AIR Currents

May 19, 2021

Editor's Note: AIR plans to release a substantially updated earthquake model for Japan this summer. This article is the third and final one in our series on managing Japan earthquake risk; it describes our updates to modeling tsunami risk. Part II described how the M9.0 Tohoku earthquake informed our view of hazard from megathrust earthquakes in Japan and how we are updating the vulnerability component of our model. Part I provided a detailed discussion on the time-dependent rupture probability forecasting of large earthquakes on the Nankai Trough and faults in Japan and explained why the updated AIR Earthquake Model for Japan incorporates HERP’s view of seismic risk in some cases and diverges in others. (All magnitudes in this article are expressed in moment magnitude.)

While risk from large tsunamis has been recognized in the modern era since the 1960s—when the M9.5 Valdivia, Chile, earthquake and the M9.2 Great Alaska earthquake occurred—it was the undersea M9.1 Sumatra earthquake in 2004 that caught the world’s attention. The Sumatra quake generated tsunami waves that devastated coastlines in many countries around the Indian Ocean, killing more than 225,000 people and demonstrating their far-reaching and catastrophic impact. Just over six years later, the world witnessed a tsunami’s destructive power again, this time in Japan. Despite early warning, sea defense structures, and evacuation plans, more than 15,000 people were killed within a matter of hours after tsunami waves as high as ~40 meters inundated coastal areas in eastern Honshu following the 2011 M9.0 Tohoku earthquake, which had ruptured a ~500 km-long segment of the subduction interface beneath the Pacific Ocean along the Japan Trench. Both the 2004 Sumatra and 2011 Tohoku earthquakes were very large and occurred at subduction zones where the characteristic earthquake magnitude(s)—based on a handful of historic records—were thought to be significantly smaller (e.g., ~M8.0 for the Japan Trench). Prior to 2011, many in the scientific community believed that only a few subduction zones—depending on their geometry, plate convergence rate(s), age, etc.—could produce great M9.0-class earthquakes. Tohoku led to a revision of our understanding of the recurrence patterns of great tsunamigenic earthquakes.

Could Another Large Tsunamigenic Earthquake Strike Japan?

Japan lies in one of the most seismically active regions on the planet, close to not one but two large subduction zones (Figure 1). Along the Kuril and Japan trenches offshore Hokkaido and eastern Honshu the Pacific Plate is subducting beneath the Okhotsk Plate at a rate of ~8–9 cm/year and has produced two great M9.0-class earthquakes over the last 100 years—Tohoku in 2011 and the M9.0 Kamchatka earthquake in 1952. In central and southern Japan, along the Sagami and Nankai troughs offshore southern Honshu and Shikoku, the Philippine Sea Plate is subducting beneath the Eurasian Plate at a rate of ~4-5 cm/year and has produced several large (>M8.0) tsunamigenic earthquakes throughout recorded history.

Figure 1
Figure 1. Japan's tectonic setting, illustrating plate boundaries (in white), active crustal faults (in red) and large subduction–interface sources (in yellow). (Source: AIR; data source: Headquarters for Earthquake Research Promotion)

Tsunami Heights Continue to Increase with Magnitude

Following the Tohoku earthquake in 2011, estimates of upper bound magnitude for large, full-interface rupturing scenarios along the Kuril Trench and Nankai Trough were revised to include M9.0-class earthquakes. This has important implications for tsunami risk in Japan. Unlike damaging short-period ground motion, which saturates with earthquake magnitudes around M8-M8.5, tsunami wave height(s) continue to increase. In fact, average wave height(s) from an M9.0 earthquake could potentially be twice as high as those for an M8.8 event. Part of the reason why many in the scientific community were surprised by the scale of devastation resulting from the tsunami that followed the Tohoku earthquake is because its magnitude wasn’t correctly anticipated in prior hazard models, which relied heavily on regional historical data spanning just a few hundred years. And rather than the ground motion intensities being significantly underestimated, the tsunami wave heights were, as they are very sensitive to the total seismic moment and therefore the moment magnitude (Mw). For example, ground motion intensities at the Fukushima Daiichi nuclear power station were nearly equivalent to the designed value of 6-upper on the 7-level Japanese seismic intensity scale, but the tsunami wave height was more than double the seawall design height of ~6 meters. This seawall was easily overtopped, compromising critical power and safety equipment, which eventually led to core damage and the release of radioactive material. Besides seawalls, many dikes and breakwaters along northeastern Honshu—including the mile-long world’s deepest breakwater near the port of Kamaishi—were either partially damaged or completely destroyed from the tsunami waves due to excessive hydrodynamic forces and/or scouring, leaving coastlines defenseless.

Tabrez AliTabrez Ali, Ph.D.
Senior Scientist II and Manager

Edited by Sara Gambrill, CEEM

Latest HERP Model Shows Nankai Trough Poses the Greatest Tsunami Risk to Japan

The Headquarters for Earthquake Research Promotion (HERP), responsible for publishing national seismic hazard maps, annually updates its long-term forecasts of earthquakes in and around Japan. According to the latest HERP model, which incorporates the post-Tohoku view of seismic hazard along subduction zones surrounding Japan, there is a ~70% probability of an M8.4–M9.1 earthquake along the Nankai Trough within the next 30 years and a ~25% probability that its magnitude will exceed M8.8. Similarly, there is a ~10% probability of a large M8.7–9.2 earthquake along the Kuril Trench near northeastern Japan. Given the population distribution in the region and the time-dependent probability of large earthquakes, the subduction interface along the Nankai Trough poses the greatest tsunami risk to the country. It has a recurrence rate of ~90 years and parts of it last ruptured ~75 years ago in 1944 and 1946 during the ~M8.0-class Tonankai and Nankai earthquakes, producing tsunamis as high as 10 meters and 7 meters, respectively. The last full segment–rupturing earthquake occurred more than 300 years ago in 1707.

While large M9.0-class subduction zone interface earthquakes produce massive tsunami waves that devastate entire coastlines, relatively smaller, moderate size shallow M7.0–8.0 tsunamigenic earthquakes can also generate locally large tsunami waves that can cause significant destruction and losses, especially if the fault is close to the shoreline and a large population center. The ~M7.9 Great Kanto earthquake of 1923 generated tsunami waves as high as 12 meters that claimed thousands of lives and caused major destruction along the Sagami Bay and Boso and Izu peninsulas. Similarly, the ~M8.0 Sanriku earthquake in 1896, which ruptured a shallow segment of subduction interface, produced a large tsunami that destroyed nearly 9,000 homes and claimed more than 22,000 lives. It is therefore important to account for such relatively smaller earthquakes in tsunami hazard analysis as well.

Capturing Tsunami Risk in Japan with AIR’s Updated Japan Earthquake Model

AIR has been at the forefront of modeling tsunami risk from earthquakes. In 2013 we released the industry’s first fully probabilistic tsunami model, based on the latest scientific research and engineering expertise, to estimate potential tsunami damage and losses from tsunamigenic earthquakes of various magnitudes throughout Japan. While shake intensity can easily be calculated using empirical ground motion prediction equations (GMPEs) derived through regression analysis of instrumentally recorded strong-motion data, the propagation of tsunami waves and subsequent inundation are calculated using physics-based computer simulations. This is due to non-linear as well as significant directivity and path (or bathymetric) effects. Unlike shear wave velocities of rock(s), which impact the propagation of elastodynamic waves, variation in bathymetry (the topography of the ocean floor, which affects water depths) can be more than three orders of magnitude.

In the updated AIR Earthquake Model for Japan, anticipated for release this summer, we numerically simulate tsunamis generated by all M>7.0 tsunamigenic earthquakes from our time-dependent, stochastic earthquake catalog to determine coastal inundation and flow velocities. The stochastic catalog reflects AIR’s view on time-dependent rupture probabilities for different sources, including tsunamigenic ones such as those along Nankai. Our model incorporates the latest data sets, including up-to-date:

  • Detailed three-dimensional subduction interface/fault geometries
  • High-resolution bathymetry/elevation data with coastal levees, dikes and sea walls
  • Regional tidal solutions to account for the effect of astronomical tides
  • Land use/land cover (LULC) data to capture land friction characteristics that impact runup extent

To generate the tsunami, we first created single/multi-segment fault rupture scenarios for all tsunamigenic earthquakes with random slip (assuming a k-2 rupture model), which were subsequently used to calculate seafloor deformation using elastic dislocation theory. GPS-derived interface coupling ratios or slip deficit rates, if available, were used during the slip generation process. To account for source uncertainty, we allowed for up to ±0.24-unit variation in characteristic magnitude, as well as the hypocentral location. This allowed us to better capture a wide variability in losses. Figure 2 shows sample slip distributions for two such earthquakes based on HERP scenario ANN20 for Nankai, which has a characteristic magnitude of M8.7. Ground-up tsunami losses from the M8.9 source on the left are ~600% higher than the one on the right with M8.6. Even for characteristic earthquakes with the same magnitude but different slip distributions and origin times (which impact tide levels), losses can vary by up to 200–300%.

Figure 2
Figure 2. Slip distributions of two realizations of the M8.7 ANN20 HERP scenario along the Nankai Trough; the earthquake on the left has a magnitude of M8.9, whereas the one on the right has a magnitude of M8.6. (Source: AIR)

To simulate tsunami propagation and inundation, we used a staggered, finite-difference, nonlinear shallow-water equation solver with multi-resolution nested grids covering the entire coastline of Japan, including all its islands. Shallow-water finite-difference solvers have been extensively validated for modeling tsunami propagation. They are computationally efficient compared to fully three-dimensional models and therefore ideal for probabilistic tsunami hazard analysis. Figure 3 shows our simulated maximum wave heights following the 2011 Tohoku earthquake throughout the Northwest Pacific Basin using the source model of Fujii et al., along with simulated and observed waveforms recorded by three DART buoys and a tide gauge in Hilo, Hawaii, ~6250 km away. We find excellent agreement, both for wave amplitude(s) as well as phase at all four locations. We also find good general agreement between simulated and observed maximum wave heights along eastern Honshu. Some of the mismatch around 39.5°N–40°N latitude near Iwate has been attributed to a large submarine mass failure (up to ~500 km3) east of the central Sanriku coast that was triggered by the Tohoku earthquake.

Figure 3
Figure 3. Simulated maximum wave height(s) and DART buoy/tide-gauge observations along the eastern coastline of Japan, following the March 11, 2011, Tohoku earthquake. (Source: AIR)

The coastline of Japan is one of the most developed in the world with a large percentage of the country’s population living in urban areas along or near the coast. To reduce the risk of flooding in low-lying areas and reclaimed land from tsunamis and typhoon-driven storm surge, a large and complex network of seawalls, breakwaters, coastal levees/dikes, water gates, and other defensive infrastructure have been built. In the aftermath of the Tohoku earthquake, a significant investment has been made to improve the seawalls/levees along the northeastern coast of Japan as well to protect against future large tsunamis. To accurately model coastal inundation, it is important to account for these structures. Some of them, such as levees and dikes, are implicitly captured in our model because they exist in the high-resolution Digital Elevation Maps (DEMs) used in the simulations (e.g., Figure 4, right). Others, such as seawalls (e.g., Figure 4, left), which are too narrow to be captured by DEMs, are explicitly accounted for; they are digitized and merged into DEMs used in the model.

Figure 4
Figure 4. Left: A seawall near the inlet of Kinokawa River in Wakayama, Japan (Source: Google Street Map); right: Dike/wall outlined in dashed black line near Shinchi Thermal Power Station in Fukushima captured in high resolution (~5 m) DEM. (Source: AIR)

What if a Large Nankai Earthquake Were to Occur Today?

A significant part of the Japanese coastline is at risk from large (M>8.0) tsunamigenic earthquakes. The greatest risk, however, is from large subduction interface earthquakes along the Nankai Trough, followed by those along the Kuril Trench and offshore northern Sanriku. According to our updated model’s results, all of the top 10 and ~70% of the top 100 most damaging tsunamigenic earthquakes in our stochastic catalog, in terms of tsunami losses, were associated with Nankai, which would impact the coastlines of southern Honshu and Shikoku; those along the Kuril Trench and offshore northern Sanriku would cause significant damage in Hokkaido and northern Honshu, respectively. In general, the scale of damage increased with increasing magnitude and all top 10 damaging earthquakes along the Nankai Trough had magnitudes greater than M8.8. For earthquakes that do not occur along the Nankai Trough, 9 out of the top 10 damaging earthquakes have magnitudes greater than M8.8 and occur along the Kuril Trench. We also find that tsunami losses from earthquakes along Nankai, depending on the scenario, ranging from relatively smaller ~M8.0 single Tokai segment ruptures to ~M9.0 full (Tokai–Tonankai–Nankai–Hyuga-nada) interface-rupturing scenarios, can vary by two to three orders of magnitude.

Figure 5 shows maximum wave heights throughout Japan following a relatively large, simulated M8.8 earthquake along the Nankai Trough. This particular earthquake ruptures the Tokai, Tonankai, and Nankai segments resulting in wave heights as high as ~35 meters near Muroto in Shikoku, although most of the damage and loss (up to 50%) occurs in the prefecture of Wakayama, where 20- to 40-foot high waves flood low-lying areas near the shore.

Figure 5
Figure 5. Maximum wave heights above mean sea level (Zmax) in meters around Japan and on land (left-inset) in southeastern Japan following an M8.8 earthquake along the Nankai Trough (slip distribution shown in the right-inset). (Source AIR)

Figure 6 shows AIR’s probabilistic tsunami hazard map for Japan with maximum wave height exceeding a return period of 475 years. Hazard is relatively higher along the coastline(s) of Hokkaido, northeastern Honshu, southeastern Honshu, and Shikoku primarily due to subduction earthquakes along the Nankai Trough and the Japan and Kuril trenches. In general, large earthquakes along Nankai cause relatively higher losses in Wakayama, followed by Osaka-Hyogo, Tokushima, Mie, Kochi and Shizuoka prefectures whereas large earthquakes along the Japan and Kuril trenches cause relatively higher losses in Hokkaido, followed by prefectures in Northern Honshu including, Aomori, Miyagi and Iwate.

Figure 6
Figure 6. Probabilistic tsunami hazard map for Japan with maximum wave height exceedance levels for 475-year return period. (Source: AIR)

Figure 7 shows total tsunami losses from all earthquakes in our model, broken down by prefecture. Tsunami hazard in Tokyo, the most populated city/prefecture in Japan, is low due to its location along Tokyo Bay and lower time-dependent rupture probabilities for Great Kanto/Genruko-type earthquakes along with their relatively low magnitudes.

Figure 7
Figure 7. Contribution to total tsunami loss by prefecture from all tsunamigenic earthquakes in our model. (Source: AIR)

While tsunami losses typically don't contribute greatly to the average annual loss from earthquakes in Japan, they can contribute greatly to losses for individual events. Their contribution to damage and insured loss is most significant in the tail end of the earthquake distribution.

Managing Japan Tsunami Risk

Ten years ago this year, the M9.0 megathrust Tohoku earthquake ruptured part of the Pacific-Okhotsk plates and relieved stress on the subduction interface along the Japan Trench; this earthquake generated tsunami waves as high as ~40 meters, which inundated coastal areas in eastern Honshu. In the south, however, the Nankai Trough has been locked with the Amurian Plate since the M8.1 Tonankai quake in 1944 and the M8.3 Nankai quake in 1946 and currently poses serious earthquake risk to southeastern Japan. Our catalog reflects AIR’s view on time-dependent rupture probabilities for different sources, including tsunamigenic ones such as those along Nankai, and the model explicitly and implicitly accounts for Japan’s sea defenses. While no model can predict when or if the next large tsunamigenic earthquake in Japan will occur, using the updated AIR Earthquake Model for Japan, anticipated for release this summer, can help you prepare for tsunami losses by providing the most detailed and accurate view of tsunami risk in Japan.


Fujii, Y., Satake, K., Sakai, S. et al. Tsunami source of the 2011 off the Pacific coast of Tohoku Earthquake. Earth Planet Sp 63, 55 (2011).

David R. Tappin, Stephan T. Grilli, Jeffrey C. Harris, Robert J. Geller, Timothy Masterlark, James T. Kirby, Fengyan Shi, Gangfeng Ma, K.K.S. Thingbaijam, P. Martin Mai, Did a submarine landslide contribute to the 2011 Tohoku tsunami?, Marine Geology, Volume 357, 2014, Pages 344-361, ISSN 0025-3227.



You’re almost done.
We need to confirm your email address.
To complete the registration process, please click the link in the email we just sent you.

Unable to subscribe at this moment. Please try again after some time. Contact us if the issue persists.

The email address  is already subscribed.

You are subscribing to AIR Blogs. In order to proceed complete the captcha below.