Foreshocks, Mainshocks, and Aftershocks: What to Expect After a Large Earthquake

Quick Answer: What Are Foreshocks and Aftershocks?

They are positions in a sequence, not types of earthquake. The same rupture gets a different name depending on what happens next.

Label	What it means	When it is assigned
Mainshock	The largest earthquake in the sequence, so far	Provisionally, and it can be revoked
Foreshock	An earthquake that turned out to precede a larger one nearby	Only after the larger one arrives
Aftershock	A smaller earthquake following the mainshock in the same area	Days to years afterward

USGS puts the crucial limitation plainly: "An earthquake cannot be identified as a foreshock until after a larger earthquake in the same area occurs."

And the single most important thing to know, because almost everyone gets it backwards:

Aftershocks become less frequent over time. They do not become smaller. USGS: "The magnitudes of the aftershocks do not get smaller with time, only their rate changes."

A damaging aftershock weeks or months into a sequence is not a surprise. It is exactly what the statistics allow.

The Labels Are Assigned Backwards

When a moderate earthquake happens, nothing in the seismic record marks it as a foreshock. It looks like an ordinary earthquake, because that is what it is until something bigger follows. The label is applied retroactively, at which point its warning value is precisely zero.

This is not a gap in the monitoring network. It is a property of the phenomenon. During any unfolding sequence, nobody, including USGS, knows whether the largest event has already happened.

The 5 percent rule, stated carefully

There is a number for this, and it comes with three qualifiers that matter as much as the figure:

"Around the world, the likelihood that an earthquake will be followed by a larger earthquake nearby and within a week is about 5%."

That is worldwide, nearby, and within a week. Drop the week and it becomes a lifetime risk. Drop "worldwide" and it becomes a local number it is not. USGS deliberately leaves "nearby" unquantified, so any specific radius you see attached to it was invented somewhere downstream.

USGS also stresses that 5 percent is a baseline that moves: "That 5% foreshock probability varies with the activity level of an aftershock sequence." Unusually productive sequences carry a higher chance, quiet ones lower.

Do not flip it into reassurance either. "95 percent chance it is over" is the same statistic pointed the wrong way, and the residual 5 percent is the part that matters.

Ridgecrest, 2019: the textbook case

On the morning of 4 July 2019, an M6.4 struck near Ridgecrest, California. It was the mainshock. It was reported as the mainshock, and for about 33 hours that is what it was.

Then an M7.1 arrived on the evening of 5 July, and the record rewrote itself. The M7.1 became the mainshock. The M6.4 became a foreshock. USGS has recorded more than 3,500 earthquakes above magnitude 2 in the sequence, and the M7.1 was the largest earthquake to strike Southern California in twenty years. USGS notes it damaged communities already hit by the earthquake the day before.

Tohoku, 2011: when even a major earthquake is a foreshock

USGS uses the 2011 Japan sequence as its own illustration. An M7.3 struck on 9 March 2011. It was called the mainshock.

Fifty-one hours later, the M9.1 Tohoku earthquake and tsunami struck, and the M7.3 was reclassified as a foreshock.

An M7.3 is a serious earthquake anywhere on Earth. This is not a story about a subtle warning sign being overlooked. It is a story about a major earthquake providing no way whatsoever to know that something roughly 500 times more energetic was two days out.

Note the direction of the relabelling, which is commonly reported backwards: when a larger event arrives, the earlier events become foreshocks and the new event becomes the mainshock. The original mainshock is not demoted to an aftershock.

Two Foreshock Questions That Sound Identical

Consumer coverage muddles these constantly, and the confusion produces a genuinely dangerous impression.

The question	The answer
What fraction of large earthquakes were preceded by foreshocks?	Somewhere in the 10 to 50 percent range, depending on catalog sensitivity
What fraction of ordinary earthquakes turn out to be foreshocks?	About 5 percent

Both are true. They are about different populations, and only the second one is useful when a small earthquake happens near you.

The asymmetry exists because small earthquakes vastly outnumber large ones. Perhaps a third of big earthquakes had some warning tremor, and yet only about one tremor in twenty precedes anything bigger.

The research here is genuinely contested. A 2019 study by Trugman and Ross, using an ultra-sensitive southern California catalog of more than 1.8 million earthquakes, reported that 72 percent of mainshocks were preceded by significantly elevated foreshock activity. Within a year, van den Ende and Ampuero reanalysed the same 46 mainshocks and found only 33 percent, falling to 18 percent once background fluctuations were accounted for.

Even the optimistic side does not claim this is actionable. Trugman and Ross write that foreshock sequences are "diverse in character, a fact which may preclude real-time forecasting based on foreshock activity."

And the arithmetic finishes the argument. If about 5 percent of earthquakes precede something larger, treating every moderate earthquake as a warning generates roughly 19 false alarms for every hit. USGS is blunt about where this leaves prediction: "Neither the USGS nor any other scientists have ever predicted a major earthquake."

What Actually Happens After the Mainshock

The rate falls off as a power law

Fusakichi Omori worked this out in 1894 from felt reports of the 1891 Nobi earthquake. Aftershock frequency decays as roughly one over the time since the mainshock. The modern form adds an exponent, p, which is usually somewhere between 0.9 and 1.5 with a median near 1.1, and varies from sequence to sequence.

USGS states the practical version: "there are about 10 times as many aftershocks on the first day as on the tenth day."

The important structural point is that this is a power law, not an exponential. Exponential decay dies quickly. Power-law decay has a long, heavy tail, which is why aftershock sequences last so absurdly long.

One thing p does not depend on is the size of the mainshock. A bigger mainshock does not produce a slower-fading sequence. It produces a more numerous one.

The magnitudes do not shrink

This is the corrective worth carrying away from the whole article.

The rate drops. The magnitude distribution does not change shape. Aftershocks follow the same Gutenberg-Richter frequency-magnitude relationship as earthquakes generally, with a b-value near 1, and no clear tendency has been found for larger aftershocks to die off faster than small ones.

So the picture of a sequence gently tapering into harmlessness is wrong. What actually happens is that earthquakes of every size become rarer at the same rate. A large one late in a sequence is not anomalous, just less likely on any given day.

How big can the largest aftershock be?

You will see Båth's law quoted: the largest aftershock averages about 1.2 magnitude units below the mainshock. Treat that with real caution.

It is an ensemble average across many sequences, and the scatter is enormous. Observed differences run from 0 to 3 magnitude units from one sequence to another. A careful study of ten California sequences found a mean of 1.16 with a standard deviation of 0.46, with individual values from 0.6 to 1.9. For any single earthquake, Båth's law has essentially no predictive power.

There is a deeper problem too. Vere-Jones showed that if you simply draw the two largest events from one Gutenberg-Richter distribution, the expected gap is about 0.43, not 1.2. Later work concluded the observed 1.2 arises largely from how seismologists select mainshocks and aftershocks out of catalogs, not from any physical difference between them. It is closer to an artifact of bookkeeping than a law of rupture mechanics.

And certainly do not treat it as a ceiling on what comes next. The scatter alone reaches 0, meaning an aftershock essentially equal to its mainshock. More to the point, the earthquake you just felt may not be the mainshock at all: at Ridgecrest an M6.4 was followed by an M7.1, and at Tohoku an M7.3 by an M9.1. In both cases the sequence was relabelled rather than capped, and a 1.2-unit rule of thumb gives you no warning of that whatsoever.

Turkey, 2023: an "aftershock" that was a major earthquake

On 6 February 2023 an M7.8 struck near Pazarcık in southern Türkiye at 4:17 a.m. local time. An M6.7 followed eleven minutes later. About nine hours after that, an M7.5 struck near Elbistan.

USGS classifies that M7.5 as an aftershock. It ruptured roughly 161 km, about 100 miles, of the Çardak fault, a different fault from the M7.8, which broke nearly 354 km of the East Anatolian fault. Offsets on the Çardak fault reached as much as nine metres.

If your mental model of "aftershock" is a rattle that knocks a picture off the wall, that event should replace it.

More than 50,000 people were killed, with totals commonly cited between about 55,000 and 60,000 and counts widely believed to be undercounts. It is worth being precise about why: USGS attributes the severity primarily to a building stock that was extremely vulnerable to begin with, mainly older low- to mid-rise concrete frames with masonry infill, rather than to the two-earthquake structure of the sequence.

How Long Does a Sequence Last?

Longer than almost anyone expects, and the honest answer is that there is no agreed definition of when one ends.

The standard definition is the time for activity to fall back to the local background rate, at which point aftershocks may well continue but become indistinguishable from ordinary seismicity. Because that depends on the background rate and on how you draw the aftershock zone, published durations for the same earthquake can differ by an order of magnitude.

A concrete illustration: for the 2018 M7.1 Anchorage earthquake, an extremely well-instrumented modern sequence, a USGS report estimated it would take between 2.5 years and 3 decades for the rate to decay back to background. That factor-of-twelve range is not sloppiness. It is the actual state of knowledge.

Sequences running ten years or more have been documented since the beginning of the twentieth century. Omori himself reported his formula still fitting the 1891 Nobi aftershocks a decade later.

Duration also seems to depend on how fast the fault is being reloaded. Work by Stein and Liu found plate-boundary faults moving faster than about 10 mm per year produce sequences lasting on the order of ten years, while continental interiors deforming at under a millimetre per year can run for hundreds of years or longer. Their proposal that modern New Madrid seismicity represents aftershocks of the 1811 to 1812 earthquakes is a live scientific argument rather than a settled fact: Page and Hough countered that the required productivity would have produced far more magnitude 6 events than the historical record shows.

Where and How Many

Where. USGS defines aftershocks as occurring within one to two fault lengths of the mainshock. That is an operational window used to pull sequences out of catalogs rather than an observed physical boundary, and triggered earthquakes demonstrably occur beyond it. Measured aftershock zones are usually larger than the area that actually slipped, though the observed ratio spans roughly 0.5 to 5, and they expand outward roughly logarithmically as the sequence continues.

Depth matters here too: USGS notes that earthquakes deeper than 30 km are much less likely to be followed by aftershocks than shallow ones.

How many. Productivity rises roughly tenfold per magnitude unit of mainshock, but the variability around that is the real headline: the number of aftershocks varies by factors of 100 between mainshocks of comparable magnitude. Depth, plate-boundary type and faulting style all correlate with productivity. This is why no honest article can tell you how many aftershocks to expect from a given magnitude.

USGS Aftershock Forecasts Are Forecasts, Not Predictions

USGS issues public aftershock forecasts for mainshocks of magnitude 4 and above in the contiguous United States, and magnitude 5 and above in other US states and territories. It also computes them for some smaller earthquakes of particular public interest, such as events in densely populated areas. The first forecast typically appears about 20 minutes after the mainshock and is updated frequently through the first day, then on a slowing schedule.

Each forecast gives, over one-day, one-week, one-month and one-year windows:

• the expected number of smaller aftershocks likely to be felt (M3+ and M4+)
• the probability of aftershocks large enough to do damage (M5+)
• the probability of future moderate (M6+) to large (M7+) earthquakes

Note that the windows nest: the one-year figure includes the day, week and month.

After the 2025 M8.8 Kamchatka earthquake, the USGS one-week forecast gave a 2 percent chance of another M8+, 24 percent for M7+, and 96 percent for M6+. Those are probabilities of rates in a window, for that sequence at that moment. They are not predictions of specific earthquakes, and USGS defines a genuine prediction as requiring date and time, location, and magnitude, which nobody can supply.

The model behind most of these forecasts assumes every aftershock is triggered by the mainshock. USGS flags the limitation itself: it works well when only smaller aftershocks follow, and may perform poorly when a large aftershock triggers its own sequence, which is exactly what happened at Ridgecrest and in Türkiye.

What about swarms?

A swarm is a sequence of mostly small earthquakes with no identifiable mainshock. The largest event may fall in the middle rather than at the start, many events are of similar size, and activity may stay flat or even increase over weeks rather than decaying.

Swarms are common in volcanic and geothermal areas, where moving magma or fluids change stress in the crust, but they also occur on faults with no volcano involved. USGS is candid that there is no precise definition of when a mainshock-aftershock sequence becomes a swarm.

Staying Safe During a Sequence

The protective action during an aftershock is the same as during any earthquake. There is no special aftershock variant, and you should be suspicious of anyone who invents one.

Drop, Cover, and Hold On. Drop onto your hands and knees. Cover your head and neck with one arm, crawling under a sturdy table if one is close or next to an interior wall away from windows if not. Hold on until the shaking stops. Ready.gov states it directly: "Expect aftershocks to follow the main shock of an earthquake. Be ready to Drop, Cover and Hold On if you feel an aftershock."

Do not run outside during shaking, and skip the doorway. Official guidance is to stay put until the shaking stops. Doorways are not safer.

What genuinely changes during a sequence is the context, not the action. You may now be in or near a structure that has already been damaged, with objects and building elements loosened by the mainshock.

Damaged buildings are the real aftershock hazard

Ready.gov's rule is blunt and unqualified: "If you are in a damaged building, go outside and quickly move away from the building. Do not enter damaged buildings."

Engineering guidance goes further, and it is worth knowing where it comes from. The Applied Technology Council, in a technical brief funded by USGS, states that aftershock collapse risk "is highest for previously damaged buildings" and gives a specific first-day rule:

"During the first 24 hours, entry into seriously damaged buildings should be avoided in case the damaging shock is a foreshock and a subsequent event is the main shock."

That is professional guidance written for building officials rather than a public safety instruction, but the reasoning behind it applies to anyone: in the first day, you do not yet know that the biggest earthquake has happened.

In most Californian jurisdictions, buildings get inspected and posted under the ATC-20 system as green (inspected, no restriction), yellow (restricted use), or red (unsafe). The posting criteria explicitly account for aftershocks aggravating existing damage. This is a US system, not a universal one.

A green placard is worth understanding correctly. It means a rapid screening found no obvious safety hazard. It is not a structural certification, and postings can change after detailed inspection.

The Canterbury Earthquakes Royal Commission examined the CTV building in Christchurch, which was green-placarded the day after the September 2010 Darfield earthquake and collapsed in the February 2011 earthquake, killing 115 people. It is important to be accurate about what the Commission concluded: it attributed the collapse chiefly to design and construction deficiencies plus unusually intense February shaking, not to a building weakened in September and finished off later. The building was critically under-designed.

The transferable lesson is narrower and still valuable. The Commission found that damage visibly worsened between the two earthquakes, that the building manager assumed widening cracks were normal, and that the right response would have been to bring the engineer back to re-inspect. Occupants also reported the building felt noticeably "livelier" after September.

So: cracks that keep widening, a lean that is still progressing, or a building that starts moving differently are reasons to get a professional back in, not signs that things have settled.

The practical list

• Wear sturdy shoes. Broken glass and debris cause a large share of post-earthquake injuries.
• Expect aftershocks, and know where you would take cover in the room you are in.
• Shut off the gas only if you actually suspect a leak, since only the gas company can restore it.
• Check chimneys before use. They are a classic aftershock falling hazard.
• Outside a tsunami zone, the Earthquake Country Alliance advises evacuating only if the building is damaged or the surroundings are unsafe. An undamaged building can be a reasonable place to stay even without power.
• If you are trapped, text or bang on a pipe, cover your mouth, and use a whistle rather than shouting to conserve air and avoid inhaling dust.

If you are in a tsunami evacuation zone and feel strong or long shaking, do not wait for an official alert. Move inland or to high ground.

FAQ

Do aftershocks get weaker over time?

No, and this is the most common misconception about them. USGS states that aftershock magnitudes do not get smaller with time, only their rate changes. Aftershocks become less frequent, roughly ten times rarer on day ten than on day one, but a large one remains possible late in a sequence.

How long do aftershocks last?

Days to years, and there is no agreed definition of when a sequence ends. For the 2018 M7.1 Anchorage earthquake, a USGS report estimated 2.5 years to 3 decades before the rate returned to background. Sequences in slowly deforming continental interiors can run far longer.

Can an aftershock be bigger than the mainshock?

Yes, about 5 percent of the time worldwide within a week. When it happens the labels are reassigned: the new event becomes the mainshock and the earlier ones become foreshocks. The 2019 Ridgecrest M6.4 became a foreshock when the M7.1 followed 33 hours later.

Can scientists tell if an earthquake is a foreshock?

No. USGS states that an earthquake cannot be identified as a foreshock until a larger one occurs in the same area. During any sequence, nobody knows whether the largest event has already happened.

What is the difference between an aftershock and an earthquake swarm?

A mainshock-aftershock sequence has a clear largest event at the start and decays over time. A swarm has no identifiable mainshock, contains many events of similar size, and may stay flat or intensify for weeks. USGS notes there is no precise boundary between the two.

Is it safe to go back into my house after an earthquake?

Not if it is damaged. Ready.gov's guidance is to leave a damaged building and not re-enter. Engineering guidance from the Applied Technology Council adds that aftershock collapse risk is highest for previously damaged buildings and that entry into seriously damaged buildings should be avoided in the first 24 hours, because the earthquake may turn out to be a foreshock.

How many aftershocks will there be?

There is no reliable answer for an individual earthquake. Productivity rises roughly tenfold per magnitude unit on average, but varies by factors of 100 between mainshocks of similar size. For US events of magnitude 4 and above in the contiguous states, and magnitude 5 and above elsewhere in the US, USGS publishes an event-specific forecast, which is far more useful than any general rule.

Always Verify with Official Sources

Aftershock forecasts are updated as a sequence evolves. For event-specific forecasts and authoritative earthquake details, use the USGS Earthquake Hazards Program. For preparedness guidance see Ready.gov, for tsunami alerts see the tsunami warning centres, and follow your local emergency management agency during an active sequence.

Follow a Sequence with WeatherAI

When a sequence is unfolding, the useful thing is seeing the events accumulate in one place:

• Nearby and global earthquake notifications with magnitude thresholds you set, so a busy sequence does not bury you
• Magnitude and distance filtering on maps and lists
• Full event detail including depth and magnitude type
• USGS or EMSC as your data source
• Earthquake widgets for Home Screen and Lock Screen with your own minimum magnitude

WeatherAI reports earthquakes that have already been detected. It is monitoring rather than earthquake early warning, it is not a substitute for official emergency alerts, and no app can tell you whether the earthquake you just felt was a foreshock.

Download WeatherAI for iPhone →

---

Keep reading: why earthquake depth decides how much damage an earthquake does, why earthquake magnitudes change after the first report, and Richter vs moment magnitude. Definitions for aftershock, foreshock and earthquake swarm are in our glossary.