USING INDEPENDENT COMPONENT ANALYSIS AND IMAGE SEGMENTATION TO IDENTIFY ATMOSPHERIC FEATURES IN TIME SERIES OF INTERFEROMETRIC UAVSAR DATA

Coastal wetlands play a crucial role in supporting diverse ecosystems and providing numerous ecosystem services. The monitoring of wetland hydrodynamics is essential for understanding and assessing their vulnerability to environmental stressors. In recent years, InSAR (Interferometric Synthetic Aperture Radar) time series analysis has emerged as a valuable tool for studying wetland hydrodynamics. However, accurate wetland water level change monitoring is occasionally hindered by the presence of high amounts of atmospheric water vapor over coastal areas, which mislead the interpretation of InSAR retrievals. In the scope of NASA’s Delta-X airborne mission, data collected by UAVSAR’s L-band Synthetic Aperture Radar (SAR) instrument are used to measure water level change (WLC) within Atchafalya, Louisiana. The UAVSAR instrument flies at ∼ 12.5 km of altitude, and the primary artifact in the resulting time series is a delay caused by high water vapor content in the wet troposphere. This delay biases the InSAR measurements to create apparent ”false” WLC and leads to misinterpretation of the phenomena of interest. In this paper, we present a methodological approach to identify and treat these wet tropospheric delay features, here referred to as ’cloud-related features’. We employ Independent Component Analysis (ICA) combined with image segmentation as a blind source separation technique to discriminate between water level change-related signals and cloud-related signals in the InSAR time series data. Our findings provide a specific methodological case study towards addressing the challenges associated with wet tropospheric delay for Airborne InSAR, and a potential alternative solution for accurate and reliable water level change monitoring in coastal wetlands.