Philippine Earthquakes, 2000–2026
Every earthquake of magnitude 4.5 or greater recorded in the Philippine region since 2000 — 9,203 of them — from the USGS global catalogue. The page is deliberately narrow: it covers seismicity and nothing else, and it argues that the headline count most sources quote is the wrong number to quote.
The USGS catalogue holds 9,203 Philippine earthquakes of M4.5 or greater since 2000. It also holds far fewer small ones than physics allows, which is why every rate on this page is quoted at M4.5+ and not at the M2.5 the catalogue nominally starts at.
- Counts rise from 6 events in the M3.0 bin to 9,818 in the M4.5 bin. Small earthquakes outnumber large ones roughly tenfold per magnitude unit, so a rising count is the seismometer network failing to hear, not the ground going quiet. Below M4.5 this catalogue measures instruments.
- The strongest event in the record is M7.8, and 19 events reached M7.0 or more.
- 74.4% of M4.5+ events are shallower than 70 km — but USGS pins depth to exactly 10.0 km when it cannot resolve it, and 23.1% of events carry that placeholder. The shallow share is softer than it looks.
- December appears to hold 13.03% of annual seismicity against an even-split expectation of 8.33%. Remove the thirty days after each M7+ and it falls to 8.86%. There is no earthquake season; there was one large earthquake.
The Number You Cannot Use
Before any count means anything, the catalogue has to be checked against the one law seismology is sure of: small earthquakes are far more common than large ones. This catalogue says the opposite, and that tells you where it stops being usable.
Why this matters more than it sounds
The full catalogue holds 16,489 events at M2.5+ over 27 years, and that is the number a page like this normally leads with. It is not a count of Philippine earthquakes. It is a count of Philippine earthquakes that a mostly-foreign seismometer network happened to hear, and that network got substantially better during the period. Quoting it as a trend measures the instruments.
Events in the M3.0 bin
Gutenberg-Richter puts roughly ten times as many M3 events as M4 events. This catalogue has six.
Events in the M4.5 bin
More than a thousand times the M3.0 count. The rise is instrumental: the global network simply does not detect small Philippine events.
Working threshold
Above this the counts finally behave. Every rate, share and trend on this page is computed at M4.5+ for that reason, and never at M2.5.
How Many Per Year
Annual M4.5+ counts, 2000–2026. The final year is partial.
Mean per full year
Averaged over complete years only; 2026 is excluded because it is still running.
Busiest year
800 events at M4.5+ — and most of that is one aftershock sequence rather than a busy year, which the aftershock section takes apart.
Peak count
Against a mean of 334. A single mainshock can more than double a year's total.
Magnitude Distribution
How M4.5+ events divide by size. Shares are of the M4.5+ population, not of the whole catalogue.
M4.5–4.9
7186 events. Felt locally, rarely damaging.
M5.0–5.9
1862 events.
M6.0–6.9
136 events, plus 19 at M7.0 or above — 0.2% of the population.
How Deep
Depth decides how much shaking reaches the surface: a shallow M6 does more damage than a deep M7. This is also the section where the catalogue is least trustworthy, and it says so.
Shallow, under 70 km
6847 events. These are the destructive ones.
Intermediate, 70–300 km
2142 events, mostly along the subducting slabs.
Depth is a placeholder
Of all M4.5+ events, this share carries a depth of exactly 10.0 km — the value USGS assigns when depth cannot be resolved. They are counted as shallow above, so treat that first figure as an upper bound.
Where They Happen
USGS labels each event with the nearest settlement, not a region, so the only grouping that is not invented here is by latitude. South of 10°N is Mindanao and the Sulu and Celebes seas; 10–13°N is the Visayas; north of 13°N is Luzon and the Philippine Sea.
South, under 10°N
5361 events, and the strongest in the record at M7.8.
North, 13°N and above
2394 events. Holds most of the population, including Metro Manila.
Central, 10–13°N
1448 events — the quietest band by count, which is not the same as the safest. The 2013 Bohol earthquake sits here.
The Strongest on Record
The twelve largest events in the catalogue. Depth is what separates the ones people remember from the ones they do not.
| Time | Magnitude | Depth | Location |
|---|---|---|---|
| 2026-06-07 23:37:41 UTC | M7.8 | 56.99 km | 25 km SW of Kablalan, Philippines |
| 2010-07-23 22:51:11 UTC | M7.6 | 578.0 km | 61 km W of Bantogon, Philippines |
| 2012-08-31 12:47:33 UTC | M7.6 | 28.0 km | 89 km E of Sulangan, Philippines |
| 2023-12-02 14:37:04 UTC | M7.6 | 40.0 km | 19 km E of Gamut, Philippines |
| 2001-01-01 06:57:04 UTC | M7.5 | 33.0 km | 21 km SE of Lukatan, Philippines |
| 2002-03-05 21:16:09 UTC | M7.5 | 31.0 km | Mindanao, Philippines |
| 2010-07-23 23:15:10 UTC | M7.5 | 640.6 km | 72 km S of Panubigan, Philippines |
| 2025-10-10 01:43:59 UTC | M7.4 | 59.42 km | 12 km E of Santiago, Philippines |
| 2010-07-23 22:08:11 UTC | M7.3 | 607.1 km | 68 km W of Gadung, Philippines |
| 2017-01-10 06:13:48 UTC | M7.3 | 627.17 km | 189 km SSE of Tabiauan, Philippines |
| 2005-02-05 12:23:18 UTC | M7.1 | 525.0 km | 138 km SW of Palimbang, Philippines |
| 2006-12-26 12:26:21 UTC | M7.1 | 10.0 km | 30 km SW of Hengchun, Taiwan |
Strongest
25 km SW of Kablalan, Philippines, 2026-06-07.
At M7.0 or above
Over 27 years — about one every eighteen months.
At M6.0 or above
Frequent enough that building standards, not luck, decide the outcome.
Magnitude Against Depth
A common claim is that bigger Philippine earthquakes are deeper ones. Across M4.5+ events the correlation is essentially nil, and the small print explains why the averages look like it is not.
Correlation
Pearson r between magnitude and depth over all M4.5+ events with a resolved depth. That is no relationship.
Mean depth, M7.0+
Against 60.1 km for M4.5–4.9 — which looks like a strong trend until you read the next card.
Median depth, M7.0+
The median barely moves across bands. The mean is dragged by a handful of very deep large events, one at 578 km. Nineteen events cannot support a trend.
What One Large Earthquake Does
Every M7.0+ in the record, and how many M4.5+ events followed it within thirty days, against that year's ordinary thirty-day rate. This section exists because the next one is wrong without it.
2 December 2023, M7.6
M4.5+ events in the following thirty days, against a normal thirty-day count of 65.8 for that year.
As a multiple
The largest aftershock response in the record, and the reason December looks like a season.
Not every large event does this
The February 2026 M7.1 was followed by fewer M4.5+ events than a normal month. Aftershock productivity varies enormously, so it cannot be predicted from magnitude alone.
Is There an Earthquake Season?
There is a persistent belief that Philippine earthquakes cluster in certain months. The raw counts appear to agree. They are agreeing with one earthquake.
December, as recorded
Against an even-split expectation of 8.33%. On its face, a strong season.
December, aftershocks removed
Drop the thirty days after each M7+ and the excess disappears almost exactly. The 2023 sequence alone supplied 510 events.
Spread across all twelve months
Largest monthly share minus smallest, aftershocks removed. That is what no season looks like — and it is the answer to the question in the heading.
Decade by Decade
M4.5+ rates by period. The rise is real in the record and is not safely readable as more earthquakes.
2000–2009
M4.5+ events per year.
2020–2026
Per year, over a partial period. Nearly 1.8× the 2000s rate.
The part that is harder to dismiss
M6.0+ events in 2020–2026, versus 47 in the whole of 2000–2009. M6 events were detected reliably throughout, so this is not a detection artefact — but seven years of a Poisson process is a small sample, and no trend is claimed from it here.
What This Page Does Not Cover
An earlier version of this page carried charts of volcano alert levels, Taal and Mayon eruption histories, tsunami warnings, fault-line lengths and peso damage totals. Every one of those numbers was invented. They have been removed rather than corrected, because correcting them would need sources this repository does not have.
Volcanoes
Alert levels, eruption chronologies and evacuation figures come from PHIVOLCS. Nothing here is derived from them, so no volcano number appears on this page.
Damage and casualties
NDRRMC situational reports carry these. They are published as PDFs per event and are not aggregated in this repository.
Faults and tsunami history
Fault geometry belongs to PHIVOLCS mapping; tsunami records to the NOAA/NGDC database. Both are open. Neither has been ingested here yet, and until they are the page stays quiet about them.
Method
Everything above is reproducible from two scripts and eleven CSVs in this repository.
Source
USGS FDSN event/1/query, GeoJSON, no API key. Fetched one
year at a time because the service caps a response at 20,000 events and
returns exactly that many rather than erroring — a truncated year
would otherwise look like a complete one. Both the fetcher and
checks.sql assert no year sits at the cap.
What counts as Philippine
A bounding box, 4–22°N and 116–128°E, wider than the land area because the trenches that produce the large events are offshore. The box is a choice that moves every number on this page, so it is carried on every row of every CSV rather than mentioned once in a header.
Threshold
M4.5+ for every rate, share and trend, because the catalogue is demonstrably incomplete below that — see section 01. The M2.5+ totals are published in the annual CSV so the difference can be inspected, and are used nowhere as a trend.
Aftershock handling
A 30-day window after each M7.0+ mainshock. Crude compared with a proper declustering algorithm, and deliberately so: it is simple enough to check by hand, and the December effect it removes is large enough that no subtler method is needed to see it.
Depth caveat
USGS fixes depth at exactly 10.0 km when it cannot resolve it. Rather than dropping or hiding those rows, the depth table reports the share carrying the placeholder as its own line.
Verification
Fifteen assertions in checks.sql cover box containment,
duplicate ids, threshold nesting, share sums and response truncation.
Every figure quoted above is bound to a SQL query in
facts.sql and re-checked against the CSVs on every build,
so a number that does not round-trip to a row cannot be published.
Key Findings & Summary
- The catalogue holds 9,203 M4.5+ Philippine earthquakes across 27 years, a mean of 334 per year.
- It is not usable below about M4.5. Counts rise from 6 events in the M3.0 bin to 9,818 in the M4.5 bin, which inverts the one distribution seismology is confident about. Any "earthquakes are increasing" claim built on small-magnitude counts is measuring seismometers.
- 58.3% of M4.5+ activity sits south of 10°N, against 26.0% north of 13°N — which is where most of the population is.
- Magnitude and depth are uncorrelated: r=0.064 across all M4.5+ events. The mean depth per band rises with magnitude, but the median does not; a few very deep large events are doing all the work.
- There is no earthquake season. December's apparent 13.03% share falls to 8.86% against an 8.33% expectation once the thirty days after each M7+ are removed, and the whole-year spread collapses to 2.38 pts. One M7.6 on 2 December 2023 was followed by 510 M4.5+ events, 7.8× the normal rate.
- M6.0+ counts rise from 4.7 to 8.6 per year between the 2000s and the 2020s. M6 events are detected reliably throughout, so this one is not an instrument artefact — but seven years is a small sample and this page does not call it a trend.
Sources & Citations
Every figure on this page traces to one of these, through a CSV in
data/ph-earthquake/. Each is checked against its source query on every
build.
USGS FDSN event servicePrimary
Every catalogued earthquake in 4-22N/116-128E, 2000 onward — Queried one year at a time; the service caps a response at 20000 events
Let's Discuss This Analysis
Interested in geoscience data analytics or disaster risk reduction research?
Have a dataset you'd like analyzed or need a mini AI project as a starter guide? Send me your suggestions!
Prefer the plain-English version?
I wrote a companion post — “The Number I Had to Throw Away” — in simple, everyday words.