document Type: Research Article
Department of Oceanography, Federal University of Rio Grande, Av. Italia, Km 8, Rio Grande RS, Brazil, 96201‐900
NASA Goddard Space Flight Center
Department of Physics, Federal University of Rio Grande, Brazil
Publication date: 2006-04-20
More about this publication?