The use of modern statistical theories in the assessment of earthquake hazard, with application to quiet regions of eastern North America

Abstract It is generally accepted that both deterministic and statistical approaches are useful for the characterization of earthquake hazard. Although the most reliable estimates of seismic hazard can only be based on an improved understanding of the earthquake mechanism, efficient utilization of the appropriate methods provided by recent statistical theories is also important in seismic risk analysis. This is especially true in regions where the connection between seismicity and geologic structure is tenuous at best. We are particularly interested in developing better statistical treatments of data for regions with little known seismic activity. To this end, we have applied three statistical methods to the historical record of seismicity in relatively quiet regions of eastern North America. These are: (1) the threshold method for tail inference, a new theory for modeling earthquakes with sizes above a given threshold, (2) the ‘bootstrap’ technique in which the characteristics of an unknown population are simulated by replacing the true population by an estimated one, and (3) a technique to estimate the number of earthquakes below a given size, in order to compensate for the under-reporting of small earthquakes in most catalogs. A combination of these techniques has been used to estimate the probabilities of future large earthquakes for the regions studied. Because of limitations imposed by existing catalogs, the size estimate used has been maximum intensity.