disaster deaths

disaster diplomacy







Disaster Deaths

Disaster Deaths

This page originated in:
Kelman, I. and S.N. Jonkman. 2007. "Disaster Deaths Research Challenges". Presentation at the Hazards and Disasters Researchers Meeting, Boulder, Colorado, USA, 11-12 July 2007, full text (223 kb in PDF).

Plaque near Eastbourne, New Zealand commemorating the Wahine disaster.

Plaque near Eastbourne, New Zealand commemorating the Wahine disaster. A hazard of wind, storm, waves, rocks, a weak ship, or poor human decisions and reactions?
(Copyright Ilan Kelman 2004.)


"If a man drowns in flood waters, or is killed by a building collapse during an earthquake the cause of death is reasonably unambiguous, but when a dying villager in Bihar is found to have a severe case of cholera, to have suffered chronic malnutrition for years, and not to have eaten for two weeks since no food was available due to the drought and the inadequate distribution of relief supplies, are we to say that he will die of cholera, malnutrition, drought, or administrative inefficiency? Some such combination of 'causes' is true in all cases." From pages 8-9 in: Hewitt, K. and I. Burton. 1971. The Hazardousness of a Place: A Regional Ecology of Damaging Events. University of Toronto Press, Toronto, Ontario, Canada.

Yet for the flood death, did administrative inefficiency lead to lack of warning and lack of rescue? Did the person stay behind to protect possessions because insurance was not available or not affordable? Had malnutrition weakened the person, making it harder to evacuate? For the earthquake death, did administrative inefficiency lead to lack of monitoring and enforcement of building codes? Did bidding for building contracts force the construction company to cut their bid price and short-change the structure on materials? Did the person have a disability without their specific needs being accounted for?

All disaster deaths are ambiguous.

Disaster mortality studies tend to focus on people who died immediately from medical causes which can be attributed to a specific hazard. Many deaths occur long after the hazard has passed. The cause might be physical, for example someone in a coma, or mental, due to stress, lack of post-disaster support, or lack of resources (not just financial resources) to deal with the difficulties. Because establishing cause and effect is not always straightforward, these deaths frequently remain unnoticed and become part of a disaster's hidden or unadmitted cost. Often, these deaths are inappropriately described as "indirect", "secondary", or "peripheral". Any death from a disaster is part of the toll even if it is difficult to include in the numbers.

Therefore, any study of disaster deaths has the inherent limitation of being potentially incomplete without a method for proving completeness. Nonetheless, to develop appropriate policies and practice for reducing disaster-related fatalities, it is important to understand who is dying in disasters and how and why they are dying. This topic has a rich history and foundation, as indicated by the list of references on Medical Emergencies and Mass Casualties (284 kb in Word) which was compiled by David Alexander as of 10 August 2003. This material, publications since then, and the understanding gained from this disaster-related work need to be linked to wider topics, notably the non-fatal consequences of disasters along with non-disaster mortality and morbidity.

Memorial to the 1954 Hurricane Hazel deaths in Toronto, Canada.

Memorial to the 1954 Hurricane Hazel deaths in Toronto, Canada.
(Copyright Ilan Kelman 2016.)

Considerations in compiling and calculating disaster deaths

A. Definitions

1. Disaster

Books have been written about trying to define "disaster" with no resolution. With regards to disaster deaths, the fundamental practical question is which fatal events might not be considered disasters? An example is fifteen solo snowmobilers dying in fifteen separate avalanches compared to three neighbouring families of five people each dying in the same avalanche. Are they all avalanche disaster fatalities or do we make judgements regarding which are disaster deaths and which are not? Some examples of Disaster Definitions.

2. Event

Consistently defining the spatial and temporal scope of disasters is difficult, with the ultimate conclusion that disasters are not events but are processes. Hazards are sometimes events, but delineating start and stop times for linked hazards can be difficult. For example, for Mount Pinatubo in the Philippines in 1991, the fatal ashfalls had relatively clear start and stop times, but lahars continued killing more than a decade after. Is each ashfall and each lahar a separate hazard event or are they all part of the 1991 Mount Pinatubo eruption as a hazard event or long-term volcanic hazard process? Similarly, delineating geographical boundaries for a hazard's spatial extent can be inconsistent. When a single tropical cyclone (a hazard event) hits multiple countries, some databases list it as multiple disasters, one per country.

3. Hazard categories

Hazard categories frequently overlap and are not always consistently defined. Tropical cyclones in Bangladesh are sometimes listed as wind storms, even though most studies show that most of the deaths are water-related, from drowning in storm surge or freshwater flooding. That is, the storm might be considered to be more wind-related but the disaster deaths tend to be more flood-related (and different types of floods). Similarly, deaths in earthquake-induced landslides deaths have been considered either earthquake deaths or landslide deaths depending on the author. Tsunamis have many origins, yet tsunami-related deaths are often pooled as a single category of tsunami deaths. Others might be ambiguous regarding whether or not they always belong as disaster deaths such as:

  • Macrobiological hazards.

  • Unusual Floods and Drownings.

  • Oesterhelweg, L. and K. Püschel. 2008. "'Death may come on like a stroke of lightening...': Phenomenological and morphological aspects of fatalities caused by manure gas". International Journal of Legal Medicine, vol. 122, pp. 101-107.

4. Causality

Why do disaster deaths happen? Definitionally, it is not because of the hazards--events or processes--but because of vulnerabilities. Vulnerability is the long-term process of social values, behaviours, actions, values, and decisions, which mean that some people cannot deal with potential hazards, so they become harmed and disasters occur. While the physical forces and energies of hazard parameters lead to medical problems which can induce death, the fundamental cause is that the people could not avoid those hazard parameters. This cause of death is vulnerability, not the hazard.

B. Data

5. Skewed statistics

Hazards and events, small and large, can skew statistics in three ways. First, a single large event could radically alter long-term trends. The literature does not yet report any verifiable human deaths from a meteorite strike in recorded history, but a single large event could dwarf the total death toll from all disasters over the past millennium. Second, underreported small events have less influence on overall statistics than they should have, even though their cumulative impact might be far greater than the events garnering attention, which is termed "invisible disasters" problem. Third, hazard and vulnerability baselines are always changing, so establishing trends is not as simple as tracking numbers over time.

6. Counting

Counting fatalities can be inconsistent. After Hurricane Katrina hit the USA in 2005, deaths amongst evacuees in Harris County, Texas were compiled with a pregnant woman listed as two separate fatalities. Disaster death compilations vary about whether or not they include homicides, suicides, and vehicle crashes within the tally of disaster-related deaths. A similar discrepancy arises from vehicle crashes related to wildfire smoke or evacuees falling asleep while driving.

7. Prevented deaths

Deaths can be prevented due to a hazard, such as fewer vehicle crashes if people do not drive in a blizzard or a storm. Should disaster deaths researchers calculate background rates of all "normal" deaths and add or subtract any differences following a disaster? Or should the focus be identifying only who is clearly killed in a disaster rather than considering overall rates? As well, some studies have noted that in the months and years following a disaster, the background rate of deaths decreases, because the disaster killed the most vulnerable members of the population who would have soon succumbed to "normal" death causes without the disaster. This observation is termed "the harvesting effect" and is particularly notable for heat waves and cold waves.

8. Non-immediate deaths

Non-immediate deaths from disasters can occur months or years after a hazard. A person might have long-term physical injuries or mental health impacts, eventually leading to death. This point connects to the question regarding when a disaster or hazard stops.

C. People's behaviour

9. Risk judgements

Judging and misjudging risks, including possible consequences, occurs prior to and during hazards, often influencing whether or not an individual is killed or survives. Judgements might or might not be influenced by different levels of information, from formal education over years to warnings being issued now. Understanding how and why these risk judgements occur is not straightforward, especially if the individual is killed.

10. Risk-related actions

Once a judgment is made regarding risks, action (or inaction) will be taken. The form of risk-taking or risk-avoiding (in)actions influences fatalities. Actions might or might not be influenced by different levels of information, from formal education over years to warnings being issued now. Different people have different options available, such as the difference between climbing a volcano for photography or gas samples compared to poverty forcing people to live in informal settlements on a volcano's slope. Specific (in)actions do not necessarily follow specific risk judgements and determining why (in)actions were taken is not easy, especially if the individual is killed.

11. Rescue and response

Search, rescue, and medical efforts--or lack thereof--frequently make the difference between survival and fatality. Where resources are put into emergency response and medical treatment (as is appropriate) while vulnerability reduction is neglected (which is not appropriate), disaster death tallies might not reflect the chronic conditions in which people live, always leaving them at risk. Prevention being better than cure does not mean to diminish the role of cures, but successful cure can lead to views that prevention is less important.

Grouping D: Data analysis

12. Proportional and absolute numbers

Both numbers of deaths and rates of deaths are important to calculate. Proportionality of fatalities within a population has different manners of calculation, such as death rates within the entire population and death rates within subgroups, whether defined through demographic characteristics or other categorisations.

13. Each factor's importance

The relative importance of factors listed here can vary, especially the sensitivity of results to choices made during the analysis. Conducting multiple analyses simultaneously would help to check the results and to determine their sensitivities to choices made.

14. Analysis scale

Comparing disaster death analyses at different spatial and temporal scales could provide insights into what the data can and cannot explain.

A sign on a bridge near Sendai, Japan indicating the limit of the past tsunami inundation zone.

How useful is the past as a guide to the future? (near Sendai, Japan).
(Copyright Ilan Kelman 2014.)


Strong connections occur among the points, but no ranking of importance is implied in the order in which they are listed. Two main conclusions are that for disaster deaths research:

1. Basic definitional and methodological choices influence the results.
2. Consistency in disaster deaths compilations is not always evident or easy.

Some disaster deaths data might not be collectable, although not all policies to reduce disaster deaths might need complete data or detailed science to support their implementation. Six main impediments to disaster deaths research explain the inconsistencies and the difficulties inherent in resolving the concerns, because some disaster deaths data might not be collectable or comparable:
1. Collecting detailed fatality data is not always a post-event priority.
2. Formal death records with all the information requested might not always be available.
3. Treating bodies and the bereaved with proper respect is important, and, in some situations, might preclude collecting desired data.
4. Disaster deaths data can be influenced by political agendas that inflate numbers to attract help or that reduce numbers to avoid outside attention and intervention or to minimize compensation.
5. Determining the decision making process of each individual fatality--for example, understanding how a warning was received (or not received) and acted upon (or not acted upon) due to their vulnerabilities--is challenging. Techniques for doing so exist, but are not always transferable across different event types or circumstances.
6. For establishing long-term trends, much data has been irreversibly lost.

Additionally, policies that are known--or just assumed--to be effective are often difficult to prove through disaster deaths data. For example, a significant proportion of flash flood deaths in the USA are said to occur in vehicles and the "Turn Around, Don't Drown" campaign is based on that premise. Anecdotally, this campaign saves hundreds of lives each year, or more. But data on vulnerabilities, decision-making process, blood alcohol content, and vehicle type rarely appears in studies. Yet we know for certain that alcohol impairs judgment and reaction time, so do we really need to calculate the percentage of vehicle-based flash flood drownings who were drunk? Identifying drunk drivers who drown might also add to the grief of the bereaved. Similarly, discussions continue regarding the safety of cars compared to mobile homes in a tornado. Finally, after the research-related deaths of a dozen volcanologists in the early 1990's, Codes of Conduct were developed for volcano research. Given the small sample size of volcanologist deaths, it would be challenging to prove that the codes save lives. Yet is such proof relevant to implementing the codes of conduct? These examples suggest that certain aspects of disaster deaths might represent cases where policies and practices can be developed and implemented without solid scientific research or detailed data as a basis.

Overall, the severe limitations of disaster deaths data and analyses, as many authors correctly do, should always be acknowledged and accepted. That does not mean stopping the work, either the scientific publications or the policy influence. More cross-hazard work would be most important, such as by pooling data, rather than being isolated with one's preferred hazard, in order to continue exploring vulnerabilities as the causes of disaster deaths. This approach includes applying the papers that propose disaster deaths frameworks and seeing if common data collection methods and categories might be helpful across hazards. Disaster deaths research should move away from the tendency to focus on hazard parameters and to compartmentalize research by hazards, instead understanding better how to resolve vulnerability characteristics, irrespective of the hazard.

Mt. Vesuvius.

Mt. Vesuvius in southern Italy has been the subject of several casualty studies, related to past events and future possible eruptions. (Copyright Ilan Kelman 2003.)

Case study: Astronomical hazards

Causes and circumstances of deaths from astronomical phenomena are not well-studied, providing an interesting research area into deaths and potential deaths from NEO (Near-Earth Objects such as comets and asteroids) impacts along with geomagnetic storms and other forms of space weather. Numerous websites cite cases of meteorite-related casualties, but none could be verified until 15 February 2013. In the morning local time, as caught by numerous videos and photos, a meteorite burned through the atmosphere over the Chelyabinsk region, Russia. The shockwave shattered glass in thousands of buildings, injured over 1,200 people, and collapsed a factory roof. It appears as if no large fragments fell in populated areas and that no one was hit by any meteorite fragments.

The casualties, it seems, were all injuries with only a handful being serious, so no fatalities, similar to other cases. The negative, though, is hard to prove: how could we be certain that no one has died? The best feasible statement might be that no death in human history has yet been verifiably attributed to space objects.

NASA around 2000 stated on one of their websites that "It is a fact that there is no record in modern times of any person being killed by a meteorite". Papers give other potential examples, none of which are confirmed. Pickering (1919) quotes the Bible for one example of deaths and also suggests one story of a man killed by a meteorite in India in 1827. Yamamoto (1928) describes a report of a 3-year-old girl in Japan being hit by a meteorite and suffering minor injuries. Yau et al. (1994) and International Comet Quarterly list several examples throughout history where meteorites are stated as having killed or injured people. On 9 February 2016, BBC reported a potential meteorite death.

With reference to the 30 June 1908 Tunguska explosion in Siberia, Cohen (2008) writes "Many Evenki, a tent-dwelling nomadic people indigenous to the area, told of animals, their homes and even fellow tribespeople being hurled into the air by a shock wave. An unfortunate few were incinerated". It is unclear what evidence exists to support that contention or if it is part of the local folklore. As well, an ongoing debate surrounds the date of the Kaali meteorite strike on Saaremaa, Estonia within the last several thousand years and the extent of human inhabitation at the time. If the island was inhabited during the strike, fatalities would have been likely, but almost impossible to prove or disprove.

Examples of Kaali meteorite strike references (210 kb in Word).

The Kaali meteorite crater on the island of Saaremaa, Estonia.

The Kaali meteorite crater on the island of Saaremaa, Estonia, created within the last several thousand years. Did the impact kill anyone?
(Copyright Ilan Kelman 2008.)

Chapman (2004; see also Chapman and Morrison, 1994) estimates the annual number of deaths internationally from NEO impacts as averaging between 300 and 3,000, based on the annual probability of different sized impacts, the time available to respond to specific threats, and the predicted consequences of events. These results are naturally dominated by low-probability high-consequence events meaning that year-to-year deaths are usually zero. Plenty of literature (e.g Kuypers et al., 1999 and Xie et al., 2005) exists on mass extinctions due to astronomical phenomena before human beings existed.

The threat to Earth from NEOs has led to monitoring and warning programmes. Several search programmes exist in Europe, the USA, and Japan (Fulchignoni and Barucci, 2005; Thuillot et al., 2005). Plenty of international discussion has occurred regarding how to avert or prepare for a collision once a threat has been identified, but more operational planning and testing of countermeasures is needed (Carusi et al., 2005).

Tsunamis could be caused by asteroid or comet impacts. Kharif and Pelinovsky (2005) report that the most recent known ocean impact occurred approximately 2.15 million years ago, although traces from more recent, smaller events might have vanished. It is possible that no human fatalities have yet resulted from a space-impact-related tsunami.

An individual's characteristics leading to fatalities from astronomical phenomena can include belief systems. Many cultures have interpreted comets and bright meteors as portending calamity (Brown, 1973), likely leading to suicides. In 1997, one sect saw Comet Hale-Bopp's arrival as an opportunity to board a spaceship to the Promised Planet and 39 people committed suicide in a group (Mancinelli et al., 2002). As well, fatalities from heart attacks might have occurred due to meteorite strikes or sudden meteor flashes. Individual characteristics which increase or decrease the potential for astronomical hazard-related suicide or heart attack could be a research area.

The physical mechanism of most space impact deaths would be crushing or physical trauma by the object if it lands, burning from an atmospheric explosion, trauma due to physical forces from the pressure wave or shock wave, or similar mechanisms to earthquakes (including tsunamis, if they occur) as the impact waves radiate outwards. An object could potentially skim the atmosphere without a direct impact, resulting in a regional superfire and/or pressure wave catastrophe along with the consequences of accompanying changes in atmospheric chemistry leading to acid rain and ozone layer destruction (e.g. Munich Re, 2001). Therefore, the principal space object hazard characteristics related to fatalities would be momentum (the product of mass and velocity) and trajectory (which would be included within the velocity vector).

Extraterrestrial radiation is an astronomical phenomenon related to cancer fatalities, particularly solar radiation exposure exacerbated by the ozone hole as a factor in skin cancer rates (e.g. Green et al., 1999; Oikarinen and Raitio, 2000). These studies also discuss an individual's characteristics which could increase their vulnerability to skin cancer. Other radiation phenomena such as giant flares from other stars--for instance, the 27 December 2004 event reported by Palmer et al. (2005)--lead to the scientists commenting in the media that similar events within several light years of Earth could cause a mass extinction.

Significant scope exists for further research into the causes and circumstances of fatalities from the environmental hazard of astronomical phenomena.

Brown, P.L. 1973. Comets, Meteorites and Men. Robert Hale and Co., London, U.K.

Carusi, A., E. Perozzi, and H. Scholl. 2005. "The Near Earth Objects: possible impactors of the Earth. Mitigation strategy". Comptes Rendus Physique, vol. 6, no. 3, pp. 367-374.

Chapman, C.R. 2004. "The hazard of near-Earth asteroid impacts on earth". Earth and Planetary Science Letters, vol. 222, issue 1, pp. 1-15.

Chapman, C.R. and D. Morrison. 1994. "Impacts on the Earth by asteroids and comets: assessing the hazard". Nature, vol. 367, pp. 33-40.

Fulchignoni, M. and M.A. Barucci. 2005. "The Near Earth Objects: possible impactors of the Earth. NEO Impact Consequences and Hazards". Comptes Rendus Physique, vol. 6, no. 3, pp. 283-289.

Green A, D. Whiteman, C. Frost, and D. Battistutta. 1999. "Sun Exposure, Skin Cancers and Related Skin Conditions". Journal of Epidemiology, vol. 9(no. 6 Supplement), pp. S7-S13.

Kharif, C. and E. Pelinovsky. 2005. "The Near Earth Objects: possible impactors of the Earth. Asteroid impact tsunamis". Comptes Rendus Physique, vol. 6, no. 3, pp. 361-366.

Kuypers, M.M.M., R.D. Pancost, and J.S. Sinninghe Damste. 1999. "A large and abrupt fall in atmospheric CO2 concentration during Cretaceous times". Nature, vol. 399, pp. 342-345.

Mancinelli, I., A. Comparelli, P. Girardi, and R. Tatarelli. 2002. "Mass Suicide: Historical and Psychodynamic Considerations". Suicide and Life-Threatening Behavior, vol. 32, no. 1, pp. 91-100.

Munich Re. 2001. Topics – Annual review: Natural catastrophes 2001. Munich Re Group, Munich, Germany.

Oikarinen, A. and A. Raitio. 2000. "Melanoma and other skin cancers in circumpolar areas". International Journal of Circumpolar Health, vol. 59, no. 1, pp. 52-6.

Palmer, D.M., S. Barthelmy, N. Gehrels, R.M. Kippen, T. Cayton, C. Kouveliotou, D. Eichler, R.A.M.J. Wijers, P.M. Woods, J. Granot, Y.E. Lyubarsky, E. Ramirez-Ruiz, L. Barbier, M. Chester, J. Cummings, E.E. Fenimore, M.H. Finger, B.M. Gaensler, D. Hullinger, H. Krimm, C.B. Markwardt, J.A. Nousek, A. Parsons, S. Patel, T. Sakamoto, G. Sato, M. Suzuki, and J. Tueller. 2005. "A giant gamma-ray flare from the magnetar SGR 1806-20". Nature, vol. 434, pp. 1107-1109.

Pickering, W.H. 1919. "Meteorites and Meteors". Popular Astronomy, vol. 27, no. 4, pp. 203-209.

Thuillot, W., J. Vaubaillon, H. Scholl, F. Colas, P. Rocher, M. Birlan, and J.-E. Arlot. 2005. "The Near Earth Objects: possible impactors of the Earth. Relevance of the NEO dedicated observing programs". Comptes Rendus Physique, vol. 6, no. 3, pp. 327-335.

Xie, S., R.D. Pancost, H. Yin, H. Wang, and R.P. Evershed. 2005. "Two episodes of microbial change coupled with Permo/Triassic faunal mass extinction." Nature, vol. 434, 494-497.

Yamamoto, I. 1928. "Probability of Injury by a Meteorite". Popular Astronomy, vol. 36, no. 3, pp. 207-208.

Yau, K., P. Weissman, and D. Yeomans. 1994. "Meteorite falls in China and some related human casualty events". Meteoritics, vol. 29, no. 6, pp. 864–871.

Sunset on Inis Oirr, Aran Islands, Ireland.

Are the sun's rays an astronomical environmental hazard?
Sunset on Inis Oirr, Aran Islands, Ireland.
(Copyright Ilan Kelman 1997.)

Case study: Water hazards

Water-related hazards are complex, with examples:

  • Flash flooding, which does not necessarily happen in a floodplain.

  • Fluvial/river flooding.

  • Jökulhlaups, often classified as a volcanic hazards.

  • Glacial lake outburst floods (GLOFs), with some parallels to jökulhlaups.

  • Storm surge, one form of sea flooding.

  • Tsunamis, another form of sea flooding, often classified as being an earthquake-, landslide-, volcano-, or NEO-related hazard, but also linked to seiching. As with seiching, tsunamis can happen in freshwater or saltwater.

  • Many windstorms bring heavy flooding, so deaths might be from wind, from water, or from a combination.

  • Avalanches are simply solid water yet are rarely classified with floods or storms.

  • Mudflows including lahars are a combination of water and solids.

As well, many water-related deaths might not be linked to a specific water-related hazard, with examples:

  • Boat crashes and sinkings.

  • Swimming pool, bucket, and toilet drownings.

  • Vehicles going into water.

  • Water sport fatalities.

The understanding of flood fatalities must be placed in the overall context of the health effects of floods (lethal and non-lethal effects), drownings (flood-related or otherwise), water safety, and Unusual Floods and Drownings.

Cross memorial to low-water crossing victim, Upolu, Samoa. Low-water crossing, Upolu, Samoa.

Roadside cross memorial for a victim of this low-water crossing on Upolu, Samoa.
The story is that, after heavy rainfall producing flash flooding, the river was overflowing this crossing, a van tried to cross it, and the vehicle was swept away. Several people in the van escaped, but one died.
(Copyright Ilan Kelman 2004.)

Some further documents on flood fatalities:

As an example of what an individual can achieve on this issue in the field, not only for water-related lifesaving but also for international development, download Lisa Mitchell's report on Lifesaving in Bolivia (232 kb in Word). Despite the challenges involved in setting up a lifesaving training course, "One month after the course volunteers...rescued 4 people during a swim race".

Some examples of earlier studies with methods to consider:

  • Gruntfest, E.C. 1977 (August). What People Did During the Big Thompson Flood. Working Paper 32, Prepared for the Urban Drainage and Flood Control District, Big Thompson, Colorado, U.S.A., full text (2,060 kb in PDF)

  • Sommer, A. and W.H. Mosley. 1971. Survey of 1970 Cyclone-Affected Region of East Pakistan, February-March 1971: Final Report, full text (2,108 kb in PDF). Parts of this report were published as A. Sommer and W.H. Mosley, 1972, "East Bengal Cyclone of November, 1970: Epidemiological Approach to Disaster Assessment", The Lancet, vol. 299, 13 May 1972, pp. 1029-1036.

Driving through floodwater during the Cambridgeshire, U.K. floods, February 2001.

If you want to die in a flood, drive through floodwater (Cambridgeshire, U.K.; the car's number plate has been deliberately blurred).
(Copyright Ilan Kelman 2001.)


The main recommendation is that disaster deaths research should focus less on partitioning data and analyses by hazard and instead aim to improve understanding of vulnerability characteristics for reducing disaster deaths, irrespective of the hazard. Despite impressive and interdisciplinary work regarding the causes and circumstances of deaths from disasters, this field requires plenty more work, particularly to ensure that policies and practices are based on robust and comparable evidence while acknowledging what is not known. Links to other topics would assist, such as loss of life modelling and non-fatal disaster-related health impacts including physical injuries and mental health.

Tornado shelter sign leading to the men's toilet at Denver International Airport.

Tornado shelter in the toilet at Denver International Airport. People are familiar with the room and there are no worries about emergency supplies of toilet paper.
(Copyright Ilan Kelman 2006.)

Contact Ilan.

The material on this website is provided for information only. Neither definitive advice nor recommendations are implied. Each person or organisation accessing the website is responsible for making their own assessment of the topics discussed and are strongly advised to verify all information. No liability will be accepted for loss or damage incurred as a result of using the material on this website. The appearance of external links on this website does not constitute endorsement of the organisations, information, products, or services contained on that external website.