A generic approach to parameterize the scaling properties of atmospheric delays in InSAR time-series

The observed phase in time-series of Interferometric Synthetic Aperture Radar (InSAR) products is a combination of primarily differential topography, line-of-sight displacement and atmospheric delay contributions. These components need to be disentangled to derive accurate atmospherical products from InSAR. However, isolating the atmospheric component from InSAR has been proven difficult as it is spatiotemporally highly dynamic and a superposition of two atmospheric states.

Here we propose an approach to parameterize the stochastic properties of the single-epoch atmospheric delay field as a way to represent the atmospheric signal in InSAR.

We found that the atmospheric signal of a time-series of interferograms can be characterized by structure functions, which can be used to isolate the single-epoch structure functions. Using two isotropic and three anisotropic scaling parameters it is then possible to construct a structure function characterizing the atmospheric signal per SAR acquisition. Especially, the isotropic parameters for the small scale and large scale atmospheric delay variations, can be used to characterize the atmospheric signal. For a test set of 150 Sentinel-1 acquisitions, this results in a difference in signal strength of the InSAR atmospheric signal with a factor of about 10 for small scale and 50 for large scale variations.

Our parametrization demonstrates that the scaling properties of the InSAR atmospheric signal for different SAR acquisitions are very similar and can be described using only five parameters. After parameter estimation we can then provide time-series of the expected atmospheric signal using distance and direction only for any combination of points within the InSAR image.