August 19, 2026
By Burak Oktenli
Key Takeaways:
Post-earthquake ionospheric disturbances detected by GNSS are established science, but claims of pre-earthquake TEC precursors remain unproven and face high scrutiny.
Apparent pre-quake signals often arise from analysis choices, natural ionospheric variability, or traveling disturbances that also occur on non-earthquake days, as shown by large control studies and re-examinations.
Genuine precursors require frozen analysis rules set in advance, rigorous negative controls, and prospective testing that distinguishes them from ordinary background behavior before they can be treated as early-warning tools.
Post-earthquake ionospheric disturbances are established science. Pre-earthquake precursor claims face a harder test: they must beat ordinary ionospheric variability, analysis choices, and untouched control data.
A large earthquake does not stop at the ground. The rupture launches acoustic and gravity waves upward through the atmosphere, and minutes later the ionosphere can register the event as traveling disturbances in electron density. Dense GNSS networks have imaged these effects after major earthquakes, including the 2011 Tohoku-Oki event. In some cases, the ionosphere has even been used to infer properties of the seismic source after rupture.
That part is established science. The scientific trouble begins when the arrow of time is reversed.
If the ionosphere responds after a major earthquake, could it also change before one? Could total electron content, or TEC, measured from GNSS signals become an earthquake precursor? The possibility is attractive for an obvious reason: short-term earthquake prediction remains one of geophysics’ most difficult problems, while GNSS networks continuously monitor the ionosphere over large areas.
The debate became especially visible after the 2011 Tohoku-Oki earthquake. Later that year, geodesist Kosuke Heki reported a positive TEC anomaly beginning roughly 40 minutes before the magnitude-9 event and argued that similar behavior might occur before other very large earthquakes. The claim was unusually testable because the underlying GNSS observations were public and the proposed anomaly had a specific temporal and spatial form.
Then came the difficulty that defines precursor science: the same data can look different depending on how the background is constructed. Masashi Kamogawa and Yoshihiro Kakinami argued that the apparent pre-earthquake enhancement could instead arise from the choice of reference curve and from a post-earthquake ionospheric depletion associated with the tsunami. Heki and others continued to defend a precursor interpretation. The disagreement was not about whether the ionosphere changed. It was about what portion of that change belonged before the earthquake, what counted as the baseline, and whether the feature was physically exceptional.
This distinction matters because an anomaly is not yet a precursor. A precursor is a prediction claim.
To deserve that name, a signal must do more than look unusual when plotted around a known earthquake. It must occur often enough before earthquakes, rarely enough when earthquakes do not follow, and under an analysis rule fixed before the outcome is known. Otherwise a researcher is asking a retrospective question: can I find something unusual before this event? Given the natural variability of the ionosphere, the answer will often be yes.
The ionosphere is a difficult place to make retrospective anomalies carry predictive meaning. Solar activity, geomagnetic storms, atmospheric waves, tides, traveling ionospheric disturbances, local time, season, latitude, and the geometry of the GNSS measurement can all change TEC. An apparent pre-earthquake feature therefore has to beat not only random noise but a moving background full of real geophysical structure.
That is why large-control studies matter. In 2017, Jeremy Thomas and colleagues examined Global Ionospheric Map TEC around 1,279 magnitude-6-or-larger earthquakes between 2000 and 2014. Their analysis did not find statistically significant evidence that TEC changes before earthquakes differed from the control behavior strongly enough to establish a useful global precursor relation. That result did not prove that no seismo-ionospheric precursor could ever exist. It showed how much weaker a pattern can become when the denominator includes many earthquakes and many ordinary days.
The issue remains active. Researchers continue to publish studies reporting statistical or event-specific ionospheric anomalies before earthquakes. But a 2026 re-examination by Kai Koyama and Yoshihiro Kaneko illustrates why the burden of proof remains high. They revisited reported TEC anomalies before large earthquakes and found that the signal observed before the Tohoku-Oki event was more plausibly explained as a large-scale traveling ionospheric disturbance passing over Japan than as a localized anomaly generated by the future rupture. Their analysis also found similar disturbances on days without a major earthquake.
That last comparison is the critical one. A signal that appears before an earthquake can still be scientifically interesting. But prediction requires knowing how often the same signal appears when nothing happens.
Three rules would make this field easier to interpret.
First, freeze the detector before the earthquake sample is evaluated. The time window, spatial window, smoothing, detrending method, geomagnetic exclusions, anomaly threshold, and decision rule should be declared using development data or prior theory. If those choices change after the target earthquake is inspected, the resulting pattern is exploratory, not confirmatory.
Second, publish untouched negative controls. An earthquake day should be evaluated beside ordinary days matched for local time, season, solar and geomagnetic conditions, and measurement geometry. The key number is not simply how often an anomaly appears before an earthquake. It is how much better the rule discriminates earthquake-linked periods from the ordinary behavior of the ionosphere.
Third, keep post-event sensing and pre-event prediction separate. Coseismic ionospheric disturbances have real scientific and potentially operational value. They can illuminate atmospheric coupling and, under appropriate conditions, contribute information about the event after it begins. None of that evidence automatically raises the probability that a similar-looking feature before the rupture was a precursor.
These standards are not designed to kill an unconventional hypothesis. They are how an unconventional hypothesis earns the right to become useful. A genuine precursor should become stronger when tested prospectively, when the analyst loses the freedom to choose the best-looking baseline after the fact, and when the method is forced to survive ordinary non-earthquake days.
The distinction is important beyond one corner of geophysics. Modern sensors and machine-learning systems can scan enormous physical datasets for anomalies. That makes it easier than ever to discover correlations near consequential events. It also makes control data, frozen analysis windows, and prospective validation more important. The more places a system can look, the less impressive one retrospective anomaly becomes.
The ionosphere clearly feels major earthquakes after they happen. Whether it reliably whispers beforehand remains unresolved. Until a proposed precursor survives prospective testing against the ordinary behavior of the ionosphere, it should be described for what it is: a hypothesis, not an early-warning signal.
About Burak Oktenli
Burak Oktenli holds an MBA and a Master of Professional Studies in Applied Intelligence from Georgetown University. His research addresses the governance of authority in autonomous and AI-enabled systems, and his writing has appeared at the Modern War Institute at West Point, RUSI, RealClearDefense, RealClearMarkets, and Geopolitical Monitor. He is the author of Authority Architectures for Autonomous Systems, a ten-volume series on how authority in autonomous systems is delegated, monitored and recovered, at authority-architecture.me.
View all posts by Burak Oktenli →
No comments:
Post a Comment