Do the recent South Korean magnitude-5.4 earthquakes imperil the safety of its nuclear power plants?

By Ross Stein, Ph.D., Temblor, Inc., and Jaesung Park, Ph.D., Nephila Advisors, LLC.


Busan, South Korea


On 15 November 2017, a M=5.4 shock struck the southeast portion of the Republic of Korea. This event followed another M=5.4 shock on 12 September 2016, only 15 mi (25 km) to the south. These are the two largest South Korean shocks in the Global CMT and USGS ANSS catalogs, which began recording in 1977. Both shocks appear to lie on or close to the likely 120-km (70-mi) long active Yangsan Fault, a length that would make possible a future M~7.3 earthquake. The slip rate along the fault is likely low, and so while such earthquakes would be possible, they also would be rare.


A range of evidence points to a long and active fault


We examined the seismic catalogs, Google Earth imagery of the faulted landforms, the Global Earthquake Activity Rate (GEAR) model, and the GPS-derived crustal strain rates (GSRM2.0), to assess the significance of the seismic threat. We believe all these data are consistent with an active fault (with oblique right-lateral reverse slip, and dipping to the southeast) capable of rupturing in a large earthquake. The slip rate is probably low (


Our interpretation of the Yangsan and Ulsan Faults. Okada et al (1994) found a right-lateral/reverse fault motion on the Yangsan Fault, and inferred a slip rate of ~0.1 mm/yr. By comparison, the San Andreas slip rate is 200 times higher.


We could trace the fault for about 85 km, through the cities of Yangsan and Gyeongju. For much of this route, rivers have incised the valleys and altered fault-formed scarps, and so we are not certain that the fault is a truly continuous feature. Irrespective, our analysis is no substitute for fieldwork combined with LiDAR imagery (aerial laser relief mapping). But based on the morphology we see above, and the earthquake focal mechanisms below, we believe the fault is dominantly right-lateral (whichever side you are on, the other side moves to the right), with some compression (leading to thrust slip).


Gyeongju, near the 2016 epicenter. Gyeongju was the capital of the ancient kingdom of Silla (57 BC – 935 AD). which ruled about two-thirds of the Korean Peninsula between the 7th and 9th centuries. 


The recent earthquakes confirm the fault is active


Both of the M=5.4 shocks and their aftershocks lie on or near the candidate active fault, and their focal mechanisms (there are always two possible rupture surfaces for each quake) are both consistent with a fault that dips (or is inclined to) the southeast by 50°-70°. It is possible that neither of these small shocks ruptured the fault itself, but instead slipped secondary features, a behavior common on the San Andreas and other faults.


The mainshocks and their aftershocks, as well as the focal mechanisms (the ‘beachballs’) of the mainshocks. The mechanisms are also consistent with the NIED Aqua CMT (Japan) catalog and the Global CMT (Columbia University) catalog. The Korean Meteorological Administration locates the 15 November 2017 earthquakes about 5 km (3 mi) east of the pictured Yangsan Fault (red line).


Global Earthquake Activity Rate (GEAR) model forecasts that M=5.4 shocks are not rare


The Temblor app uses the Global Earthquake Activity Rate (GEAR) model of Bird et al. (2015), which blends the Global CMT earthquakes with GPS-derived crustal strain to forecast the rate of M=5 to M=9 earthquakes everywhere on Earth in a uniform and consistent manner. The GEAR model was submitted by its authors on October 2015 to independent testing by the Collaboratory for the Study of Earthquake Predictability (CSEP) for M≥6.0 shocks. In the first 18 months of test results (which captured 285 M≥6 shocks), GEAR outperformed its competitors, and its own model forecast. The value of GEAR is that any locations on Earth can be inter-compared, because the data and methods are common to both. Here is the GEAR forecast for Korea.


The GEAR model (color bands above) suggests that a M=5.4 shock would be expected in a typical lifetime. Put another way, a shock of that magnitude would have a 60% chance of occurring in an 85-year lifetime, or a 1% chance of striking per year. Is this correct, or are such shocks more common?


GEAR is based on M≥5.77 shocks since 1977 (there were 4,600 of them through 2014), and GPS-derived strain rates, which have been measured continuously since the late 1990’s over most of the world (there are 22,400 of them). How well does this work for South Korea?


GEAR could be augmented to provide a better forecast


There were no M≥5.77 shocks since 1977 in South Korea, and so no earthquakes were used in this portion of GEAR. But there have been seven M≥4.7 shocks since 1977, which perhaps could be used in a hybrid version of GEAR.

The GPS data in South Korea contains about 40 stations with velocities, and so in South Korea, GEAR is based exclusively on this data. Nevertheless, the signal-to-noise ratio in South Korea is much lower than in neighboring Japan, and so additional GPS velocities—which likely exist—could greatly enhance the model. Here is an analysis of the Korean portion of GEAR.


These map excerpts are from Kreemer et al. (2014). Notice that the sense of slip (bottom panel) is consistent with the two M=5.4 shocks, even though they struck after the model was published and submitted for testing. Kreemer’s model is called the GEM Strain Rate Model 2.0, or GSRM2.0.


What next?


We recommend that the seismic hazard be reassessed in light of the two recent M=5.4 shocks, because their locations and focal mechanisms are consistent with the existence of an active fault capable of much larger earthquakes, whose shaking could be larger than was anticipated in the design of the nuclear power plants. An enhanced GEAR model could be used to develop a probabilistic hazard model that forecasts the intensity of shaking, rather than only the occurrence of earthquakes. This would provide the immediate input to earthquake risk mitigation planning, and risk assessment, of the nuclear power plants near the active faults of the Korean peninsula.


Busan at night




NIED Aqua CMT catalog

Korean Meteorological Administration

Global CMT Catalog of Columbia University

Collaboratory for the Study of Earthquake Predictability (CSEP)

Bird, P., D.D. Jackson, Y.Y. Kagan, C. Kreemer, and R.S. Stein (2015), GEAR1: A Global Earthquake Activity Rate Model Constructed from Geodetic Strain Rates and Smoothed Seismicity, Bull. Seismol. Soc. Amer., 105, 2538–2554, doi: 10.1785/0120150058.

Kreemer, C., G. Blewitt, and E.C. Klein (2014). A geodetic plate motion and Global Strain Rate Model, Geochem. Geophys. Geosyst. 15, 3849–3889, doi: 10.1002/2014GC005407.

Atsumasa OKADA, Mitsuhisa WATANABE, Hiroshi SATO, Myung-Soon JUN, Wha-Ryong JO, Sung-Kyun KIM, Jeong-Soo JEON, Heon-Cheol CHI, Kazuo OIKE (1994), Active Fault Topography and Trench Survey in the Central Part of the Yangsan Fault, Southeast Korea, Journal of Geography (Chigaku Zasshi) 103, 111-126,