Taiwan’s double earthquake may have launched a ‘killer pulse’

The shaking from the magnitude-6.8 mainshock was larger than expected in 2500 years. The foreshock may have left a seismic warning that another quake was coming.

By Wei-An Chen, B.S., and Chung-Han Chan, Ph.D., Earthquake Disaster & Risk Evaluation and Management Center (E-DREaM)

Citation: Chen, W., Chan, C., 2022, Taiwan’s double earthquake appears to have launched a ‘killer pulse’ that toppled buildings, Temblor, http://doi.org/10.32858/temblor.275

This article is also available in Traditional Chinese.

A magnitude-6.4 earthquake hit Taitung in southeastern Taiwan just after 9:30 p.m. on September 17. Dozens of aftershocks over magnitude 4.0 struck within 2 hours of the mainshock. Seventeen hours after the event, a larger, magnitude-6.8 earthquake rattled Taitung again, causing severe damage.

Technically, the magnitude-6.4 earthquake is now regarded as a foreshock and the larger magnitude-6.8 event is the mainshock, assuming there is no larger earthquake in the near future in this region. Although such foreshock-mainshock sequences do occur in Taiwan and elsewhere, here we explore whether there were any precursory signals before the magnitude-6.8 earthquake. We also address whether the mainshock was, broadly, forecast by the 2020 seismic hazard assessment for Taiwan.

An earthquake traffic light

The ratio of small to large earthquakes that occur in a region over a given time period, known as the b-value, is thought by some researchers to indicate the amount of stress in the crust (Schorlemmer et al., 2005). A low b-value means that large events are more common than normal compared to smaller ones. This might mean that a region hasn’t released enough stress through small events, and so could be accumulating stress for a large earthquake.

Some researchers have extended this theory, arguing that the variation in b-values over the course of an earthquake sequence may portend an imminent large event (Gulia and Wiemer, 2019). Scientists use a Foreshock Traffic Light System (Gulia and Wiemer, 2019) to monitor clusters of earthquakes. A green traffic light indicates b-values are high, which could suggest that the seismicity behaves more like an aftershock sequence and no larger event is expected. Conversely, a red light indicates b-values are low, which they interpret to mean that stress is accumulating and a larger earthquake could take place soon.

We tested the traffic light during this earthquake sequence by examining data from earthquakes that struck nearby in the past five years, including the most recent events. The b-value, calculated from a complete catalog, was stable through the past five years until it dropped by 14% after the magnitude-6.4 foreshock — from 0.78 during 2017-2022 to 0.67 on Sept. 17-18. The traffic light threshold between red and green is a 10 % variation in b-values, suggesting the traffic light would have been red in this case. The b-value variation is larger than the uncertainty (dashed lines in image below), confirming the significance of temporal change in b-values. We further analyzed the aftershock sequence after the Sept. 18 mainshock and found the b-value had increased to 0.71, suggesting the traffic would have turned to yellow.

b-values in eastern Taiwan before and after the September earthquake sequence
b-values in eastern Taiwan before and after the September earthquake sequence


Based on the location and the calculated orientation of fault slip, the mainshock likely struck along the Central Range Structure (No. 34 in image below), a 86-kilometer (53-mile) long structure that parallels the Longitudinal Valley Fault (No. 33 in image below) — the boundary between the Eurasian and Philippine Sea tectonic plates. The Taiwan Earthquake Model (TEM; Chan et al, 2020) indicates that the rupture probability of the Central Range Structure is relatively high: 9 % in the coming 25 years. This fault has one of the highest earthquake hazards in Taiwan, according to Chan et al. (2020).

Likelihood of large earthquakes on Taiwan’s major faults.
Likelihood of large earthquakes on Taiwan’s major faults.


To understand the impact of the magnitude-6.8 earthquake on nearby structures, we analyzed the potential damage from its seismic waves. Those recorded at a seismic station 26 kilometers (16 miles) north of the epicenter were particularly strong due to a phenomenon known as directivity , or “killer pulse.” These destructive waves occur when energy is focused in the direction of fault rupture.


The magnitude of an earthquake is proportional to the area of the fault surface that slips, or “ruptures.” As an earthquake progresses, the ruptured area grows along the fault plane. The seismic waves traveling in the direction of the propagating rupture pile up, similar to the Doppler effect, creating larger waves that would occur otherwise.

The Central Range Structure ruptured from south to north, toward the northern seismic station, according to U.S. Geological Survey calculations. Killer pulse waves could be associated with a three-story building collapse in the vicinity of this seismic station, but more analyses would be necessary to confirm.

Buildings have a natural period, or resonance, over which they sway; taller buildings take longer to complete one back and forth motion than shorter buildings. When passing seismic waves have the same period as a building’s natural resonance, constructive interference causes the wave amplitude to increase, forcing buildings to drastically sway.

In the case of the magnitude-6.8 mainshock, the seismic waves had periods of 0.2-0.6 seconds, which is similar to the natural resonant period of many low rise buildings. The force the seismic waves imparted on some buildings exceeded both the design earthquake (a theoretical earthquake event that engineers use to check the resilience of a new structure) and maximum credible earthquake (the most severe earthquake believed to be possible based on geologic and seismological evidence) from the Taiwan building code, suggesting low rise buildings may have been strongly affected by this earthquake.


Even though earthquakes are unpredictable, careful analysis of seismic data can provide scientists hints of future events. Armed with these tools, those in earthquake-prone regions can better protect themselves and authorities may be able to warn residents of potential threats.


Gulia, L., & Wiemer, S. (2019). Real-time discrimination of earthquake foreshocks and aftershocks. Nature, 574(7777), 193-199.

Schorlemmer, D., Wiemer, S., & Wyss, M. (2005). Variations in earthquake-size distribution across different stress regimes. Nature, 437(7058), 539-542.

Chen, Z. Y., & Liu, Z. Q. (2019). Effects of pulse-like earthquake motions on a typical subway station structure obtained in shaking-table tests. Engineering Structures, 198, 109557.