This paper gets a bit more complicated, but I can still follow it, since I did some of this fault slip modeling myself. Basically, you set up a fault plane with distributed stress near failure, and you use wave propagation computer code. Then you let is snap in one spot. If you are lucky it spreads into a big giant earthquake! I did it with discrete elements and some fundamental physics. I don't know if these guys are using some fancy integrated shortcuts.
Click on it to see something readable! Once you set up the model, you let it rip! Here are the distributed slips for their model.
You can see that some areas on the fault have slipped more than others, and this is realistic. The next step (on another paper) is not to assume a single planar fault, but assemble it from multiple planes. This gives a lower PGV along the fault, but about the same in the big soil basins.
Members can see the video on-line, but that seems like too much work for me! Here is a snapshot of the PGV shooting away from the fault and hitting the Los Angeles Basin. This is a directivity blast, which is much the same as my thrust 'Fist of God', but is a lateral pulse along a strike-slip fault.
You can see the PGV pulse zooms on past LA, but starts the whole valley shaking. It wouldn't be pleasant there! Finally, we see the maximum PGV's generated by this 'Big One'.
Remember, it takes 0.5 m/s to damage a house, about 2 m/s to collapse things, and 4 m/s for things thrown in the air! (along with total destruction!). Still, we only expect 10-20% of total building stock destruction, since many structures are strong, and on solid foundations.
I love papers like this, since they make reporting in PGA, or 'spectral acceleration', or 'response spectrum', a total farce! The decent structural analysis, with absorbing elements uses explicit code with direct velocity time histories. Nevertheless, expect the Nuclear Industry to continue to use acceleration, since they are all very old guys, and we must take pity on them. :)