A strip adaptive processing approach for the SMOS space mission.
Résumé
SMOS is a mission of the European Space Agency (ESA) devoted to the monitoring of Soil Moisture and Ocean Salinity maps at global scale from L-band space borne radiometric observations. It will be the first attempt to apply to remote sensing of the Earth surface, the concept of Synthetic Aperture Imaging Radiometers (SAIR), initially developed for radio astronomy. The single payload of the mission is a Microwave Imaging Radiometer with Aperture Synthesis (MIRAS) operating at 1.415 GHz. Interferometer measurements, also called complex visibilities, are obtained by cross-correlating the signals collected by every pairs of antennae of a Y-shaped interferometric array. They are related to the radiometric brightness temperature of the observed scene by a spatial FOURIER-like integral. A regularized reconstruction process provides a band-limited solution to the corresponding inverse problem where the GIBBS oscillations caused by the truncation of the complex visibilities to the star-shaped frequency coverage are damped with an appropriate windowing function. The reconstructed brightness temperature map is obtained in the reference frame of the antenna. Up to now a single isotropic windowing function has been used in simulating this process; consequently this map exhibits a constant and isotropic angular resolution over the instrument field of view. However, owing to the geometry of the problem, these properties are no longer true for its projection onto the Earth's surface. Since the retrieval of physical quantities implies to combine several measurements for various incidence angles, the variations in shape and size of the weighting function at surface level (the "SMOS pixel") may result in significant errors when the SMOS pixel features strong heterogeneities. It has been suggested that this problem could be solved by the so called "strip adaptive" processing, i.e. using a specific windowing function for every location in the SMOS antenna frame of reference, in order to achieve a close to uniform pixel at surface level. In this paper, taking advantage of the band-limited solution to the reconstruction problem, various windowing functions are used in order to provide maps which exhibit to various extents a constant and isotropic spatial resolution at ground level. This contribution describes the properties of these windowing functions and how the corresponding multi-windows apodization (i.e. strip adaptive processing) is performed. To support the theory and to illustrate the performances of this imaging method, numerical simulations are presented within the frame of the SMOS project. Various options are illustrated in terms of their consequences for soil moisture measurements retrieved from future SMOS data.