A surface reflectance model for aerosol remote sensing over land

MERIS aerosol remote sensing over land is based on the use of pixels covered by vegetation. Dense Dark Vegetation pixels are selected using the Atmospheric Resistant Vegetation Index as spectral index. Above an ARVI threshold, TARVI, at a given wavelength, a standard DDV reflectance rho(DDV) is set to a constant value. Initially, 11 biomes and 20 DDV models have been selected from the POLDER 1 imagery. A clear limitation to the initial process was the limited spatial coverage of DDV pixels. That the reason why, the DDV concept has been extended to include less dark pixels. Preliminary results indicated that a simple linear regression between surface reflectance rho and ARVI applies. The goal of this paper is to investigate the scope for such linear relationship. The first step is to start from the current DDV models and to complementary define the slope x which describes the linear decrease of r with ARVI and the ARVI range on which it applies. An extensive archive of MERIS images has been fully corrected from the atmosphere using ground based measurements. The outputs are the surface reflectances from which the reflectance model can be built. An independent validation on the surface reflectance model has been conducted on a large SeaWiFS archive on which the MERIS algorithm has been processed. The outputs of this work are new monthly Look Up Tables in the spectral bands used for aerosol remote sensing: 412 nm, 443nm and 670 nm. For each of the 11 biomes used in MERIS, the slope x and the ARVI domain on which the linear fit applied, are used as auxiliary data in the new version of the MERIS ground segment. The second MERIS processing includes this new reflectance model. We investigate the limitations of this model in two aspects. First, locally, we use ground based solar extinction measurements to validate the aerosol products. The aerosol optical thicknesses are quite well retrieved in the blue while MERIS underestimate by a factor two the aerosol optical thickness at 670 nm. We also analysed MERIS level 3 aerosol products on which the spatial continuity of the DDV models is poor between biomes. In both cases, the new surface reflectance model does not appear responsible but instead the initial DDV model has to be consolidated through more accurate values of the DDV surface albedo and through a better global mapping.