Averaging properties of systematic errors undergone in synthetic aperture 2D radiometry.

The SMOS mission, currently under phase A study, is an ESA (European Space Agency) led project aimed at global monitoring of surface soil moisture and sea surface salinity from radiometric L-band observations. The MIRAS radiometer onboard SMOS uses a 2D synthetic aperture concept, in order to achieve satisfactory spatial resolution performances for a minimal cost in terms of payload mass and volume. One counterpart of this advantage, however, is an increased intrinsic complexity of the instrument as well as data processing. As the satellite moves ahead, every area on Earth's surface is seen for a variety of incidence angles ; the principle for the retrieval of physical parameters consists of fitting model computed brightness temperatures to the radiometric (dual polarization) observations, over the range of available incidence angles. Over the open ocean, a major issue for obtaining useful measurements of the surface salinity is the radiometric accuracy, since the overall dynamic range due to salinity variations only extends over a few Kelvins. While random errors, mostly due to radiometric sensitivity, are of the order of unity on the Practical Salinity Scale (PSC), it is expected that averaging independent data over a sizable space/time domain (according to GODAE requirements) should bring this kind of error down to below 0.1 PSC. On the other hand, the synthetic aperture radiometer will be specifically affected by many non random errors, which arise from channel errors (imperfect knowledge of each element antenna radiation pattern, misalignment of phases and gains in each receiver...) as well as baseline errors (which are introduced when combining signals collected from every couple of antenna elements in order to determine the visibility functions). Finally, the reconstruction of brightness temperature fields from the visibility data introduces specific errors due to the finite extent of the domain within which visibility functions are available. It is therefore of prime importance to assess to which extent these errors can be averaged out when considering brightness temperatures reconstructed from various regions within the field of view of the radiometer and building space time averages of the retrieved salinity values. This paper reports the results of a study which addresses this issue, considering also possible dependences of the errors on the observed scene itself