Ten-Year Empirical Analysis (2016-2026) | 206 Earthquakes M5.0+ | 2026-04-05
USGS Earthquake Database
Open-Meteo Weather API
Claude Opus 4.6 Analysis
Limitations: This is an exploratory data analysis, not a rigorous academic paper. The sample size is limited (only 18 M6.0+ events), and conclusions are for reference only — they do not constitute predictive evidence.
*2026 data as of April 5. The 2024 spike is mainly from the April 3 Hualien earthquake and its numerous aftershocks.
The April spike (71 events) is mainly from the 2024 April 3 Hualien earthquake sequence. Excluding April 2024, the monthly distribution is more even, though winter-spring (Dec–Apr) is still slightly higher than summer-autumn.
| Date | Mag. | Depth | Location | Notes |
|---|---|---|---|---|
| 2024-04-02 | 7.4 | 40 km | Hualien | Largest in a decade, April 3 Hualien earthquake |
| 2022-09-18 | 6.9 | 10 km | Taitung | Sept. 18 Chishang earthquake |
| 2022-03-22 | 6.7 | 24 km | Hualien | — |
| 2025-12-27 | 6.6 | 63 km | Yilan | — |
| 2022-09-17 | 6.5 | 10 km | Taitung | Sept. 18 foreshock |
| 2016-02-05 | 6.4 | 23 km | Tainan | Feb. 6 Meinong earthquake |
| 2018-02-06 | 6.4 | 17 km | Hualien | Feb. 6 Hualien earthquake |
| 2024-04-03 | 6.4 | 14 km | Hualien | April 3 aftershock |
| 2021-10-24 | 6.2 | 69 km | Yilan | Deep earthquake |
| 2022-01-03 | 6.2 | 19 km | Hualien | — |
| 2022-05-09 | 6.2 | 21 km | Hualien offshore | — |
| 2018-02-04 | 6.1 | 12 km | Hualien | Feb. 6 foreshock |
| 2019-04-18 | 6.1 | 20 km | Hualien | — |
| 2020-12-10 | 6.1 | 71 km | Yilan | Deep earthquake |
| 2024-04-22 | 6.1 | 10 km | Hualien | April 3 aftershock sequence |
| 2024-04-22 | 6.1 | 9 km | Hualien | Same-day double quake |
| 2024-08-15 | 6.1 | 14 km | Hualien | — |
| 2025-01-20 | 6.0 | 16 km | Tainan | — |
| 2025-12-24 | 6.0 | 8 km | Taitung | — |
Below are the weather statistics for the 15 days preceding each M6.0+ earthquake (day -15 to day -1):
| Earthquake Date | Mag. | Location | Avg Temp °C | Total Rain mm | Rainy Days | Heavy Rain | Avg Press. hPa | Press. Range hPa | Avg Humid. % |
|---|---|---|---|---|---|---|---|---|---|
| 2016-02-05 | 6.4 | Tainan | 15.1 | 125.4 | 10 | 2 | 1022.2 | 12.2 | 91.1 |
| 2018-02-04 | 6.1 | Hualien | 18.6 | 120.1 | 15 | 0 | 1018.9 | 10.2 | 84.7 |
| 2018-02-06 | 6.4 | Hualien | 17.5 | 138.7 | 15 | 0 | 1020.0 | 10.7 | 83.5 |
| 2019-04-18 | 6.1 | Hualien | 24.6 | 74.1 | 13 | 0 | 1014.1 | 7.0 | 85.3 |
| 2020-12-10 | 6.1 | Yilan | 20.6 | 254.9 | 15 | 5 | 1022.0 | 5.8 | 85.2 |
| 2021-10-24 | 6.2 | Yilan | 25.8 | 319.8 | 12 | 5 | 1014.4 | 15.6 | 84.9 |
| 2022-01-03 | 6.2 | Hualien | 20.1 | 29.3 | 5 | 0 | 1022.3 | 13.7 | 80.7 |
| 2022-03-22 | 6.7 | Hualien | 23.1 | 33.0 | 5 | 0 | 1014.5 | 10.7 | 80.3 |
| 2022-05-09 | 6.2 | Hualien | 26.8 | 75.0 | 8 | 1 | 1012.7 | 8.4 | 84.9 |
| 2022-09-17 | 6.5 | Taitung | 22.9 | 155.9 | 13 | 2 | 1009.0 | 17.8 | 91.5 |
| 2022-09-18 | 6.9 | Taitung | 22.8 | 142.9 | 12 | 2 | 1009.5 | 17.8 | 91.3 |
| 2024-04-02 | 7.4 | Hualien | 23.3 | 16.7 | 7 | 0 | 1017.0 | 13.6 | 82.2 |
| 2024-04-03 | 6.4 | Hualien | 23.8 | 14.2 | 6 | 0 | 1016.6 | 13.6 | 81.4 |
| 2024-04-22 | 6.1 | Hualien | 26.9 | 27.1 | 6 | 0 | 1013.0 | 8.6 | 82.3 |
| 2024-08-15 | 6.1 | Hualien | 31.1 | 28.6 | 7 | 0 | 1007.1 | 5.9 | 79.5 |
| 2025-01-20 | 6.0 | Tainan | 15.9 | 13.1 | 5 | 0 | 1020.7 | 9.5 | 79.3 |
| 2025-12-24 | 6.0 | Taitung | 19.3 | 11.3 | 3 | 0 | 1019.1 | 3.0 | 79.8 |
| 2025-12-27 | 6.6 | Yilan | 20.9 | 68.6 | 9 | 0 | 1020.3 | 9.2 | 81.9 |
The control group consists of 18 randomly selected "no major earthquake" dates (evenly distributed across seasons from 2016–2025), using the same 15-day weather window.
Of 18 M6.0+ earthquakes, 8 occurred in winter (44%) and 6 in spring (33%), totaling 78%. Only 1 occurred in summer (6%).
This means the differences in temperature, pressure, and rainfall may simply reflect seasonal effects rather than independent weather factors.
However, the global seismological community still debates whether earthquakes have seasonality. Some research suggests that the crust is affected by tidal forces and atmospheric pressure loading, and that winter high-pressure systems may exert additional stress on shallow faults.
Average rainfall in the 15 days before earthquakes was 91.6mm, compared to 59.3mm for the control group — 54% higher. This is the most significant difference.
Even accounting for seasonal factors, some cases had anomalously high rainfall: 254.9mm before the 2020/12 Yilan M6.1 and 319.8mm before the 2021/10 Yilan M6.2.
Academic background: "Rainfall-induced seismicity" is a known geophysical phenomenon. Heavy rainfall infiltrating the ground increases pore water pressure, potentially reducing effective normal stress on faults and making those already near critical state more likely to slip. This effect is more pronounced for shallow earthquakes (<20km).
83% (15/18) of major earthquakes had an average humidity above 80% in the preceding 15 days, compared to only 56% for the control group.
However, Taiwan is inherently a high-humidity environment, and the winter-spring northeast monsoon brings additional moisture. A larger sample is needed to confirm whether this difference is statistically significant.
The 15-day pressure range (max minus min) before earthquakes averaged 10.7 hPa, compared to 9.3 hPa for the control group. The gap is modest, but several of the largest earthquakes showed notably high pressure variation:
Academic background: The hypothesis that atmospheric pressure changes trigger earthquakes has a long history. Rapid pressure changes (such as typhoon low-pressure systems passing) can theoretically exert small but measurable forces on the crust. The 2022 Sept. 18 Taitung earthquake was preceded by a typhoon approaching Taiwan.
Of 18 M6.0+ earthquakes, not a single one had total rainfall below 10mm in the preceding 15 days. The control group also had none (Taiwan rarely goes 15 days without rain), but the lowest-rainfall earthquakes still recorded 11–16mm.
Viewed from another angle: Taiwan's major earthquakes never occur during completely dry periods. This "absence" itself is a pattern worth noting.
The preceding 15 days saw 319.8mm of rainfall (highest of all 18), with 5 heavy-rain days (>20mm/day). Pressure range was 15.6 hPa. This is the case most consistent with the "rainfall-induced seismicity" hypothesis.
However, this earthquake had a depth of 69km (deep), and rainfall infiltration theoretically cannot influence such depths.
Preceding 15-day rainfall was 155.9mm, pressure range was 17.8 hPa (highest), humidity was 91.5% (highest). A typhoon system was approaching Taiwan during this period, causing dramatic pressure and weather changes. Two M6+ earthquakes struck within two days.
Preceding 15-day rainfall was only 16.7mm, with relatively calm weather. Pressure range of 13.6 hPa was slightly elevated but not extreme. This is the best counterexample for weather-earthquake correlation — the decade's largest earthquake occurred during the most unremarkable weather period.
The observed differences (54% more rainfall, 3.8% higher humidity, 4.3°C lower temperature) are most likely driven by seasonal bias — Taiwan's major earthquakes statistically favor winter and spring, which inherently feature lower temperatures, higher humidity, and more rainfall.
The strongest counterevidence is the 2024 Hualien M7.4 (the decade's strongest), which occurred during an extremely calm weather period, demonstrating that weather is not a necessary condition for major earthquakes.
Data sources: USGS Earthquake Catalog API | Open-Meteo Historical Weather API | Central Weather Administration Seismological Center