Where earthquakes break the rules
In 1944, Beno Gutenberg and Charles Richter discovered something remarkable: the relationship between earthquake magnitude and frequency follows a power law. For every unit increase in magnitude, there are roughly ten times fewer earthquakes. A magnitude 5 is ten times rarer than a magnitude 4, a magnitude 6 is a hundred times rarer than a 4, and so on.
The law is written as log₁₀(N) = a - bM, where N is the number of earthquakes
at or above magnitude M, a is a constant describing the overall rate, and b —
the "b-value" — describes the ratio of small to large earthquakes. Globally, b is approximately
1.0. But not everywhere, and not always.
I downloaded the complete USGS earthquake catalog — 1,007,324 earthquakes from 1970 to March 2026 — and asked: where and when does the law hold, and where does it break down? The b-value turns out to be a seismological thermometer: it measures stress. When b drops, stress is accumulating. When b rises, energy is being released in many small ruptures rather than building toward a big one.
First, let's verify that it works. The plot below shows the frequency-magnitude distribution for all one million earthquakes. The x-axis is magnitude, the y-axis is the cumulative count (how many earthquakes at or above that magnitude) on a log scale.
The fit is extraordinary: R² = 0.995. Across six orders of magnitude — from M4.7 earthquakes that happen thousands of times per year to M9+ events that occur once per decade — the relationship is almost perfectly linear on a log scale. The MLE b-value is 1.048 ± 0.002, right at the canonical value of 1.0.
This is one of the most robust power laws in nature. It spans from tiny tremors you'd never feel to the 2011 Tohoku earthquake that shifted Japan's main island eight feet east. The same law governs them all.
But the global average conceals enormous regional variation. The b-value is not a universal constant — it depends on the tectonic setting, the stress regime, and the local geology. Low b-values (more large earthquakes relative to small ones) tend to occur in regions of high tectonic stress. High b-values (many small earthquakes, fewer large ones) are associated with volcanic regions and areas where strain is released gradually.
The map below shows b-values computed on a 5°×5° grid. Red cells have low b-values (high stress, fewer small earthquakes relative to large ones). Blue cells have high b-values (lower stress, dominated by small events). Hover for details.
The pattern is striking: subduction zones (where one plate dives under another) tend toward redder hues — lower b-values — while spreading ridges and volcanic arcs trend blue. The Pacific Ring of Fire forms a near-continuous band of intermediate-to-low b-values, tracing the planet's most seismically active boundary.
Using a consistent completeness threshold (Mc = 4.5) for fair comparison, Alaska and California have the lowest b-values (~0.93) — slightly fewer small earthquakes relative to large ones, consistent with the high-stress subduction and transform boundaries of the Pacific Rim. The Himalaya has the highest (b = 1.17), suggesting strain is being released through many small events along the complex collision zone. The spread is narrow — 0.92 to 1.18 — but systematic: subduction zones run low, collision and volcanic zones run high.
The b-value also varies with depth. Shallow earthquakes (0-35 km) occur in the brittle crust where faults can lock and build stress. Deeper earthquakes occur in the mantle where deformation is more ductile.
The b-value is not just a spatial quantity — it evolves over time. Using a sliding 2-year window, we can watch the global b-value fluctuate from 1970 to the present. Major earthquakes appear as perturbations: they temporarily increase the b-value (by adding many small aftershocks) before the system relaxes back.
The most striking feature is the stability. Despite massive earthquakes — the 2004 M9.1 Sumatra event that killed 230,000 people, the 2011 M9.1 Tohoku earthquake that triggered a nuclear disaster — the global b-value stays within a narrow band around 1.0. The law is remarkably resilient.
When a major earthquake strikes, it triggers a cascade of aftershocks that follow their
own power law: Omori's law (1894). The rate of aftershocks decays as a power
of time: n(t) ∝ 1/(t + c)^p, where p is typically close to 1. Select a major
earthquake below to see its aftershock sequence.
The Omori exponent p ranges from 0.73 (2004 Sumatra — slow decay, persistent aftershocks) to 1.18 (2012 Wharton Basin — rapid decay). This variation reflects the rheology of the surrounding rock: hotter, more ductile material produces faster aftershock decay.
The most tantalizing aspect of b-value research is its potential as a precursor. The idea, supported by laboratory rock fracture experiments, is that b decreases before large earthquakes as stress builds on a fault. When the crust is highly stressed, it can only sustain large cracks, so the ratio shifts toward bigger events.
Select a major earthquake to see how the regional b-value evolved in the years before it struck.
The signal is noisy and retrospective — you can always find a dip if you look hard enough. The real test would be prospective prediction, and that remains elusive. But the pattern is suggestive: several of these major earthquakes were preceded by a measurable decline in the regional b-value. The 2010 M8.8 Chile earthquake and the 2011 M9.1 Tohoku earthquake both show clear b-value decreases in the years before they struck.
The Gutenberg-Richter law is not just a statistical curiosity. It emerges from the self-organized criticality of the Earth's crust — a system perpetually on the edge of instability, where stress accumulates slowly and is released suddenly across all scales. The b-value is the fingerprint of this critical state, varying with stress, temperature, rock type, and tectonic setting.
The law tells us that magnitude 9 earthquakes are inevitable. Not where, not when, but that they will happen. The same physics that produces the thousands of magnitude 3 earthquakes happening right now, somewhere beneath your feet, guarantees that once every few decades, the Earth will move in a way that reshapes coastlines and redirects history.
We cannot predict earthquakes. But we can measure the stress they leave behind, and the stress building toward the next one. The b-value is not a crystal ball — it's a thermometer. And in some parts of the world, the temperature is rising.