Analysis of afterslip and viscoelastic relaxation following the 2004 Sumatra-Andaman earthquake from GPS observations in northern Sumatra

Global Positioning System (GPS) data in northern Sumatra, Andaman Islands and Thailand provides an opportunity to understand the postseismic deformation associated with the 2004 Sumatra-Andaman earthquake. Previous study of postseismic deformation after the 2004 earthquake reached different conclusions, which attributed to different assumption on the postseismic deformation mechanisms. Also, previous studies use only a certain part (e.g. Andaman Islands only) of observations without investigating postseismic deformation data in northern Sumatra, which is close to the largest coseismic slip rupture of 2004 SAE. In this study, we tackle this problem by taking GPS data in northern Sumatra, Aceh GPS Network for Sumatran Fault System (AGNeSS), into account. AGNeSS data are important because of the following reasons: (1) The network is located in the near-field of the main slip patch of the 2004 SAE. (2) GPS measurements started a few months after the main shock, providing information of early postseismic deformation after the 2004 SAE. (3) Continuous GPS measurements provide a good control on both horizontal and vertical components. (4) GPS data have never been used in analyzing postseismic deformation. The continuous GPS data in northern Sumatra suggest that there are multiple physical mechanisms controls the postseismic deformation, that is afterslip and viscoelastic relaxation. In the first step, we search for the optimum rheology model by fitting vertical component with viscoelastic relaxation model, and consider residual is caused by an afterslip. Using the estimated rheology model, we calculate 'afterslip' displacements by subtracting predicted viscoelastic displacements from observed GPS data, and with these 'afterslip' displacements, we estimate afterslip distribution on the plate interface. Since in the first estimation of the rheology model analyzed displacement data contain afterslip effects, we then correct the afterslip contribution from the original deformation data and estimate the rheology model again. We iterate this calculation process until we obtain the optimum model. By using this strategy, we can calculate joint mechanisms of viscoelastic relaxation and afterslip simultaneously. For viscoelastic relaxation, our best solution yields a rheological structure with elastic layer depth of 65×5 km and a Maxwell viscosity of 8.0×1 x 1e18 Pa s, respectively. The afterslip inversion results shows that major afterslip occurred during two years after the 2004 Sumatra-Andaman earthquake where the maximum slip was approximately 0.9 m occurred between 20 to 40 km depth. Our afterslip patch result showed that it is consistent with the idea of afterslip is driven by the stress change due to the mainshock. We also find that the dominance of viscoelastic relaxation significantly increased, as the afterslip continuously decreased during the four years time periods between 2005.91~2009.87. Finally, our results suggest that our model satisfied observation data in northern Sumatra, however, it poorly fit GPS data in Thailand, respectively.