Abstract:
A statistical analysis of the relative distance and time delay separating pairs of earthquakes is performed, to determine how a seismicity system evolves dynamically following a stress perturbation (earthquake). We analyse the space-time correlations of a worldwide seismicity data set (CNSS catalogue, 1963-1998, $M ≥ 5, depth ≤70 km$), and show that seismic activity diffuses away from an earthquake as the time delay increases following its occurrence. Two regimes are observed: a slow diffusion at short timescales (up to 10 days) during which the mean distance $R (Δt )$ between the initial earthquake and the subsequent earthquakes grows as $R (Δt ) ~ΔtH $, with $H = 0.19$, and a second regime at longer timescales with $H = 0.4$. The growth exponent H in the first regime increases systematically with the size of the initial earthquake, but no such dependence is observed for the second regime. Associating the latter with the viscous diffusion of stress in the upper mantle, we obtain an estimate of about 1017 Pa s for the asthenospheric lateral viscosity. This diffusion indicates that the relaxation response of the crust to a stress step depends non-linearly, in respect of both its intensity and general form, on the perturbation. A positive correlation between the regional heat flow and diffusion exponent is found, suggesting a strong thermal control on the diffusion. The overall auto-decorrelation and this diffusion process are the two major mechanisms by which seismicity systems relax stress concentrations, by redistributing them in space and in time. Both processes exhibit typical power-law behaviours, supporting the notion of space-time scale invariance of stress exchanges between seismically active faults.
