by Franklin Wolfe
figures by Abagail Burrus

Over the past half-century, earthquakes have been the leading cause of death from natural disasters and have imposed dramatic cultural, economic, and political impacts on society. Compounding their inherent physical hazard is how they strike suddenly without obvious warning, and how they possess a ‘fatal attraction‘ for humans—most of the world’s largest cities lie in areas of major seismic activity.

As a doctoral candidate in the Department of Earth and Planetary Sciences at Harvard University with a focus on understanding earthquake phenomena, the most common question people ask me is, “So, when is the next big earthquake going to happen?” Unfortunately, the answer is somewhat dissatisfying because the truth is, we really don’t know.

Why is this the case? What is so special about earthquakes that make them so difficult to forecast? Will we ever uncover a reliable prediction method? Answers to these questions are the Holy Grail of earthquake science and, if answered, could protect society from the next big earthquake catastrophe.

What is an earthquake?

The Earth’s rigid outermost layer consists of seven major pieces called plates. These plates move around because they are riding on top of a slowly-flowing layer called the mantle, similar to saltine crackers floating on top of a bowl of thick soup. These plates move at about the same rate as your fingernails grow (i.e., centimeters per year). When these plates collide with, pull apart from, or slide against each other, energy gets stored in the rigid outer layer—this is a similar phenomenon to compressing a spring. This energy is released when it surpasses a certain threshold, causing the ground to shake. This shaking of the Earth’s surface from a sudden release of energy is called an earthquake.

The majority of the world’s earthquakes occur along the plates’ boundaries, such as along the outer edges of the Pacific Ocean, colloquially known as the Ring of Fire (Figure 1). The tremendous forces with which these plates crunch past each other, however, extend beyond the immediate vicinity of the boundary. For example, the crust beneath Los Angeles, which lies ~30 miles from the active plate boundary (i.e., the San Andreas Fault), is riddled with cracks. These cracks, which are called faults, store and release energy in the form of earthquakes. California’s crust is similar to a smashed glass window pane in which the San Andreas Fault is just the most prominent crack.

Figure 1: The Ring of Fire. An area of particularly high earthquake activity (orange) along the outer edges of the Pacific Ocean is known as the ‘Ring of Fire.’ As seen in the inset, California’s crust is riddled with cracks known as faults.

Why is it so difficult to forecast the next big ‘quake?

Let’s start with a simpler part of the question: where will the next ‘big’ earthquake occur?

It has been shown that larger faults typically give rise to larger earthquakes. Theoretically, if all of the faults were mapped, then we should be able to put constraints on the strongest possible earthquakes a given region would experience. This is important because the energy released by earthquakes can vary by a factor of quadrillions.

However, estimating fault size and the corresponding energy released is not always so simple. Faults often exhibit complex geometries, which makes it complicated to model the area of the fault. Additionally, faults can rupture in tandem: thirteen different faults failed at the same time during the Kaikōura Earthquake of 2016 in New Zealand. Furthermore, recent history has demonstrated that earthquake size does not always correlate with damage; depending on where it occurs, a moderate-magnitude earthquake can be more devastating than a ‘big’ one. For example, the 1994 Northridge, CA magnitude 6.7 earthquake resulted in major property damage and loss of life, whereas the 2018 Fiji magnitude 8.2 earthquake, which was 178 times stronger, did not cause any damage. Thus, the magnitude of earthquake does not tell the whole story.

Now for the more complicated part of the question: when will the next big earthquake occur?

Timing is the most difficult challenge in the game of earthquake prediction. In fact, two of the theories that inform our (best) predictions could be flawed. The first theory is called Elastic Rebound Theory, which states that the Earth’s crust will bend and deform under intense stress until, eventually, it breaks under the strain. Slippage along the break (i.e., an earthquake) allows the rock on each side to rebound to a less deformed state and release the stored energy, allowing the process of accumulating strain to begin anew. The second theory is called the Characteristic Earthquake, which describes how the most studied earthquake-generating faults seem to have distinct segments. These segments repeatedly rupture to produce earthquakes of similar magnitudes after accumulating a similar amount of strain in the intervening period between earthquake events. Assuming these two theories always held, you could predict when the next earthquake would happen based on 1) the location of greatest unrelieved strain, 2) the time since the last earthquake on the fault, and 3) precise knowledge of the fault zone (which we may never achieve for many areas).

In 1985, this framework led earthquake scientists to believe that the Parkfield segment of the San Andreas Fault was overdue for an earthquake. The most sophisticated monitoring effort in the world was deployed to capture it in action. Scientists stated confidently that the next earthquake would hit by 1993 at the latest. However, the earthquake occurred over a decade later in 2004 and without any warnings. This type of discordance between predictions and data is common, and no reliable prediction method has been discovered despite numerous attempts. Thus, even the two most prevalent working theories of earthquake prediction, which are grounded in science and useful for understanding earthquake phenomena, are not fully reliable for prediction purposes.

What can we do?

Today, prediction methods are primarily focused on probabilistic earthquake forecasting, which is the statistical assessment of general earthquake hazard in a given area over a certain time frame. Probabilistic forecasting concerns the odds at which an earthquake might occur, while the earlier technique of deterministic prediction involves specifying exactly when an earthquake will occur. Probabilistic forecasting can provide warnings to areas that may be more prone to earthquake risk, allowing them to bolster their earthquake resistance with improved infrastructural designs and emergency measures before a potential future ‘quake occurs.

The probabilistic earthquake prediction technique is currently being employed in California through a model called the Uniform California Earthquake Rupture Forecast 3, which provides authoritative estimates on the likelihood of earthquake fault ruptures throughout the state. As inputs, the tool includes the geometries of all major faults in the region, as well as the most current data on how often earthquakes occur and how strong they are when they do.

Figure 2: An earthquake early warning system in action. When an earthquake begins, fast-moving P waves are detected, and emergency information can be disseminated before the arrival of slower-moving surface waves.

A second promising development is the earthquake early warning system. Upon detection of an earthquake, it provides a real-time warning of seconds to minutes for neighboring regions that might be affected. This system takes advantage of the different speeds of seismic waves that make up the energy radiating from an earthquake. In short, if the system detects the first arrival of the fastest waves, known as P waves, before the arrival of the more-dangerous, slower-moving surface waves, an alarm can be triggered (Figure 2). Using high-speed automation, even a few seconds of warning could be enough to stop machinery, such as trains and elevators, and alert people to get to safety.

Currently, the rail system in the San Francisco Bay Area (BART) uses an earthquake early warning system to automatically slow trains when an earthquake occurs. The system, called ShakeAlert, is now being implemented in California, Oregon, and Washington. Dr. Richard Allen, a seismologist at University of California Berkeley is one of the leaders in this field and helped develop the ShakeAlert system. Allen is developing a smart phone-based detection system that has shown promising early results in densely populated regions.

Mexico City residents are alerted by an Earthquake Early Warning System siren to exit a building just before it collapses.

However, falsely warning of an earthquake that never occurs is costly due to the unnecessary activation of emergency measures and the potential disruptions to commerce and everyday life. And similar to crying wolf, it could undermine the credibility and effectiveness of future warnings. To circumvent these issues, artificial intelligence (AI) technologies could be implemented to detect patterns and signals in the earthquake prediction data that humans cannot see. According to Dr. Brendan Meade, a professor in the Earth and Planetary Sciences Department at Harvard University, the way we approach problems in the geosciences may be reversed in the future. Instead of using equations to model a given system, we may use AI to search for an answer and then try to understand what it means. The application of AI to earthquake prediction is still in its infancy, so time will tell whether this approach is effective.

So, when is the next big earthquake going to happen? We may never know, but I remain optimistic that we can limit earthquake hazards to society. Whether the solution is found through AI, earthquake early warning, or continually improving probabilistic models through better understanding of the earthquake process, there is much improvement that can be made to increase our resiliency.

Franklin Wolfe is a doctoral candidate in the Earth and Planetary Sciences program at Harvard University. Connect with him about earthquake hazards, energy resources, or NBA playoff basketball on LinkedIn.

Abagail Burrus is a third-year Organismic and Evolutionary Biology Ph.D. student who studies elaiophore development.

For more information:

  • Earthquake Debates: In the 1970s, scientists were optimistic that a method for earthquake prediction would soon be found, by 1997, the consensus had turned. As a result, a debate among researchers in the journal Nature ensured, which can be found here. The participants concluded that deterministic, short-term earthquake prediction is unrealistic, but agreed that time-dependent seismic hazard could be justified on both physical and observational grounds.
  • Scientists Prosecuted for False Prediction: Despite the challenges of earthquake prediction, this has not stopped legal authorities from prosecuting scientists and politicians for failing to provide adequate warning. After the L’Aquila earthquake of 2009 in Italy, seven scientists and technicians in Italy were convicted of manslaughter for failing to classify the hazard of the region. They have since been acquitted after a seven-year appeal process.
  • United States Geological Survey Earthquakes Page: The USGS and its partners monitor and report earthquakes, assess earthquake impacts and hazards, and perform research into the causes and effects of earthquakes. Click here for maps of the latest earthquakes, compilations of information about significant earthquakes, data on historic seismicity, frequently asked questions, summary posters, and more!

One thought on “Predicting the Next Big Earthquake

  1. The complexity is unimaginable, though we may have not gotten the solution over decades, but we shall unravel the mystery soon with our complex divine intelligence and with the help of artificial intelligence. Rid on… Please I need help Admin, I wish I could do my in public health at the Harvard University because I deserve a great knowledge impact by proof. Franklin wolf. God bless your effort in trying to make the earth less harful and habitable.

Leave a Reply

Your email address will not be published. Required fields are marked *