It’s currently impossible to predict earthquakes before they occur, but scientists are inching closer to doing so with new and innovative ways to monitor the movement of the Earth’s crust. In this excerpt from When Worlds Quake: The Quest to Understand the Interior of Earth and Beyond (Princeton University Press, 2026), author Hrvoje Tokarčić, Chair of Geophysics at the Australian National University, delves into why earthquake prediction is so difficult, highlighting the Parkfield experiment, in which scientists waited nearly 20 years for an earthquake to occur on the San Andreas Fault.
Approximate answers to these comments can be given with the following targeted questions. “We still cannot defeat malignant diseases, but should we stop researching because of that?”
you may like
We are used to discussing the causes of earthquakes after every event, especially in places around the world where earthquakes occur. The frequency of earthquakes is debated, and some people often claim that they can recognize an upcoming earthquake by something else. People tend to believe that earthquakes are easier to explain and, of course, more predictable than physical forces within the Earth: full moons, planetary conjunctions, heavy rainfall, bone pain, over-exploitation and greed of the Earth’s resources.
Visit Parkfield, a small and picturesque town of just 18 people, located near the center of the San Andreas Fault, between San Francisco and Los Angeles, California in the 1970s and 1980s. You’re probably wondering why. Well, this small town is known in the seismological world for its turbulent geological history. This means that Parkfield has experienced a significant earthquake on average every 22 years since the mid-18th century.
However, it was interesting to note that the seismograms recorded for the 1922, 1934, and 1966 earthquakes were nearly identical, with the lines in one seismogram oscillating from those in the other. Additionally, the 1934 and 1966 earthquakes had foreshocks that occurred about 17 minutes before the main shock, and their seismic records were very similar.
You wonder how that is possible. Such similarity in seismic records is only possible if the same fault plane is always activated and recorded in sufficiently long waves using the same equipment. Of course, the shorter the wave, the greater the difference. In other words, the source, the earthquake, and the receiver, the seismometer, are in fixed locations, and the waves propagate between them through the same material. In other words, we have a perfect natural laboratory in which experiments are carried out. You just have to wait long enough.
Scientists therefore had a good map to study the mechanisms of earthquakes that sometimes recur on well-monitored active faults. Since the mid-1980s, they have installed all kinds of equipment near Parkfield and along the fault. A powerful seismometer, then a strain meter to measure the deformation of rock 650 feet (about 200 meters) deep along the fault, a magnetometer to measure the strength of the magnetic field, a creep meter to measure surface displacement along the fault, and other scientific “weapons”. They predicted with 90-95% confidence that the next earthquake would occur between 1985 and 1993. Some of the main questions are:
1. How is stress on a fault distributed spatially and temporally due to crustal deformation before and after an earthquake?
2. Do earthquakes repeat at an average time interval, or is each earthquake unique and has its own story?
you may like
3. How do the structures of faults and surrounding rocks affect the nucleation of small earthquakes and the probability of larger earthquakes and their temporal and spatial distribution?
They wondered what the surface deformation we measured could tell us about the stress distribution on the fault, and were hoping for a positive result that would confirm their predictions of earthquake occurrence from 1985 to 1993. They kept waiting. At the time, I was working once a week with a colleague at the U.S. Geological Survey’s California office in Menlo Park, northwestern Silicon Valley, where I could observe scientists participating in experiments.
Eventually, a magnitude 6.0 earthquake occurred in Parkfield, but not until 2004. We greeted this earthquake, which has received the most attention and research in human history, with a big question mark hanging over our heads. It occurred 11 years after the predicted time. It’s devastating. That’s why the “Parkfield Experiment” left a bitter taste of disappointment in my mouth. However, as the saying goes, only those who are not afraid of failure will ultimately succeed. Research continued.
Why is earthquake prediction so difficult? Each fault is different, some of which we know about, many of which we don’t. Earthquake catalogs do not go back far enough, and after all, underground structures are completely invisible to us.
We do not know how deep the fault extends, whether it is flat or curved, whether its surface is smooth or rough, whether and where it contacts other faults, the chemical composition of the rocks on one side of the fault and the other, or its physical properties such as strength and porosity. We do not know exactly how the deformation of the Earth’s surface that we observe is related to deformation and stress at the depth of faults. Many other factors are also unknown. Although predictions are possible, they are probabilistic in nature and must be taken with a grain of salt. So how should we proceed?
Not everything is so negative. The first good news is that most countries have seismic hazard maps. They are well made, but of course they need constant updating. The other good news is that based on basic knowledge of physics and the propagation of seismic waves throughout the Earth’s interior and surface, we can predict how the ground and some buildings will behave during an earthquake, which is already a huge advantage.
This is possible thanks to basic science and seismological research into the nature of the underground, just as radiologists are able to irradiate the inside of the human body. Ironically, earthquakes help us because they act as a source of waves that illuminate the Earth’s interior. Advances in engineering, construction, computer science, and numerical methods have made it possible to predict the behavior of infrastructure during earthquakes. In any case, these hazard maps serve as input to engineers, builders, and insurance companies.
After all, the most positive thing is that modern research, including laboratory models and artificial intelligence, is being carried out around the world in the direction of one day being able to predict earthquakes. Of course, this does not mean that there is no significant investment in science and technology, and it will need to continue to develop. This may get us to the point where thousands or even millions of microsensors must be installed on every fault in the Earth’s interior to monitor strains in real time.
In a sense, we will have a “crystal ball” – insight into the dynamics and future behavior of the fault. In fact, we are already doing it today, but only by tracing the surface of the Earth with the help of satellites. InSAR, LIDAR, and GPS are just some of the networks and techniques used to learn where the Earth’s crust is most stressed by surface deformation.
The mechanism of stress or tension build-up in faults is still under investigation. Presumably, the hot rocks of the Earth’s continental crust below a depth of about 9.3 miles (15 kilometers) are ductile, and this rock mass “flows” at a faster rate than at the surface, but earthquakes do not occur, thus bending the top of the Earth’s crust and increasing stress along fault surfaces. However, it is not yet known how this stress is distributed in space.
In addition, laboratory experiments at high pressures and temperatures provide insight into how hard rocks are and how strain and stress are related. By excavating around the fault, the chemical and physical structure of the soil is examined. Old tree trunks will be explored and excavations will be carried out to detect past earthquakes from rock samples.
Investments are being made in using seismic waves and tomography to study the Earth’s deep interior and the mechanisms of earthquakes. Investments are also being made in improving mathematical geophysics, machine learning, and techniques for processing vast amounts of digital data. Investments are also being made in warning systems based on P-wave detection. Even seconds of warning before an S-wave arrives is critical to saving people and infrastructure. Similarly, investments are being made in modern architecture that is earthquake resistant.
But the bottom line is that unless we want to move to a stable part of the continent, somewhere in Siberia, the permafrost regions of northernmost Canada, or remote parts of the Australian outback where earthquakes are rare, we need to learn to live with earthquakes.
Adapted from When Worlds Quake: The Quest to Understand the Interior of Earth and Beyond. Copyright © 2026 by Hrvoje Tkalčić. Reprinted with permission of Princeton University Press.
Source link
