Geometrical performances of MERIS, VEGETATION and MODIS fAPAR products
Marie Weiss(1), Fréderic BARET(1) and Jérôme DEMARTY(1)
(1) INRA-EMMAH, Site Agroparc, Domaine Saint-Paul, 84914 Avignon Cédex 9, France
To properly use products derived from several sensors, geometrical characteristics have to be explicitly taken into account. As an example, when assimilating medium resolution fAPAR products into canopy functioning models running per plant functional types (PFT), pixel composition in terms of PFT must be accurately known. Further, this is mandatory to compare satellite derived products with ground based measurements, as well as when inter-comparing several satellite derived products: the comparison should be achieved over the same spatial support area.
Apart from the spatial sampling interval, geometrical performances of a given sensor mainly depend on its Point Spread Function (PSF) and geolocation accuracy. When considering fAPAR products, the sensor PSF is degraded by additional terms such as geo-location uncertainties (increased by the fact that the product may be derived from compositing), spatial resampling (closely linked to the projection system), and atmospheric scattering.
In this study, we propose to determine the PSF and geolocation accuracy of fAPAR products derived from 3 medium resolution sensors: (i) MERIS, (http://merci-srv.eo.esa.int/merci/) generated using the BEAM 3.7 (http://184.108.40.206/downloads.html ) TOAVEG processor and registration tool on level 1b MERIS data, (ii) VEGETATION (CYCLOPES version 3, http://postel.mediasfrance.org/fr/PROJETS/R&D/CYCLOPES/) and (iii) MODIS collection 5 (http://edcimswww.cr.usgs.gov/pub/imswelcome/ ).
The PSF is determined using a high spatial resolution SPOT image (20m) considered as the ground truth. The SPOT fAPAR is generated using a similar algorithm to the MERIS TOA-VEG processor (neural nets applied on top of canopy reflectance data). The aggregation of the SPOT image is performed by assuming a Gaussian PSF for each sensor with the same parameters in the compas directions. The parameters of the Gaussian functions are then determined by maximizing the correlation coefficient between aggregated high spatial resolution (SPOT) image and medium spatial resolution fAPAR product. In the fiiting process, the medium resolution image is concurrently co-registered to the high spatial resolution one, yielding estimates of geolocation accuracy.
This work is performed over 6 sites corresponding to different latitudes and a range of surfaces in terms of canopy functioning and landscape heterogeneity. The geometrical performances of the three fAPAR products are then compared.
Keywords: ESA European
Space Agency - Agence spatiale europeenne,
observation de la terre, earth observation,
satellite remote sensing,
teledetection, geophysique, altimetrie, radar,
chimique atmospherique, geophysics, altimetry, radar,