Every year the planet’s seismometers locate somewhere around 12,000 to 14,000 earthquakes, and the great majority are small. Why the numbers fall away so steeply as magnitude climbs has been a long-running puzzle for seismologists, and a University of Michigan team now argues that part of the answer sits in rock that most models leave out: the cracked, weakened zone on either side of a fault.
Read next on Morning Overview: Pieces of a Learjet’s tail landed miles from the Florida crash that killed all nine aboard
The work, published in Science Advances on Nov. 18, 2026, was led by Peng Zhai, who built the model during his doctoral work at Michigan and is now a postdoctoral researcher at Utah State University, with Yihe Huang, a Michigan professor of earth and environmental sciences, as senior author. Their paper is titled “Small and Large Earthquakes Driven by Inelastic Deformation in Weak Fault Zones.”
The gap in single-fault models
According to the university’s news release, models that predict earthquakes mostly count only the slip along a main fault. The Michigan model also includes the neighboring rock that becomes damaged during earthquakes, and it follows the full earthquake cycle: stress buildup, nucleation, rupture and the healing of the damaged rock, over timescales from seconds to centuries.
Huang said the damaged rock is rarely considered in such models. Zhai gave a practical reason, saying of the simpler approach: “It’s easier to compute.”
The release says the model reproduces three features of real earthquakes at once. It explains why small earthquakes are far more frequent than large ones, why quakes of different sizes release about the same stress per unit of fault area, and why the energy needed to rupture a unit of fault area rises with earthquake size. A single framework producing all three is the claim that sets it apart from models built to explain one at a time.
How cracked rock caps a rupture
The mechanism the authors propose is that the surrounding rock soaks up energy. A large rupture that would otherwise run unchecked pushes into damaged material, which deforms inelastically and absorbs some of that energy, so the rupture can stall as a smaller quake. The release describes the results as consistent with the long-standing weak-fault-zone hypothesis, and the authors write in the paper that their results suggest the strength of the fault zone can be very low.
An earlier abstract from the same group, presented as a poster at the 2025 Southern California Earthquake Center annual meeting by Zhai, Huang and Jean-Paul Ampuero, lays out the same logic in more technical terms. It states that “low material strength within the fault zone is required for generating multiscale seismicity” and says the simulated fault zones also yield cascades of foreshocks, mainshocks and aftershocks, plus a power-law distribution of magnitudes. That abstract is a conference poster with no quantitative results on its page, so its details are not a substitute for the journal paper.
The magnitude pattern the model targets is among the best-known in seismology. An IRIS fact sheet notes that the yearly count falls about tenfold with each step up in magnitude, while each step releases roughly 30 times more energy. The model aims to explain why that ratio holds from tiny tremors to destructive ones.
What Michigan found in Japan earlier
The same group has found weak, damaged rock in the field. A February 2025 Science Advances study of the eastern Kanto region of Japan, which includes Tokyo, was led by Huang with Satoshi Ide, Aitaro Kato, Keisuke Yoshida, Chengxin Jiang and Zhai. According to a ScienceDaily report, it found that the plate boundary there holds foliated, damaged rock, and Huang called that “the main takeaway”: “the subduction zone plate boundary contains much more foliated, damaged rocks.”
The earthquakes in that region are small and occur about 60 to 70 kilometers deep, and the region produces roughly half of the quakes felt in Tokyo. The authors suggested that rock properties may contribute about as much to earthquake generation as the stress on a fault does. Huang also argued that mapping specific active spots could eventually improve earthquake warning systems.
Limits of the fault-zone model
A better physical model of the earthquake cycle does not by itself produce a new hazard map. The California long-term forecast, UCERF3, produced by the California Geological Survey with the U.S. Geological Survey and the Southern California Earthquake Center, is described by the state as a rupture forecast that estimates where and when the ground might slip, not a ground-shaking model. Nothing in the Michigan release says that forecast or any other official one has been changed to use the new physics.
The release also does not report real-world tests of the new model against a specific fault. Zhai’s own framing in the release is the modest one: better understanding of the small earthquakes could clarify the link between earthquakes of different sizes.
Whether other groups can reproduce the size distribution with the same weak-fault-zone assumption, and whether field measurements like those in Kanto confirm it elsewhere, will decide how far the idea travels. The paper, with the DOI 10.1126/sciadv.aef7106, is the place to see the model’s parameters.
This article was produced with the assistance of AI and reviewed by Morning Overview editors prior to publication.
More from Morning Overview
- 11 engines built to run well past 200,000 miles
- A dangerous unpatched Windows exploit is now public and can hand attackers full control
- Microsoft says its November Windows security updates are breaking printing for some users
- Older Teslas mostly keep their batteries, and here’s what wears out instead