Improvement of Envisat RA2-MWR Performances on Coastal Areas.
Picard, Bruno1; Obligis, Estelle1; Frery, Marie-Laure1; Féménias, Pierre2
1CLS, FRANCE; 2ESA/ESRIN, ITALY

Envisat ESA programm launched with success on March 2002 embeds the RA2 altimetry mission. By the RA2 radar altimeter side, the so-called RA2-MWR microwave two-channels radiometer is dedicated to the estimation of the wet tropospheric correction directly related to the column-integrated atmosphere water vapor also named precipitable water. The simultaneous measurement of the atmospheric path delay with the altimetric range allows a high-performance correction with a spatial and temporal resolution unreachable by the models. On coastal areas, radiometric measurements are contaminated by signal coming from land through both the main lobe and side-lobes of the antenna pattern. Contamination through side-lobes can be mitigated using external look-up tables as proposed by Obligis et al 2007. The correction is established for each pixel of a 1°x1° grid by averaging the brightness temperatures on a 3000 km-radius disk and applying a given constant weighting for the side-lobes. Contamination through the main lobe can be taken into account as an input of the retrieval algorithm as proposed by Brown 2010, using the land proportion in the main lobe field of view. It has been yet implemented for Jason-1 and Jason-2 missions.
We propose to assess the improvement of RA2-MWR performances on coastal areas obtained by the implementation of a new method for computing the side-lobes correction tables and the improvement of Brown's method to a neural network approach.
When Obligis et al 2006 simulates the influence of side-lobes using a constant weighting on a 3000 km-radius disk, we have used profiles of actual antenna patterns measured on ground. The influence of side-lobes with this method is now reduced to 300 km and the weighting is computed using the measured side-lobes levels. Thanks to these updated correction tables, geographical patterns observed on the difference between AMSRE and RA2-MWR water vapour have been reduced and various other metrics quantify an improvement of the radiometer performances. The on-ground measured antenna patterns have also been used to compute land proportion in the main lobe. Then, the Coastal Neural Network algorithm has been defined using the same method developed for the present geophysical parameters retrieval (see Obligis et al 2009). Adding land proportion to the classical inputs, the two-channels brightness temperatures and the altimetric sigma-naught, improves also the radiometer performances. Results for RA2-MWR have been compared to Jason-2 and Jason-1.
Finally, we have quantified the improvements provided by the combination of these two algorithms.