Basis of the algorithm for remote sensing over Land Surfaces

Aerosol remote sensing over land from visible radiance measurements is more difficult than over ocean because the surface reflectances are generally much greater than the aerosol ones, except over dark surfaces (vegetation in the blue channel, lakes in near infrared). Airborne experiments have shown that the relative contribution of the surface compared to the atmosphere, is less important in polarized light than in total light. So, the aerosol algorithm over land is based on a best fit between polarized POLDER measurements and data simulated for different atmospheres (aerosol models and optical thickness) and ground surfaces conditions. In a given direction, the simulated polarized radiances are computed as follow:

Lp(s, v, ) = Lpatm(, n, a, m, s, v, , z) +Lpsurf(s, v, ) exp[-m(caa + cmm)] Eq.1

Where s is the solar angle, v the zenith angle, the relative azimuth angle and m = 1/cos s + 1/cos v the air mass. is the wavelength, n the considered aerosol model, a the aerosol optical thickness, m the molecular optical thickness and z the pixel elevation.
Lpsurf is the surface contribution depending on the kind of surface (Nadal and Bréon, 1999). The exponential factor, in which ca(n) and cm are empirical coefficients, corresponds to its attenuation through the total atmosphere. The atmospheric term Lpatm is interpolated in Look Up Tables computed using a multiple scattering code.

Over land, ground based measurements show that the aerosol polarization mainly comes from the small spherical particles (Vermeulen et al., 2000) with radii less than about 0.5 µm corresponding to the accumulation mode. So, the aerosol models used in the algorithm consist in lognormal size distributions of spherical particles: their characteristics are close to the oceanic small mode ones.

Knowing the super-pixel characteristics (altitude, surface classification, NDVI), the surface (Lpsurf) polarized radiances are computed in the 865 and 670 nm channels, for the viewing directions.Given an aerosol model n and an aerosol optical thickness 865, the directional polarized radiances are calculated, using Eq. 1, at the 865 and 670 nm wavelengths. Then, 865 value is adjusted to fit the polarized measurements. The root mean square n indicates the fit quality. The rms n are computed for the set of aerosol models and compared. The min minimum value gives the best model, characterized by its Angstrom exponent, and the associated optical thickness. A quality index IQinv, using min, indicates the confidence degree on the results.