Is Australia prepared for the next earthquake? Reflections of Peter McBean

This briefing is excerpted with the kind permission of Ruth Cooper, Digital Editor at Mahlab, from an article by Charlotte Barkla that appears on, a website of Engineers Australia. It is an interview with Peter McBean, FIEAust CPEng, who was awarded the prestigious 2020 John Connell Medal by Engineers Australia for his outstanding contribution to structural engineering. Peter McBean succeeded Paul Somerville, Chief Geoscientist of Risk Frontiers, as President of the Australian Earthquake Engineering Society for the 2016 – 2019 term.

Peter McBean FIEAust CPEng.

On 22 February 2011, an earthquake rocked Christchurch on New Zealand’s South Island, killing 185 people and causing widespread damage. In the decade since, structural engineer Peter McBean has been working to ensure a similar disaster doesn’t befall Australia.  “The lived experience of people here is that we really don’t have earthquakes … But earthquakes remain a serious risk in this country,” McBean, Joint Managing Director of consulting firm Wallbridge Gilbert Aztec, told create.

“In reality, we’ve been lucky so far, given that earthquakes have generally avoided areas with significant populations, other than the 1989 Newcastle [New South Wales] event, which claimed 13 lives. “Australia experiences a Newcastle-sized earthquake about every two years, and a Christchurch-sized earthquake once a decade … But one day our luck will run out.” McBean said there are about a dozen known faults around the Adelaide region, and a number around Melbourne, while “a background level of seismicity extends across the continent”.

McBean was awarded the prestigious 2020 John Connell Medal by Engineers Australia for his outstanding contribution to structural engineering, but it was circumstance, rather than choice, that first sparked his interest in earthquake engineering. “I graduated university in the early 1980s, shortly after the first Earthquake Design Standard in Australia was published,” McBean said. “As the graduate in my team, the standard was given to me to have a look at… I went off and did that and found it really interesting.”

It’s an interest that hasn’t dampened over the past 35 years of his career, which has included the structural design of major projects, including the new Royal Adelaide Hospital. “When you get involved in earthquake engineering, you have to look at buildings or structures in a holistic sense — you have to consider them as a whole organism,” McBean said.

“People generally view buildings as being lifeless and static, but when you start thinking about their earthquake performance, you have to consider their dynamic behaviour and how everything moves, which, to me, brings the building to life. Also of interest is the need to consider interactions with non-structural components and the fact that you’re pushing materials to, or beyond, their limits, and that for typical structures our primary focus is about saving lives.”

On the ground in Christchurch

McBean had the opportunity to experience this firsthand in the aftermath of the Christchurch earthquake in 2011, as one of two engineers embedded within the Australian Urban Search and Rescue team sent to the region.  “For two weeks we lived in a tent in the middle of the red zone in the city,” McBean said.  “Each day we were given a task, to search a particular building or structure … It was very much like being on a post-apocalyptic movie set; broken buildings and no people. “Entering a hotel, we saw a credit card still in the machine and people’s luggage waiting at the reception counter. Someone checking in had run for their life during the earthquake and left everything where it was.”

Christchurch Cathedral after the 22 March 2011 earthquake. Source: Margaret Low, GNS Science. (not part of the create article).

He was awarded the Humanitarian Overseas Service Medal by the Australian Government for this work, and the experience had an influential effect on McBean’s career, and on him personally. “A lot of the construction techniques used [in Christchurch], particularly in many older buildings, are similar to those used in Adelaide and other parts of Australia,” he said.  “So to see these buildings so badly damaged and broken was a real wake-up call for me.” Interestingly, it was the performance of some of the newer buildings that came as a surprise. “What shocked me most was that most of the people who died were victims of the collapse of two relatively modern buildings. Those structures shouldn’t have catastrophically failed. They were preventable failures, in my view,” he said. “Coming back to my hometown of Adelaide, for the first couple of weeks I was walking around thinking to myself, ‘That’s going to come down, that building’s not going to survive [an earthquake]’. “Since then, I’ve been on a bit of a crusade to do what I can to prevent such collapses from happening here in Australia.”

Incorporating lessons learned

Following his experience in Christchurch, McBean saw the opportunity to incorporate some of these lessons into the design of the new Royal Adelaide Hospital, a project he had started work on prior to attending the New Zealand disaster.  “I came back from Christchurch knowing the importance of redundant load paths, of the vulnerability of transverse structures, of the need to design robust columns, and the need to detail and design structural walls better than we had been doing,” he said.  “And, just as important for a major hospital with a post-disaster role, I saw the need to influence the design of non-structural components.” The new hospital, completed in 2017, is designed to withstand a 1-in-1500-year earthquake without collapse and to remain fully operational after a 1-in-500-year earthquake event.

The new Royal Adelaide Hospital is designed to remain fully operational after a 1-in-500-year earthquake event.

“We made sure, for example, that the restraints for ceilings and services weren’t just compliant with the loads you’d expect, but detailed properly so they could survive the earthquake and remain operational,” McBean said. “For a major hospital, you need the power to stay on, the water to continue to flow in the pipes, the partitions to remain standing, and so on, right down to the dispensary in the pharmacy, which is obviously critical for a hospital. The attention to detail needed during design, construction and commissioning for such facilities is extraordinarily high.”

Making change

Many designers are unaware of the fundamental differences between wind and earthquake design philosophies, McBean said, noting there is significant complacency surrounding the design for earthquakes in Australia, largely due to the low frequency with which they have historically occurred in populated areas. Keen to improve the way earthquakes are perceived and treated by designers, McBean joined the Standards Australia Committees for both AS1170.4, Earthquake Actions in Australia, and AS3600, Concrete Structures, and served as President of the Australian Earthquake Engineering Society from 2016 to 2019. “We were able to mount a compelling argument for substantive change to the seismic provisions of the Concrete Structures code,” McBean said. “This was extensively updated and a revision published in 2018, which is now called up under the National Construction Code. “A lot of the practices that we saw being done that needed fixing are now effectively law as a minimum standard in the country. That’s been a real achievement, and it started on the journey after Christchurch.”

Looking to the future

While buildings designed to the updated 2018 standard should perform well during an earthquake, it’s older structures that remain problematic. “If we are lucky and don’t get a major earthquake in a populated area for 50 years, the benefits of the changes made in 2018 will be profound… [but] many structures will take a generation or two to replace or upgrade,” McBean said.

It’s also the long-term recovery he finds concerning. “The current earthquake approach to design in the country, almost universally, is to allow the structure to dissipate the earthquake energy by inelastic deformation,” McBean said. “If there’s a hit on a major city at the moment, even if the buildings were designed well, we’d be bulldozing most of the city and rebuilding it. The economic and social dislocation of that would be with us for a decade or more.”

New techniques and construction methodologies being explored could change this.

“You can design buildings in such a way that they fail in predetermined locations, and then those parts can be selectively replaced,” McBean said. “This costs a little bit more at this stage, but if you’re comparing it to the cost of pulling a building down, getting a new one designed and built, then you might be interested in paying a few per cent more upfront as insurance.”


A view of Australia’s climate in 2020

A view of Australia’s climate in 2020

Stuart Browning, Tom Mortlock, Andrew Gissing and Paul Somerville, Risk Frontiers

The recently released biennial BoM/CSIRO State of the Climate 2020 Report (SoC 2020) provides a summary of the changes that have occurred in Australia over the past century, including projections for the future. Australia’s climate has warmed on average by 1.44 ± 0.24 degrees Celsius since national records began in 1910, which is shown to be linked to an increase in the frequency of extreme heat events. This observed increase is within error of the increase of 1.5 degrees relative to pre-industrial levels recommended as a global (best-case) target under the 2015 Paris Climate Accord, to which Australia is a signatory.

Long-term warming leading to higher probabilities of extremes

As with previous reports, SoC 2020 has a primary focus on the increasing frequency and severity of extremes. This is of particular relevance to insurance and emergency management as most natural disasters affecting Australia are weather or climate related. The report concludes that the science has been broadly consistent and largely accurate in the way that it has described and projected the climate system for the last several decades. Australia is experiencing climate change now, and the warming trend is continuing (Figure 1).

Figure 1. The number of days each year in which the Australia area averaged daily mean temperature each month is extreme (warmest 1% of days for each month). Source: BOM/CSIRO (2020)
Figure 1. The number of days each year in which the Australia area averaged daily mean temperature each month is extreme (warmest 1% of days for each month). Source: BOM/CSIRO (2020)

In the two years since the SoC 2018 report, Australia has experienced its hottest and driest year on record: 2019. These extremes, combined with anomalously windy conditions, played a pivotal role in the disastrous Black Summer fires. SoC 2020 explains that these conditions would not be considered extreme in a 1.5-degree warmer world (best case scenario), let alone the 6 degree warmer world projected by some simulations under a ‘business as usual’ scenario.

Climate extremes experienced in Australia, and globally during 2019, are part of a continuing trend where each decade is hotter than the previous one—this warming is entirely consistent with global climate model projections produced more than 20 years ago (Hausfather et al 2020). Global climate models continue to paint a disturbing picture of future climate under business as usual scenarios. One of the key arguments for inaction on addressing climate risk has been perceived uncertainty around climate model skill. As we move further into the 21st century we are now able to assess the performance of previous generations of climate models. This proof-of-skill comes at a critical time, as projections from the latest generation of the Coupled Model Intercomparison Project (CMIP) models are increasingly incorporated into climate related decision making and disclosures across the global financial sector.

SoC 2020 also includes a brief discussion on the role of climate trends on compound disasters. Compound disasters describe a situation where successive events occur over a short timeframe. Due to their role in amplifying losses and straining resilience, Risk Frontiers and the Bushfire and Natural Hazards CRC have recently completed a comprehensive research piece on compound disasters (Gissing et al., 2020). Natural climate drivers, such as El Niño Southern Oscillation (ENSO), strongly influence the type of events that comprise compound disasters. For example, during El Niño years, compound disasters usually comprise of heatwave and fire, and are often superimposed on drought stressors, whereas during La Niña years compound disasters are more likely to comprise storm, flood, and cyclone. This helps to explain why the approaching 2020-21 summer season will be very different from last year. While swings from drought to flood are a natural and defining characteristic of Australian climate, SoC 2020 describes how long-term warming trends amplify this natural variability and will continue to do so in coming decades.

Long-term warming increases the likelihood of extreme events beyond our historical experience, calling into question the applicability of historical records to accurately project the risk environment of the future. For this reason, Risk Frontiers has now ‘climate-enabled’ their catastrophe loss models using the latest CMIP6 simulations, meaning estimated financial losses due to extreme weather events can be modelled for a range of future climate scenarios and time horizons by blending and downscaling historical and future climate model data.


BOM/CSIRO (2020). State of the Climate 2020.

Gissing, Andrew, Matthew Timms, Stuart Browning, Lucinda Coates, Ryan Crompton and John McAneney (2020). Compound Natural Disasters in Australia: A Historical Analysis. [Available online]

Hausfather, Z., Drake, H. F., Abbott, T. and Schmidt, G. A.: Evaluating the Performance of Past Climate Model Projections, Geophys. Res. Lett., 47(1), 1535, doi:10.1029/2019GL085378, 2020.

Newsletter Volume 19, Issue 4

A national view of storm tide flooding in Australia

Thomas Mortlock

As the summer of 2020/21 fast approaches, the strengthening La Niña in the Pacific not only signals heightened flood risk along the east coast (Ward et al., 2014), but also the probability of above-average tropical cyclone activity (Kuleshov et al., 2008). Storm surge associated with tropical cyclones and extra-tropical storms such as East Coast Lows can cause significant damage to property and loss of life when they coincide with high tides. For example, during tropical cyclone Yasi in 2011 (which coincided with the last major La Niña event in Australia), a 5-metre storm tide (storm surge plus tide) was observed at Cardwell – 2.3 metres above the Highest Astronomical Tide. The Port of Hinchinbrook Marina was destroyed along with millions of dollars’ worth of boats (Figure 1) and the Lucinda Bulk Sugar Terminal. Between Cairns and Townsville, damage was reported to roads, sea walls, buildings and vegetation, with significant beach erosion (BoM, 2020).

Figure 1. Damage to boats at Hinchinbrook Marina in the aftermath of Yasi’s storm surge. Source:

Recent updates to Risk Frontiers’ Address-Based Risk Rating Database means that a risk rating for storm tide inundation, resulting from both tropical cyclones and extra-tropical storms, is now available for every address in Australia, for both present-day and future climate scenarios. This provides an up-to-date national view of storm tide flood risk for the first time since the National Coastal Risk Assessment looked at the impact of sea level rise (and, in some states, storm tide) in 2011 (DCCEE, 2011).


Storm tide modelling is based on a long-term sea level hindcast (Pattiaratchi et al., 2018) combined with tropical cyclone storm tide exceedances generated in a study using Risk Frontiers’ tropical cyclone wind loss model, CyclAUS (Haigh et al., 2014). Tidal attenuation and amplification of the open-coast tide in rivers and estuaries is accounted for after Hanslow et al. (2018). Storm tide flood depths are mapped onto the coast using high-resolution (5 m) digital elevation data and a hydraulic connectivity routine. Hazard data are intersected with Geocoded National Address File (G-NAF) information to estimate storm tide flood depths at the address level.

The present-day storm tide flood hazard is defined as the 1% Annual Exceedance Probability (AEP) combined tide and storm surge water depth. Relative rates of sea level rise from McInnes et al. (2015) are applied to the 1% AEP storm tide to model future changes to exposures under Representative Concentration Pathways (RCP) 2.6 (low emissions scenario) and 8.5 (high emissions scenario), for the 2030s, 2050s and 2090s. Here, sea level rise is regarded as the primary driver of future changes to storm tide risk; future changes to tropical cyclone and extra-tropical storm activity were not included. In this article, present day exposure is compared to changes modelled for RCP 2.6 and 8.5 by the 2090s.

Storm tide risk highest in Queensland

There are approximately 179,000 addresses at risk of the 1% AEP storm tide flood (i.e. properties that have a 1% probability of coastal flooding in any given year under current climate conditions) around Australia. This increases to around 243,000 (+ 36%) by the 2090s with sea level rise projected under a low emissions scenario (RCP 2.6) and 271,000 (+ 51%) under a high emissions scenario (RCP 8.5) (assuming present-day building stock and with no coastal defences in place).

This is comparable to earlier findings by the National Coastal Risk Assessment (DCCEE, 2011), which identified between 157,000 and 247,600 addresses potentially at risk of inundation by sea level rise by the end of the century. However, the DCCEE study did not include storm tide effects on a national basis and for this reason, exposure to coastal flooding is likely to have been underestimated. The increase in building stock over the past decade is likely to be another reason for the slightly lower numbers compared to this present study.

As shown in Figure 2, most of the present-day storm tide risk is concentrated in Queensland (almost two-thirds of the national total) and NSW (about one-fifth). The highest at-risk suburbs are located in the Gold Coast area, with almost 32,000 properties currently at risk of storm tide flooding (in Gold Coast, Palm Beach, South Stradbroke, Broadbeach and Southport) and an average increase of 60% in exposed addresses projected across these suburbs for the 2090s. In fact, nine of the top 10 at-risk postcodes in the country are found in Queensland (Mandurah in WA is ranked second, Table 1).

Figure 2. Present day exposure to 1 % AEP storm tide flooding by postcode.

Over the last 50 years, most tropical cyclones in Australia have remained equatorward of 25° South but occasionally they track lower into South East Queensland. Risk Frontiers maintains a database of natural hazards in Australia dating back to 1788 (PerilAUS). The PerilAUS database shows there have been 46 tropical cyclones near Brisbane and the Gold Coast. Most have been minor; the most damaging were tropical cyclones Dinah in January 1967 and Wanda in 1974. The large increase in coastal development on the Gold Coast since the 1970s drives present-day storm tide risk much more so than when tropical cyclones Dinah and Wanda impacted this area.

Table 1.Top 10 postcodes by number of properties with a 1 % probability of storm tide flooding in any given year. All postcodes experience the same level of change under RCP 2.6 and 8.5 by the 2090s because of non-sea level factors such as location of addresses and coastal topography.

Sydney and Brisbane potential future hotspots for storm tide flooding

While the Gold Coast has the highest concentration of properties at risk of storm tide flooding, the largest increases by the end of the century are projected for Sydney and Brisbane suburbs. In Sydney, the Botany Bay and Tempe areas, and Rozelle and Canada Bay on the Parramatta River, show some of the largest increases in the country, while the Brisbane CBD and Launceston in Tasmania also exhibit a substantive rise in exposure (Table 2). Port Hedland (WA) and Port Adelaide (SA) are other industry hubs where results suggest increases of over 400 % in storm tide risk by the 2090s.

Table 2. Top 10 postcodes by magnitude of change by the 2090s.

Sydney may benefit more than most from lower rates of sea level rise

As shown in Table 2, there is a big difference in storm tide flooding between low and high future emissions scenarios in Sydney. All Sydney suburbs showing the largest increases in exposure by the 2090s relative to present day under RCP 8.5 (Botany Bay, Rozelle, Tempe, Canada Bay) show no increase under RCP 2.6. This suggests that Sydney may be one of the greatest beneficiaries of emissions reductions by the end of the century in terms of reduced exposure to storm tide flooding in low-lying city suburbs around Botany Bay and along the Parramatta River.

At present, sea level rise of about 0.9 m by 2090 is used as the planning benchmark, synonymous with an RCP 8.5 emissions trajectory. However, it is important to note that the amount of sea level rise we can expect over the coming century is deeply uncertain because of major unknowns around ice sheet dynamics. Some studies (Bakker et al., 2017) suggest if substantive glacial basins of the West Antarctic Ice Sheet were to collapse, it could contribute at least a further two metres to global sea levels.

A national picture with considerable regional variation

This new national picture of storm tide risk exposes a large amount of regional variation, reflective of coastal topography, development, storm surge climate and relative rates of sea level rise. It shows that, while most of the present day risk is concentrated in Southeast Queensland, suburbs in Sydney and Brisbane may see significant increases in future storm tide flooding, depending on the rates of sea level rise that occur over the coming century, which, as noted, is deeply uncertain.

This dataset forms part of Risk Frontiers’ multi-hazard Risk Rating Database, and is suitable for national to suburb-level risk scoping. It does not replace the need for hydraulic modelling required to understand the complexities of localised storm tide flooding.


Bakker, A. et al. (2017). Sea-level projections representing the deeply uncertain contribution of the West Antarctic ice sheet. Sci Rep, 7, 3880.
Bureau of Meteorology [BoM] (2020). Understanding storm surge – Social Media Release. Bureau of Meteorology, October 2020.
Department of Climate Change and Energy Efficiency [DCCEE] (2011). Climate change risk to coastal buildings and infrastructure – a supplement to the first pass national assessment.
Haigh, I., et al. (2014). Estimating present day extreme water level exceedance probabilities around the coastline of Australia: tropical cyclone-induced storm surges. Clim Dyn, 42(1-2), 39-157.
Hanslow, D.J., et al. (2018). A regional scale approach to assessing current and potential future exposure to tidal inundation in different types of estuaries. Sci Rep, 8, 7065.
Kuleshov, Y., et al. (2008). On tropical cyclone activity in the Southern Hemisphere: Trends and the ENSO connection. Geophys. Res. Lett., 35, L14S08.
McInnes, K.L., al. (2015) Sea-level rise projections for Australia: information for impact and adaptation planning. Austr Met Ocean J, 65, 127–149.
Pattiaratchi, C., et al. (2018). Developing better predictions for extreme water levels: Final data report. University of Western Australia and Bushfire and Natural Hazards CRC, May 2018, pp 35.
Ward, P.J., et al. (2014). Strong influence of El Niño Southern Oscillation on flood risk around the world. PNAS, 111(44), 15659–15664.

From drought to flood: Australia’s seasonal forecast for summer 2021

Stuart Browning

Given the confronting bushfire weather experienced last summer there is strong interest in the climate outlook for summer 2021. At seasonal timescales there is a spectrum of predictability, from the immutable fact that summer will be warmer than winter (radiation increases due to Earth’s solar orbit and axial tilt), to the complete impossibility of predicting New Year’s Eve weather at the start of December (due to the stochastic variability of day to day weather). In between these endpoints are slow-changing components of the climate system, mostly tropical sea surface temperatures (SST), that provide us with some predictability on how each season might unfold.

The Australian Bureau of Meteorology (BOM) recently declared a La Niña event. This is significant for seasonal forecasting as La Niña events, when they occur, tend to persist though most of summer and usually result in wetter than average conditions for much of Australia. La Niña is one state of the world’s leading climate driver: the El Niño Southern Oscillation (ENSO). ENSO is usually defined by SST in the Tropical Pacific (a region called Niño 3.4 shown in Figure 1) and influences weather around the globe, including the Americas, Africa, New Zealand and Australia. The latest multi-model ENSO outlook (Niño 3.4 SST) shows moderate to strong La Niña conditions will persist for the remainder of 2020, and probably though until Autumn 2021. For Australian seasonal forecasts, the BOM also monitors tropical SST in the Indian Ocean, described by the Indian Ocean Dipole (IOD; also shown in Figure 1) and winds over the Southern Ocean, described by the Southern Annular Mode (SAM).

Figure 1 ECWMF seasonal forecast for global sea surface temperature anomalies (SSTa) during January 2021 showing a classical La Niña cold tongue extending across the tropical Pacific (source: C3S seasonal forecast). Boxes show the regions typically used for climate driver index calculations: in the central tropical Pacific, SSTa in the Nino 3.4 region are used to represent ENSO; and in the tropical Indian Ocean, the difference between SSTa in the IODW and IODE regions are used to represent the Indian Ocean Dipole. Warmer SSTa in the Australian region are generally associated with lower atmospheric pressure and increased precipitation.

The BOM’s seasonal forecast model, ACCESS-S, provides outlooks for a range of useful parameters including rainfall and temperature: Figure 2 shows this coming summer will be wetter than normal for most of Australia, and cooler than normal for some regions including coastal NSW. Forecasts for more specific parameters, such as streamflow, fire danger, and tropical cyclone activity can be derived from forecasts models, but are often developed using statistical methods based on historical relationships with major climate drivers. Seasonal forecasting of specific perils is an area of current research focus with a range of statistical, dynamical, and machine leaning approaches being explored.

Past La Niña events provide a useful analogue for what we can expect this coming summer. Tropical cyclone activity has historically been higher during La Niña seasons, with some of the most destructive cyclones on record, such as Tropical Cyclones Yasi (2011) and Wanda (1974) occurring during strong La Niña years. The BOM tropical cyclone outlook, based on the historical relationship with ENSO, is predicting 11 cyclones this season, which is slightly above average. Riverine flooding is also expected to have a higher probability this summer due to model forecasts for high precipitation (Figure 2), and that historically flooding occurs more often during La Niña events, such as 2011 when widespread flooding occurred across the eastern states including the overflow of Wivenhoe Dam and subsequent Brisbane floods. Rainfall during the 2011 La Nina transported so much water from the oceans to inland Australia that global sea levels lowered by ~5mm (Boening et al 2012). Overall increases in rainfall and storm activity mean that previous La Niña events have seen an increase in the frequency of compound disasters and overall insured losses.

Figure 2: ACCESS-S outlook for summer rainfall and temperature (source: BOM)

La Niña-related rainfall also has the unfortunate consequence of increasing mosquito populations and the spread of vector borne disease. The 1974 La Niña saw Australia’s wettest year on record contribute to widespread outbreaks of Murry Valley Encephalitis and Ross River virus, while the recent 2010-12 La Niña saw over 1000 cases of Ross River Virus and an outbreak of Barmah Forest Virus in Victoria.

The other major tropical climate driver to influence Australia is the IOD. ENSO and IOD are usually coupled, where El Niño couples with IOD positive and La Niña couples with IOD negative. Anomalously strong IOD positive last year was one of the main climate factors contributing to the Black Summer bushfires. We are facing the opposite situation this year, with IOD negative during spring 2020 having contributed to recent drought-breaking rains across south eastern Australia. Most models are currently forecasting neutral to negative IOD conditions for the remainder of spring and early summer indicating much lower fire risk for this season relative to 2019/20. The unfortunate caveat is that increased springtime precipitation encourages vegetative growth, so that if we do experience a dry period over summer, grasses can cure quickly raising the risk of grass fires. One of the largest areas burned on record for Australia (~1.5 million Ha) occurred during the 1974-75 summer when vast grassfires burned across central NSW: this was during one of the longest and most sustained La Niña periods in the past century.

By this time last year the fire season was well underway with out-of-control blazes in southeast Queensland; the same regions are now being battered by successive storms while Victoria and NSW are dealing with floods. Tumbarumba, for example, was devastated by last summers fires and is already under flood waters this spring. Such seemingly wild swings from one year to the next are a familiar and defining characteristic of Australia’s climate. There are numerous examples where a strong El Niño is followed by a strong La Niña, such as the 1982-83 El Niño and Ash Wednesday fires which were followed by La Niña and flooding in 1984, or the more recent 2009 El Niño and Black Saturday fires which were followed by La Niña and widespread floods in 2010-12. While paleoclimate research shows this is a typical feature of Australia’s climate, there is some evidence that ENSO variability (swings from one extreme to the next) has increased in recent centuries and will continue to strengthen in coming decades (Cai et al 2015).

Seasonal outlooks for La Niña and associated increased rainfall might be welcome news for many Australian communities that have been dealing with protracted drought. A milder bushfire season should also be appreciated by those still recovering from the devastation of last summer. However, if this year is similar to previous La Niña events then we should prepare for more rainy days and an overall increase in weather related risk and insured losses.


Boening, C.: Oceanography: Detecting sea-level rise, Nature Publishing Group, 4(5), 327–328, doi:10.1038/nclimate2205, 2014.
Cai, W., Santoso, A., Wang, G., Yeh, S.-W., An, S.-I., Cobb, K. M., Collins, M., Guilyardi, E., Jin, F.-F., Kug, J.-S., Lengaigne, M., Mcphaden, M. J., Takahashi, K., Timmermann, A., Vecchi, G., Watanabe, M. and Wu, L.: ENSO and greenhouse warming, Nature Climate Change, 5(9), 849–859, doi:10.1038/nclimate2743, 2015.

How Overdispersion Drives the Spread of the SARS-CoV-2 Pandemic

Paul Somerville, Risk Frontiers

Even after months of extensive research by the global scientific community, many questions about the spread of the SARS-Covid-2 pandemic remain unanswered. Widespread expectations of catastrophic outbreaks in China, South Korea and Japan were not realized. In the early months of 2020, a few cities such as New York accounted for a substantial portion of global deaths, while many others with similar population density, weather, age distribution, and travel patterns were spared. There was an enormous death toll in northern Italy, but not the rest of the country. In Guayaquil, Ecuador, so many people died so quickly in April that bodies were abandoned in streets. A snapshot of this heterogeneity in the average daily number of cases per 100,000 people by country and within the United States in the week of October 18-24 is shown in Figures 1 and 2 respectively.

Population-level analyses often use average quantities to describe heterogeneous systems, particularly when variation does not arise from identifiable groups. A prominent example, central to the understanding of epidemic spread, is the basic reproductive number, R0, which is defined as the mean number of infections caused by an infected individual in a susceptible population. Most of the discussion about the spread of SARS-CoV-2 has concentrated on R0 which, without social distancing, is about 3. In real life, however, some people infect many others and others do not spread the disease at all, so the most common individual number is zero. Hence population estimates of R0 can obscure considerable individual variation in infectiousness, as highlighted during the global emergence of severe acute respiratory syndrome (SARS) and SARS-CoV-2 by numerous ‘superspreading events’ in which certain individuals infected unusually large numbers of secondary cases.

Lloyd-Smith et al. (2005) introduced the ‘individual reproductive number’, ν, as a random variable representing the expected number of secondary cases caused by a particular infected individual. Values for ν are drawn from a continuous probability distribution with population mean R0 that encodes all variation in infectious histories of individuals, including properties of the host and pathogen and environmental circumstances. In this framework, superspreading events are not exceptional events, but important realizations from the right-hand tail of a distribution of ν. Stochastic effects in transmission are modelled using a Poisson process, so that the number of secondary infections caused by each case, Z, is described by an ‘offspring distribution’ Pr(Z = k) where Z∼Poisson(ν). In their preferred model, Lloyd-Smith et al. (2005) let ν be gamma-distributed with mean R0 and dispersion parameter k, yielding Z∼negative binomial(R0,k). The negative binomial model includes the conventional Poisson (k → ∞) and geometric (k = 1) models as special cases. It has variance R0(1 + R0/k), so smaller values of k indicate greater heterogeneity. The parameter k is referred to as the overdispersion, and represents the random variability in the mean of the Poisson distribution. An analogous departure from a conventional Poisson model was encountered in our briefing on the temporal clustering of large earthquakes (Briefing note 412).

Figure 1. Average daily cases per 100,000 people by country in the week of October 18-24, 2020. Source: New York Times.
Figure 2. Average daily cases per 100,000 people in the United States in the week of October 18-24, 2020. Source: New York Times

The role of overdispersion in the spread of SARS-Covid2 has been described by Kupferschmidt (2020) and Tufekci (2020). Lloyd-Smith et al. (2005) estimated that SARS – in which superspreading played a major role – had a k of 0.16. The estimated k for MERS, which emerged in 2012, is about 0.25. In the flu pandemic of 1918, in contrast, the value was about one, indicating that spatiotemporal clusters played less of a role.  The most recent estimate of k for SARS-CoV-2 is about 0.1 (Endo et al., 2020), indicating a higher spatiotemporal clustering rate than for all the other pandemics, with 10% of the cases leading to about 80% of the spread. This kind of behavior, alternating between being super infectious and fairly noninfectious, is what k captures and what focusing solely on R0 hides. This has presented a large challenge, especially for health authorities in Western societies, where the approach to controlling the pandemic has been modeled on the flu.

Disease patterns can be thought of as having deterministic or stochastic trends. In the deterministic case, an outbreak’s distribution is more linear and predictable, while in the stochastic case, randomness plays a much larger role and predictions are hard, if not impossible, to make. In deterministic trajectories, what happened yesterday is expected to give a good estimate of what to expect tomorrow. Diseases like the flu are practically deterministic and adequately represented by R0, and they are nearly impossible to stop until there is a vaccine. Stochastic phenomena, however, do not operate that predictably, and random variations can rapidly tip conditions from one state to another.

The highly skewed, imbalanced distribution of k in stochastically spreading cases like SARS-CoV-2 means that an early run of bad luck with a few super-spreading events, or clusters, can produce dramatically different outcomes even for otherwise similar countries, as occurred in Italy and South Korea (Figure 3). Scientists who have looked globally at known early-introduction events, in which an infected person comes into a country, found that in some places, such imported cases led to no deaths or known infections, while in others, they sparked sizable outbreaks. This could explain some puzzling aspects of this pandemic, including why the virus did not rapidly spread around the world sooner after it emerged in China, and why some very early cases elsewhere – such as one in France in late December 2019, apparently failed to ignite a wider outbreak. If k is really 0.1, then most chains of infection die out by themselves and SARS-CoV-2 needs to be introduced undetected into a new country at least four times to have an even chance of establishing itself. Most of the cases that left China simply fizzled out. Using genomic analysis, researchers in New Zealand looked at more than half the confirmed cases in the country and found 277 separate introductions in the early months, but that only 19 percent of introductions led to more than one additional case. This may even be true in congregate living spaces, such as nursing homes, and multiple introductions may be necessary before an outbreak takes off.

Meanwhile, in Daegu, South Korea, just one woman, dubbed Patient 31, generated more than 5,000 known cases in a megachurch cluster. Nevertheless, overdispersion is also a cause for hope, as demonstrated by South Korea’s aggressive and successful response to that outbreak with a massive testing, tracing, and isolating regime. Since then, South Korea has also been practicing sustained vigilance, and has demonstrated the importance of backward tracing. When a series of clusters linked to nightclubs broke out in Seoul recently, health authorities aggressively traced and tested tens of thousands of people linked to the venues.

Figure 3. Seven-day rolling average of new cases in selected countries, by number of days since 10 average daily cases first recorded, October 26, 2020.  Source: Financial Times.

Australia targeted mass suppression of the virus with a focus on clusters, and generally succeeded in this approach, with the temporary exception of widespread community transmission in Melbourne due to failure of quarantine measures. Australia invested in widespread testing early on and used mechanisms to monitor contact tracing.

One of the most interesting cases has been Japan, a country that was hit early on, like Australia, by passengers from a cruise ship. Japan followed what appeared to be an unconventional model, not deploying mass testing and never fully shutting down. By the end of March, influential economists were publishing reports with dire warnings, predicting overloads in the hospital system and huge spikes in deaths. The predicted catastrophe never came to be, however, and although it faced some future waves, there was never a large spike in deaths despite its aging population, uninterrupted use of mass transportation, dense cities, and lack of a formal lockdown.

In the beginning, Japan was no better situated than other countries such as the United States and those in Western Europe that now have much worse outcomes. Like them, Japan did not initially have the capacity to do widespread testing. Nor could Japan impose a full lockdown or strict stay-at-home orders, because even if that had been desirable, it would not have been legally possible. Instead, the Japanese specialists had noticed the overdispersion characteristics of COVID-19 as early as February, and so they created a strategy focusing mostly on cluster-busting, which tries to prevent one cluster from igniting another. This entailed aggressive backward tracing to uncover clusters. Japan also focused on ventilation and counseling its population to avoid places where the three C’s come together – crowds in closed spaces in close contact. Cultural factors including wearing masks in public especially during the flu season, and consideration of the welfare of others.

This strategy is in contrast with the Western response, which tries to eliminate the disease case by case, when that is not necessarily the main way it spreads. Japan did get its cases down and kept up its vigilance. When the government started noticing an uptick in community cases, it initiated a state of emergency in April and tried hard to incentivize the kinds of businesses that could lead to super-spreading events, such as theatres, music venues, and sports stadiums, to close down temporarily. Now schools are back in session in person, and even stadiums are open, but without chanting.

Countries that have ignored super-spreading have risked getting the worst of both worlds: burdensome restrictions that fail to achieve substantial mitigation. The UK’s recent decision to limit outdoor gatherings to six people while allowing pubs and bars to remain open is just one of many such examples. Many studies have shown that super-spreading clusters of COVID-19 almost overwhelmingly occur in poorly ventilated, indoor environments where many people congregate over time – weddings, churches, choirs, gyms, funerals, restaurants, and such – especially when there is loud talking or singing without masks. For super-spreading events to occur, multiple things have to be happening at the same time, and the risk is not the same in every setting and activity.

If public health workers know where clusters are likely to happen, they can try to prevent them and avoid shutting down broad swaths of society. Lockdowns are a blunt tool, because they concede that not enough is known about where transmission is happening to be able to target it, so everything is indiscriminately targeted. Unfortunately, studying large COVID-19 clusters is harder than it seems. Many countries have not collected the kind of detailed contact tracing data needed, and the lockdowns have been so effective that they also robbed researchers of a chance to study superspreading events. Before the lockdowns, there was probably a 2-week window of opportunity when a lot of these data could have been collected.


Endo A, Centre for the Mathematical Modelling of Infectious Diseases COVID-19 Working Group, Abbott S et al. Estimating the overdispersion in COVID-19 transmission using outbreak sizes outside China [version 3; peer review: 2 approved]. Wellcome Open Res 2020, 5:67 (

Financial Times (2020).

Kupferschmidt, Kai (2020). Why do some COVID-19 patients infect many others, whereas most don’t spread the virus at all? Science Magazine, May 19, 2020.

Lloyd-Smith, J., Schreiber, S., Kopp, P. et al. Superspreading and the effect of individual variation on disease emergence. Nature 438, 355–359 (2005).

Risk Frontiers (2020). Devil’s Staircase of Earthquake Occurrence: Implications for Seismic Hazard in Australia and New Zealand. Briefing note 412, April 2020.

Tufekci, Zeynep (2020). This Overlooked Variable Is the Key to the Pandemic It’s not R. The Atlantic, September 30, 2020


Tracing tsunami impacts back to their source across ocean basins

Paul Somerville, Chief Geoscientist, Risk Frontiers

Large transoceanic tsunamis have impacts that extend far from the earthquake that cause them, and so it often occurs that tsunami impacts that are recorded in historical documents or in geological sand deposits cannot be easily associated with their source. In the course of preparing this Briefing, I found a possible association of a tsunami deposit in Tanzania and fourteen tsunami deposits in South Asia with the AD 900 tsunami at Nagapattinam, India.

The 1700 “orphan” tsunami in Japan traced back to Cascadia, western North America

On the night of January 27, 1700, a mysterious tsunami flooded fields and washed away houses in Japan. It arrived without the warning that a nearby earthquake usually provides. Samurai, merchants and villagers recorded the event, but nearly three centuries passed before discoveries in North America revealed the tsunami’s source: an earthquake of estimated Mw 9.0 that occurred on the Cascadia subduction zone off the coast of Washington and Oregon. The evidence there consisted of trees that were swamped by the coastal subsidence caused by the earthquake, with tree rings indicating their approximate time of death (Atwater et al., 2016).

Figure 1. The 1700 Cascadia tsunami.  Source: Atwater et al., 2016.

Tracing tsunami deposits in Tanzania back to their likely tsunami source in Sumatra-Andaman

The December 2004 Mw 9.1 Sumatra-Andaman earthquake generated a tsunami whose greatest impact was felt in Indonesia, Sri Lanka, India, and Thailand, where more than 200,000 people lost their lives. Nine hours later, more than 5,000 km from the earthquake’s epicenter, the tsunami reached the coastline of Eastern Africa (Figure 2). Runup heights of almost 10m were measured in Somalia, where 298 fatalities occurred. Reduced impacts were observed farther south, along the African coast of Kenya and Tanzania, probably because it struck on a low tide. There was no tsunami warning system in the Indian Ocean at the time of the 2004 event, but one has now been implemented.

Figure 2. The 2004 Sumatra-Andaman Tsunami. Source: NOAA.

Learning more about tsunami risks has global implications, but the December 2004 tsunami has been predominantly seen as an eastern Indian Ocean event, and as a consequence, much of the work to find ancient tsunami deposits and to understand the recurrence intervals of such catastrophic events has been focused on that region. Tsunami risk has been thought to be low in East African countries, mainly because of the limited damage caused by the 2004 tsunami.

However, in Pangani Bay, Tanzania, Maselli et al. (2020) identified evidence of a deadly tsunami that they conclude occurred about 1,000 years ago, suggesting that the tsunami risk in East Africa could be higher than previously thought. They visited a field site close to Pangani Bay, where they discovered the tsunami deposit. At a depth of about 1.5 meters, they found a sand layer (Figure 3) hosting human remains lacking traditional funerary burial. The bones did not present any evidence of disease or trauma due to battle. Tanzania is subject to cyclones, and as noted by Shanmugam (2012), there can be considerable ambiguity in distinguishing between paleo-tsunami deposits and paleo-cyclone deposits using sedimentological criteria, and considerable uncertainty in the age dating of these deposits.

The mixed fossil assemblage indicative of continental, estuarine, and marine habitats was present within the sand layer. The occurrence of marine shells supported the hypothesis that a tsunami may have impacted the area, although a cyclone could have done the same. Radiocarbon dating indicated that the event that deposited the sand layer in Pangani Bay occurred about 1,000 years ago.

Sedimentary evidence of paleo-tsunami deposits of the same age were reported from Thailand, India, Indonesia, southern Sri Lanka and the Maldives (Figure 4), pointing to an event in about the year 950.

Figure 3. Left: Tsunami deposit in Pangani Bay indicated by the sand layers. Credit: Davide Oppo. Right: Skeletal remains of a victim of the 1,000-year-old Indian Ocean tsunami. Credit: Vittorio Maselli.
Figure 4. Tsunami deposits found in Indian Ocean coastal sites. The horizontal red line highlights the timing of the inferred Tanzania tsunami at 950 AD. Source: Maselli et al., 2020.

Although Maselli et al. (2020) did not associate the Pangani Bay, Tanzania tsunami deposit with a historical earthquake, the NOAA Global Historical Tsunami Database (NOAA, 2020) contains an event at Nagapattinam, India in 900 AD that is estimated to have killed several hundred people, based on Rastogi and Jaiswal (2006) who note that:

“There is mention of tsunami effect in scriptures at Nagapattinam in 900 AD that destroyed a Buddhist monastery. According to literature available in the library of Thondaiman kingdom in Puduckottai, Tamilnadu, it was during the reign of Raja Raja Chola that waves had washed away the monastery and several temples and killed hundreds of people. There is evidence of this in Kalaki Krishnamurty’s book ‘Ponniyin Selvan-The Pinacle of Sacrifice,’ In the chapter ‘The Sea Rises’, the author explains how the sea had risen very high and the black mountain of water moved forward. The sea inundated warehouses and sheds and began to flow into the streets. Ships and boats seemed suspended in mid-air, precariously poised on the water peaks. The book also describes how an elephant was swallowed by the gushing water.”

It seems likely that this is the event that Maselli et al. (2020) identified in Pangani Bay, Tanzania. Further, considering the tsunamigenic earthquake sources that are present in the Indian Ocean (Jaiswal et al., 2008; Schafer and Friedmann, 2019, Figure 5), it is evident that the Sumatra-Andaman Sea Subduction zone is the only one with the potential to generate large transoceanic tsunamis in the Indian Ocean, so we sought published evidence from other geological  records of large tsunamis in that region to see if the AD 900 event is recorded in them.

Figure 5. Map of maximum magnitude estimates for the world’s major subduction zones. The inset obscures most of the southern Indian Ocean, where there are no subduction zones. Source: Schäfer and Friedemann, 2019.

The AD 900 event does not appear on the lists of events inferred to have occurred at sites in Aceh (Rubin et al., 2017) and the Andaman Islands (Malik et al., 2019), but in both cases it occurred at a time when conditions were such that a record would not have been made.  In the case of Aceh, Rubin et al. (2017) found evidence for at least 11 prehistoric tsunamis that struck the Aceh coast between 7,400 and 2,900 years ago, and state:

“The cave probably contained stratigraphic evidence of recent historic tsunamis from 2,900 years BP to the 2004 Indian Ocean tsunami that have been identified elsewhere in the region, but these were most likely removed by subsequent tsunamis inundating the cave as indicated by the erosional unconformity beneath the 2004 deposit.”

In the case of the Andaman Islands, Malik et al. (2019) describe evidence for 7 events between 1881 and before 5600– 5300 BCE, and state:

“The sequence includes an unexplained hiatus of two or three millennia ending around 1400 CE, which could be attributed to accelerated erosion due to Relative Sea-Level (RSL) fall at ~3500 BP.”

From this we conclude that the absence of observations of the AD 900 event at Aceh and the Andaman Islands does not preclude the likelihood that it occurred somewhere in this region.

How often do tsunamis like the Mw 9.1 2004 Sumatra-Andaman event occur?

The NOAA (2020) historical record of tsunamis in the Indian Ocean is very sparse before about 1681; with only the AD 416 event recorded in Java, the AD 900 event recorded in Nagapattinam, and the 1524 event recorded in Dabhol, India appearing in the catalogue before then. However, since 1681, there have been numerous events: 1 of Mw >9.0; 6 events of Mw 8.5-8.9; 8 events of Mw 8.0-8.4; 25 events of Mw 7.5-7.9, and 18 events of Mw 7.0-7.4, most of which have occurred in Indonesia, with many of those occurring east of Sumatra and not greatly affecting the Indian Ocean. Jaiswal et al. (2008) list 7 events affecting India and the surrounding region before 1668, and 14 events since then.

In Aceh, Rubin et al. (2017) conclude that the average time between tsunamis is about 450 years with intervals ranging from a long, dormant period of over 2,000 years, to multiple tsunamis within the span of a century. In the Andaman Islands, Malik et al. (2019) suggest a recurrence of 420–750 years for mega-earthquakes (magnitude Mw about 9), and a shorter interval of 80–120 years for large magnitude earthquakes (magnitude Mw about 8). Taken together, these studies suggest a recurrence interval of about 500 to 750 years for earthquakes like the Mw 9.1 2004 Sumatra-Andaman earthquake.

The Mw 9.1 2004 Sumatra-Andaman earthquake had little impact on Australia, with most of the impact occurring along the northwest coast of Western Australia. Localised inundation and tidal surges lasted for several hours along the Western Australian coast, resulting in boats losing their moorings and being damaged in marinas, and thirty swimmers needing to be rescued. The Mw 7.7 Pangandaran, west Java earthquake of 17 July 2006 generated a tsunami that affected several areas from north of Geraldton to Port Hedland, and inundated and destroyed a camp in the Steep Point area, where one family was fortunately able to move quickly from their camp to safe ground and another held onto their vehicle as it was moved 10 metres by the tsunami. Fortunately, the main sources of tsunami-generating earthquakes are sufficiently distant from Australia that the Australian Tsunami Warning System that is now in place can provide at least 1.5 hours of warning before the arrival of the tsunami onshore, providing adequate time for evacuation in most situations.

Given the locations of tsunami sources in the Indian Ocean (Figure 5) and the estimated recurrence interval of very large earthquakes on these sources, we expect the tsunami hazard in Australia to be concentrated along the northwest coast. This expectation is borne out in the probabilistic tsunami hazard map for Australia for an AEP of 1:500 (Figure 6, Davies and Griffin, 2018). The catastrophic Indonesian volcanic eruptions that occurred in Tambora in 1815 and Krakatoa in 1833 indicate the presence of a potentially more dangerous tsunami source if a larger eruption were to occur. The Krakatoa eruption produced a tsunami whose runup on the coast of Western Australia was in the range of 0.5-2 m (Allport & Blong, 1995).

Figure 6. Probabilistic near shore tsunami wave height in Australia for a water depth of 100m. Source: Davies and Griffin, 2018.


Tsunami impacts that are recorded in historical documents or in geological sand deposits often cannot be easily associated with their source. However, notwithstanding the ambiguity in distinguishing between tsunami and cyclone sources (e.g. Shanmugam, 2012), it seems likely that a 1,000 year old inferred deposit recently discovered in Tanzania can be associated with the AD 900 tsunami at Nagapattinam, India and 14 tsunami deposits in South Asia. This discovery enhances our understanding of what appears to be a major historical transoceanic tsunami in the Indian Ocean. The source of this tsunami is likely to have been the Sumatra-Andaman subduction zone, which is the main source of the moderate level of tsunami hazard in northwestern Western Australia.


Allport, J.K. and Blong, R., 1995. The Australian Tsunami Database.

Atwater, Brian, Musumi-Rokkaku Satoko. Satake Kenji, Tsuji Yoshinobu, Ueda Kazue, and David K. Yamaguchi (2016). The Orphan Tsunami of 1700: Japanese Clues to a Parent Earthquake in North America. University of Washington Press.

Davies, G., Griffin, J. 2018. The 2018 Australian Probabilistic Tsunami Hazard Assessment. Record 2018/41. Geoscience Australia, Canberra.

Jaiswal, R.K., B.K. Rastogi and T.S. Murty (2008). Tsunamigenic sources in the Indian Ocean. Science of Tsunami Hazards, Vol. 27, No. 2, page 47 (2008)

Malik, J.N., Johnson, F.C., Khan, A. et al. Tsunami records of the last 8000 years in the Andaman Island, India, from mega and large earthquakes: Insights on recurrence interval. Sci Rep 9, 18463 (2019).

Maselli, Vittorio, David Oppo, Andrew. Moore, Aditya Riadi Gusman, Cassy Mtelela, David Iacopini, Marco Taviani, Elinaza Mjema, Ernest Mulaya, Melody Che, Ai Lena Tomioka, Elisante Mshiu and Joseph D. Ortiz (2020). A 1000-yr-old tsunami in the Indian Ocean points to greater risk for East Africa. Geology, v. 48, p. 808–813.

NOAA (2020). NGDC/WDS Global Historical Tsunami Database, 2100 BC to Present.

Rastogi, B.K. and R.K. Jaiswal (2006). A catalog of tsunamis in the Indian Ocean. Science of Tsunami Hazards, Volume 25, Number 3, p. 128-143.

Rubin, C. M. et al. (2017). Highly variable recurrence of tsunamis in the 7,400 years before the 2004 Indian Ocean tsunami. Nat. Commun. 8, 16019,

Schäfer, Andreas and Wenzel Friedemann (2019). Global Megathrust Earthquake Hazard—Maximum Magnitude Assessment Using Multi-Variate Machine Learning. Frontiers in Earth Science 7, 136,

Shanmugam, G. (2012). Process-sedimentological challenges in distinguishing paleo-tsunami deposits. Nat Hazards 635–30.


Accidental Ammonium Nitrate Explosions

Paul Somerville and Ryan Crompton, Risk Frontiers

Insurers are gearing up for what is likely to be one of the most expensive insured cargo and port infrastructure losses ever from the Beirut explosion, on a scale at least as large as the one resulting from the explosions at the Chinese port of Tianjin in 2015 (800 tonnes, 173 deaths). It is expected that Lebanon’s second port of Tripoli, believed to be operating at just 40% capacity on account of COVID-19, will become the country’s main gateway for both emergency supplies and normal trading.

The accidental ammonium nitrate explosion in Beirut serves as a reminder of how frequent and deadly these events are. A timeline and description of events since 2000 is shown in Figure 1. The Wyandra, Northern Territory event of 2014 is shown in Figure 1 and is one of three Australian events, described later, that appear in the Han (2016) catalogue.

Timeline of accidental ammonium nitrate explosions.
Figure 1. Timeline of the largest accidental ammonium nitrate explosions in the world since 2000.  Source: VisualCapitalist.

We analysed Han’s (2016) catalogue which lists 79 events since 1896, 42 of which occurred in the United States, to assess their frequency of occurrence. Since 1900, they have occurred at a uniform rate of about 0.75 per year, with an increase to about 1 per year since 2000. To the extent that Han’s list is incomplete, these rates are underestimated. Han (2016) notes that the ammonium nitrate that exploded in the 1947 Texas City (Galveston) event was coated with wax to prevent caking.  Practices introduced in the 1950s eliminating the use of wax coatings yield ammonium nitrate, used in fertilisers, that contain less than 0.2 percent combustible material. This practice does not appear to have impacted the frequency of events.

We also analysed the Wikipedia catalogue of 36 events, which lists both size (tonnes) and deaths, to assess the relation between size (tonnes) and number of deaths, shown in Figure 2. To first order, Log10 Deaths = 0.85 log10 Tonnes. Four notable events on the left panel of Figure 2, clockwise from top left, and labelled by numbers of deaths, are the 1921 Oppau, Germany event (450 tonnes, 561 deaths), the 1947 Texas City (Galveston, U.S.) event (2906 tonnes, 581 deaths),  the 2020 Beirut event (2750 tonnes, 220 deaths), and the 1947 Brest, France event (3000 tonnes, 29 deaths).

The relationship between size and deaths from ammonium nitrate explosions.
Figure 2. Relation between size (tonnes) and deaths from accidental ammonium nitrate explosions on linear (left) and log (right) scales. Several zero values on the axes of the log plot actually represent zero values: the 2004 North Korean event of 162 tonnes had no reported deaths.

Australian-based company, Orica, the world’s largest provider of commercial explosives and blasting systems to the mining, quarrying, oil and gas and construction markets, has a stockpile of ammonium nitrate up to four times the size of the one in Beirut. There are many stockpiles in Australia but Orica’s Kooragang Island plant has received a lot of media attention.

Between 6,000 to 12,000 tonnes are currently stored at Orica’s Kooragang Island plant in the Port of Newcastle, which produces approximately 400,000 tonnes each year. This plant is located 3 km from Newcastle’s CBD and 800 m from residents in Stockton. Up to 40,000 people live in what would be the ‘blast zone’ if there were to be an explosion. Orica state that they follow strict safety protocols and ensure that the ammonium nitrate storage areas are fire resistant and built exclusively from non-flammable materials, with no flammable sources within designated exclusion zones. The operations on the Kooragang Island site, which has been in operation for 51 years, are highly regulated to state and federal standards. The site’s safety management systems, security arrangements, and emergency response procedures undergo a strict auditing and verification process by SafeWork NSW. The Kooragang Precinct Emergency Sub Plan can be found here.

The safety of ammonium nitrate was previously highlighted in South Australia in 2013, when concerns were raised about the location of the Incitec Pivot fertiliser plant at Port Adelaide following the West, Texas, explosion of 2013 involving 240 tonnes of chemical that killed 15 people in a 50 unit apartment block. In 2013, the South Australian Government made an agreement with Incitec Pivot to move its plant away from the heart of Port Adelaide, because it posed an unacceptable risk to residents of a proposed major development there. The company moved its operations to a location further from the centre of Port Adelaide to Gillman in 2018. According to SafeWork SA, all 170 of the ammonium nitrate storages in the state are heavily regulated, heavily controlled and monitored.

Figure 3. Left: Incitec Pivot plant, Port of Adelaide; Right: Orica Kooragang Island plant.

In the remainder of this briefing we describe three Australian accidental explosions, all involving trucks.

Taroom, Queensland, 30 August 1972

A truck explosion occurred near Stonecroft Station on Fitzroy Development Road in August 1972. The truck and trailer were carrying 12 tonnes of ammonium nitrate. The truck experienced an electrical fault and caught fire north of Taroom. After the driver stopped and parked the burning truck, two brothers from a nearby cattle property who saw the fire rode up on motorbikes to assist. The three men were killed when the truck exploded at around 18:15. The explosion destroyed the prime mover and trailer, leaving a crater in the road 2 m deep, 5 m wide, 20 m long. Parts of the truck and trailer were scattered up to 2 km away. The explosion burnt out more than 800 hectares (2,000 acres) of surrounding bushland. The explosion was heard and shook houses 88 km away in Moura and 55 km away in Theodore.

Wyandra, Queensland, 5 September 2014

On September 5, 2014, an ammonium nitrate truck explosion (Figure 4) occurred near Wyandra, about 75 km south of Charleville in south-west Queensland, Australia. The truck carrying 56 tonnes of ammonium nitrate for making explosives rolled over a bridge and exploded, injuring eight persons including the driver, a police officer, and six firefighters. Rescue crews were trying to extract the driver from the truck when they found out there was ammonium nitrate inside. They were making a mad dash from the truck when it exploded.

The prime mover caught fire about 9.50pm and the driver steered off the highway, causing it to hit a guard rail near the Angellala Creek Bridge and roll onto its side in the dry creek bed. The crash led to two explosions occurring at 10.11pm and 10.12pm. The blast was so powerful that the truck disintegrated, destroying two firefighting vehicles along with it and causing catastrophic damage to the Mitchell Highway. Two road bridges were destroyed (Figure 5), one of the railway bridge spans was thrown 20 m through the air, and a major section of the highway was missing. Geoscience Australia recorded the explosion as a magnitude 2.0 event, and coincidentally, 20 minutes after the explosion, a magnitude 2 earthquake was recorded 55 km south of Charleville.

Emergency vehicles damaged by the Wyandra truck explosion. Queensland Police Service
Figure 4. Emergency vehicles damaged by the Wyandra truck explosion. Queensland Police Service.

The dangers posed by the remaining ammonium nitrate led to a 2km exclusion zone around the site for a number of days. The large crater formed by the blast closed the highway necessitating detours of up to 600 km, including a 100 km detour to Cunnamulla along the Charleville-Bollon Road. In April 2015, the $10 million tender to reconstruct the highway and bridges were awarded and the construction work took place between June and November 2015.

Damage to bridges caused by the Wyandra explosion.
Figure 5. Damage to bridges caused by the Wyandra explosion. Queensland Police Service

Queensland Transport Minister Scott Emerson noted that there are rules in place relating to signage and the particular routes that are allowed to carry dangerous goods and that he would be talking to police about whether anything was done wrongly. However, Assistant Fire Commissioner Dawson dismissed concerns that such a volatile material was being carried in trucks. “Not so much a worry; this product – and trucks like this very same truck – travel these roads every day,” he said. “Every day they’re out there and they don’t go bang. Something’s happened to bring this truck in a situation, which has possibly mixed the product on the back of the truck – maybe with the diesel fuel, the impact of the initial [crash] when it goes off the road – so those circumstances have had more of a connection to the end result. You’d be surprised – there’s a lot of these trucks – they do it very safely and very effectively.

On January 10, 2019, the Queensland State Government launched a lawsuit in the Brisbane Supreme Court claiming more than $7.8 million in damages, the estimated cost of building a temporary detour, and inspected the area to ensure it was safe as well as replacing the road and railway bridge. It was holding the trucking company, Kalari Proprietary Limited, road train driver Anthony David Eden and insurer Dornoch Limited responsible for the repair bill.

Ti Tree, Northern Territory, 18 November 2014

A road train consisting of three flat-bed trailers carrying ammonium nitrate fertiliser exploded in Ti Tree, NT on November 19, 2014 (Figure 6). Witnesses at the Ti Tree roadhouse, 200 km north of Alice Springs, saw a fire igniting on the left-hand side of the rear axle of the rear trailer. The road train driver inhaled fumes as he desperately unhooked the burning trailer of explosive ammonium nitrate from his truck on the Stuart Highway at Ti Tree. Moments later the trailer exploded with a loud bang, startling residents more than several hundred metres. The driver had towed away the two other trailers of ammonium nitrate. No-one was injured.

Police went door-to-door to evacuate residents to the school and establish a 1 km exclusion zone. Sixty to eighty people were evacuated to the school at the northern end of town at 10:30 pm, and were allowed to go home at 1:30 am but there was still a 300 m exclusion zone. At 2:00 am the fire crew declared the fire ‘safe’ and Stuart Highway was reopened.

Figure 6. Ti Tree explosion (left, Nicolai Bangsagaard) near the Ti Tree Roadhouse (right, Olivia Ryder).


From the Vault: What we knew about a future pandemic in 2005

Paul Somerville, Briefings editor

The following briefing is a reproduction of the article entitled “A Future Pandemic” that was published in our Quarterly Newsletter Volume 5 Issue 2, December 2005. It was written by Risk Frontiers’ former employee Jeffrey Fisher and Peter Curson, Emeritus Professor at Macquarie University, and edited by John McAneney. In the light of the current coronavirus pandemic, the article was very insightful and needs no further introduction. Additional Briefing Notes on this theme are numbers 121 and 173 (available on request). From time-to-time we will return to our Insights “vault” to assess how well our understanding of natural hazards and other extremes stands up to the test of actual events.

It is difficult to pick up a paper or watch television today without seeing some reference to bird flu, H5N1, or a possible influenza pandemic and the world’s lack of preparedness for it. One thing seems clear: in an increasingly interconnected world, where 1.5 billion people cross international borders by air every year, a virus could circle the globe very rapidly, possibly even before it was detected. In contrast, the so-called 1918-19 ‘Spanish Influenza’ took some 18 months to circle the globe and about four to six months to do its damage in any one country. This article examines some implications of such an event for the life insurance business and the wider economy.

Some insurance and reinsurance companies have prepared for the eventuality of a pandemic, assessing their risk and taking steps to offset expected losses. Financial instruments such as mortality bonds, the life insurance equivalent of catastrophe bonds, have been used to transfer some of the risk to the capital markets. However, catastrophe modelling, now standard for non-life lines of business, seems far less sophisticated in the case of life insurance. Many companies still appear to be working out what their losses might be.

Some Basic Numbers for Australia

So how many people are likely to die if a flu pandemic reaches Australia? If there were a repeat of the ‘Spanish Influenza’ pandemic, the death toll in Australia could be somewhere between 60,000 and 80,000. To put this in some context, some 130,000 Australians die in any one year, a sum that includes about 2,000 from influenza. Thus, a repeat of the 1918-19 scenario would represent an increase in the annual death toll of over 50%.

Circumstances today, however, are very different from 1918. Medical and community health standards have improved dramatically. In 1918, intensive care wards had yet to be developed; there were no effective drug therapies for pneumonia; knowledge of viruses was rudimentary; and doctors had no antibiotics or antiviral drugs. Taken together, these factors could reduce the death rate considerably below that experienced in 1918-19.

On the other hand, the mobility of people today could allow the disease to spread very rapidly causing a dramatic increase in patient numbers in a short space of time. This has the potential to overwhelm the health system and reduce the benefits of modern medicine because of a shortage of drugs and hospital beds. Thus, while the death rate is unlikely to be as high as for the 1918-19 pandemic, it nonetheless remains a good benchmark as a plausible worst-case scenario.

An Optimisation Problem

The current strain of bird flu has killed over half the people known to have become infected with it. While this is cause for concern, a flu pandemic could not develop with a mortality rate this high. The mortality rate is defined as the proportion of the infected population that dies from the disease.

Influenza viruses have an initial period when an infected person exhibits no symptoms. This is the virus’s window of opportunity to spread; once symptoms appear, it is fairly easy to isolate cases and prevent further transmission. Furthermore, approximately half of all people who catch the virus get only a mild case with no obvious symptoms while still being infectious. These two attributes allow a flu virus to spread throughout a population. An influenza virus that kills its host too quickly will die out before it can cause a global epidemic.

Figure 1: Modelled deaths in a population of 10,000 as a function of mortality rate.

Risk Frontiers has a simple simulation model to examine this issue. A typical simulation deals with a population of 10,000 people. They are assumed to be a fairly homogenous group in the same geographic area – imagine a small Australian town or suburb. One way of incorporating the viral attributes mentioned above is to assume some degree of negative correlation between the length of the infectious period and the mortality rate. Negatively correlating these variables means that, on average, as the mortality rate increases, the length of the infectious period decreases and so less people catch the disease. Figure 1 shows this trade-off. Initially, as the mortality rate increases, more people die. Beyond a rate of around 1.6%, however, the tide turns as the likelihood of someone dying after catching the disease increases but the total number of people infected goes down.

The real concern is that the current strain of bird flu will combine with human flu viruses or develop the ability to jump directly to humans. If this were to happen then the global death toll could be very high. The 1918-19 flu virus killed, depending upon different reports, somewhere between 1.2% and 2.8% of people who contracted it, i.e. close to the optimum shown in Figure 1. A very well-designed bug!


Clearly, targeted vaccination at the source of an outbreak is likely to be the best means of avoiding its wider dissemination. In a recent article, Ferguson et al. (2005) explores the efficacy of using such targeted preventative medicine. These authors argue that if good detection measures are in place, if anti-viral drugs are stockpiled appropriately and deployed quickly, then the chances of containing an outbreak at source by treating everyone in the vicinity would be greater than 90%. While this conclusion is encouraging, neither sufficiently rapid detection nor efficient implementation of preventative measures can be taken for granted in the countries where outbreaks are most likely to occur.

Australia is an unlikely source of the disease and it is far more likely that the general population would have to be vaccinated. Let’s assume for the moment that this is possible. The proportion of the general population that must be vaccinated to stop an epidemic depends on the Basic Infection Rate (BIR), essentially a measure of how easily transmissible the virus is. The BIR is the average number of new cases caused by each virus-infected person in a population with no immunity to that virus.  For the current strain of bird flu, we might prudently assume that no one has immunity. According to Ferguson et al. (2005), a typical pandemic strain of influenza would likely have a BIR of around 1.8, a figure that implies the need to vaccinate roughly one half of the population in order to arrest the spread of the disease (see Inset). In other words, about nine million people in Australia.

All this presupposes the availability of a vaccine. In fact, it would take about six months to isolate a particular strain and produce a vaccine in sufficient numbers.  By this time the pandemic would be over.  So, will there be enough vaccine to go around? Given current global development and production capabilities, the answer is no.

Insurance Costs

What would a modern-day pandemic cost the Australian life insurance industry? If we assume a very rough estimate of $300,000 for a life insurance payout, then it is simply a matter of counting the dead or, at least, those with life insurance.  The current level of life insurance penetration is around 30% of the adult workforce, who in turn comprise about 65% of the entire population. This being the case, a pandemic comparable to the 1918-19 influenza outbreak would lead to a total insured loss of around $4.1 billion. This sum is of the same order as a repeat of Cyclone Tracy that destroyed Darwin in 1974 (see the next issue of Risk Frontiers’ Quarterly Newsletter.)

There are, however, other complications not considered in the above calculation. For reasons that are still not entirely clear, the 1918-19 epidemic preferentially killed people between the ages of 25 and 40, i.e. those normally at the lowest risk of dying from influenza.  Thus, usual actuarial assumptions about expected age at death may not apply in the case of a pandemic.  Moreover, people in this target age group are more likely to have life insurance and will tend to be insured for relatively higher amounts.

There may be other calls on insurance caused by the failure of some businesses to fulfil critical supply contracts due to workers being afraid to, or prevented by Government decree, from turning up to work. Private medical insurance could be another source of losses for the insurance industry.

Social and Economic Consequences

While our analyses suggest that the implications of a 1918-19-type pandemic could be significant for the insurance industry, insured losses will represent only a tiny fraction of the wider economic losses borne by society.

The recent SARS epidemic gives us some clues to the likely magnitude of these losses.  The province of Ontario, for example, suffered an estimated loss of more than C$2 billion due to reductions in tourism, including lost income and jobs. Hotels in Toronto remained two-thirds empty during the peak of the epidemic and cost the hotel industry more than C$125 million. More than 15,000 people were quarantined at home for at least 10 days. If nothing else, SARS demonstrated the impact that a short-lived epidemic can have on consumer confidence, investment and consumer spending. Some sources have estimated the total global economic cost of SARS at $US 30 – 50 billion (Financial Times, 14/11/05).

A major flu pandemic would be much more significant than SARS. Businesses could be confronted by 25-30% absenteeism as home quarantine removed many from the workforce for up to two months; people would avoid shops, restaurants, hotels, places of recreation and public transport. There would be a run on basic foodstuffs, medications, masks and gloves. As there is little surge capacity in our hospitals, temporary hospitals would need to be established. Schools, childcare centres, theatres, not to mention pubs and race meetings – the fundamental heartbeat of our nation – would be closed or cancelled. Government imposed quarantine and absenteeism would severely disrupt interstate and international trade. All this would produce a decline in consumer confidence leading to significant reductions in consumption spending.

Some Other Issues

Let’s return now to the question of preventative medicine. As has already been explained, there is simply not going to be enough anti-viral drugs, vaccines and other preventative measures to go around. The current stockpile of anti-viral drugs could be insufficient even for just all essential health care workers, emergency service workers – and politicians? And what about me?  Yes moi!

Assuming Australia has the luxury of time to become better prepared, then difficult choices still remain.  For example, who will get the extra supply after the needs of essential workers are met? Would they be handed out by lottery, should they go to the elderly and young, would people be able to buy them?  Public outcry might prevent a scheme where they were sold to the highest bidder, but it is easy to imagine somebody risking the small chance of personal death and selling their vaccine shots on e-bay for large sums of money.  The problem could be an administrative and ethical nightmare.

Final Thoughts

So, where does all this leave us? As far as preventing an outbreak, the only place that this can be done is in the place of origin, most probably somewhere in Asia. If a pandemic does occur, then it is going to inevitably affect Australia. Quarantine measures that the government will feel obliged to put in place might delay its development but are unlikely to prevent it from reaching us. Given a lead-time of six months to develop an effective vaccine, society and the government will be faced with some difficult choices about who gets access to limited supplies of anti-viral drugs. And for the life insurance industry, our admittedly rough calculations suggest that it is not good news. However only a minor proportion of the economic costs will be borne by the insurance sector.  And underlying all this is a fundamental truth – a healthy population represents the human capital necessary for productivity, innovation and economic growth.

Calculating the proportion of people to vaccinate

The relationship between the Basic Infection Rate (BIR) and the proportion of people who need to be vaccinated to contain or prevent an epidemic is a relatively simple one. In order for the virus to propagate through a population, an infected person must infect at least one other person.  Thus for a vaccine program to be effective, it must lower the effective BIR of the virus to below 1.0. Assuming no immunity within population, the proportion that needs vaccination is given by the formula:

Proportion = (BIR – 1.0)/(BIR)

Given a BIR > 1.0, then vaccinating this proportion of the population will stop an epidemic from gaining hold, although small outbreaks are still possible. With a typical value for a pandemic-type strain of 1.8 (Fergusan et al., 2005), the formula suggests 44% of the population will need to be vaccinated.

If a virus is currently in circulation, then people with it already or having low level infections can be assumed to be immune and not require vaccination. This will reduce the quantity of vaccine needed.

If the virus is sufficiently widespread, however, it will still take a long time to die out and so vaccinating as large a proportion of the population as is feasible is the best defence. Moreover, we will not know the actual BIR for some time and so once again assuming a 1918-19-like worst-case scenario may be the only prudent policy.


Ferguson, Cummings, Couchemez, Fraser, Riley, Meeyai, Lamsirithaworn and Burke, Strategies for containing an influenza pandemic in Southeast Asia, Nature, Volume 437, 2005, pp 209 – 213.

Harris, Melling, Borsay, ed. The Spanish Influenza Pandemic of 1918 – 1919: New Perspectives. Routledge, 2003.

Crosby, America’s Forgotten Pandemic: The Influenza of 1918. Cambridge University Press, 1989.



Heatwave poses challenge to Japanese medical system already stressed by virus

Paul Somerville and Andrew Gissing, Risk Frontiers

In recent years, eastern Australia, like Japan, has experienced extremely high maximum temperatures that are consistent with patterns of global changes in climate. Fortunately, last summer’s heatwaves in Australia occurred before the prevalence of COVID-19, and if Australia is able to maintain its suppression of the virus, it may be able to avoid the compounding effects of those conditions. This briefing demonstrates that even with the low prevalence of the virus in Japan, these compounding effects can be significant.

The number of people showing signs of heatstroke or heat exhaustion has sharply increased recently. Temperatures soared to 41.1 C in Hamamatsu in central Japan on Monday (Mainichi Shimbun, 2020a), tying with the country’s highest-ever temperature, marked in Kumagaya near Tokyo in 2018.

The 2018 Heatwave

During the 2018 heatwave, Mainichi Shimbun (2018) showed that the 94 people who died included 26 fatalities in Tokyo, where the heat reached 40.8 degrees in the suburban city of Ome. Saitama Prefecture also reported nine deaths, while in the western part of the country, Osaka Prefecture had six, Mie and Hyogo five each, and Hiroshima saw four. Aichi Prefecture in central Japan also announced four deaths. (According to Slate (2020), more than a thousand people died from heat-related illnesses over the course of those few weeks).

When broken down by the gender of the victims, there were 52 women and 42 men (Mainichi Shimbun, 2018). All of them were 40 years old or older. Those in their 80s constituted the largest group with 37 deaths, followed by 22 in their 70s, 15 in their 60s, 10 in their 90s, five in their 50s and four in their 40s.

Among the victims, 28 fell ill while they were outside, and many were farming in their fields. As many as 36 people were found ill or unconscious while they were inside, due in several cases to broken air conditioners or electric fans. In some cities such as Yamato, elderly residents who live alone are monitored day and night by an elaborate system of motion sensors and communication protocols between city officials, residents and their relatives.

Older people tend to have difficulty recognizing when they are dehydrated. They face the risk of their conditions deteriorating before realizing it, even when they are not subject to searing heat. Lowering temperatures inside using air conditioning is important, but not all homes have air conditioners.

2020 Heatwave – Distinguishing heatwave symptoms from corona virus symptoms

On August 19, 2020, officials in Tokyo reported that 28 people died of heatstroke in the city during the eight-day period from August 12 to August 19, bringing the total number of fatalities in Tokyo in August to 131 (NHK, 2020). The Medical Examiner’s Office said that 11 of the 28 victims were in their 70s, ten were in their 80s, and about 80 percent of the victims were at least 70 years old. Eleven of the victims died at night and 27 died indoors, of whom 25 were not using air conditioners.

In the midst of this year’s heatwave, it is reported that medical workers worry that the similarity of heat stress symptoms to COVID-19 may place extra pressure on a health care system already creaking under the strain of the coronavirus pandemic (Mainichi Shimbun, 2020a).  There are times when medical personnel cannot immediately distinguish those suffering from heat-related illness from those with COVID-19 when the patient is feeling unwell with high fever because that is a symptom they have in common. Japan has a relatively small number of COVID-19 cases (Figure 1), with only 1,169 deaths so far.  The Japanese Health Ministry reported no evidence of excess deaths during April and May (the latest months for which data are available), and it is likely that undetected COVID-19 cases are contributing significantly to the numbers of heatwave deaths that are being reported.

Figure 1. COVID-19 cases in Japan.  Cases: 62,507; Deaths: 1,181; Recovered: 49.340.  Source: Worldometers (2020), 25 August 2020.

The problem posed by the pandemic is that treatment has to take account of the possibilities of both COVID-19 and heat-related conditions when staff cannot rule out the possibility of coronavirus infection. Amid reported public fears that mask-wearing to prevent the spread of the novel coronavirus could itself cause heatstroke or heat exhaustion, 12,804 people were taken to hospital across Japan between Aug. 10 and Aug. 16 for heat-related conditions, up from 6,664 people the previous week, according to the Fire and Disaster Management Agency.  There is a concern that this large number of patients being taken to the hospital may cause the hospital system to collapse if the heatwave continues.

Recent heatwave conditions in the United States have also seen authorities needing to adapt plans to account for the risks of COVID-19, with fears that people may be reluctant to leave their homes to seek cooler shelter due to infection risks. Adaptions have included restricting the number of people accommodated within cooling centres to allow social distancing.

Some resources complied by the Global Heat Health Information Network on COVID-19 and heatwaves are available here:

Public Information on Heat Stress

The Ministry of the Environment is providing English-language information about the heat stress index on its website in a bid to prevent illnesses caused by intense heat, which has become a major threat to health and even life in Japan in recent summers (Mainichi Shimbun, 2020c).

The website, designed for viewing by both smartphones and personal computers, indicates the intensity of the heat effect throughout the country in five colors, from blue (almost safe) to red (danger). It also provides two-day predictions for the heat stress index, as well as data for each observation point nationwide.

The heat stress index, also called the Wet Bulb Globe Temperature (WBGT), is one of the empirical indices showing the heat stress an individual is exposed to. It is calculated incorporating factors such as humidity, sunlight and reflection intensities and atmospheric temperature.

According to the ministry website, the number of people suffering from heatstroke shoots up rapidly when the WBGT, which is denoted in degrees but is different from normal air temperature, exceeds the upper threshold of the “Warning” level (25-28 degrees), when the air temperature is between 28 and 31 degrees Celsius.

For the warning level indicated in yellow, people are advised to rest often. When the index is at the “Severe Warning” level of orange, people are advised to refrain from heavy exercise. At the “Danger” level shown in red, people should stop all exercise.

Figure 2. Screen capture showing the Ministry of Environment website providing heat stress index information. Mainichi Shimbun (2020d).



Mainichi Shimbun:





NHK (2020):

Slate (2020):

Worldometers (2020):


California Bushfires, August 2020

Paul Somerville, Chief Geoscientist, Risk Frontiers

Nearly 771,000 acres of largely unpopulated land have burned across California during the past week as dozens of lightning-sparked wildfires moved quickly through dry vegetation and threatened the edges of cities and towns. The fires have been most severe in the state’s northern and central regions, where about 600,000 acres have burned in the past week (Figure 1).

Evacuations surged on August 18 and 19 as authorities worried that high heat and gusty winds could cause the fires to spread rapidly. The resulting fires – and complexes of many small fires – have merged into major conflagrations in many parts of the state. By August 20, several of the major fires had more than doubled in size, in some cases jumping across major highways, as crews struggled to contain the blazes. By August 21, the two largest blazes, the SCU[1] and LNU[2] Lightning Complexes, had charred 340,000 and 325,000 acres respectively, becoming the second and third largest fires in California history (Table 1). The CZU[3] Lightning Fire forced the evacuation of more than 64,000 people, some of whom may not be able to return to their homes for weeks. Five people have died and about 1,000 structures have burned.

Figure 1. Left: Fire locations in California using Active Fire Data (hotspots) derived from the VIIRS for the last 7 days. Right: Satellite Image on August 19. Source: Washington Post.
Figure 2. Left: Fire in Napa, California. Right: Fire in Lassen County, California. Source: Washington Post.

The California wildfires, along with other blazes in the West, have sent a blanket of smoke across at st 10 states and southwestern Canada, with smoke extending over the Pacific Ocean as well (Figure 1, right panel). Air quality alerts are in effect for parts of California, where the tiny particles in the dense smoke are aggravating respiratory conditions and worsening preexisting health conditions that are already threatened by the coronavirus. The cloth masks that have now become a habit for many Californians when they venture outside are largely ineffective against the tiny smoke particles filling the air, and doctors recommend using N95 masks with vents. People are being asked to shelter in place, staying at home with their windows closed and ventilation systems set to recirculate air, which is difficult during a heatwave in areas such as San Francisco where many people do not have air conditioning.

A rare mix of ingredients came together in central and northern California to produce fast-moving, explosively growing wildfires that are powerful enough to create their own weather. Doppler radar revealed at least five tornado-strength rotational signatures inside the smoke plume in Lassen County, California. The record heat reached astonishing levels during the past two weeks as a massive “heat dome” parked itself over the West. On August 16, Death Valley, California, reached 130 degrees Fahrenheit (54 degrees Celsius). The combination of an intense, long-lasting heatwave, dry vegetation at the end of the summer, and a rare outbreak of August thunderstorms led to these blazes. Fueled by the heat, thunderstorms broke out on Sunday Aug 15 as a surge of tropical moisture pushed inland. The storms’ 20,000 lightning strikes (Fig. 3), including dry lightning storms, sparked more than two dozen blazes over a period of 3 days. An ancient stand of the world’s tallest trees has fallen victim to California’s raging wildfires. The CZU and SCU complex fires near Santa Cruz have ravaged Big Basin State Park, California’s oldest state park, some of whose giant redwoods are more than 50 feet around and 1,000 to 1,800 years old (Fig. 4).

Figure 3. Lightning storms in San Francisco and Healdsburg. Source: Washington Post.
Figure 4. Giant redwoods in Big Basin State Park. Source: Washington Post.

This is just the beginning of the state’s wildfire season, something that has been a constant threat during the past four years of blazes, some sparked by downed powerlines, that have set records for size and lethality. Despite the familiarity, the current fires and their speed and thick smoke have presented a new terror amid a global pandemic – poor air quality and concerns about evacuating masses of people to crowded shelters, and that some might not heed the warnings. Tens of thousands of people have been asked to evacuate and make difficult decisions about where to go. In the past, they might have stayed with friends or family, but now they need to calculate the risk of exposure to the novel coronavirus. Wherever people go, they are likely to face other hardships. California has been enduring a record-breaking heatwave that has prompted rolling blackouts because of high electricity demands for air conditioning and other uses. Most of the area is also experiencing severe or moderate drought.

In Santa Cruz and San Mateo Counties, south of San Francisco, about 48,000 people were ordered to evacuate because of a fire, part of the CZU Lightning Complex, that is threatening communities there. The blaze has already burned 50 structures. On the evening of August 20, the University of California at Santa Cruz was under mandatory evacuation and had declared a state of emergency.

The largest of the lightning-related fires was north of San Francisco, covering Napa and Sonoma counties. On August 20, that mass of fires, the LNU Lightning Complex, had grown to 219,000 acres and was uncontained. Approximately 30,000 structures were at risk of burning and 480 had been destroyed.

The blaze near Vacaville, known as the Hennessey Fire and part of the LNU Lightning Complex, has been one of the most destructive, burning down homes and claiming the life of a PG&E worker who was assisting first responders. This blaze burned down the La Borgata Winery and Distillery in Vacaville. Mandatory evacuations remained in effect for the north part of the city on August 20, and CalFire reported three additional civilian fatalities associated with the LNU Lightning Complex.

CalFire is at normal staffing levels, with approximately 12,000 firefighters working on August 21. Additional firefighters are being sought from other states and from Australia. In Central California, a pilot on a firefighting flight near Fresno died when his helicopter crashed.

Overall losses include 5 deaths, 64,000 people evacuated, over one thousand structures burned, 31,000 structures threatened, and approximately one million acres burned as of August 21.

The 2019/20 bushfires in eastern Australia were fought under dire conditions, but the presence of the coronavirus in California has made fire-fighting conditions there even more dire, especially those relating to evacuation.  There are 665,000 coronavirus cases in the state, growing by 5,000 a day, and 12,000 deaths, growing by 150 a day.

Table 1. 20 largest wildfires in California since 1932. Only 3 occurred before 2000. (Source: Updated from Cal Fire)


[1] SCU Lightning Complex Fire: Contra Costa, Alameda, Santa Clara, Stanislaus and San Joaquin counties

[2] LNU Lightning Complex Fire: Napa, Sonoma, Solano, Yolo and Lake counties

[3] CZU Lightning Complex Fire: San Mateo and Santa Cruz counties

Risk based earthquake pricing using catastrophe model output

Paul Somerville and Valentina Koschatzky,  Risk Frontiers

As the insurance market trends toward more analytical and data-driven decisions, insurers are continually exploring ways to rate risk better and more precisely. For the case of earthquake risk, this means an enhanced understanding of the relationship between event location, frequency, severity, how buildings respond to an event and the ensuing financial costs. The increased quantity, quality and granularity (resolution) of the available underwriting data and highly refined rating engines give insurers the opportunity to become extremely risk-specific in their pricing. Risk-based pricing – charging different rates depending on different risk characteristics of specific policies and in contrast to portfolio underwriting – leads to stability and confidence in pricing.  Risk based pricing aims to ensure that premium levels are commensurate with individual property risk profiles, with those in highly exposed areas experiencing a specific rate on the earthquake component of their coverage.  This seems to be a fairer and more equitable way of pricing risk. The ability to differentiate between perceived risk and actual risk affords insurers a better way to achieve their financial goals, allocate capital and meet client needs for coverage.

Several features of earthquake hazards and risks render them readily amenable to risk-based pricing.  First, the level of seismic hazard is not uniformly distributed across a country. New Zealand is an extreme example in which Wellington is located directly on a tectonic plate boundary having extremely high seismic hazard, whereas Auckland is remote from the plate boundary and has a seismic hazard level comparable to that of Australia (Figure 1).  However, even in Australia, the seismic hazard level also varies by an order on magnitude between relatively high levels in northwestern Western Australia, the Yilgarn region east of Perth, Adelaide, and southeastern Australia on the one hand and the extremely low levels in Queensland.

Figure 1. Peak acceleration maps for 1:500 AEP on Risk Frontiers’ Variable Resolution Grid for Australia and New Zealand.
Figure 1. Peak acceleration maps for 1:500 AEP on Risk Frontiers’ Variable Resolution Grid for Australia and New Zealand.

Second, the factors that increase the level of the hazard are well understood and mapped.  These include the presence of soils that amplify the level of ground shaking compared with that on rock, and the presence of saturated sands that can be liquefied during earthquake shaking, as occurred in Christchurch during the 2010-2011 Canterbury earthquake sequence.

Third, we are able to quantify the variations in building vulnerability to earthquake damage due to different building types, heights, ages of construction, and whether seismic building code provisions were used in design on a very specific basis.  G-NAF (Geocoded National Address File) is a geocoded address index listing all valid physical addresses in Australia. NEXIS (National Exposure Information System) is a database developed by Geoscience Australia containing building details for residential commercial and industrial buildings in Australia at a Statistical Area 1 (SA1) level.  There are 57,523 SA1 in Australia. These datasets allows wood, Mid-rise Steel, Concrete, and Reinforced Masonry and low-rise Unreinforced Masonry buildings damage ratios to be modelled and enable customised underwriting in Australia at the location, SA1 or postcode level. For New Zealand, the use of a variable resolution grid created using the Linz NZ street addresses database enables us to calculate the ground shaking hazard at a resolution as fine as 500 m while the liquefaction hazard is  calculated at the address level with a resolution of 16 m.

Finally, Risk Frontiers’ QuakeAUS and QuakeNZ models use a level of refinement in property damage estimation that is unique in the worldwide catastrophe loss modelling industry. Conventional earthquake loss estimation uses building fragility functions that are pre-computed using standard capacity curves for each building category of interest with a simplified representation of the building demand curve in response to ground-shaking. We instead account for the entire response spectral shape of the ground motion, which varies with many factors, including the earthquake magnitude, earthquake distance, and soil category at the risk location. Accordingly, our loss model dynamically calculates fragility curves for each building category at each site for each earthquake in the event set.  This produces building- and event-specific damages for each building category for each event, enhancing the accuracy and reliability of the loss calculation.

These four categories of information are combined to make detailed estimates of losses for each building that are then aggregated to obtain portfolio loss estimates.  However, it is an easy step to use this detailed information to quantify potential losses for any soil type or building category at any desired level of spatial resolution. For example, our model output can provide postcode level risk premiums (average annual loss AAL’s) for all of Australia and New Zealand for a nominal risk to estimate loss rate due to earthquakes for the following building modifiers, as shown in the example in Table 1:

  • Structure Type: Unknown, Wood/Light Frame low-rise, Steel Moment Frame mid-rise, Concrete Moment Frame mid-rise, Reinforced Masonry Bearing Walls mid-rise, unreinforced masonry low-rise. Mid-rise is defined as 4+floors, low rise as 1-2 floors
  • Year-built: pre-code (1980) and post-code
  • Damage calculated separately for buildings and contents
  • Separate estimates of direct damage and demand surge
Risk Premium
Postcode Structure Type Construction Date Building Contents
2294 Light Wood Unknown 74 64
2294 Mid-rise Steel Moment Frame after 1981 26 3
2294 Mid-rise Concrete Moment Frame before 1981 44 7
2294 Mid-rise reinforced Masonry Bearing Walls Unknown 35 25
2294 Low-rise Unreinforced masonry bearing Walls after 1981 226 58
2294 Unknown before 1981 111 71
2294 Low-rise Unreinforced masonry bearing Walls after 1981 141 58
2294 Low-rise Unreinforced masonry bearing Walls unknown 142 59
2286 Low-rise Unreinforced masonry bearing Walls before 1981 51 10
2295 Low-rise Unreinforced masonry bearing Walls before 1981 168 42
2291 Low-rise Unreinforced masonry bearing Walls before 1981 90 21

Table 1. Newcastle region risk premiums for building and contents with a nominal sum-insured. Earthquake risk based on location (postcode), construction type and year of construction can inform better underwriting decisions.

This bottom-up understanding of risk and pricing will also lead to better alignment of risk-premium and capital management.

Figure 2. By introducing earthquake risk pricing, insurers have an opportunity to align portfolio risk (left) and capital management with original premium rating and risk selection (right).
Figure 2. By introducing earthquake risk pricing, insurers have an opportunity to align portfolio risk (left) and capital management with original premium rating and risk selection (right).