[1] The M w 6.6, 26 December 2003 Bam (Iran) earthquake was one of the first earthquakes for which Envisat advanced synthetic aperture radar (ASAR) data were available. Using interferograms and azimuth offsets from ascending and descending tracks, we construct a three-dimensional displacement field of the deformation due to the earthquake. Elastic dislocation modeling shows that the observed deformation pattern cannot be explained by slip on a single planar fault, which significantly underestimates eastward and upward motions SE of Bam. We find that the deformation pattern observed can be best explained by slip on two subparallel faults. Eighty-five percent of moment release occurred on a previously unknown strike-slip fault running into the center of Bam, with peak slip of over 2 m occurring at a depth of $5 km. The remainder occurred as a combination of strike-slip and thrusting motion on a southward extension of the previously mapped Bam Fault $5 km to the east.

S U M M A R YOn November 8 1997 a M w 7.6 earthquake occurred in the Manyi region of northern Tibet, near the western end of the Kunlun Fault. Over 7 m of left-lateral slip occurred on a 200-kmlong fault. Here we use InSAR observations of post-seismic surface deformation following the Manyi earthquake to investigate possible causal mechanisms. Time-series of deformation are constructed from 26 interferograms, covering the entire length of the fault for nearly 4 yr after the earthquake. Three different modelling approaches are used to try and understand the observed variations in surface displacement. First order poroelastic models predict displacement fields which do not match those observed. Modelling of viscoelastic stress relaxation in a half-space with standard linear solid rheology beneath an elastic lid provides a reasonable fit to the observations, and demonstrates that two relaxation times are needed to characterize the post-seismic transient. The best-fitting viscosity is 4 × 10 18 Pa s, and the fully-relaxed shear modulus is 32 per cent of the purely elastic value. A model with a Maxwell viscoelastic halfspace underlying the lid cannot explain the observations: at later times, larger viscosities are required than at earlier times. This increase in effective viscosity with time may be consistent with stress relaxation occurring in a power-law rheology. The time-series are also inverted for distributed afterslip on an extension of the coseismic rupture. Along-strike correlation of coseismic and post-seismic slip maxima suggests that afterslip is a plausible mechanism. The maximum slip in the afterslip model is 0.72 m after 3 yr, and the equivalent moment release is approximately 20 per cent of the coseismic moment. This study shows that it is possible to rule out certain mechanisms as the dominant post-seismic process, but with the current InSAR data set it is difficult to distinguish between other plausible options.

An Mw 6.5 earthquake devastated the town of Bam in southeast Iran on 26 December 2003. Surface displacements and decorrelation effects, mapped using Envisat radar data, reveal that over 2 m of slip occurred at depth on a fault that had not previously been identified. It is common for earthquakes to occur on blind faults which, despite their name, usually produce long‐term surface effects by which their existence may be recognised. However, in this case there is a complete absence of morphological features associated with the seismogenic fault that destroyed Bam.