On Friday 11 March 2011 a Magnitude 9.0 undersea megathrust earthquake triggered a tsunami off the Pacific coast of Tohoku (Japan) and started a cascading chain of events, including the Fukushima nuclear radiation disaster, which became the costliest natural disaster to date. It was the most powerful earthquake ever recorded in Japan and triggered a powerful tsunami wave that may have reached heights of up to 40.5 meters and which travelled at 700 km/h and up to 10 km inland. Residents of the Tohoku region including Sendai, the largest city in the area, had less than fifteen minutes of warning before more than a 100 evacuation sites were washed away. The tsunami swept the north-eastern mainland leading to US$210 billion economic losses according to a World Bank report and almost 20,000 people losing their life, mainly through drowning.
The event also had an impact on the financial system, with estimates of insured losses from the earthquake alone between USD30 and USD39 billion as of 2018 (Source: Swiss Re Sigma Report and Munich Re NatCatService 2018), leading the Bank of Japan to intervene in an effort to normalize market conditions. It posed some fundamental questions to those affected: have we witnessed the worst that could be? or Is the largest and most catastrophic earthquake-triggered tsunami yet to happen? And how ready and resilient are our interconnected societies to the consequences from these natural phenomena?
Since it was first created in 2006, the Willis Research Network (WRN) has strived to ask thought-provoking questions in the understanding of extreme events, their impact on society and on the insurance industry and the wider financial system. For nearly 15 years the WRN has collaborated with academics at the forefront of science and pioneered research to challenge assumptions and existing models to ensure the future doesn’t take us off guard. As a testament to this, in 2010, prior to the 2011 Tohoku disaster, the WRN, in collaboration with Willis Re, Tohoku University and UCL began a research project to build what would be known as one of the first tsunami catastrophe models in the industry, acknowledging the then lack of tsunami risk models to estimate the losses from this so-called “secondary” peril.
Tohoku University, world leaders in the study and modelling of tsunami hazard, led by Prof Imamura and his team, developed a physics-based numerical model to feed the shake hazard module. The UCL EPICentre, a world leader in the study of tsunami vulnerability, led by Prof Tiziana Rossetto, developed the vulnerability module, and Willis Re built a complete tsunami risk model with added financial component. To this date, the WRN continues to work with these institutions to further our knowledge of secondary perils, widening the collaboration to account for newer areas of research, such as Coulomb stress transfer and aftershock modelling with Dr Ross S. Stein, at Temblor, Inc., and Prof Shinji Toda, from IRIDeS at Tohoku University, and 3D modelling of earthquake wave propagation with Prof Kim Olsen and Dr Daniel Roten from San Diego State University.
In this article we show the WRN learning curve on earthquake and tsunami risk over the past 10 years as a leading example of the great progress that applied research collaborations have brought to the (re)insurance and disaster risk management sectors.
Bridging the gap between science and industry
Catastrophic tsunami losses have traditionally not been able to be well quantified by the insurance market. Tsunamis and other less-modelled components of earthquakes (such as fire following earthquake and landslides) have all emphasized the need to understand and quantify the risks from such so-called secondary perils (“sub-perils”).
The WRN and Willis Re co-developed a tsunami model that enables our clients to quantify and manage the risks from these extreme events using our natural catastrophe modelling expertise and insurance market knowledge.
The WRN and Willis Re co-developed a tsunami model that enables our clients to quantify and manage the risks from these extreme events using our natural catastrophe modelling expertise and insurance market knowledge. In particular, the Willis Re Japan Tsunami Model is a probabilistic and deterministic model of the country, that quantifies tsunami losses from a catalogue of tsunamigenic earthquake sources. It can also combine the shaking damage output from a third party vendor model with tsunami losses allowing for a more complete view of risk, and use custom vulnerability functions developed in conjunction with WRN partner UCL EPICentre, and an in-house expert view of exposure data unique to Willis Re, to integrate into a catastrophe modelling platform.
The model uniquely quantifies client risk from catastrophic tsunamis affecting Japan, and provides inputs to the Willis Re financial analysis. It significantly contributes to discussions on natural catastrophe risk with reinsurers and rating agencies with Willis Re expertise.
As science evolves, the WRN has strengthened its earthquake and tsunami-related research programme to understand external agents that might play a fundamental role in triggering these events: e.g. looking at seismic event rates post-Tohoku, incorporating static (Coulomb) stress. By pushing research boundaries, we help anticipate the previously unforeseeable. In the following sections, we present some of these initiatives that continue to inform the Willis Re View of Risk, in collaboration with the research groups we have the privilege to count among our WRN members.
Understanding and Modelling Tsunami Hazard: A Tohoku University collaboration
In the 21st century, damaging tsunamis have happened more frequently in the world, not because of an increasing frequency, but largely driven by increased coastal exposure. To improve our understanding of this secondary peril the WRN began a collaboration with Tohoku University’s International Research Institute of Disaster Science (IRIDeS), world leaders in the area of tsunami numerical simulation techniques and led by Professor Fumihiko Imamura. The aim was to develop a new research method for the global tsunami risk assessment considering the characteristic of regions and probability of occurrence. Under this collaborative research, the tsunami hazard was to be modelled using tsunami sources and performing numerical simulations based on historical tsunami databases. By the time the 2011 tsunami happened, the WRN had already been working on this research topic for some time.
During the 2011 Tohoku event, a maximum tsunami runup height of about 40m was observed on the Sanriku coast of Iwate Prefecture. Since then, scientists have been attempting to reproduce the 2011 Tohoku tsunami to understand such large tsunami generation mechanism. There are two possible reasons for this "silent tsunami": 1) a tsunami earthquake, a slow earthquake with shallow depth which makes less ground shaking but generates a large tsunami and 2) a submarine landslide tsunami, as in the 2011 event. In fact, a tsunami earthquake occurred in the same area in 1896 and resulted in more than 20,000 deaths mostly from the unexpectedly large tsunami rather than from the limited ground shaking. It is also possible that a large submarine landslide could have occurred together with the earthquake so that the observed tsunami exceeded the seismic energy.
The IRIDeS is able to generate mechanisms of such silent tsunamis by numerical simulation, demonstrating the underrated risk of “silent tsunamis”.
The IRIDeS is able to generate mechanisms of such silent tsunamis by numerical simulation, demonstrating the underrated risk of “silent tsunamis” (Fig 1). "Silent tsunamis" are still very difficult to predict using the current technology but real-time observations of deep sea could help improve the accuracy of tsunami warning. Such tsunamis are not well considered in the insurance industry and the inclusion of such events may be necessary for a portfolio stress-test.
Assessment of the tsunami mitigation effect of coastal defence structures
As we reflect on the 10th anniversary of the 2011 Tohoku event, the IRIDeS research focus continues on the assessment of tsunami countermeasures in Japan. Based on lessons from the 2011 event, tsunami numerical simulations were applied to assess the performance of multi-layered infrastructure as a structural tsunami countermeasure in Sendai and support the reconstruction decision making process. There are five components of the multi-layered infrastructure: the existing seawall, the reconstructed seawall, the greenbelt area (park and coastal forest), the elevated road, and the existing highway. The performance of multi-layered structures for tsunami mitigation was evaluated by inundation area and maximum flow depth. Figure 2 shows an example of the simulation results in Sendai. The simulations show that with the present infrastructure, the tsunami affected area in Sendai by the 2011 Tohoku event could have been reduced by 68%. The large buffer areas in yellow shown along the coast are due a new elevated road built by Sendai City which withholds tsunami water to the east. To the west of the road, they are designed to limit the inundation depth to below 2m.
IRIDeS has since expanded their research to the rest of Japan to assess the potential impact of such infrastructure on future tsunami events; with countermeasures in place, a far-future assessment incorporated a view of climate change through sea level rise (SLR), derived from the RCP 8.5 scenario of the IPCC report. The new SLR scenario demonstrated that for Sendai-City, if the countermeasures reduced the impacted inundation area by approximately 80% compared to the actual event, then with an additional SLR component, the impacted area was still reduced by approximately 63%. Whilst SLR will provide an increase to the area where buildings take damage from a tsunami, the new countermeasure can still prove to be effective mitigation.
A global assessment of tsunami hazards over the last 400 years
World Tsunami Awareness Day was designated by the United Nations General Assembly in 2015, calling all counties, international bodies and civil society to raise tsunami awareness and share innovative approaches to risk reduction. To commemorate, IRIDeS contributed to a report featuring a global tsunami hazard assessment based on a 400-year database. Numerical models for tsunami propagation were created based on more than 100 earthquakes from around the world. Information going back to 1600 AD show that tsunamis have occurred all over the world, not just along the Pacific Rim, as shown in Figure 3. The most damaging ones were documented between 1970-2016 in the Indian Ocean and East Japan regions. However, poor documentation from 1600-1969 AD might leave other areas of high tsunami risk exposed if not properly studied. This assessment aims to draw attention to these areas and highlight that just because we have not witnessed it in recent times it does not mean it cannot happen.
Quantifying tsunami and earthquake vulnerability: A UCL EPICentre collaboration
UCL EPICentre have been members of the WRN since 2006. They are a multidisciplinary research group that investigates risk to society and infrastructure from earthquakes and other natural hazards. Our collaboration with them looks at developing a novel, robust, unified framework for assessing the vulnerability of critical urban infrastructure to the combined effects of earthquake ground shaking, tsunamis and induced soil liquefaction. The research is not only looking at these cascading hazards but is also taking into account the effect of infrastructure interdependence across various systems and how this affects the infrastructure’s ability to provide its services. To estimate tsunami risk, we need reliable tools for assessing the damage to coastal structures from tsunami inundation. To this extent, the EPICentre, under the leadership of Professor Tiziana Rossetto, has been conducting physical experiments, field work and numerical analyses to understand how tsunami flows interact with buildings, towards their fragility assessment. Inspired by the Tohoku event, it has also investigated the cumulative effect of earthquake ground shaking and tsunami.
The EPICentre, under the leadership of Professor Tiziana Rossetto, has been conducting physical experiments, field work and numerical analyses to understand how tsunami flows interact with buildings, towards their fragility assessment.
A New Analysis method for Assessing Buildings Under Tsunami Loading
UCL EPICentre developed a new advanced structural analysis procedure, called the Variable Depth Push Over (VDPO), first used in 2017 to evaluate the response of a tsunami evacuation building in Japan. The VDPO analysis showed that although the building was seismically designed, it was still vulnerable to the failure of its ground storey from tsunami. Simply enhancing seismic design does not provide tsunami resistance, as tsunami induce significantly different failure mechanisms in structures. Building response to tsunami is almost binary, i.e. the structure either survives structurally intact, or collapses, with only very few cases of intermediate damage seen. This is explained by the long duration of the tsunami loading, not previously appropriately accounted for.
The VDPO approach has since been extended and has been used in the assessment of buildings in Chile, USA, Sri Lanka and Indonesia. It is also being referred to as an accepted structural analysis approach in the next revision of the ASCE7 Standard for Tsunami that will be published in 2022.
Buildings Under Sequential Earthquakes and Tsunami
Earthquake ground shaking often precedes tsunami inundation, yet after a tsunami it is not possible to distinguish how much damage is from the earthquake and how much from the tsunami. A number of papers were published jointly with the WRN where buildings were assessed under sequential earthquakes and tsunami, with fragility surfaces being derived for the two hazards. Contrary to general perception, the research consistently found that the preceding earthquake ground motion only slightly influenced the tsunami resistance of the building. This is due to the fundamentally different response of the structure to the two perils. These studies showed that fragility of buildings can be approximated by assessing the earthquake and tsunami response separately, with the worst damage occurring from the two hazards dominating the final damage state.
Accounting for aftershock impact and Coulomb stress through Temblor, Inc collaborations
The 2011 Tohoku earthquake was a cruel reminder of the seismic risk Japan faces. The subduction of the Pacific Plate beneath northern Honshu along the Japan Trench—which slipped up to 80 m in about 120 seconds during the M9.0 shock—and the subduction of the Philippine Sea Plate beneath southern Honshu along the Nankai Trough, are the principal drivers of its outsized earthquake hazard (Fig 4). But Japan also has its own San Andreas-like ‘transform’ fault, the Median Tectonic Line, that extends from southern Honshu through Kyushu, and ruptured in the 1995 M6.9 Kobe shock at its northern end, and the 2016 M7.0 Kumamoto shock at its southern end. A series of blind thrust faults also underly the Japan Sea coastline.
The WRN has been collaborating with Dr Ross S. Stein, at Temblor, Inc., and Prof Shinji Toda, from IRIDeS at Tohoku University for a number of years to further our understanding of earthquake risk and how it varies in time and space due to static (Coulomb) stress transfer.
In early 2020, the Temblor team identified a mild swarm of M≥5 shocks ranging from the Greater Tokyo area, and extending up the Tohoku coastline. With the prior Tohoku and Kumamoto shocks, along with this newly observed mild swarm, they built new forecasts which are used to inform their view of risk.
On 13th February 2021, a M7.2 struck offshore Fukushima at a depth of 50 km. Why did it strike and what does it foretell? According to Dr. Stein and Prof Toda, The M7.2 shock is an aftershock of the M9.0 Tohoku mainshock, which testifies to the huge lasting impact of the 2011 shock. It also struck in areas they previously forecasted as having a very high chance of M≥7 shocks. This shock has further transferred stress closer to the Japanese mainland and away from the M9.0 rupture; Temblor used this new information to update their models.
However large, one earthquake cannot prove or disprove a forecast. So, can we understand the broader pattern of seismicity of central Japan, and how it has changed since the M9.0 earthquake struck?
Two features stand out in figure 5: The site of the high slip in the M9.0 earthquake has a profoundly reduced seismicity rate today (blue) because stress there was relieved by the earthquake. But there is a massive surrounding zone into which stress was transferred (the 'Zone of influence'), about ten times larger than the zone of decrease. This zone has a much higher seismicity rate (yellow-red) than it did before the M9.0 Tohoku earthquake struck. A similar but smaller effect is seen for the M7.9 aftershock, which struck 30 minutes after the 2011 M9.0. That's because a M9.0 earthquake releases 40 times the energy of a M7.9.
So, the impact of the M9.0 Tohoku earthquake is huge and long-lasting. That means that to forecast earthquake hazard—and associated damage and loss—it is important to consider stress transfer in our view of Risk.
Use of 3D ground motion simulation to reduce the uncertainty in loss estimates, with the collaboration of San Diego State University
Tokyo is home to more than 13 million people within an area characterized by high seismic risk and exposure for the insurance industry. This high seismic hazard originates from the complicated tectonic relation and relative movements between the Pacific Plate, Philippine Sea Plate, and the continental North American plate in the triple junction setting near Tokyo (see Fig. 6). The Japanese Earthquake Research Committee estimated a 70% probability of a M7 class event over the next 30 years, and ~11,000 fatalities and US$1.1trillion in damage in case of a M7 class earthquake scenario. It is therefore critical for societal resilience and financial stability that premiums accurately reflect expected losses in a catastrophic event near Tokyo, exemplified by the 2011 Tohoku earthquake. But, the difficulty of quantifying the tail risk from these extreme events often pose Reinsurance, Capital and Regulatory challenges. Industry probable maximum loss (PML) estimates can sometimes be as different as $30bn to $300bn as in the Pacific North-West / British Columbia. When this is the case, which to trust?
One way to reduce earthquake tail risk uncertainty is to complement traditional models with sophisticated techniques to model ground motion. Three-dimensional (3D) ground motion simulation techniques have been widely used in academia since the early 90s to assess the ground shaking resulting from an earthquake scenario. Since this early breakthrough, many studies have shown the power of numerical simulation of 3D wave propagation to more accurately predict ground motions, whenever the earth structure and material parameters are reasonably well constrained. However, while 3D earthquake simulations have been facilitated by the emergence of supercomputers, the computational cost of the numerical approach is considerable as the resolution of the 3D models increases.
The WRN has been collaborating since 2017 with Prof Kim Olsen and Dr Daniel Roten, from San Diego State University, USA, who are pioneers of this methodology since the early ‘90s (See Olsen et al, 1995), looking at implementing this ground-breaking research in the (re)insurance industry. To date, such advanced ground motion simulation techniques have been used in a loss framework, with a specific focus on M9 megathrust scenarios in the Cascadia Subduction Zone, Pacific Northwest, USA (Fig 7). These results have been presented at the WRN Seismic Seminar in 2019, were we discussed how such next generation modelling can influence loss volatility and better inform decision making.
The refined seismic footprints and resulting updated financial exposure based on the 3D simulations have already been used to inform our view of risk for the Pacific Northwest, with Chile and Peru to follow next. Considering its substantial seismic hazard and risk, the Tokyo prefecture presents itself as an additional location where 3D modeling may lead to refined estimates of financial exposure for the insurance industry. In addition, the 3D models have the potential to provide important refinements to the sea bottom displacements in a submarine megathrust scenario earthquake, potentially leading to more accurate estimates of financial losses due to tsunami and fatalities if incorporated into separate tsunami models. Such efforts are currently underway for the west coast of Latin America between WRN, UCL, Tohoku University and SDSU.
Closing Remarks: “The whole is greater than the sum of the parts”
Helping society better prepare for and cope with the aftermath of catastrophic events such as the 2011 event is one of the aims of the WRN. These efforts go beyond natural disasters, accounting for People, Technological and Emerging Risks. Just because it has not been observed historically it does not mean it cannot happen in the future, and the current COVID-19 pandemic is testament of the need to continuously push the boundaries of our understanding of risk, to try and be one step ahead of the unforeseeable. To this extent, acknowledging the interconnectivity of catastrophic events and its global knock-on economic impact, our next generation of leading research aims to combine individual expertise from WRN members into wider, more complete collaborations, to gain deeper understanding of the risks we face. The synergy among research groups will bring us closer to understanding low probability and high consequence events, and help our clients in their endeavour to close the protection gap and create a world more resilient to catastrophes. The WRN continues its mission to support a realistic, well informed, updated, forward-looking view of risk.
With sea levels rising due to climate change, coastal communities will be facing an increasing tsunami risk, on top of greater chances of flooding during high tides and gradual loss of land. As exposure and populations in coastal areas are also increasing, the knock-on effect of this slow, chronic, irreversible change in natural capital, especially for island communities, calls for adaptation strategies and raised attention to tsunami research and related hazards. As part of the WRN we run several programs on climate related resilience initiatives. This article shows a small part of these. Check our website for the recent 2021 Annual Review and further details.
Other articles of interest
Increasing societal resilience to tsunami risk
Acknowledgements
This article was possible thanks to the kind contributions from Prof Tiziana Rossetto, Prof Kim Olsen, Dr Ross Stein, Prof Shinji Toda, Dr. Pakoksung Kwanchai, Prof Anawat Suppasri, and Professor Fumihiko Imamura.
References
Olsen, K.B., R.J. Archuleta, and J.R. Matarese (1995). Three-dimensional simulation of a magnitude 7.75 earthquake on the San Andreas fault, Science, 270, 1628-1632.
Satake, K. (2017). Geological and historical evidence of irregular recurrent earthquakes in Japan, Phil. Trans. R. Soc. A 373: 20140375. http://dx.doi.org/10.1098/rsta.2014.0375
