Seismic Hazard Deaggregation

Epsilon measures how many standard deviations a target ground motion lies above the median predicted by the ground-motion equation. Bazzurro and Cornell showed that at long return periods the hazard is dominated by rare, above-median shaking, so the deaggregation typically reveals a large positive epsilon. This matters for record selection: it tells you the design motion comes not from average shaking of the controlling earthquake but from an unusually strong realization of it. Ignoring epsilon and matching only magnitude and distance can select records that systematically underrepresent the spectral shape associated with high-epsilon motions, which is why the conditional mean spectrum was developed.

The controlling earthquake is the magnitude-distance-epsilon scenario most responsible for exceeding a chosen ground-motion level. It can be summarized as the mean of the deaggregation distribution or as its modal (most probable) bin. Bazzurro and Cornell warn that the mean is unreliable when the hazard is multimodal — for instance when a nearby moderate source and a distant large source both contribute — because averaging them yields an intermediate scenario that physically does not exist. Best practice is to inspect the full distribution and report the modal scenario, or several representative scenarios, rather than a single mean.

No. Deaggregation is a post-processing step that requires a completed PSHA; it decomposes the hazard the PSHA already computed and adds no new information about sources or rates. It inherits all of the parent model's assumptions, so an inadequate source model, recurrence rate, or ground-motion equation will produce a misleading controlling scenario. Deaggregation should be run at the specific return periods and spectral periods of engineering interest, because the controlling earthquake shifts across both, and it is typically presented alongside the hazard curves and uniform-hazard spectrum it is derived from.