hal-05564501 Towards an RFI Mitigation Algorithm in Synthetic Aperture Interferometer (…)

Despite operating in the protected L-band, interferometric measurements or visibilities from European Space Agency's soil moisture and ocean salinity SMOS mission are significantly affected by radio frequency interference (RFI), primarily from unauthorized man-made emissions. This contamination degrades the reconstructed brightness temperature (BT) maps, impacting the retrieval of key geophysical parameters, such as soil moisture and sea surface salinity. While extensive efforts-ranging from detection and localization algorithms to global shutdown campaigns-have led to substantial improvements in data quality, RFI remains a persistent challenge. A recent RFI mitigation algorithm was implemented in the SMOS operational processor, modeling each source as a Dirac delta function whose exact position and intensity are estimated to subtract its contribution from the visibilities. However, the algorithm has proven inadequate due to frequent overcorrection and undercorrection, introducing artifacts, such as ripples and oscillations. Its reliance on a static list also limits its effectiveness, as it fails to detect all interfering sources. This article introduces a two-step mitigation algorithm that addresses these shortcomings. The first step applies singular value decomposition filtering to attenuate RFI intensity and reduce propagation artifacts. The second step employs a modified version of the previous algorithm, incorporating a dynamic, source-specific search to detect and correct residual interferences. Evaluation using real SMOS data-over individual snapshots, half-orbit segments, regional BT maps, and long-term time series-shows substantial improvements in reducing artifacts and missed sources. Comparison with ocean model BT confirms that the corrected BT remains physically consistent and aligned with model expectations. Overall, the proposed method enhances the robustness and reliability of RFI mitigation in SMOS processing.