Earthquake Prediction

by Brian Dunning
February 28, 2023

Is it possible to predict earthquakes, and if so, how is it done? Are there more than just one way to do it? Is it something deep underground, or is it weather related, or does it have something to do with gravitational kneading of the Earth by other planets? Considering their profound impact as natural disasters, earthquakes are a serious topic. In parts of the world where they're common, they can be either deadly or relatively harmless, depending on the preparedness level of the people who live there, particularly on how well they've constructed their buildings for seismic resistance. There are some out there who claim the ability to predict specific earthquakes, to enough detail that, if correct, people can evacuate when needed — an extraordinarily valuable service — if it's true. Today we're going to look at the evidence and the history.

There has probably never been a serious earthquake in modern history that wasn't followed with someone claiming to have predicted it. Case in point, perhaps obviously: on February 6, 2023, a magnitude 7.8 earthquake struck along the East Anatolian Fault System, a confluence of the Anatolian, Arabian, and African tectonic plates. Tens of thousands of people in Turkey and Syria lost their lives; at the time of this writing, we're not close to having a total count. That part of the world is well known to be a seismically hazardous area; however, the East Anatolian Fault itself is relatively quiet, this particular quake being a rare event. It's usually the other faults around that are the culprits in major quakes in the region.

Perhaps the most visible claimant to having predicted it was a Dutch YouTuber who believes planetary alignments are the true cause of earthquakes, and that by paying attention to the positions of the planets, one can predict earthquakes — watching something he cryptically calls "planetary constellations" and "lunar geometry". When I last checked, a tweet he made after the quake boasting of his prediction had more than 12 million views. People were replying, asking for advice, would there be aftershocks, would there be tsunamis. His tweet was also followed by savage criticism of his false claim of prediction ability during an event where so many had been killed, and were currently still dying under rubble. Followers of both sides engaged in explosive and vituperous argument.

Now it's important to note that his earlier tweet didn't actually predict anything at all — he simply said that sooner or later a quake would happen in this, one of the world's most active seismic areas. That's no different from saying sooner or later there will be big waves at Waimea Bay.

The scientific fact — as every legitimate seismologist on the planet will tell you — is that specific earthquakes cannot be predicted. We do know a lot about seismically active areas. Geologists are well aware of the parts of the world where faults are active and earthquakes are more common, and they have access to a certain amount of data about the current state of a given fault — but that information is largely limited to what can be measured on the surface. Extending deep underground beneath every fault is a generally understood geological structure, but with countless details and complexities that are unknowable. Every time a fault moves, the two plates end up in a slightly new position, with an unknowable new amount of friction holding them in place, and unknowable pressures from continued plate movement.

There is an excellent YouTube demonstration by seismologist Dr. Ross Stein, in which he places a brick on high-friction sandpaper, then attaches a rubber band to it, connected to a line on a reel. When he cranks the reel (representing the constant underground movement of plates), it stretches the rubber band (representing the elasticity of the Earth's plates), until the brick overcomes the friction and skids forward a little bit. As he constantly cranks, the brick makes inconsistent moves forward. Sometimes it moves a lot, and it takes more time for the band to stretch before it moves again; sometimes it takes several little steps in closer succession. Every time the brick stops, it has a slightly varying new amount of friction holding it in place, with a new amount of strain on the rubber band. We can see general trends in the brick's overall progress across the table, but every specific jump it takes is unpredictable.

Consider how oversimplified Dr. Stein's tabletop example is compared to a fault in the real world, with its countless convolutions; every point of contact between the two plates with a different amount of strain and a different amount of friction. And, unlike Dr. Stein's constant turning of the crank, the pressures underlying Earth's plates are always changing. Some faults may never move again, and others may open in new places. Such changes are over very long timescales, but they are yet another layer of complexity about which it's possible to know only so much. It is a system in which chaos plays a dominant role.

A proper seismologist can tell you a lot about a seismically active region, but the one thing she cannot ever tell you is the size and time of the next slip of that fault. Fault zones are chaotic systems, not ordered ones. And in their scientific definition, chaotic systems are unpredictable — at least in the lack of an omniscient understanding of every particle involved, and more supercomputers than have ever existed.

Nevertheless, there exists among the general public a sort of vague assumption that earthquakes are predictable, to some degree or another. There is a belief that animals can tell when they're coming — a claim that we conclusively disproved in Skeptoid episode #646. There is also the existence of earthquake warning systems, which are real — but they don't involve prediction, only the detection of an active earthquake once it starts. Earthquake warning systems detect the P-waves, which are faster and typically less destructive than the much slower and more destructive S-waves that follow. Those S-waves are typically 2 km/sec slower than the P-waves that preceded them, so depending how far you are from the detection, an earthquake warning system can send you a text message, or trigger a siren in your neighborhood, or activate the emergency broadcast system on your television, a matter of seconds before you're in real danger. This can also contribute to a general belief in the predictability of earthquakes.

This belief among the public that earthquake prediction is possible led to one of the great travesties in the history of science communication. In 2009, a magnitude 6.3 earthquake struck L'Aquila, Italy. 308 people died. Officials charged and convicted six scientists and one former government official with manslaughter for having failed to predict the earthquake, and in 2012 they were sentenced to six years in prison. Two years later, six of the seven had their convictions reversed on appeal, but the convictions remain one of the most outrageous examples of science illiteracy at a high government level. When prosecutors and judges misunderstand science to the point that they go around prosecuting researchers for not having magical insight into things that are beyond our ability to know, it puts scientists everywhere in danger. Why would one continue doing their job trying to better understand risks, and help the country better prepare, when a possible consequence is imprisonment?

For reasons that are obvious, people have long sought to develop ways to predict earthquakes in order to avoid great loss of life:

• As far back as 1880, the British geologist John Milne worked with colleagues in Japan and developed the world's first practical seismograph, following a destructive earthquake in Yokohama. He proposed a number of phenomena that could be studied to see if they correlated with an upcoming earthquake, including small earthquakes, various types of weather, predictive animal behavior, tides, earthquake lights, and temperature changes of hot springs.
  
  
• In a paper following the catastrophic 1906 earthquake in San Francisco, geophysicist Harry Fielding Reid proposed a sort of primitive strain gauge. Build a series of piers 1 km apart perpendicular to a fault. Through observation, the angle could be studied; and if it changed by more than a factor of 1 in 2000, it might mean that a "strong shock" was imminent.
  
  
• While the majority of seismologists were writing that prediction was not and would probably not ever be possible, others still put their heads to it. Throughout the first half of the 20th century, many seismologists published proposed precursor events. None had been observed to correlate yet, but research was still advocated on animal behavior, strain gauges, and earthquake lights.
  
  
• In the 1960s, a broad movement emerged, driven by recommendations of the Ad Hoc Committee in the United States (1965), advocating wide implementation of testing instruments. Strain gauges, seismographs, volcano monitoring, radon gas emission monitoring, in addition to logging tides and weather — even fluctuations in the Earth's magnetic field — to look for correlations, all along the world's primary seismic belts (high-risk earthquake zones): "all possible indicators foretelling the occurrence of earthquakes" with the "greatest achievable sensitivity".
  
  
• In 1976, the US National Research Council recommended "The United States should now make a national commitment to a long-term program aimed at developing a reliable and effective operational earthquake-prediction capability."

For the next quarter century, enthusiasm grew and shrank, as so much data was collected, and so many potential predictors were identified; only for each of them to be lost in the noise as more and more data poured in, proving that none of them were reliable. To this day, of all the countless earthquake precursors studied, not a single one has been found to be dependable. Again, in all the staggering amount of data collected, only thing is for sure: chaos.

There is one type of earthquake foreknowledge that actually is real, and that's what seismologists today actually practice. It's called earthquake forecasting — although the terms forecasting and prediction sound like they mean the same thing, in seismology they have very different meanings. Earthquake forecasting means assessing the general seismic hazard level of a region, ideally working out the probable frequency and magnitude of likely earthquakes. If you're wondering if this is a good place to build, say, a hospital or a power plant, earthquake forecasting can tell you whether it is or not. Forecasting is also how we establish what the building codes should be in an area for earthquake resistance.

The more seismic data you have in a region, the better you can forecast its hazard level. Not only do you want to know its earthquake history, going as far back as you can, you also want to understand the geology, you want to understand where the faults are, how the plates are moving, what the elasticity is, a whole range of specialized knowledge that geologists spend entire careers trying to characterize.

Every time you don't die in an earthquake because you're in a building that was properly built to a well-informed building code, you can thank generations of geologists who dug holes and took core samples and pored over decades of seismographs.

Who you shouldn't thank is a nut on YouTube who is contemptuous of geology, and instead spreads misinformation according to his own meaningless notions.

By Brian Dunning
Follow @BrianDunning