Good suggestion, Chris. But, since we're assigning Roger various tasks to test his statistical prowess, I have a suggestion that has the virtues of being both difficult and controversial. In a couple of sister forums, earthquake prediction sites where fringe science is allowed to run rampant and the skepticism quotient is very low, the "103-degree ring" hypothesis is being promulgated. This is, in effect, a prediction scheme which claims that even moderate earthquakes increase the probability of subsequent earthquakes along or close to a great circle 103 arc degrees distant from the original quake.

In theory, then, one could plot the intersections of that ring with seismogenic zones (there typically would be only a few of those), and predict that quakes at those locations are likely to occur within a short period of time. Since 103 degrees of arc takes one quite a large distance from the originating event, I think we would have to rule out any chance of changes in static stress being involved. That leaves, then, only dynamic stress as the responsible triggering agent. Consequently, the time period involved would be rather short. Perhaps a few days to a couple of weeks. Some research on this arcane subject might be required.

Now, I don't entirely rule out the possibility of that hypothesis having some value. But I am extremely dubious. Those who advance it point to various quakes which they claim are at the 103-degree distance from an earlier quake (I haven't made the calculations myself), as evidence for its predictive value.

I don't know how this idea got started, but it will not have escaped some of you that 103 degrees is one of the boundaries of the S-wave shadow zone - the other being 180 degrees. Within those limits, S-waves are not registered by seismographic stations.

The main challenge would be to select several largish quakes, and then figure out how to plot regions 103 degrees distant and then determine the statistical significance of earthquake occurrence over time at those locations. Or something like that. One would have to subtract any possible effect due to various seismogenic zones being coincidentally at that distance from each other. Or - perhaps it would be enough to just establish that that relationship exists naturally, and propose that that alone accounts for any observed correlations.

Mike Williams