North Korea’s Nuclear Tests Left Faults Still Moving

North Korea’s final known nuclear test produced a magnitude 6.3 seismic signal beneath Mount Mantap on September 3, 2017. The mountain has not returned to its earlier seismic state.

A study published in Science on September 17, 2026, reports that earthquake activity around the Punggye-ri test site increased over subsequent years and spread along pre-existing weaknesses in the crust. Researchers identified 1,399 local earthquakes in seismic records covering 2008 through May 2025. That count spans years before and after the final explosion, so it should not be read as 1,399 earthquakes caused by the 2017 test.

The more consequential finding concerns the sequence in which the earthquakes developed. Sustained tectonic seismicity began around September 23, about three weeks after the explosion, and later increased in both rate and total seismic energy release. The researchers wrote that activity “expanded gradually as preexisting faults became active,” according to coverage quoting the paper.

That delayed, growing pattern differs from the short aftershock sequences previously documented at major Cold War test sites. It also complicates a basic task of nuclear monitoring: deciding whether a seismic signal marks a new explosion, the collapse of an old test cavity or an earthquake on a fault destabilized years earlier.

Six Explosions Changed the Starting Conditions

North Korea conducted six underground nuclear tests at Punggye-ri between October 2006 and September 2017. Tunnels beneath the 2,205-metre Mount Mantap placed the devices under hundreds of metres of rock, which helped contain the explosions.

The last test was the largest. Its estimated yield was 100 to 250 kilotons of TNT, and its seismic signal resembled a magnitude 6.3 earthquake. An underground explosion vaporizes and melts nearby rock, creates a cavity and transfers stress into the surrounding crust. As the cavity cools, its roof can fail.

In 2017, instruments detected a magnitude 4.1 event 8.5 minutes after the explosion. Researchers interpret that signal as the test cavity collapsing. Satellite radar also recorded substantial deformation of the mountain. Those observations describe an immediate mechanical aftermath: a large blast, a damaged cavity and a visibly altered peak.

The later earthquakes followed another timetable. Study co-author Kwang-Hee Kim told ScienceAlert’s account of the research that only a few events were reported immediately after the sixth test, while sustained tectonic activity began about 20 days later and intensified over the following years.

That distinction matters because seismic waves from an explosion pass quickly. Persistent earthquakes require a mechanism within the crust. The study’s proposed mechanism is stress redistribution: repeated explosions damaged rock around the test site and altered the forces acting on faults that already existed.

How Researchers Found Earthquakes Routine Monitoring Missed

The team examined 17 years of recordings from seismic stations in China and South Korea. It used matched-filter detection, a technique that searches continuous records for signals resembling known earthquakes. This can recover small events that routine detection systems overlook.

Researchers located 955 of the 1,399 detected earthquakes with enough precision to examine their spatial pattern. The events formed two roughly parallel structures extending north-northwest from Punggye-ri. One aligned with the projected continuation of a fault mapped south of the site. The other did not match a known surface fault, although its geometry resembled one.

Several observations therefore point in the same direction: sparse activity before the later tests, major deformation in 2017, delayed seismicity after the final explosion, rising activity through May 2025 and earthquake locations tracing fault-like structures. Together, they support the researchers’ interpretation that nuclear testing reactivated crustal weaknesses.

The evidence remains observational. Scientists cannot rerun six nuclear explosions under the same mountain while holding every geological variable constant, a limitation that geology shares with most investigations of rare disasters. Historical earthquake catalogues are also poor at capturing tiny events in remote areas, and the matched-filter method is more sensitive than older monitoring. Part of the large event count therefore reflects improved detection.

Even with that qualification, the baseline appears unusually quiet. Historical records identified one large crustal earthquake within 200 kilometres between 23 BCE and 1903 CE, a magnitude 6.7 event in 1810 about 100 kilometres away. Major catalogues recorded no crustal earthquakes within 50 kilometres from January 1904 until the September 2017 test. The more sensitive analysis did find sparse local activity beginning in 2013, after North Korea’s third test.

The cautious reading is that Punggye-ri was never guaranteed to be perfectly inert. The evidence instead indicates that its seismic behaviour changed across the testing period, especially after the largest explosion.

Why the Cold War Comparison Has Limits

Underground nuclear tests at the Nevada National Security Site and the former Soviet Union’s Semipalatinsk site in Kazakhstan also produced earthquakes. Those sequences were generally concentrated near the explosion and diminished over days or weeks. Mount Mantap’s delayed onset, expansion along fault-like structures and growth over several years depart from that familiar template.

The comparison helps define the anomaly, but it does not establish a universal rule for test sites. Rock type, fault orientation, prior damage, explosion yield and burial geometry differ from one location to another. Mount Mantap endured six tests under one mountain, culminating in an explosion large enough to deform the summit. Its response cannot simply be transferred to Nevada, Kazakhstan or a hypothetical future test site.

Cold War specialists used the term “tired mountain” for a test site fractured so extensively that it could no longer reliably contain radioactive gases. Scientific American reported that Mount Mantap now fits that description. It also quoted Columbia University seismologist Paul Richards, who was not involved in the study, asking whether the continuing activity could lead to larger earthquakes.

The study does not provide a forecast of when, where or how large such an earthquake might be. Rising seismicity and reactivated faults indicate a changed hazard, but they do not supply a deterministic countdown. Public information about the underground fault network is incomplete, and North Korea releases few details about Punggye-ri.

A Longer Memory for Nuclear Monitoring

Seismic monitoring has long relied on differences between explosions and tectonic earthquakes. An explosion begins with an outward pressure pulse; an earthquake usually begins with rocks shearing along a fault. Location, depth, waveform and aftershock behaviour add further clues.

Mount Mantap introduces a longer chain of attribution. A future signal near Punggye-ri could be a conventional earthquake on a regional fault, continuing activity induced by earlier tests, a cavity collapse or a new explosion. One event can also create conditions for another, leaving monitors with a geological history to reconstruct rather than a single waveform to classify.

That does not make verification futile. It makes long baseline records, improved event locations and physical models of damaged test sites more important. A new explosion would still carry diagnostic seismic features. Persistent induced earthquakes add background complexity, particularly when the country operating the site supplies little direct information.

The broader inference is limited but useful: underground nuclear testing can leave a seismic legacy longer than the aftershock periods observed at better-known Cold War sites. At Mount Mantap, the monitoring record suggests that the relevant unit of time is measured in years. The next suspicious signal from Punggye-ri will arrive with nine years of altered crust beneath it.