📄 earth science

Magnitude Compression at Two Completeness Thresholds and the Likelihood of Large Earthquakes: Pre-registered Tests of Blue-Dominance and Red-Persistence Rules

This pre-registered study analyzing earthquake catalogues from Japan, Chile, and Turkey found that while a "blue-dominance" rule (where magnitude compression at a lower threshold exceeds that at a higher threshold) successfully identified periods of significantly reduced large earthquake likelihood, none of the tested rules could reliably predict the approach of large earthquakes.

Original authors: RamaKrishna Pasupuleti

Published 2026-09-30

📖 5 min read🧠 Deep dive

CC BY 4.0

Original authors: RamaKrishna Pasupuleti

Original paper licensed under CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/). ✨ This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer

Earthquakes are the planet's way of releasing built-up stress along the cracks in its crust, and for over a century, scientists have studied the catalogs of every tremor recorded by sensitive instruments. These records contain a vast amount of data, listing the time, location, and strength of millions of events, from the faintest shivers barely felt by humans to the massive ruptures that reshape landscapes. A central challenge in this field is distinguishing the background noise of daily seismic activity from the specific patterns that might signal a large, dangerous earthquake is imminent. While it is well established that earthquakes tend to cluster in time and space, with smaller shocks often following larger ones, finding a reliable short-term warning for a major event has remained elusive. Most proposed signals have failed to hold up when tested rigorously against new data, leaving the scientific community without a dependable method to predict when the ground will next shake violently.

In a recent study, a researcher set out to test a specific idea about how the size of earthquakes changes over time, using a method designed to avoid the common pitfalls of looking for patterns after they have already appeared. The study focused on three distinct regions known for their seismic activity: the Noto peninsula in Japan, the Iquique region in Chile, and the Kahramanmaras area in Turkey. The researcher examined the relationship between two different ways of measuring the "compression" of earthquake magnitudes. One measurement looked at all earthquakes larger than a magnitude of 1.5, which are frequent and small, while the other looked only at those larger than 3.0, which are rarer and more energetic. By tracking these two measurements side by side over time, the researcher tested three specific rules to see if they could predict when a large earthquake would occur within the next three months. To ensure the results were not a lucky guess, the rules were written down and locked in a digital timestamp before any data was analyzed, a practice known as pre-registration that prevents scientists from changing their criteria after seeing the results.

The study found that one of the three rules offered a clear and statistically significant insight, while the other two did not. The successful rule involved a period where the measurement of the smaller, more frequent earthquakes remained consistently higher than the measurement of the larger, rarer ones. When this "blue dominance" occurred, meaning the curve for the small events stayed above the curve for the large events without interruption, the likelihood of a major earthquake in the following 90 days dropped dramatically. In the three regions studied, this state was followed by a large earthquake only about one-fifth as often as on an average day. This result held true even after the researchers accounted for the total number of earthquakes happening in the area, suggesting that the pattern of sizes, rather than just the sheer volume of activity, carried the information. The finding implies that when the ground is producing small tremors that are very uniform in size, without the larger events becoming similarly uniform, the region is in a quiet phase where a major rupture is unlikely.

Conversely, the study explicitly ruled out the idea that a rise in the larger earthquakes could serve as a warning sign. One of the tested rules suggested that if the curve for the larger earthquakes crossed above the curve for the smaller ones repeatedly, or if it stayed above for a week or more, a large earthquake would be more likely to follow. The data showed that while repeated crossings did correlate with higher activity, this was simply because the area was already busy with many earthquakes, not because the pattern itself predicted a big event. Similarly, when the curve for the larger earthquakes stayed above the smaller one for at least seven days, it did not reliably precede a major quake. In fact, only four out of the fourteen large earthquakes analyzed in the study were preceded by such a sustained period. This result challenges the intuitive notion that a long stretch of uniform, high-energy small quakes signals that the crust is accumulating the stress necessary for a massive release.

The researchers were careful to note that their findings are based on a relatively small number of large earthquakes—fifteen in total across the three zones—and that the results need to be confirmed on independent data sets that were not part of this initial analysis. The study does not claim to have solved the problem of earthquake prediction, nor does it offer a tool that can tell people exactly when to evacuate. Instead, it identifies a specific condition where the risk of a major earthquake is significantly lower than usual. This distinction could be practically useful for emergency managers and scientists, helping them identify periods when routine monitoring is sufficient, rather than when heightened alertness is necessary. By freezing the rules and code used in the study, the researcher has made it possible for others to test these same patterns on new data from Japan and California, ensuring that the scientific process continues to refine our understanding of how the earth behaves before it shakes.

Technical Summary: Magnitude Compression at Two Completeness Thresholds and the Likelihood of Large Earthquakes

Problem Statement
Earthquake catalogues contain vast amounts of data, yet few derived statistics have demonstrated reproducible short-term forecasting skill for large earthquakes. While indicators such as changes in the $b$ -value, seismic quiescence, and accelerating moment release have been proposed, they often fail when subjected to pre-registered tests against independent data. The dominant source of reliable short-term information remains earthquake clustering, as described by the Epidemic-Type Aftershock Sequence (ETAS) model. Previous work by the author (Pasupuleti, 2026) utilizing the CSQ indicators (Magnitude Compression $C$ , Quiescence $Q$ , and Skewness Anomaly $S$ ) found no reproducible short-term forecasting skill when computed at a single magnitude threshold. This study investigates whether the relationship between magnitude compression computed at two different completeness thresholds—low ( $M \ge 1.5$ , "blue") and high ( $M \ge 3.0$ , "red")—can identify periods of varying likelihood for large earthquakes.

Methodology
The study analyzed three well-recorded seismic zones: Noto (Japan), Iquique (Chile), and Kahramanmaras (Turkey), containing a total of 15 large earthquakes. The analysis relied on pre-registered rules, with specifications cryptographically time-stamped (SHA-256) prior to computation to prevent bias.

Indicators: The study utilized the CSQ indicators ( $C, Q, S$ ) calculated from rolling 20-event windows. $C$ measures the uniformity of recent magnitudes relative to the catalogue-wide baseline.
Dual-Threshold Approach: Two curves were generated: $blue(t)$ (compression at $M \ge 1.5$ ) and $red(t)$ (compression at $M \ge 3.0$ ). The difference $D(t) = red(t) - blue(t)$ was used to define state transitions.
Pre-Registered Rules: Four specific rules were tested:
- Blue Dominance: No red-over-blue crossing in the last 30 days, with red above blue on $\le 10\%$ of those days.
- Red Activity: $\ge 2$ crossings of red above blue in the last 30 days.
- Red Persistence: Red above blue continuously for at least 7 days.
- Red Persistence (21-day): Top 10% of the area where $D > 0$ in the last 15/20/30 days.
Statistical Evaluation: The probability of a large earthquake occurring within 90 days following a specific state was compared to the baseline probability. Significance was assessed via permutation tests (1,000 replicates) that preserved the autocorrelated nature of seismic runs. An activity adjustment was applied post-hoc using logistic models to control for the 30-day counts of events at both thresholds.

Key Results
The study yielded one statistically significant finding and several negative results:

Blue Dominance (Success): Persistent blue dominance met its pre-registered criterion. Following blue-dominant days, the probability of a large earthquake within 90 days was approximately five times lower than the average day (Pooled Ratio $\times 0.22$ , $p = 0.005$ ). This association held across all three zones and persisted after adjusting for recent seismic activity (Odds Ratio 0.18, 95% CI 0.14–0.24).
Red Activity (Failure): While repeated red crossings initially showed a high pooled ratio ( $\times 2.79$ , $p = 0.001$ ), the rule failed the pre-registered criterion because it did not perform consistently across all zones (specifically failing in Kahramanmaras). Furthermore, the effect disappeared entirely after adjusting for seismic activity (Odds Ratio 1.12, $p = 0.33$ ), indicating the signal was driven by general clustering rather than a specific precursor state.
Red Persistence (Failure): Sustained red dominance (continuous for $\ge 7$ days) showed modest within-zone increases in likelihood but failed the pooled criterion (Ratio $\times 1.00$ , $p = 0.17$ ). This was attributed to Simpson's paradox, where zones with low base rates contributed most of the "red-sustained" days. Crucially, only 4 of 14 evaluable large earthquakes were preceded by such a phase.
Preceding States: Analysis of the 15 large earthquakes revealed that most were not preceded by a sustained red phase. One large earthquake occurred during a blue-dominant period.

Significance and Claims
The paper claims that persistent blue dominance is the first rule derived from the CSQ indicators to meet a pre-registered criterion in this research program. The finding suggests that periods where small earthquakes ( $M \ge 1.5$ ) are more uniform in size relative to their long-term behavior, and where larger events ( $M \ge 3.0$ ) do not exhibit a distinct shift in uniformity, may mark intervals of significantly reduced likelihood for large earthquakes.

The author emphasizes that no rule identified the approach of a large earthquake. The study explicitly rejects the hypothesis that sustained high-threshold compression (red dominance) marks energy accumulation preceding large events. The significance of the blue-dominance rule is framed as a potential practical tool for identifying periods where routine monitoring (rather than heightened alertness) might be appropriate, rather than as a predictive forecast for imminent rupture.

Limitations and Future Work
The author notes several limitations: the small sample size (15 earthquakes, 7 from a single swarm), the post-hoc nature of the activity adjustment, and the sensitivity of pooled metrics to base-rate differences between zones. The rules and code have been "frozen" and are currently being tested on independent catalogues (Japanese and Californian) to confirm the findings. The paper concludes that the blue-dominance result should be regarded as a hypothesis generated and first-tested in these specific catalogues, requiring confirmation on independent data before broader application.

Technical Summary: Magnitude Compression at Two Completeness Thresholds and the Likelihood of Large Earthquakes

More like this