You can treat earthquakes, in effect, as cuts in a big slab of rubber. You imagine you have this big slab sitting on your desk. You grab a carving knife and make a cut in the rubber, and then you take the two sides of the rubber and pull them across one another, slide them across. Then imagine you take Crazy Glue and squirt it in along the cut and wait until it dries. Now let go. Obviously you have some displacement across the cut and some distortions in the rubber. If you know how stiff the rubber is, you can convert the distortions to stresses and those can be modeled on the computer. That’s basically all we do. We look at how one earthquake changes the stress around it. Our idea is embarrassingly simple: An earthquake is going to reduce the stress on the fault that just slipped, but it’s also going to increase the stress in other places in the neighborhood of the fault. All things being equal, the next earthquakes, both little and big, should be in the places where the stress was jacked up.

Well, I can answer that in two ways. First, how did people respond to this idea? And the answer is they were revolted by it. Even though we could show that aftershocks tend to occur in regions where we calculate that the stress is jacked up, and they tend to stay away from places where the stress dropped, the actual changes in stress are very small. We’re talking about the amount of stress you put in a car tire—a couple of bars. We were claiming that changing the stress by even a quarter of a bar is enough to turn earthquakes on and off. People said it’s preposterous that such small changes could be responsible. Then things proceeded as they often do in earthquake research: All new ideas look demonstrably wrong until people keep seeing the phenomena over and over again. Plate tectonics, for example, looked obviously wrong until the evidence became overwhelming.

The second part of the answer to your question is that the prevailing view then, and perhaps less so but still today, is that earthquakes are essentially random independent phenomena. If you’re throwing darts onto a board, no matter how good your last shot was, your next one is uncorrelated with it. Each shot is based instead on your general level of skill, not on the last shot. The only difference between that random view and what most people say is the case for earthquake occurrence is that more darts hit major faults like the San Andreas than elsewhere. It’s as if you had magnets behind the dartboard aligned along the major faults. Our view is that that idea cannot be right. Earthquakes, for example, trigger abundant aftershocks, and so they strongly condition the subsequent throws after big events. We view earthquakes as being in a kind of conversation, promoting and inhibiting each other by the transfer of stress. And unless we can figure out how that works, we will never be able to forecast earthquake occurrence.

In this case, we published an article in Science first, because we wanted a fast, high-profile journal, and because several other groups were heading in the same direction and we wanted to make sure we were not left behind. It would be dishonest to say science is not a competitive game. Once that 1992 paper was behind us, however, we wanted to contribute to a collection of papers about the Landers earthquake, because our sense was that these collections have a longer lifespan than individual papers. Further, Science papers are a poem, not a tomb, making it almost impossible for readers to fully understand your thinking. So a year later, we had an opportunity to do that with the Bulletin. The King et al. 1994 paper was intended to be comprehensive enough to really permit people to understand it, and persuasive enough to encourage others to reproduce it.

By degree. There is still, however, a fairly large chorus of complaints about our theory. But people have looked for all the holes in the 1994 paper; they have checked the math and reproduced its findings and figures many times over. This means that, while we may not be right, we are not obviously wrong. Anyway, we sleep better.

One thing we did that has been a research catalyst is we took the software created to make these kinds of calculations, called ‘Coulomb,’ and we built a users’ guide and tutorials, and put them on the Web and made them available to everyone. So rather than ask someone to commit six months of their time to create the tools we had already built for ourselves, we simply made them available. We thought that would accelerate the pace of research. In addition to us, several other groups have since made their software available on the Web. It means a researcher or student can jump in when they see a possible connection they want to pursue. We also teach the software in short courses from time to time. We want there to be nothing that keeps you from giving stress-triggering a try. If it works, you’ll use it on a regular basis. If it doesn’t, you can quickly move on to something else.

Since 1997, Shinji Toda, Tom Parsons, Jim Dieterich, Massimo Cocco, and I have focused on moving from stress changes to seismicity rates and earthquake probabilities. This is akin to pulling a magnet across a sandbox; we accumulate numerous parameters and assumptions to make the transition, all of them uncertain. So the transition comes at a steep cost, but it also promises to explain many more features of the time-dependence of earthquake occurrence. And we must do this if we are to transform this work into a societally useful tool to forecast earthquakes.

Landers made things very, very clear. Here we had a magnitude 7.3 earthquake, which was followed three hours later by the Big Bear magnitude 6.5 shock. Big Bear didn’t occur on the fault that ruptured at Landers, but on another one 40 kilometers away. Either that’s highly coincidental or there’s some relationship between the two shocks. The Big Bear earthquake is not what is traditionally described as an "aftershock," because it wasn’t on the Landers fault. Our calculation showed that the Landers earthquake caused a very large increase in stress at Big Bear. About 75% of the 10,000 or so small aftershocks of Landers also struck where the stress is calculated to have risen. Then the next magnitude 7 earthquake that occurred, in 1999, was in what we calculated to be the next largest lobe of fault-stress increase. Equally exciting, earthquakes were far less common in the "stress shadows," the regions where the stress had dropped.

The Landers shock is like the Velveteen Rabbit of earthquake science; we’ve loved it and fussed over it until its eyes have popped out. We just can’t give up on it. There are other large and well-recorded earthquakes that rise to that standard, but you can count them on one or two hands. We’re scientific scavengers; we’ve studied all of them. But now we are saying to people, "Hey, take a look at your earthquake, your site, your data, and see if it’s consistent or inconsistent with this idea." Our bet was is that if things looked different with other earthquakes, then we could quickly revise or enhance the theory. If they looked similar, we’d be building a case. No matter what happened, by making the tools software available for everyone to use, we’d be moving faster than we could have moved if we kept the tools to ourselves.

One difficulty is that it’s very hard to be fully objective. The cases you choose to work with tend to be the ones that support your hypothesis, and consciously or not, you tend to discard the ones that don’t get you anywhere. This occurs because the scientific rewards are restricted to where we’re able to make a strong and compelling case. Further, journals generally don’t relish work that is equivocal or ambiguous. The risk is that we will delude ourselves, particularly when stress-triggering has become a well-etched hypothesis. So there is always the chance that you’re building a case that’s illusory. You have to be vigilant to make sure that’s not happening. Some brilliant and thoughtful people have voiced very strong criticisms about what we’ve done, and we have to listen to them and figure out how to appraise the validity of their criticisms. Although it’s hard to admit, reviewers and critics keep you honest and push you harder.

Ross S. Stein
U.S. Geological Survey
Menlo Park, CA, USA