(I will not be checking this page for the next 2 weeks). Here is what I would do to understand earthquakes better. Make digital structure contour maps of faults using all kinds of data, including precisely relocated slip planes of earthquakes, studied in 3D. I would then watch the occurence of earthquakes in an area (e.g., California, or west Turkey) as they occur within the 3D fault model. I think that continuously-recorded GPS data would have to be modeled together with the fault geometry and the seismicity as it occurs. Then, wait 50 years to understand the patterns as they occur.

In retrospect, the second Turkey earthquake of 1999-the Duzce quake, which killed a couple of thousand, could have been predicted if the baseline data were in place-see Seeber and Armbruster AGU abtract-I forget if this was 2000 or 2001....

I am working on part of this problem: the 3D fault maps....
The whole approach would take some number of millions of $, plus a concerted effort.
Chris