Anisotropy-corrected vegetation indexes derived from POLDER/ADEOS

A key originality of the polarization and directionality of Earth reflectances (POLDER) instrument is its ability to measure the bidirectional reflectance distribution function (BRDF) in a few days assuming cloud-free conditions. This ability can be used to produce time series of vegetation indexes corrected for directional effects. This paper presents the first results of such vegetation indexes during a period of four months over the continent of Africa, and an assessment of the noise level of their spatial and temporal variations. The results show that the vegetation indexes derived from the atmospherically corrected directional daily measurements depend significantly on the viewing direction, with variations of indexes in the range 0.05-0.2. Composited vegetation indexes corrected for directional effects (called anisotropy-corrected indexes) are produced at ten-day intervals as normalized differences of spectral hemispherical reflectances at 865 and 670 nm. The hemispherical reflectances are calculated by directional integration of a three-parameter semi-empirical BRDF model adjusted against time series of daily reflectance data. The values of anisotropy-corrected indexes, evaluated at the scale of Africa at two different dates, differ substantially (0.1-0.25), both in space and time, from maximum value composite (MVC) vegetation indexes calculated every ten days using a selection of POLDER data acquired in the vertical across-track acquisition plane. It is argued that the anisotropy-corrected indexes have a clearer physical meaning than MVC indexes and should lead to a better quantitative description of the biosphere. A comparative characterization of space and time behavior of anisotropy-corrected and MVC vegetation indexes shows that anisotropy-corrected indexes have more noisy single date images than the MVC indexes on vegetated areas, but also contain less noiselike oscillations in their temporal profiles