Consideration of the three-dimensional heterogeneity of mantle rheology allows models of viscoelastic relaxation following the 2004 Sumatra-Andaman earthquake to simultaneously fit both the observed far-field and near-field postseismic deformation. We use horizontal and vertical campaign and continuous GPS observations from the Andaman, Nicobar, and Sumatran forearc islands, mainland Sumatra, Thailand, the Malay Peninsula, the Indian Ocean, and southern India, spanning the first five years of postseismic deformation. The postseismic relaxation models consider contributions from the 2004 Mw 9.2 Sumatra-Andaman, the 2005 Mw 8.7 Nias, and 2007 Mw 8.4 Bengkulu earthquakes. Far-field motions to the east of the ruptures are equally well fit by homogeneous or laterally variable earth models. However, only models with contrasting rheology across the subducting slab, a ten-times higher mantle viscosity under the Indian Ocean lithosphere than the backarc mantle, can also produce the observed enduring postseismic uplift along the forearc and lack of far-field transient displacements in southern India. While postseismic uplift of forearc stations can also be produced by rapid and enduring down-dip afterslip, the inferred rheology structure is consistent with the distribution of mantle temperature inferred from seismic tomography.

The concept of a weak asthenospheric layer underlying Earth's mobile tectonic plates is fundamental to our understanding of mantle convection and plate tectonics. However, little is known about the mechanical properties of the asthenosphere (the part of the upper mantle below the lithosphere) underlying the oceanic crust, which covers about 60 per cent of Earth's surface. Great earthquakes cause large coseismic crustal deformation in areas hundreds of kilometres away from and below the rupture area. Subsequent relaxation of the earthquake-induced stresses in the viscoelastic upper mantle leads to prolonged postseismic crustal deformation that may last several decades and can be recorded with geodetic methods. The observed postseismic deformation helps us to understand the rheological properties of the upper mantle, but so far such measurements have been limited to continental-plate boundary zones. Here we consider the postseismic deformation of the very large (moment magnitude 8.6) 2012 Indian Ocean earthquake to provide by far the most direct constraint on the structure of oceanic mantle rheology. In the first three years after the Indian Ocean earthquake, 37 continuous Global Navigation Satellite Systems stations in the region underwent horizontal northeastward displacements of up to 17 centimetres in a direction similar to that of the coseismic offsets. However, a few stations close to the rupture area that had experienced subsidence of up to about 4 centimetres during the earthquake rose by nearly 7 centimetres after the earthquake. Our three-dimensional viscoelastic finite-element models of the post-earthquake deformation show that a thin (30-200 kilometres), low-viscosity (having a steady-state Maxwell viscosity of (0.5-10) × 10¹⁸ pascal seconds) asthenospheric layer beneath the elastic oceanic lithosphere is required to produce the observed postseismic uplift.