Using teleseismic backprojection and InSAR to obtain segmentation information for large earthquakes: a case study of the 2016Mw 6.6 Muji earthquake

Abstract

Summary A good understanding of earthquake rupture segmentation is important to characterize fault geometries at depth for follow-up tectonic, stress-field or other analyses. We propose a data-driven strategy and develop pre-optimization methods to support finite fault inversions with independent prior estimates on earthquake source parameters. The first method we develop is a time-domain, multi-array and novel multi-phase backprojection of teleseismic data. This method infers the spatio-temporal evolution of the rupture process, including a potential occurrence of rupture segmentation. Secondly we apply image analysis methods on InSAR surface displacement maps to infer rupture characteristics (e.g. strike and length) and the number of potential segments. Both methods can provide model-independent constraints on fault location, dimension, orientation and rupture timing, applicable to form priors of model parameters before detailed modeling. We demonstrate and test our methods based on synthetic tests and an application to the 25.11.2016 Muji Mw 6.6 earthquake. Our results indicate segmentation and bilateral rupturing for the 2016 Muji earthquake. The results of the backprojection of the Muji Mw 6.6 earthquake using high-frequency filtered teleseismic waveforms in particular shows the capability to illuminate the rupture history with the potential to resolve the start and stop phases of individual fault segments.

Discontinuous transtensional rupture during the Mw 7.2 1995 Gulf of Aqaba earthquake

“…The simple kinematic model used to calculate the synthetics produces sharper start and stop phases than actually observed (Steinberg et al, 2022), as the modelled fault ends abruptly, whereas in nature there would be tapering. Furthermore, the synthetic semblance maps produce much sharper semblance patches in comparison to the real data backprojection.…”

“…Therefore, a mismapping of this semblance seems likely. Previous studies have shown that the choice of grid depth and the deviation of the source depth from the grid depth have a strong effect on the location of the mapped semblance (Steinberg et al, 2022;Daout et al, 2020), so this seems to be the likely cause. However, waveform coda and depth phases can also contaminate the BP results, especially in the later stages of the rupture process.…”

“…Instead of conducting a classical single largearray backprojection, we used multiple small-scale arrays for backprojection. Here, we outline the method only briefly, further details can be found in Steinberg et al (2022).…”

“…Semblance is a dimensionless mea- sure of coherence of waveforms at an array and can be multiplied, similar to a likelihood (Rössler et al, 2010). Therefore, we obtained the multi-array semblance by multiplication of all semblances (Steinberg et al, 2022).…”

“…This results in unwanted side lobes either in parallel or at an acute angle to the fault depending on the station-array geometries. This effect can only be suppressed by using multi-phase semblance (Steinberg et al, 2022), which is hindered for the 1995 Gulf of Aqaba earthquake by limited data availability. Furthermore, the depth phase separation is challenging, particularly for strike-slip earthquakes that generate relatively large sP phases.…”