Viscoelastic relaxation following subduction earthquakes and its effects on afterslip determination

Afterslip is commonly thought to be the controlling process in postseismic deformation immediately following a great megathrust earthquake and is usually inferred from geodetic observations using purely elastic models. However, observed motion reversal of the near-trench area right after the 2011 M-w 9 Tohoku-oki earthquake demonstrates the dominance of viscoelastic relaxation of coseismically induced stresses. To understand the importance of incorporating viscoelasticity in afterslip determination, we employ biviscous Burgers mantle rheology and use finite element models to explore how viscoelastic relaxation in short-term postseismic deformation is controlled by various geometrical and rheological factors. Our results indicate that immediately after large megathrust earthquakes (M-w>8.0), viscoelastic deformation should always cause opposing motion of inland and trench areas and subsidence around landward termination of the rupture, although the rate of such postseismic motion depends on local conditions such as the age and hence thickness of the slab and transient mantle viscosity values. While elastic models may be adequate for afterslip estimation for earthquakes of M-w<7.5, neglect of viscoelasticity for larger events leads to overestimate of afterslip downdip of the rupture and underestimate of afterslip at shallower depths. Reassessing shallow afterslip following the 2005 M-w 8.7 Nias earthquake using 2-D viscoelastic models suggests that the actual afterslip may be greater than that estimated using an elastic model by more than 50%. Similarly, interpreting trenchward motion of some seafloor GPS sites following the Tohoku-oki earthquake using a viscoelastic model suggests large shallow afterslip outside of the main rupture area.