Pollitz, Science 2005

Non-technical summary. Large earthquakes re-distribute stress in the crust surrounding the rupture and at deeper levels in the Earth. At depths greater than about 15 km, rocks are at high enough temperature that they flow. Consequently, any stress changes at deeper levels (whether produced by an earthquake or another source) cannot be sustained for long periods of time. The rocks must "relax" these stresses and flow in a ductile fashion, and this causes continued movements of Earth's surface that are measurable with GPS if the earthquake is sufficiently large. Many earthquakes of magnitude greater than 7 have produced such pos-tseismic motions for periods of years. Until recently the largest such earthquake that had post-seismic motions measured with GPS was the 2002 M7.9 Denali, Alaska earthquake. It produced initial post-seismic velocities of nearby GPS sites up to 300 mm/yr, or 1 mm per day.

The December 26, 2004 M9.2 Sumatra-Andaman earthquake was the largest earthquake to have occurred since 1964, and it was the largest earthquake to have occurred during the era of modern space geodesy. It produced motions exceeding 1 cm over the entire surface of the Earth. With a rupture length of about 1500 km, it is the longest known rupture to have ever occurred. The length and large slip (between 5 and 20 meters along most of the rupture) involved generated large stress changes in Earth's crust and the underlying mantle to depths of 100s of km. The initial post-seismic motions generated by the earthquake reached 2 mm per day and exceeded 1 mm per day over an area of about one million square kilometers (Figure 1). The GJI paper by Pollitz et al. re-visits the mathematical modeling of post-seismic relaxation, with additional account taken of changes in Earth's gravitational potential, and this model is compared with GPS observations of rapid post-seismic motions in Thailand, Malaysia, and Sumatra. The main conclusion is that relaxation of the asthenosphere -- the ductile rock between a depth of about 60 km and 220 km -- takes place with at least two timescales. A Burgers body rheology (Figure 2) is used to simulate this behavior. The shorter timescale of 3 months explains the rapid post-seismic motions abserved over the first 3 months, and the longer timescale of 5 years explains the more subdued post-seismic motions that continue even 1 1/2 years after the earthquake. The Burgers body rheology combined with a slip model of the Sumatra-Andaman earthquake leads to predictions of postseismic velocity and acceleration that agree well with observations (Figure 3). The existence of two timescales for relaxation of the ductile rock at depth confirms findings made after the much smaller 1999 M7.1 Hector Mine earthquake [link to Pollitz, 2003] and 2002 M7.9 Denali, Alaska earthquake [link to Pollitz, 2005].

Figure 1. Calculated postseismic velocity field at Earth's surface at selected times following the 26 December 2004 Sumatra-Andaman earthquake. We use a rheology model (Figure 2) which has a Burgers body rheology in the asthenosphere, and coseismic fault Model C of Banerjee et al. (BSSA submitted, 2006).

Figure 2. Viscoelastic structure used in this study. Density, shear modulus, and bulk modulus versus depth are plotted. In addition to upper and lower mantle Maxwellian viscosities, the viscosity structure is characterized by a Burgers body rheology in the asthenosphere with Kelvin-element viscosity of 5 x 10^17 Pa s and Maxwell-element viscosity of 10^19 Pa s.

Figure 3. Predicted initial postseismic velocity and average postseismic acceleration for the first 3 months following the 26 December 2004 Sumatra-Andaman earthquakes. Superimposed are the observed velocity and acceleration vectors for several GPS sites in southeast Asia. Note that the acceleration vectors generally point in the opposite direction to the initial postseismic velocity vectors, meaning a large deceleration in elevated crustal velocities with time.