Energy is logarithmic with magnitude, but if the underlying model is analogous to springs being loaded and suddenly released, that model is quadratic in energy with distance.

The article cites 1399 detectable follow-on earthquakes. If each of them represented a movement of earth an equal distance as compared to a single event, each of those movements would be 1/1399th in scale and therefore ~5x10^-7 in energy per event. (One two-millionth.)

I think that's clearly better.