MERIS atmospheric correction over coastal waters: validation of the MERIS aerosol models using AERONET

Abstract

Standard aerosol models (SAMs) are used for the Medium-Resolution Imaging Spectrometer (MERIS) level-2 processing over water, first to remotely sense the aerosols in the near-infrared and secondly to perform the atmospheric correction for ocean colour analysis. However, are these SAMs still suitable over coastal areas? The present work was intended to answer that question through the use of the Aerosol Robotic Network (AERONET) by selecting CIMEL radiometers operating over the sea surface or near the coastline. The current official MERIS algorithm overestimates aerosol optical thickness (AOT) over coastal waters at 865 nm. This can be related either to incorrect assumptions of the underlying surface assumption or to the assumptions of the aerosol properties (e.g. phase function). This study looks at the importance of aerosol modelling and confirms that the improved aerosol models must be used in the retrieval chain. Extinction measurements were first used to derive the aerosol optical thicknesses (AOTs). The spectral dependency of the AOTs between 670 nm and 865 nm allowed the selection of a standard aerosol model. The ability of the standard aerosol models to retrieve the AOTs at 440 nm was then analysed as a key element in the extrapolation of the aerosol path radiance from the near-infrared to the blue spectral range. The two outputs of this analysis are systematic biases in this retrieval process and accordingly they are an estimation of the dispersion. The first output can be defined as a corrective factor in the aerosol path radiance at 440 nm and the second output can be used for error analysis. A radiative transfer code was used to simulate the sky radiance in the principal plane of acquisition. Comparisons at 870 nm illustrated the ability of the standard aerosol models to retrieve the aerosol path radiances with a direct impact on the AOT retrieval from satellite observations at 865 nm.

Acknowledgments

First we thank all the PIs for the different CIMEL instruments we used in the AERONET network. This work was supported by ESA/ESRIN. We thank S. Delwart and J.P. Huot, both from ESTEC, for their helpful comments since the beginning of the present study. Many thanks to F. Lemire and particularly to F. Zagolski for computation assistance.