Measurement of interseismic strain across the Haiyuan fault (Gansu, China), by InSAR

The Haiyuan fault is part of a major left-lateral fault system at the northeastern edge of the Tibet-Qinghai plateau. Two M8 earthquakes (1920 and 1927) occurred along the fault, bracketing an unbroken section of the fault identified as the Tianzhu seismic gap. We use interferometric synthetic aperture radar data from descending orbits of the ERS satellites, acquired between 1993 and 1998 along two adjacent tracks covering the gap, to measure the current surface movements and better understand the present day mechanical behavior of this fault section. The analysis of the radar data involves first the combined correction of orbital errors and errors associated with the phase delay through the troposphere. A subset of the data is then selected based on the analysis of the residual noise spectra for each pair of data. The selected interferograms are stacked and the average phase change rate is converted in fault-parallel velocity assuming that the ground movement is horizontal and parallel to the fault. Velocity maps from both tracks show a zone of high velocity gradient across the fault, a few kilometers wide, consistent with left-lateral slip on the Haiyuan fault. The average velocity field from the two tracks in their overlapping area is well fit with a single screw dislocation model in an elastic half-space. The derived fault slip rate at depth (4.2–8 mm/yr) is consistent with recent GPS results. The corresponding shallow apparent locking depth (0–4.2 km) can be explained by a current low stress accumulation on the fault due to creep almost on the entire fault plane. However, unless it is transient, this creep would be paradoxical with the occurrence of past large earthquakes along this fault section, as revealed by paleoseismology. An alternative model, implying both shallow creep in the brittle upper crust and deep aseismic slip beneath the seismogenic layer, separated by a locked section, would be consistent with InSAR observations and with the potential for large earthquakes on the fault as well. A two-dislocation model with slip at 5 mm/yr beneath 15 km and a transient creep rate of 11 mm/yr between 2 and 7 km fits the InSAR data. However, the width and creep rate of the shallow creeping zone and their possible along-strike variations are still poorly resolved with the present data set.