A Kalman Filter approach of InSAR time series analysis
Abstract
The Sentinel 1 Interferometric Synthetic Aperture Radar (InSAR) constellation images the ground every 6 days allowing to build time series of ground deformation with a dense coverage and a fine resolution. The mission is planed for 20 years over which terra-bites of data will build up making traditional InSAR time series analysis increasingly difficult. We propose the use of data assimilation for fast update of pre-existing surface deformation model. We present a Kalman filter (KF) approach for time series analysis. The filter projects iteratively the a priori knowledge of deformation at a future time, a prediction that is later adjusted when new SAR acquisitions are available. We test our method with a set of synthetic interferograms, constructed from various functions of time simulating ground deformation. Those tests reveal that the KF performs as well as classic least-squares analysis, providing that covariances are well setup, and that it may estimate the size of unexpected transient events. In particular, we outline the critical role of a priori covariances of the state as well as process noise associated with the last added SAR acquisition. We apply this approach to reconstruct a time series of surface displacements in the vicinity of the Chaman fault (Afghanistan), a large left-lateral strike slip fault that accommodates the relative motion between India and Eurasia. We estimate the fault slip rate over the 2014-2018 period and measure surface creep rate. This method has the potential to absorb the huge flux of incoming data in real time while refining our knowledge of deformation over time.
- Publication:
-
AGU Fall Meeting Abstracts
- Pub Date:
- December 2018
- Bibcode:
- 2018AGUFM.G41B0697D
- Keywords:
-
- 1240 Satellite geodesy: results;
- GEODESY AND GRAVITYDE: 1241 Satellite geodesy: technical issues;
- GEODESY AND GRAVITYDE: 1908 Cyberinfrastructure;
- INFORMATICSDE: 1932 High-performance computing;
- INFORMATICS