Do deep earthquakes occur randomly in time and in space? Or does their occurrence exhibit regular patterns – perhaps correlated with other phenomena? If you consider the big picture and don't focus too hard on the details the answers are simple: deep earthquakes are almost random with respect to time; and they are very nonrandom in space. But what about those details? In this chapter I use a descriptive approach to characterize the most obvious features of deep earthquake clustering. Then I apply some simple statistical methods to evaluate various more subtle features.

Spatial clustering

Earthquakes and subduction zones

Geographically, the world's deep earthquakes are definitely not random; the great majority lie near deep-ocean trenches beneath well-defined subduction zones (see Figs. 2.1 and 2.2). Moreover, of those with depths exceeding 300 km, about two-thirds occur in the Tonga–Kermadec region (see Fig. 1.14). When plotted in cross sections oriented normal to the trench axis, intermediate- and deep-focus quakes commonly form approximately planar groups called Wadati–Benioff zones (Figs. 5.1 and 5.2).

What is the thickness of these planar zones? For intermediate-depth seismicity there are a few places where travel times from regional network data are available and there have been careful relative relocations of deep seismicity. At a depth of about 100–125 km beneath western Argentina, Smalley and Isacks (1987) determined that all quakes occurred within a zone approximately 20 km thick, with 90% confined to a 12 km thick region (Fig. 5.3).