The SISCAL project aims to distribute near-real-time through internet ocean color and sea surface temperature data. To ensure the compatibility of the final products between different sensors, an in house algorithm was implemented, based on the "best" state of the art. We will deal here with the atmospheric correction module for the ocean color retrieval (and the upstream classification module). SeaWiFS data in coastal areas were processed using the SISCAL algorithm. This validation exercise of the algorithm brought some improvements to the algorithm: a better selection of the suitable pixels for aerosol remote sensing, the introduction of a background correction for clear days. We also raised the problem of the aerosol climatology through the regular existence of small aerosols which are never taking into account in the standard aerosol models used for atmospheric correction.
Inland waters are an increasingly valuable natural resource with major impacts and benefits for population and environment. As the spatial resolution is improved for “ocean color” satellite sensors, such observations become relevant to monitor water quality for lakes. We first demonstrated that the required atmospheric correction cannot be conducted using the standard algorithms developed for ocean. The ocean color sensors have spectral bands that allow characterization of aerosol over dark land pixels (vegetation in the blue and in the red spectral bands). It is possible to use a representative aerosol model in the atmospheric correction over inland waters after validating the spatial homogeneity of the aerosol model in the lake vicinity. The performance of this new algorithm is illustrated on SeaWiFS scenes of the Balaton (Hungary) and the Constance (Germany) lakes. We illustrated the good spatial homogeneity of the aerosols and the meaningfulness of the water leaving radiances derived over these two lakes. We also addressed the specificity of the computation of the Fresnel reflection. The direct to diffuse term of this Fresnel contribution is reduced because of the limited size of the lake. Based on the primary scattering approximation, we propose a simple formulation of this component.
Spatial remote sensing atmospheric correction algorithms validation remains a challenge particulary over land and coastal environment. To assess this type of algorithm in the case of MERIS scheme, we propose a methodology based on the use of in-situ extinction and sky measurements from the world-wide Sun radiometer network AERONET. The spectral dependency in the blue and red derived from the extinction measurements is used to parameterize an aerosol model defined by the Jung power law size distribution in a first step and a chemical composition represented by a refractive index. This model is used to compute the phase function, a main input to a radiative transfer code (successive order of diffusion based) that allows to simulate the atmospheric parameters (radiances, transmittances). A comparison between the diffuse transmittance from sky measurements and that simulated allow to check the validity of the proposed method. The context of the study is calibration and validation in remote sensing using only the radiative properties of the atmosphere. A sensitivity study of the method to various parameters and an error budget will be reported.
Over land, the Dense Dark Vegetation is used to derive in a first stage the aerosol path radiance and in a second stage to propose an aerosol product which consists of the aerosol type and of the aerosol optical thickness. Air quality monitoring of the particles is based on measurements of PM10 and PM2.5 which are respectively the density of particles of diameter lesser than 10μm, lesser than 2.5 μm, at the surface. The satellite aerosol product can be converted into PM10 and PM2.5, based on different assumptions: particle density and vertical distribution mainly. This first attempt to monitor PM from space can be validated with in-situ data. An other approach will simply consist in using the in-situ PM measurements to calibrate the satellite imagery. With the frame of an European project, we generated, over an area centred on Lille (50'36° N, 3'08 E, North of France), a data base with the SeaWiFS archive, and the PM data collected by the regional air quality network. The above technique will be applied and validate using this data base.
Over land, the aerosol remote sensing is based on the observation of Dense Dark Vegetation (DDV) and this concept is applied on SeaWiFS with a spectral index (ARVI) to detect the DDV and the use of the bands at 412 nm, 443 nm and 670 nm to characterize the aerosols. We first extend the possibility to remote sense aerosol over less dark vegetation through a simple modeling of the vegetation reflectance ρDD.V versus the ARVI. A linear relationship exists between ρDD.V and ARVI, regardless of the geometrical conditions, but with a slight seasonal dependence. We then developed a level 3 aerosol product for global studies with 32 by 32 SeaWiFS pixels combined in a macro-pixel in order to include a significant number of DDV pixels Within these macro pixels, we documented the composition of the scene (sea, land and clouds), the mean aerosol products and the spatial dispersion of the aerosol product. This level 3 will be first used to investigate the sensitivity of the aerosol module to the inputs parameters: DDV models, assumptions on the aerosol model, accuracy of the radiometric calibration... Then, we will produce statistics on the aerosols over Europe over the SeaWiFS life time.
As the spatial resolution is improved for "ocean color" satellite sensors, such observations become relevant to monitor water quality for lakes. The required atmospheric corrections can not be conducted using the standard algorithms developed for ocean: need to account for the lake elevation, high water turbidity... The new generation of sensors has more spectral bands which allow to characterize the aerosol over dark land pixels (vegetation in the blue and in the red). Dense vegetation is identified using a spectral index and its reflectance is known from auxiliary data. We then derive, from the top of atmosphere radiances in two spectral bands, the optical thickness and the size distribution for aerosol. Knowing the aerosol model in the lake vicinity, it is then possible to apply atmospheric corrections over inland waters. A specific difficulty arises from the contamination of the photons reflected by the surrounding land and scattered towards the sensor. A simple formulation to correct this adjacency effect can be used for the Rayleigh scattering. We force the 865 nm water reflectance to be equal to zero to derive for each water pixel a function describing the aerosol adjacency effect. Assuming that the aerosol phase function does not vary much with the wavelength, we can correct all the spectral bands. The different stages of this new algorithm are illustrated on SeaWiFS.
The detection of Dense Dark Vegetation (DDV) using the Atmospherically Resistant Vegetation Index (ARVI) and then the aerosol retrieval over DDV is the critical point of the atmospheric correction scheme over land for MERIS implemented in the level 2 processor. We present here what we can expect from the MERIS product by applying a MERIS-like land algorithm to SeaWiFS data over Europe. It is shown that the DDV cover is sufficient in summer but not in winter where an extension of the concept of DDV is needed in order to enable an operational aerosol characterisation. A linear relationship between ARVI and reflectance of the extended DDV in the red should allow the use of such grey targets for the retrieval of aerosol optical properties (aerosol optical thickness at 550 nm and Angström coefficient) throughout the year with a little loss of accuracy
Aerosol remote sensing over land requires knowing the surface reflectance in some spectral bands. Dense dark vegetation can be used in the blue and in the red based on ground based measurements of their reflectances or even space measurements from a statistical analysis for clear days. An aerosol remote sensing algorithm based on DDV is available on MERIS data (Santer et al., 1999). An other alternative is to derive the surface reflectances from space as far as you have ground based characterization of the aerosols to perform suitable atmospheric correction, at least on a representative time series (Borde and Verdebout, 2001). The two algorithms, applied on SeaWiFS images, are compared over three sites (Toulouse, Ispra, Adriatic) for which ground based measurements are available.