Seismic hazard deaggregation has become a standard part of probabilistic seismic hazard assessment (PSHA). The first product of PSHA is calculation of the likely severity of ground motion at a given range of annual probability levels, and this is extremely important for seismic design of structures to be built at the site under examination. However, for full analysis of proposed structural designs, engineers also need to examine scenario events to produce detailed time histories. To select such scenarios, a deaggregation of the hazard is performed, whereby the details of sources that contribute to the annual frequency of exceeding specified levels of ground motion, or P_exc, are identified. A common format for such a deaggregation is shown in Figure 1. This relates to the 475-year peak ground acceleration (pga) at Wellington, New Zealand (41.28°S 174.77°E), and shows the distribution in magnitude and distance of sources that contribute to P_exc. Return period is approximately the reciprocal of P_exc. Stiff soil site conditions (Standards New Zealand 2004) were assumed.

The analysis in Figure 1 used the interim version of the updated seismic hazard model for New Zealand (Stirling et al. 2007), with the attenuation function developed by McVerry et al. (2007). Based on a Poisson time dependence model, a return period of 475 years corresponds to a 10% probability of exceedance in 50 years.

From Figure 1, it is apparent that for this site the main contribution to ground motion of this severity is from earthquakes of magnitude about 7.6 less than 10 km from the site (blue), and there is another strong contribution from larger events in the distance range 10 to 20 km (red). These correspond to the Wellington and Wairarapa faults, respectively (see Table 1). There are other events less than 10 km from the site and small contributions from other sources. At this site the major contributions are from specific faults nearby, which are readily identified. At sites where there is significant background seismicity, however, the plot will be much more complicated and not so easy to interpret.

Figure 1 deaggregates probabilistic pga at the site; other parameters are also commonly deaggregated in the same way, in particular response spectral acceleration at a variety of natural periods. But the figure has a major shortcoming in that it represents only one return period; to obtain a full appreciation of the various contributing sources it is necessary to perform a succession of analyses to cover the full range of return periods.