Animal Earthquake Prediction

by Brian Dunning
October 23, 2018

It's true if you've ever been to the movies or gotten your education from the University of YouTube: Animals can predict earthquakes — and they can do it hours, days, or weeks in advance. However, if you get your information from more conventional sources, you probably know that this claim is by no means widely held among seismologists. The question of whether animals can sense impending quakes seems particularly thorny, but it turns out it really isn't, after all.

But before we get too far into the subject, let's be clear on what we're not talking about, and that's the genuine ability of many animals to sense some earthquakes a few seconds before most people can. As human beings, our survival and success depends much more upon abstract thought than upon physical senses. Animals, on the other hand, have to deal with things like avoiding predators or finding prey, so their evolutionary pressures have selected for heightened sensory perception. Most animals are thus much more likely to feel earthquakes too slight for us to notice, and/or to hear the slight noises from things shifting.

There are two basic types of seismic waves: body waves that travel through the Earth, and surface waves propagating along its surface. Body waves are the ones that reach us first, and are the ones relevant to our discussion of animal sensing. There are also two types of body waves, and you're probably already familiar with them. The first is called the Primary wave, or P-wave. This wave is compressional, causing vibrations parallel to the direction of its movement. P-waves travel the fastest, between 4 and 7 km/s (about 9,000-16,000 mph). The other type is the Secondary wave, or S-wave, which causes shearing vibrations perpendicular to the direction of its movement. S-waves are much slower, traveling only 2 to 5 km/s (4,500-11,000 mph). The faster P-waves are also the only ones that propagate through water.

In most cases, P-waves are quite a lot weaker than S-waves. So although they arrive 2 km/sec sooner, humans are less likely to notice them than are animals. Thus, with their heightened sensitivity, it's expected that cats and dogs and other animals will prick up their ears or jump to their feet or otherwise express surprise a few seconds before humans notice an earthquake. This is thoroughly documented and well-supported by basic science, and is not in any dispute. It's also quite distinct from the claims made by those who believe animals have some additional sense that allows them to predict earthquakes well in advance, which is what we're going to study today. We just want to make sure we're not confusing animals' actual ability to detect tiny perturbations better and sooner than humans with the claims that they have a significant predictive ability — the non-ambiguous detection of an earthquake hours, days, or weeks in advance.

For a long time, one of the better-known methods for predicting earthquakes came from Northern California. It had been a popular belief, trusted and promoted by the local press, that when there was a significant rise in the number of "Lost Pet" reports in the San Jose Mercury News, there was an 80% chance of an earthquake with a magnitude between 3.5 and 5.5 within 110 kilometers of San Jose, within some certain amount of time. Widespread repetition of this belief led to a general acceptance among many people. Finally, in 1988, one seismologist had had enough, and published a detailed statistical analysis of the claims and found that there was not the slightest bit of truth to it.

But all this really says is that if animals can sense quakes, escaping from their owners in San Jose isn't the way they show it. An interesting proposal is the idea that animals would have evolved behavior to protect them from earthquakes. We know that many animals are able to sense P-waves a few seconds before the destructive part of the earthquake hits. What would be required for an evolutionary behavior to form would be the deaths of some animals who fail to react during those few seconds. There are not really all that many ways for an earthquake to kill an animal. Burrow collapse is one and tsunami action is another. If, over countless millennia, enough animals left burrows and ran for higher ground whenever faint vibrations were detected, it is conceivable that such a behavior could form. However, when we read the anecdotes of altered animal behavior prior to an earthquake, the reported behaviors are all over the map, but they are not either of these two behaviors.

This raises an important point, which is the inherent unreliability of almost all reports of animal behavior prior to a quake. It is that such behavior is almost never noticed, recalled, or described until after the quake, at which point the reports are uselessly biased by expectations and ex post facto analysis. There is, however, one part of the world where seismologists do pay attention to such reports, and log them, year round: Japan and China. Let's find out why.

Only once has animal behavior been claimed to have successfully predicted an earthquake: the 1975 Haicheng earthquake in China. An evacuation order was issued, and estimates have been made that as many as 100,000 lives were probably saved. There are numerous reports of animals behaving strangely in the days leading up to the quake. In nearly all its mentions in the literature, this prediction has been held up as a shining example of the plausibility of earthquake prediction, and of the role of animals in particular.

However, under closer scrutiny, the story largely falls apart. For some weeks, the region had been getting struck by a long series of foreshocks, any one of which is statistically just as likely be the last in a cluster of earthquakes. During this whole period, Chinese officials were making various predictions of larger quakes and issuing various evacuation orders. They were almost all false alarms, and this was a pattern that had long been common in China and continues through today. Nearly all the predictions are wrong. It just so happened that in this one case, a major earthquake did follow. If it hadn't been for the series of smaller quakes — which do not precede most large earthquakes and are not predictable themselves — it's unlikely that there would have been evacuation orders preceding the large quake. As far as what role the animals may have played, all information reported about this has been anecdotal. Not even the Chinese themselves have established any kind of guidelines for what constitutes predictive behavior or how it would be distinguished from non-predictive behavior.

Even today, the vast majority of research into the animal predictions comes from China, and also from Japan, which lacks even China's one example of a successful earthquake prediction. In those two countries, the belief that animal behavior is a reliable predictor of earthquakes is nearly universal, supported mainly — it appears — by little more than the confirmation bias established by the Haicheng case, as well as a lengthy cultural tradition.

The myth continues to be recycled in the news, and one strong reason is that every so often, a fringe paper gets published somewhere making highly suspect claims that animals have been been demonstrated to have foreknowledge of earthquakes. Whenever such a paper gets published, it invariably gets picked up by mass media news sources, further distorted and exaggerated, and duly impacts the public's perception of the phenomenon.

One notable example was in 2015, when headlines flashed around the world — in major news outlets and consumer magazines — that scientists have found animals can predict earthquakes up to three weeks in advance. The source was a paper published in the journal Physics and Chemistry of the Earth titled "Changes in animal activity prior to a major (M = 7) earthquake in the Peruvian Andes". I read this paper several times, and my best description of it is "miraculously bizarre". The paper centered around a statistical analysis of the number of animals captured on game cameras inside one Peruvian national park over a 30-day period prior to a magnitude 7 earthquake, and found that the number of animals spotted was lower than in other 30-day periods that did not precede large quakes. They even concluded the most likely cause for the reduced animal activity, which I must quote: "Positive airborne ion injection at ground-to-air interface." The only comparison of this time period to others was the earthquake. They did not take into account other factors that may have a much larger impact on animal activity, such as weather or season.

The authors began with two unsupported assertions, first that "positive airborne ions" are "known to be aversive to animals"; and second, they repeated a variation of a plausible notion popular among ball lightning and earthquake light enthusiasts, which claims that earthquakes produce massive electrical charges in the ground (though in this case they said that charge fluctuates and produces Ultra Low Frequency radiation). They correlated this to a space weather event, basically nighttime aurora activity, associated with Very Low Frequency radiation, and suggested some unconvincing connection between the two. This event was about a week before the quake, when they said animal activity began to be at its lowest.

In short, the entire paper was a word salad combining bits and pieces from a number of unrelated fields, many of which stem from the world of pseudoscience, and then drew a causal effect to a decline in animal activity prior to an earthquake — ironically, exactly the opposite of what nearly all other advocates claim, which is that animal activity increases before an earthquake. It was an outlier result among a community of fringe ideas, and yet it was sufficient to saturate the world's news with a report that animals can indeed predict earthquakes three weeks in advance.

This paper should stand as a powerful reminder that we should always be highly skeptical of astounding new science discoveries reported in the mass media.

What we're left with is a series of somewhat far-out answers in search of a question. The purpose of a theory is to explain an observation, and in this case, we don't really have an observation. We have countless anecdotes of unusual animal behavior, but they vary so widely — and all are well within the range of normal animal activity — that it's by no means clear that an explanation might be needed. We lack both a plausible mechanism by which animals could predict quakes, and evidence that they might do so. Until either of those changes — which, considering the enormous amount of research that's been thrown at this so far, seems unlikely — the idea that animals can predict earthquakes is on very shaky ground.

By Brian Dunning
Follow @BrianDunning