Towards a Global Strain Map from InSAR and GPS
Wright, Tim1; Walters, Richard1; Bekaert, David1; Garthwaite, Matthew2; Hussain, Ekbal1; Wang, Hua3
1University of Leeds, UNITED KINGDOM; 2Geoscience Australia, AUSTRALIA; 3Guangdong University of Technology, CHINA

By exploiting phase measurements from multiple acquisitions of Synthetic Aperture Radar (SAR) data, interferometric SAR (InSAR) can be used to measure the build up of tectonic strain around locked faults. We have recently presented a new method for combining InSAR results with GPS data to produce regional strain maps [Wang and Wright, GRL 2012]. Here we present new results from Turkey, Iran and Tibet using this method, describe our latest tests of atmospheric correction techniques, and discuss our plans for processing Sentinel-1 data for the Alpine-Himalayan belt.
Ninety percent of earthquake deaths occur in regions straining at rates higher than 1.2 x 10-8 yr-1. The accuracy of InSAR measurements is critically dependent on the length scale over which the observations are made, so we have previously defined a critical length scale of 100 km, based on the observed distribution of surface deformation around locked faults. Atmospheric errors in individual interferograms over 100 km length scales are ~25 mm (Emardson et al, JGR 2003). The error on the linear rate of deformation from an InSAR time series is proportional to the (mission length)-3/2 and (revisit time)1/2. i.e. to half the error on linear deformation rates, the revisit time must be cut by a factor of four, or the mission duration increased by ~60%. Accuracies of 1.2 mm/yr over 100 km in the satellite line of sight can be achieved with a revisit time of 12 days with a 5 year mission. These accuracies could be achieved in ~3 years if 50% of the atmospheric noise could be estimated and removed.

We have produced strain maps for parts of Turkey, Iran, and Tibet which we present here. In eastern Turkey, the strain is clearly focussed around the North and East Anatolian Faults. By contrast, the strain in Western Tibet is highest away from the major faults. We have tested atmospheric corrections using the ECMWF reanalysis data, and find that these do not work as well as has been previously reported. Only ~20% of the atmospheric noise in Eastern Turkey, for example, can be removed in this way.

If we assume coherence can be maintained in all straining areas, and ascending and descending data are both acquired, Sentinel-1A should be able to resolve tectonic strain for ~80% of the surface area straining above our target threshold of 1.2 x 10-8 yr-1 on length scales of 100 km. We are establishing an automatic Sentinel-1 processing system with the ultimate aim of producing a high-resolution strain map for the entire Alpine-Himalayan belt. The methodology could be applied globally to improve the existing strain rate model, derived from GPS alone.