Are There More Earthquakes in October?

Data retrieved September 28, 2026 · Analysis regenerated September 28, 2026

Data

I queried the USGS ANSS Comprehensive Earthquake Catalogfor every cataloged earthquake within 150 km of central Lima(-12.0464, -77.0428) from September 28, 2016 through September 28, 2026. The 100-event CSV extract is saved with this analysis so the result does not change underneath the page.

"All earthquakes" here means all events in the USGS catalog, not every physical tremor. The smallest event is M3.2, the median is M4.5, and 95 of 100 are M4.0+. USGS is therefore mostly measuring moderate earthquakes in this region; Peru's local network records smaller events that do not appear here.

Tests

The null hypothesis is that an earthquake is equally likely on every observed day. Monthly probabilities are therefore proportional to time exposed, rather than exactly 1/12 each. Event times are assigned using Peru time (America/Lima).

For Maritza's specific claim, I use a one-sided exact binomial test. Each earthquake is a trial, "occurred in October" is a success, and the null success probability is October's 8.49% share of the observed time. Because October was specified before looking at the data, this is the direct test of her observation.

I also use Pearson's chi-squared goodness-of-fit test to compare all twelve observed monthly counts with their expected counts. This broader test asks whether any month departs from uniform daily timing.

Answer

Results

Monthly counts

October and November tie for the largest count at 13. The expected values differ slightly by month length and by the exact endpoints of the ten-year window.

October by year

Five of the 13 October earthquakes occurred in 2016 and three in 2017. 7 of the ten Octobers had one or no cataloged events, so the apparent increase is concentrated in the beginning of the window rather than recurring consistently every year.

Sensitivity checks

Counting at most one earthquake per local calendar day leaves 97 event-days and produces a similar October result (13 versus 8.23, p = 0.067), so same-day clustering is not driving the result. The October signal disappears at M4.5+, which shows that the conclusion depends on the catalog's lower-magnitude events.

Interpretation

I can see why people think this is true, but it's not quite statistically significant. The data lean in Maritza's direction: October's rate is about 1.6 times the rate in other months. But a result at least this favorable to October would occur by chance about 8.1% of the time under the null. With only 100 events, that is not enough evidence to conclude that October is genuinely different.

The strongest honest conclusion is: interesting pattern, not established seasonality. Another decade, or a more complete IGP catalog of smaller local earthquakes, could make the test much more informative.

Caveats

• The 150 km circle is a transparent "near Lima" definition, not the administrative boundary of the Lima region.
• Earthquakes can cluster through aftershocks, so individual events are not perfectly independent. The event-day check reduces, but does not eliminate, that dependence.
• Detection and catalog inclusion can change over time. The magnitude sensitivity table checks whether the result survives stricter, more consistently observed thresholds.
• Statistical significance would describe cataloged event frequency, not earthquake predictability or increased danger in any particular October.