Generic interferometric synthetic aperture radar atmospheric correction model and its application to co- and post-seismic motions

Abstract

The tremendous development of Interferometric Synthetic Aperture Radar (InSAR) missions in recent years facilitates the study of smaller amplitude ground deformation over greater spatial scales using longer time series. However, this poses more challenges for correcting atmospheric effects due to the spatial-temporal variability of atmospheric delays. Previous attempts have used observations from Global Positioning System (GPS) and Numerical Weather Models (NWMs) to separate the atmospheric delays, but they are limited by (i) the availability (and distribution) of GPS stations; (ii) the time difference between NWM and radar observations; and (iii) the difficulties in quantifying their performance. To overcome the abovementioned limitations, we have developed the Iterative Tropospheric Decomposition (ITD) model to reduce the coupling effects of the troposphere turbulence and stratification and hence achieve similar performances over flat and mountainous terrains. Highresolution European Centre for Medium-Range Weather Forecasts (ECMWF) and GPS-derived tropospheric delays were properly integrated by investigating the GPS network geometry and topography variations. These led to a generic atmospheric correction model with a range of notable features: (i) global coverage, (ii) all-weather, all-time usability, (iii) available with a maximum of two-day latency, and (iv) indicators available to assess the model’s performance and feasibility. The generic atmospheric correction model enables the investigation of the small magnitude coseismic deformation of the 2017 Mw-6.4 Nyingchi earthquake from InSAR observations in spite of substantial atmospheric contamination. It can also minimize the temporal correlations of InSAR atmospheric delays so that reliable velocity maps over large spatial extents can be achieved. Its application to the post-seismic motion following the 2016 Kaikoura earthquake shows a success to recover the time-dependent afterslip distribution, which in turn evidences the deep inactive subduction slip mechanism. This procedure can be used to map surface deformation in other scenarios including volcanic eruptions, tectonic rifting, cracking, and city subsidence.