Earthquake Hazard and Risk Assessment in New Zealand by Monte Carlo Methods

Abstract

When assessing earthquake hazard and risk, we build seismicity models because the historical record is inadequate. If we had a historical catalog extending back 100,000 years or so, with each event recorded on broadband instruments in the near field, we could assess the long-term average hazard and risk without using a seismicity model. Earthquake hazard assessment would reduce to statistical analyses of the recorded events. We could determine mean rates of occurrence, distributions of likely ground motion, deaggregation of hazard estimates to show the contributions of individual sources, and a host of other variables and products directly from such a historical record. There is of course no such record, so we must build models that represent it, as far as can be ascertained from the historical, seismological, and geological data that are available. We employ elegant mathematical procedures involving integration of models and the incorporation of statistically derived uncertainties to estimate the hazard and its reliability. This note seeks in no way to invalidate these procedures but follows Musson (1998, 1999, 2000) to suggest that Monte Carlo models provide a simple method for simulating a long historical catalog, incorporating all the uncertainties in variables by allowing each to vary according to its estimated distribution, and performing statistical analysis of the resulting time series. This method is being used for assessment of both hazard and risk in New Zealand. A host of studies have proposed models for the occurrence of earthquakes. In the western USA, for example, recent studies by Frankel et al. (1996) used about 450 faults with specified length, width, strike, dip, slip rates, and magnitude distributions. Harmsen et al. (1999) have concluded that the central and eastern USA have two dominant sources of seismic hazard: source zones in Missouri and South Carolina. Contributions to seismic hazard …