Validation of empirical SeaWiFS algorithms for chlorophyll-α retrieval in the Mediterranean Sea -: A case study for oligotrophic seas

The major aim of this paper is the validation of SeaWiFS-derived chlorophyll-a concentration in the Mediterranean Sea. A data set containing in situ chlorophyll-a profiles and optical measurements of in-water and above-water radiances was used to evaluate the performances of several ocean color algorithms in the Mediterranean Sea. The analysis revealed a systematic overestimation of chlorophyll-a concentration by National Aeronautics and Space Administration (NASA) global algorithms (OC2v4 and OC4v4). The error appears to be correlated with chlorophyll-a concentration. by exhibiting marked differences at low values (C<0.15 mg/m(3)). In particular at low concentration, the bias observed for OC2v4 is about twice that observed for OC4v4. The same analysis made using the Gitelson et al. [J. Mar. Syst. 9 (1996) 283.] Coastal Zone Color Scanner (CZCS) regional algorithm (GIT) revealed that this model underestimates the pigments concentration but it does not exhibit a correlation between the error and the measures. On the other hand, when the NASA standard algorithms are applied to remotely sensed data, the behavior appears reversed: the OC2v4 algorithm exhibits better estimates than OC4v4, which is probably more affected by atmospheric correction problems. When applied to satellite data, the GIT algorithm performs better than the NASA global algorithms, although the estimates are very poor in the high chlorophyll-a range. Two new Mediterranean algorithms are then proposed by fitting our Mediterranean bio-optical data set with linear and OC2-like functional forms. The new algorithms perform well when applied either to the bio-optical measurements or to satellite data. The different behavior of the same algorithm when applied to bio-optical measurements or to remotely sensed data demonstrates that the atmospheric correction is still the main source of error in ocean color data. Due to the relatively small number of available in situ data, the algorithms that we generated have to be considered very preliminary. Discussion was carried out on the reasons of the global algorithm misfit, providing possible explanations and some preliminary result. The influence of coccolithophores and of the yellow substance on the optical response of the Mediterranean waters is investigated, showing that they can at least partially explain the systematic misfit. All the above shows that a region like the Mediterranean Sea requires an independent treatment of the atmospheric and of the in-water bio-optical term to obtain reliable estimates of phytoplankton activity. (C) 2002 Elsevier Science Inc. All rights reserved.