Temporal correlations of earthquake focal mechanisms

Abstract

We investigate the time-distance dependence of the 3-D rotation angle, Φ, for pairs of focal mechanisms in the worldwide Harvard catalogue. We study the original catalogue as well as the subcatalogue with 'well-constrained' earthquakes. Since the distribution of Φ is scale-invariant (fractal) for small angles, the statistical analysis encounters serious difficulties due to the presence of outliers. We apply several statistical techniques to overcome this problem. To investigate the influence of errors in seismic moment inversion on the Φ angle distribution, we simulate errors to derive the bounds on possible rotation angle variations. We find that the random errors reported in the Harvard CMT catalogue explain only about one-third of the difference in scalar moment and focal mechanism orientation when the Harvard moment tensor is compared to solutions in two other catalogues (USGS and ERI). This implies that systematic effects should be significantly larger than the random errors given in the Harvard solutions. Statistical analysis of the catalogue suggests that the rotation angle is less than 10°-12° for small time-distance intervals and increases with time difference between pairs of earthquakes, reaching 20°-25° for time intervals larger than 100 days for small distances. These results are important for predicting future earthquake activity and are relevant to modelling the stress influence on earthquake occurrence.