Independent Component Analysis and parametric approach for source separation in InSAR time-series at regional scale: application to 2017 - 2018 Slow Slip Event in Guerrero (Mexico)

InSAR data now allow obtaining an unprecedented spatial and temporal coverage of tectonic deformations with a short recurrence time of 6 to 12 days. However, signal separation in an InSAR time-series is notoriously challenging: discriminating long-wavelength changes in atmospheric conditions from tectonic deformations (both of which can be correlated with altitude) remains a difficulty.

In this study, we focus on the 2017-2018 Slow Slip Event (SSE) that lasted for several months on the Mexican subduction interface, in the Guerrero state area, where the permanent GPS network has a low spatial density with uneven station distribution.

In our study area, covering about 350x500 km, the atmospheric changes over the 2016-2018 Sentinel-1 observation period is the dominant signal in the InSAR time series, and show a complex temporal evolution. Signal separation is further complexified by the interplay between 3 major earthquakes (2 in September 2017 and one in February 2018), their postseismic deformation and the large slow slip event (SSE).

We extract transient SSE signal from Sentinel-1 InSAR time series with two different approaches. The first approach is a parametric method which consists of a regularized least-square inversion, imposing a functional form for each deformation and atmospheric components and using the ZTD from GPS data to characterize the seasonal signal. The second approach uses Independent Component Analysis (ICA) of the InSAR time series to decompose the signal without imposing a priori information on the SSE signal.

Both methods provide consistent results and allow the atmospheric signal to be separated from the tectonic signal without prior correction from atmospheric models. We obtain accurate SSE surface displacements maps along the radar line-of-sight, precisely matching observations on available GPS time series. We invert these surface displacements for slip on the subduction interface. Our model constrains the SSE amplitude (equivalent magnitude Mw 7.2), and temporal evolution (duration of 14 months).