Comparison of three SeaWiFS atmospheric correction algorithms for turbid waters using AERONET-OC Measurements

The use of satellites to monitor the color of the ocean requires effective removal of the atmospheric signal. This can be performed by extrapolating the atmospheric optical properties in the visible from the near-infrared spectral region assuming that seawater is a black body in this latter part of the spectrum. However, the non-negligible water-leaving radiance which characterizes turbid waters in the near infrared (NIR) may lead to an overestimate of the atmospheric radiance in the whole visible spectrum with increasing severity at shorter wavelengths. This results in severe errors, if not complete failure, of various algorithms for the retrieval of chlorophyll-a concentration, inherent optical properties and biogeochemical parameters of surface waters. This paper presents results of an inter-comparison study of three methods that deals with NIR water-leaving radiances: 1) the standard SeaWiFS algorithm (Stumpf et al. 2003; Bailey et al., 2010); 2) the algorithm developed by Ruddick et al. (2000); and 3) the algorithm of Kuchinke et al. (2009). The algorithms are compared using ground-based measurements from three AERONET-Ocean Color sites, one in the Adriatic Sea and two in the East Coast of United States of America. The focus of this study is on estimating the normalized water-leaving radiances nLw in the visible. Based on the matchup exercise, the best overall estimates of the normalized water-leaving radiances are obtained with the latest SeaWiFS standard algorithm version with relative error varying from 14.97% for lambda=490 nm to 35.27% forlambda=670 nm. The least accurate estimates are given by the algorithm of Ruddick, the relative error being between 16.36% for lambda=490 nm and 42.92% for lambda=412 nm. The algorithm of Kuchinke appears to be the most accurate algorithm at 412 nm (30.02%), 510 (15.54%) and 670 nm (32.32%) Similar conclusions are obtained for the aerosol optical properties (aerosol optical thickness ¿(865) and the Ångström exponent, alpha(510,865)). Those parameters are retrieved more accurately with the SeaWiFS standard algorithm (relative error of 33% for alpha(865) and of 54.15% for alpha(510,865)). But none of the algorithms is really able to accurately retrieve the values of alpha(510,865) as they are still moderately under-estimated and very patchy. The algorithm of Kuchinke is the less accurate at retrieving the aerosol optical thickness (relative error of 52.25%) while the algorithm of Ruddick is the less accurate at retrieving the Ångström exponent (relative error of 71.15%). Results from Kuchinke here are affected by the difference in aerosol-size distribution between the model and the measurements at the time of collection. A detailed analysis of the hypotheses of the methods is given for explaining the differences between the algorithms. The determination of the aerosol parameter is critical for the algorithm of Ruddick while it is the bio-optical model for the algorithm of Stumpf utilized in the standard SeaWiFS atmospheric correction. For the algorithm of Kuchinke, it is more difficult to find the sources of errors, as all the aerosol and oceanic parameters are simultaneously optimized. When regionally tuning the bio-optical model, the error on the nLw is reduced but the aerosol model is too simple to be applied everywhere. In conclusion, the results show that for the given atmospheric and oceanic conditions of this study, the new SeaWiFS atmospheric correction algorithm is most appropriate for studying the marine and aerosol parameters in turbid waters.JRC.DDG.H.3-Global environement monitorin