Synthetic earthquake sequences
Current numerical simulations are directed to the generation and analysis of synthetic earthquake catalogues at strike-slip faults. The main purpose of these numerical modelling is to check the behaviour of different fault structures under tectonic loading. The generated earthquake sequences are analysed and will be compared with given earthquake data.
Large earthquakes rupture commonly segmented faults which are characterised by changes in strike and dip between the participating segments. Because of this, fault models with different fault traces are simulated. Characteristics of the modelled earthquake sequences are comparable to the known features of real earthquake catalogues such as the Gutenberg-Richter distribution, the typical occurrence of fore- and aftershocks as well as sometimes a time period of seismic quiescence before a main shock happens.
The b-values of the generated earthquake sequences changes from 0.83 to 0.99 if the bend angles increases to its maximum value of 30 degrees. The deviations from linearity in the lower and higher magnitude range are well founded on model restrictions, smallest amount of the zoning length at the lower magnitude end, and limited extension of the segments at the higher end.
Additionally, the analysis of real data of large earthquakes indicate that the times between consecutive large events are in accordance with different statistical distributions. Because the available data of inter-event times at any existing fault are typically scarce (always less than ten events, but normally much less), it is not possible to decide whether of the applied theoretical distributions is able to estimate future earthquakes probabilities better. Theoretical datasets contain much more inter-event times and therefore are used to check the fits of different well-known statistical distributions.
Three empirical distributions (Gamma, Lognormal, Weibull) and one physically-based (Brownian-Passage-Time) were fitted to the modelled recurrence intervals. Generally, the interevent times show an asymmetric distribution; the ratio median/mean increases from 0.75 to 0.92 if the inclination increases from 0° to 30°. The aperiodicity (standard deviation/mean) is reduced simultaneously from 0.83 to 0.55. Despite being different, all distributions describe the dataset of interevent times rather well, but the best fits for each inclination lead to significantly different behaviour of the hazard function for large interevent times (nearly constant and memoryless or increasing with time). Besides the differences in the long-time range these distributions show also non-uniform characteristics for short inter-event times, which leads to remarkable uncertainties in the time dependent probabilistic seismic hazard assessment especially in these both time ranges.
Dr. Holger Schelle
Helmholtz Centre Potsdam
GFZ German Research Centre for Geosciences
D-14473 Potsdam Germany