In the days since the 8.9-magnitude earthquake that devastated Japan, a lot of people have been asking: “Why didn’t scientists see it coming?” Well, they did—at least insofar as recognizing that the region was capable of producing a quake of this magnitude. But despite rapid advances in earthquake science, researchers aren’t close to being able to predict when big earthquakes will strike. In “Predicting the Unpredictable: The Tumultuous Science of Earthquake Prediction” (Princeton University Press)—and in the following Failure Interview—Pasadena-based seismologist Susan Hough explains why the ability to predict earthquakes remains elusive, and figures to stay that way for the foreseeable future.
What kind of advances have we seen in recent years in terms of earthquake science?
Since the early 1990s, we’ve gotten a much better handle on how earthquakes influence other earthquakes. Now it’s clear that when a big earthquake happens, the waves travel out and they trigger other earthquakes. There is communication between earthquakes that is more interesting and varied than we realized.
What’s the difference between earthquake prediction and earthquake forecast?
A prediction has to be meaningfully narrow in terms of magnitude, location and time; there’s going to be a magnitude 7 in San Francisco on Tuesday at 4 p.m., for instance. You can get progressively more fuzzy and say, “There’s going to be an earthquake between magnitude 6 and 7 within 100 km of San Francisco in the next 3 months.” That’s still a prediction, because it’s a usefully narrow window. Then things can get fuzzier and fuzzier. I can say, “There’s going to be a magnitude 5 quake in California sometime next year.” The trick is to make a prediction where you’re saying something more than what we know about earthquake rates, because some statements you can make with fair certainty just because we know how frequently earthquakes occur.
Forecasting is the other end of the spectrum, where you are talking about how often earthquakes occur on average in a given area. But you’re not really saying anything about exactly when.
What have been some of the spectacular failures of earthquake prediction?
In the 1970s there was optimism that [accurate] prediction was right around the corner, as people were starting to get a handle on how earthquake processes worked. The “Palmdale Bulge” is a good example of a spectacular failure, if you want to pick one. There was an ominous signal—warping around the San Andreas Fault—in a part of the fault where people have been expecting the Big One for a long time. It fit in with a lot of earlier observations, and there was a theory that could plausibly explain it. Then the whole thing deflated—turning into the Palmdale soufflé, as people have said.
Are there any earthquake predictions that panned out?
Parkfield [California] is an interesting story because the original prediction  was a magnitude 6 earthquake plus or minus four years of 1988. There were various papers talking about precursors and saying a Parkfield earthquake was imminent, but 1988 and 1992 came and went and those predictions didn’t pan out. You might say that was a failure, but then the earthquake happened in 2004. The predicted part of the fault did break and it was a magnitude 6 quake. It was way late, but in earthquake science a factor of two is not a big deal. Most seismologists would say it was a partial success.
Why is it so difficult to evaluate the success of any particular prediction method?
The problem is that we can’t run experiments. We have to wait for earthquakes to happen. So seismologists catalog past earthquakes and start looking for patterns. The problem is that when you’re looking for patterns you have a lot of latitude to find them. It’s like using the Super Bowl as a predictor of presidential elections. You might find a pattern that looks good, but it isn’t meaningful.
Individuals on the fringes of the scientific community have also been known to make earthquake predictions, including Iben Browning, who predicted that a big earthquake would strike New Madrid [Missouri, near the Kentucky and Tennessee borders] on or about December 3, 1990.
Browning, who was a climatologist, was looking at high tides, and was predicting earthquakes when you had especially high tides. If you look at it scientifically, it’s wacky. Even if you suppose that tides have something to do with it—which has never been demonstrated—that doesn’t tell you where an earthquake is going to hit. Browning made statements about high tides around the time of the Loma Prieta earthquake in 1989. It’s not clear why he identified the Bay Area; it may have been that there were a couple of magnitude 5 earthquakes in the Loma Prieta region in the months before. It may have been that he was talking to a San Francisco audience and wanted to get their attention. But an earthquake did happen, so people became aware of him and started to listen. Then he made the [New Madrid] prediction, which gained a lot of steam in the media and turned into a media circus—a media quake, as some called it. Some of the [disaster] management folks got into the act and scheduled drills around the same time, which gave the prediction some measure of credibility, at least in terms of public perception.
There was also the Brady-Spence prediction in Peru [Brian Brady and William Spence predicted that a major earthquake would strike near Lima in 1981]. It was made by someone who had scientific credentials—to a point. The prediction wasn’t based on solid science, but ended up being taken seriously in an international arena, and then there were touchy political issues because people in Peru were concerned and the U.S. had to deal with it.
Which earthquakes have you experienced firsthand?
I felt one of the pre-shocks for Loma Prieta in August of 1989. The Landers earthquake (1992) was a magnitude 7.3, and then the Big Bear aftershock was a 6.5, and I was here for both of those. The Northridge earthquake (1994) was smaller but closer to where I was. The last one would be Hector Mine, a 7.1 (1999). It didn’t receive a lot of media attention but it was big enough and close enough that it was unmistakably a big earthquake. When you feel a moderate quake, if it’s far away, it might start as a gentle rumbling. But for something like Northridge or Landers, it’s instantly, <i>Boom!</i> You go from sound asleep to the ride of your life. There’s this booming energy from big earthquakes that is very impressive.
If we experienced an earthquake right now, what would you do?
Duck and cover. That’s what we’re trying to drill into people. The old advice—to get under a doorway—isn’t thought to be good advice anymore. If you can get under a sturdy piece of furniture, that’s best. But what you do depends on where you are. I’ve been in parking garages, which are known to be a disaster in terms of earthquake vulnerability, and I’ve considered what I would do if the ground started to shake. The best I can come up with is to find the biggest SUV that is close by and crouch next to that.
If you were in your office and the Big One hit southern California, what is the first thing you would do after the shaking stops?
I’d want to get to the computer; we have systems that should spit out a magnitude and location pretty fast. The reality is that my phone would be ringing off the hook almost immediately, and the demand for information from the media would be overwhelming.
And as with past earthquakes, I’d want to deploy portable seismometers to record the aftershocks, because aftershocks are fairly predictable as earthquakes go. If you have a magnitude 7 earthquake you expect, on average, one magnitude 6 aftershock and ten magnitude 5’s, so it’s an opportunity to design an experiment and record a bunch of earthquakes. The first experiment I did was in 1989. I put instruments around the Nimitz Freeway [which collapsed during the Loma Prieta earthquake], so I could do a forensic analysis and look at the pattern of damage.
In terms of your computer network, are you worried that you won’t have access to information, or that it might fail to record some data?
That’s been a concern. There’s been quite a bit of work done trying to harden the system because we don’t want the system to fail at the critical moment. There are backup generators, bracing of equipment and some redundancy. You need to get data from seismometers to the network, so you’re relying on different types of telemetry. It’s going to happen to some extent, but people have worked very hard to make sure we aren’t completely knocked off the air.
But even if things are really bad here [in southern California], there are seismometers that would be working that report to Menlo Park, and then there’s a national network in Golden [Colorado], so there are backup protocols. When there’s a big earthquake, the USGS [U.S. Geological Survey] in Golden sees the earthquake, and they wait for the relevant region to post the information. But if the region doesn’t post within five or ten minutes, then Golden will step in and take the primary responsibility.
Is earthquake risk mitigation something we should be paying more attention to here in the U.S.? In the book you note how the Trans-Alaska Pipeline [built over a series of skids where the pipeline crosses the Denali fault], survived a magnitude-7.9 quake in 2002, which averted an economic and environmental disaster.
The Trans-Alaska pipeline example is singular, but there’s a collective recognition that prediction might never work and we need to focus on hazard mitigation. And even if you can predict earthquakes, you have to design structures to withstand them. That is the focus of the earthquake hazard program and the money that is spent in the U.S. is overwhelmingly aimed in that direction. In terms of actual retrofitting, there’s always more you can do and it’s just a matter of economics. These are tough times and when people are struggling to hang onto their house, it’s not realistic for them to invest thousands of dollars for something that may never happen. But for critical structures like hospitals and schools, failure is not acceptable.
What’s the worst case scenario in terms of the Big One in California?
California is complicated—for two reasons. The faults are complicated and where people live is not immediately where the faults are. For the Bay Area, a repeat of the 1906 earthquake would be the Big One. For Los Angeles, if there is a magnitude 8 quake on the San Andreas—the fault does not run through the city, it’s on the north side of the San Gabriel Mountains—it’s a question of how hard the ground is going to shake, and what it is going to do to tall buildings. There are people doing simulations who say it could knock down tall buildings. But those are computer simulations and there are a lot of uncertainties. There are also faults that run right through the L.A. area—a half-dozen that are big enough to produce a 7.5 earthquake, essentially right under downtown Los Angeles. I personally think that would be the worst case, because that would shake the bejesus out of the entire area. But it’s a much less likely earthquake than the Big One on the San Andreas.
And, if you include the Pacific Northwest, we know the Big One is going to be a magnitude 9-ish earthquake on the [Cascadia] subduction zone. There was a quake like that in 1700.
Do you believe we’ll ever be able to predict earthquakes?
I’m inclined to doubt it, but I think it’s possible. The question is: Are we ever going to be able to identify something in the earth that tells us—unmistakably—that a “big one” is coming. It’s worth keeping the lines of investigation going, but there’s been an awful lot of work and we haven’t found anything yet.
No Great Shakes: New science suggests the earthquake hazard in the Midwestern U.S. has been greatly exaggerated
Charles Richter and the Richter Scale: Five things you didn't know about the world’s most famous seismologist