Hunt for slow slip events along the Sumatran subduction zone in a decade of continuous GPS data

Abstract

Slow slip events (SSEs) have been observed in GPS time series for many subduction zones worldwide but not in decade‐long GPS time series from the Sumatran GPS Array (SuGAr). An outstanding question has been whether SSEs have simply not occurred on the Sunda megathrust or whether they have been obscured by the prodigious number of earthquakes and their ensuing postseismic deformation within the time of geodetic observation. We remove all known tectonic signals from the time series to search for evidence of SSEs. The residuals are essentially flat at the centimeter scale. To search for signals at the millimeter scale we test various filtering and visualization techniques. Despite these efforts, we conclude that it is difficult to confirm that SSEs exist at this scale using the current data, although we do see a few suspicious signals. The lack of evidence for events may reflect SSEs occurring at a magnitude, location, or timescale that renders them undetectable with the current resolution of the SuGAr, that the properties of this megathrust are not conducive to SSEs, or because the megathrust is in an active period of the earthquake cycle.

Afterslip following the 2007 Mw 8.4 Bengkulu earthquake in Sumatra loaded the 2010 Mw 7.8 Mentawai tsunami earthquake rupture zone

“…We processed the GPS data using the GPS-Inferred Positioning System and Orbit Analysis Simulation Software (GIPSY-OASIS) version 6.2 [Zumberge et al, 1997]. Full details of the data processing are described in Feng et al [2015]. We transformed the daily positions in the ITRF2008 (International Terrestrial Reference Frame 2008) into daily positions relative to the Sunda plate, using the ITRF2008-Sunda transformation [Altamimi et al, 2012].…”

“…We transformed the daily positions in the ITRF2008 (International Terrestrial Reference Frame 2008) into daily positions relative to the Sunda plate, using the ITRF2008-Sunda transformation [Altamimi et al, 2012]. Feng et al [2015] estimated GPS coseismic offsets and postseismic decays, together with long-term rates, and annual and semiannual seasonal signals, on all three components of each station using a nonlinear least squares optimization procedure. The resulting coseismic offsets estimated from the optimization procedure represent more accurate and more spatially coherent estimates compared to previous studies.…”

“…1. we reestimate the coseismic slip distribution of the 2007 Bengkulu earthquake with improved coseismic SuGAr offsets estimated by Feng et al [2015] and a more realistic fault geometry compared to previous studies; and 2. we account for viscoelastic deformation in our models by jointly inverting for afterslip on the fault and strain within finite volume blocks located at asthenospheric depths, which accommodate ductile deformation due to viscous flow.…”

The Consideration of Formal Errors in Spatiotemporal Filtering Using Principal Component Analysis for Regional GNSS Position Time Series

“…Through the eigen-decomposition of covariance, we obtain the eigenvectors and eigenvalues and rearrange them in descending order. With the sorted eigenvectors, the principal components (PCs) calculated in Equation (6) and the time series can be completely reconstructed from the PCs using Equation (7).…”

“…Due to its existence, especially when the CME is larger than the transient signals, it obscures the detection of transient signals such as slow slip events and episodic tremors. Therefore, CME must be filtered out before the detection of transient signals [4][5][6]. In addition to spatial-correlated noise, temporal noise is also significant for each coordinate component, which is the combination of white and power law noise [7][8][9].…”