A stochastic fault model: 2. Time‐dependent case

Abstract

A random model of fault motion in an earthquake is formulated by assuming that the slip velocity is a random function of position and time truncated at zero, so that it does not have negative values. This random function is chosen to be self‐affine; that is, on change of length scale, the function is multiplied by a scale factor but is otherwise unchanged statistically. A snapshot of slip velocity at a given time resembles a cluster of islands with rough topography; the final slip function is a smoother island or cluster of islands. In the Fourier transform domain, shear traction on the fault equals the slip velocity times an impedance function. The fact that this impedance function has a pole at zero frequency implies that traction and slip velocity cannot have the same spectral dependence in space and time. To describe stress fluctuations of the order of 100 bars when smoothed over a length of kilometers and of the order of kilobars at the grain size, shear traction must have a one‐dimensional power spectrum is space proportional to the reciprocal wave number. Then the one‐dimensional power spectrum for the slip velocity is proportional to the reciprocal wave number squared and for slip to its cube. If slip velocity has the same power law spectrum in time as in space, then the spectrum of ground acceleration will be flat (white noise) both on the fault and in the far field.