Tsunami from Asteroid Impacts


      Tsunami from Asteroid/Comet Impacts

      by Michael Paine 
    See also Indian Ocean [Earthquake] Tsunami and Is Sydney at risk?
    Asteroid/Tsunami News

    Publication notes (please read)
    Nature of tsunami
    Asteroid impacts with Earth

    Tsunami Links Science of Tsunami Hazards Vol 20 No 1 (2002) 'Tsunami book gives a better understanding of ancient floods on Mars' (800K PDF).

    This page is mainly concerned with prevention of future tsunami disasters.

        Stony Asteroid Diameter  Hills & Goda (1998) (their Figure 1) Ward & Asphaug (1999) (their Figure 6) Crawford & Mader (1998) (their Table 1) 200m 1m (5m from equation) 5m negligible 500m 11m 15m <2m 1 km 35m 50m 6m How can we resolve these differences in order to carry out a risk assessment? There have been no detected asteroid impacts into an ocean on Earth so it is difficult to verify the models. However, the CTH computer code used by Crawford and Mader successfully predicted the consequences of the impact of Comet Shoemaker-Levy 9 with Jupiter. In the (fortunate) absence of experimental evidence on the Earth, the conservative results produced by Crawford & Mader have been used in the following analysis. In other words, it is assumed that asteroid impacts will generally produce non-coherent waves which dissipate quickly.There may be cases where an asteroid impact produces coherent waves but this would be due to a combination of unusual conditions, such as shallow water, rather than the norm.

        In the case of asteroids 200m and larger there is likely to be an impact into the ocean. For objects under this diameter there is a reduction in the size of the deepwater wave due to energy dissipation in the atmosphere. Speed, trajectory, density and strength of the object can affect the nature of the explosion. There does not appear to be an empirical formula available to deal with these smaller objects and it is possible that the smaller asteroids produce no appreciable waves. On the other hand, in the case of serious tsunami generated by earthquakes the energy involved is estimated to be equivalent to about 2 Megatons of TNT (Yabushita 1998). The impact by a 100m asteroid typically involves kinetic energy of about 75Mt so it would only involve the conversion of about 3% of this energy to wave energy in order to produce a serious tsunami - albeit, the tsunami could quickly dissipate, compared with an earthquake generated tsunami.

        On balance the following conservative values have been used for risk assessment. These are based on extrapolation of Crawford and Mader data. Note  that compared with Table 2, the range has been reduced to 100km to obtain reasonable values for the smaller asteroids.

        Table 3. Estimated deepwater wave height (above sea level)
        at a point 100km from asteroid impact (based on extrapolation of  Crawford & Mader)

        Asteroid Diameter (m) Deepwater Wave Height (m) 50 0.12 100 0.7 200 3 500 22 1 km 70 2 km 230

        Estimated risk to coastal locations

        Taking the New Guinea experience as a reference level, it is assumed that a tsunami with a 10m run-up will be of concern to low-lying coastal areas. This risk is estimated in the following steps:


          a) Determine the run-up factor W for the location in question.
          b) Determine the critical deepwater wave height that will produce a tsunami with a run-up height of 10m (H = 10 / W).
          c) For each size of asteroid, determine the distance over which a deepwater wave will need to travel before it has reduced in size to the critical height determined in step (b). This will be the "danger radius" for this combination of run-up factor and asteroid size.
          d) Determine the area of a semi-circular area of ocean with a radius equal to the distance derived in step (c).
          e) Calculate the probability of an impact within the area derived in step (d).

          Using a log-log plot of the Crawford and Mader data (see Appendix), the following estimates of danger radius have been derived by (gross) extrapolation.

        Using a log-log plot of the Crawford and Mader data, the following estimates of critical radius have been derived by (gross) extrapolation. Table 4
        "Danger radius": Estimated radius from impact
        for a tsunami 10m or higher at the shore Stony Asteroid Tsunami Run-up Factor

        Diameter 5 10 20 40 (m) Distance from impact (km)

        50 10 20 40 60 100 40 70 130 230 200 140 250 460 820 500 800 1400 2500 4400 1000 2800 5000 9000 16 000 It is noted that, irrespective of run-up factor,  the radius derived for a 50m asteroid  is about the same as the radius of direct devastation for the Tunguska event.

        Impacts by asteroids 2km and larger exceed the global catastrophe threshold and are disregarded for the purpose of analysing tsunami effects.

        For most coastal locations the surface area of ocean which poses a tsunami threat is a semi-circle with a radius R equivalent to the distances derived in the above table.  This radius is, however, limited by the size of the ocean. An area corresponding to 30% of the surface area of the Earth has been used for this limit (the approximate size of the Pacific Ocean). Applying equation (1) to the resulting semi-circular areas provides the following estimates of average intervals between events:

        Table 5 - Estimated interval between major tsunami events
        (tsunami height 10m or more) Stony Asteroid Tsunami Run-up Factor

        Diameter 5 10 20 40 (m) Average interval between tsunami events (years)
        for a single location ("city") on the shore of a deep ocean.

        50 - 81 million 20 million 9 million 100 - 66 million 19 million 6 million 200 83 million 26 million 8 million 2 million 500 20 million 7 million 2 million 670 000 1,000 4 million 1.3 million 400 000 330 000 All* 3 million 1 million 300 000 190 000 *All = 1/ ( 1/T50+   1/T100 + 1/T200 + 1/T500  + 1/T1000)

        In all cases it appears that risk of serious tsunami from asteroids 200m diameter and smaller  is much less than for larger objects.

        For a given coastal location the predicted average interval between 10m tsunami events (bottom row from Table 5) can be compared with the average interval between "direct" impacts (Table 1) to derive the relative risk for that location compared with an inland location (that is, a location which is not vulnerable to a 10m tsunami). Note that this is independent of the actual rate of impacts.

        Table 6 - Relative risk of coastal location compared to inland location

        Tsunami Run-up factor Relative risk due to all types of impact 0 (inland) 1 5 4 10 14 20 46 40 74

        This tentative analysis suggests that the risk to a low-lying coastal area from tsunami generated by asteroids is significantly greater than the risk from a "direct" impact by such objects. The average interval between such tsunami events is estimated to range from about 190,000 years for a location with a run-up factor of 40 to about  3 million years for a location with a run-up factor of 5. These compare with an average interval of 14 million years for a "direct hit".


        Comparison with the risk analysis by others

        In a paper titled "Asteroid impact hazard: a probabilistic hazard assessment" to be published in Icarus (and findings presented at the Tsunami Symposium in May 1999), Ward and Asphaug (1999) set out a comprehensive method of determining the impact tsunami risk. This analysis is based on methods they have developed for assessing earthquake risk. Probabilities are derived for a range of tsunami sizes striking a given coastline within a 1,000 year period. Note that in that paper tsunami height is measured just before the wave reaches the shore rather than run-up height. They assess the tsunami risk for a generic coastline and for the coastal cities San Francisco, New York, Tokyo, Hilo Harbour (Hawaii), Perth and Sydney.

        The estimates derived above indicate considerably less risk from an asteroid-generated tsunami than that derived by Ward and Asphaug. For example, they estimate the risk of a 10m tsunami inundating a generic coastline (with a semi-circular "target area" of ocean having is radius of 6,000km) is 1.1% in 1,000 years, equivalent to one event every 91,000 years and about one tenth of the risk estimated above.

        The main differences are likely to arise from assumptions about initial wave size and dispersion.

        Comparison with other asteroid impact risks

        In effect, the above analysis refers to risk of being caught in a region of direct devastation (being within the "blast area") compared with being within an area inundated by a tsunami.  In the case of an impact by a large asteroid (diameter 2km or more) it has been estimated that 25% of the human population would die - mainly from indirect effects, such as starvation. This type of event is thought to occur  with an average interval of 1 million years. The annual  risk of dying in such an event is therefore about 1 in 4 million, which is similar to the tsunami risk for a location with a run-up factor of 5 (1 in 3 million).

        Maybe it's not all bad news!

        Simulation of tsunami generated by Eltanin impact near Chile. Extract from animation by Dr C Mader.
        In some circumstances an ocean impact might even be less hazardous to mankind than a land impact because less debris will be thrown into the atmosphere and indirect effects might be reduced. For example, it has been noted by Mader (1998) that the Eltanin impact  by a large asteroid (estimates range from 1km to 4km) into the ocean near Chile some 2 million years ago did not create a crater on the seabed and apparently did not result in mass extinction. Contemplate what could have happened if the object had struck a slightly more northerly latitude and a few hours earlier - perhaps continental Africa would have been the target. Would Australopithecus, such as "Lucy", have survived? (Update: maybe it wasn't so benign - see this ABC News item about primate extinctions)
        Update 15 Feb 02: Steve Ward provided this dramatic graphic of the Eltanin impact (right click and View for a higher resolution).



        This tentative analysis suggests that the risk from asteroid tsunami has been substantially overstated - particularly in popular books about asteroid impacts with Earth. For typical coastal regions the risk of dying from an asteroid-generated tsunami is probably no greater than that of dying from the indirect effects of a large asteroid striking the Earth.

        For some coastal regions with unusual vulnerability to tsunami the risk of dying from asteroid-generated tsunami may be several times greater than that of dying from other asteroid-related causes. For these highly vulnerable areas the typical interval between asteroid tsunami events is likely to be about 200,000 years - assuming that impacts are randomly distributed in time.

        There is considerable uncertainty about most of the "input values" used in these estimates. Also it is possible that impacts are not randomly distributed in time (Steel et al, 1995) and the Earth may be subjected to a barrage of small asteroids (or comet fragments) from time to time. This may have happened over the past few thousand years and could be a source of some of the tsunami that appear to have struck Australia during this period. Until we better understand the impact threat, there is no cause for complacency over the long intervals derived above. Finally,  it is stressed that the run-up factor is not the sole issue in determining the destruction caused by a tsunami.


        Crawford D.A. and Mader C.L. (1998) "Modeling asteroid impact and tsunami", Science of Tsunami Hazards, Vol 16, No.1.

        Hills J.G. and Goda M.P (1998a) "Tsunami from asteroid and comet impacts: the vulnerability of Europe",Science of Tsunami Hazards, Vol 16, No.1.

        Hills J.G. and Goda M.P (1998b) "Damage from the impacts of small asteroids", J Planetary and Space Science, Elsevier Science,( available in PDF format )

        Mader C.L. (1998) "Modeling the Eltanin asteroid impact", Science of Tsunami Hazards, Vol 16, No.1.

        Morrison D. and Chapman C. 1995 "The Biospheric Hazard of Large Impact". Proceedings of Planetary Defense Workshop.

        Nott J. and Bryant E. (1999) "PALEOTSUNAMIS ALONG THE AUSTRALIAN COAST", Proceedings of the Tsunami Symposium,(temporary link), The Tsunami Society, May 1999.

        Rynn J. and Davidson J.(1999) "CONTEMPORARY ASSESSMENT OF TSUNAMI RISK AND IMPLICATIONS FOR EARLY WARNINGS FOR AUSTRALIA AND ITS ISLAND TERRITORIES", Proceedings of the Tsunami Symposium,(temporary link), The Tsunami Society, May 1999.

        Steel D. (1995) Rogue Asteroids and Doomsday Comets, John Wiley & Sons

        Steel D., Asher D., Napier W. and Clube S. (1995) "Are impacts correlated in time?" Hazards due to comets and asteroids,

        Ward S.N. and Asphaug E. (1999) "Asteroid impact tsunami: a probabilistic hazard assessment", Icarus, 1999 (preprint). Summary in PDF format

        Yabushita S (1997) "On the possible hazard on the major cities caused by asteroid impact in the Pacific Ocean - II", Earth, Moon and Planets. 76 (1/2):117-121.

        R.Young R.W.,Bryant E.,Price D. and Spassov E. (1996) "The imprint of tsunami in quaternary coastal sediments of Southeastern Australia"
        website http://wwwrses.anu.edu.au/~edelvays/tsunami1.html


        Vulnerability the East Coast of Australia to a 10m Tsunami

        Irrespective of the cause, there is a need to assess the risk to coastlines from tsunami. The south east coast of Australia makes an sobering case study. This coastline covers about 1,500 km from the Sunshine Coast in Queensland to Eden in New South Wales. Many low lying coastal areas along the south east coast of Australia have been intensively developed. Excluding the non-coastal suburbs of Sydney and Brisbane, the total population along this coastline is about 1.2 million.

        Consider the effects of a 10m tsunami like the one which hit northern New Guinea in July 1998. Based the topography of coastal developments  along the south east coast of Australia it is conservatively estimated that about 50,000 dwellings, containing about 140,000 people (about 12% of the population), are in areas which could be inundated by a 10m tsunami. If it is assumed that these people are in or near their dwellings (or similar vulnerable areas) for 50% of the time and that the death rate from people caught in such a tsunami is 50% then it is expected that 25% of the population would be killed#. The predicted death toll from one event which caused a 10m tsunami along the south east coast of Australia is therefore 35,000 (25% of 140,000). This could easily double during peak summer periods.

        Based on the above predictions, and assuming a  run-up factor of 10, the chances of an asteroid-generated tsunami event occurring in the next fifty years are estmated to be about 1 in 20,000 - a low risk but high consequence event. For comparison, Ward & Asphaug (1999) include a site-specific calculation of tsunami risk for Sydney. They estimate there is a 1.15% risk of a 10m or higher tsunami in the next 1,000 years - this is equivalent to a 1 in 1,700 chance in 50 years.

        Research by the University of Wollongong suggests that the New South Wales South Coast has been struck by at least six large tsunami within the last 6,000 years - a typical interval of 1,000 years - perhaps much less ( Young et al 1995). One possible cause is giant underwater "landslides" on the edge of the continental shelf but earthquakes and asteroid impacts may also be causes. Irrespective of the risk of tsunami from asteroid impact we really need to learn more about the risk to our coastlines from major tsunami.

        # For comparison, the earthquake-generated 24m tsunami which hit about 300km of the coastline of Honshu, Japan in 1896  killed 27,000 people. In vulnerable fishing villages 80% or more of residents were killed. The tsunami hit at 8pm when most people were at home.

        Update Dec 1999: The paper CONTEMPORARY ASSESSMENT OF TSUNAMI RISK AND IMPLICATIONS FOR EARLY WARNINGS FOR AUSTRALIA AND ITS ISLAND TERRITORIES by Rynn and Davidson is now available in PDF format (part of a 7.6Mb file for Vol 17 No. 2). See my review.

        Poisson Distribution - events randomly distributed in time

        In a Poisson process "events" occur at random points in time. It is possible to work out the probability of a certain number events occurring over a given interval of time. This can be useful for understanding the "uncertainty" about the time between impact events, and why the apparent lack of significant impacts over the past few hundred years (other than Tunguska) does not mean the estimates of impacts, from the known NEO population, are incorrect. This table sets out the probability for a process where the long term average interval between events is 100 years, as is estimated to be the case with the impact of 50m asteroids.  Probability of given number of events in an interval
         Number of events Interval (yrs) Nil 1 2 3 4 50 61% 30% 8% 1% 0.2% 100 (mean) 37% 37% 18% 6% 2% 200 14% 27% 27% 18% 9% 500 1% 3% 8% 14% 17%

        For example, the probability of exactly one event during a 100 year interval is 37% which is the same probability as nil events. This may seem counter-intuitive given  that the average interval between events  is 100 years but it simply results from th erandom distribution. Notice that the probabilities in this row, (and the row for 50 year interval) add up to 100% - the other rows, if extended to larger numbers of events, would also add up to 100%.
        Since both geological and historical records of a 50m NEO impact are unlikely to be reliable beyond 200 years the "fact" that just one event (Tunguska) appears to have occurred in this period is not unusual - it has a probability of 27% - the same probabilty as 2 events.

        Small time intervals
        The above table applies where the time interval under consideration (h)  is similar to the average time interval between events (T). Where the time interval under consideration is a small proportion of the average interval between events the probability of 1 event during time h is closely approximated by:
          P(one event in time h) = h /T
        Therefore, the probability of a "one in 2,500 year" event occurring during a 50 year period is 50/2500 = 2%  (that is a 1 in 50 chance of an event during a 50 year period).

        In a true poisson distribution we are never "overdue for an event" - the probability of an event occurring in the next year is the same as in the previous year and will be the same in the following year. The time since the last event has nothing to do with the timing of the next event.

        Are impacts randomly distributed in time?
        These calculations are based on the assumption that impacts are randomly distributed in time (Poission distribution). In the paper "Are impacts correlated in time?" by Steel, Asher, Napier and Clube (see the book "Hazards due to comets and asteroids"), it is proposed that impacts of objects in the range 50m to 300m arrive not randomly but in epochs of high activity.

        Notes about the analysis

        There are several uncertainties about the values used in the above calculations. These can only be resolved through more research into the near-earth asteroid population (i.e. Spaceguard Survey) and tsunami effects on the coastlines of interest.

        Population estimates

        Estimates of numbers killed depend on the number of people located in vulnerable areas when the tsunami hit and the circumstances at the time. The above population estimates for the East Coast of Australia were based on the proportion of each coastal town/city considered to be vulnerable to a 10m tsunami (roughly the proportion of dwellings in low-lying coastal areas - this varied from town to town but averaged out to about 12%) and the population of each town. The populations could easily double in peak summer periods. For a typical coastal plain it is estimated that  a 10m tsunami would penetrate about 1km inland whereas a 100m tsunami would penetrate about 22km (Based on Hills and Mader 1995).

        The basis of the estimate of the proportion killed is given in the analysis but this is highly dependent of the time of day, the season and the weather. A popular beach day would obviously be the worst scenario - perhaps ten times the estimate. Fortunately the chances of this are very slight - the total hours of popular beach days perhaps comprise 2% of the total hours in a year (mind you, Warringah's 6 beaches had an estimated 1.7 million visitors last summer!).

        Extrapolation of Crawford & Mader data

        The following graph is a log-log plot of the extrapolations used to derive Table 4, superimposed on the data from Crawford & Mader (their Table 1 on page 28). It can be seen that the extrapolations are speculative for both smaller asteroid sizes and large distances, since the Crawford and Mader data do not go below an asteroid diameter of 250m and do not go beyond a radius of 1,000km (and then only for the 1km asteroid). Strictly the extrapolations for the 50m and 100m asteroids do not take into account airburst effects but since the contribution of these impacts to overall tsunami risk turns out to be very low this will have negligible effect on the risk estimates. As a consequence of the uncertainties the risk estimates derived above should be regarded as ballpark estmates.

        The graph shows deepwater wave height (metres above sea level) by distance from impact (kilometres) for a range of asteroid diameters. The horizontal lines show the deepwater wave height which would produce a tsunami with a run-up height of 10m for a range of run-up factors (5, 10, 20 & 40).

        An estimate of "danger radius" can be derived from the intercept of these lines with the asteroid lines. For example, the lower, thick horizontal line shows a deepwater wave height of 0.25m which would produce a 10m tsunami at a location with a run-up factor of 40. This intercepts the extrapolated line for a 200m asteroid at a "distance from impact" of about 800km, suggesting that an impact by a 200m diameter asteroid anywhere within a radius of 800km would produce a tsunami 10m or higher at a location with a run-up factor of 40 (this is an unusually high factor).

      Alternative estimates of  impact tsunami wave heights

      Jack Hills from Los Almos National Laboratories is a key researcher in this field. He published several papers including "Tsunami produced by the impacts of small asteroids" published in the proceedings of  The Planetary Defence Workshop, May 1995. That paper includes an empirical formula for calculating the height of a deepwater wave (tsunami) 1,000 kilometres from the impact point. For an asteroid diameter D=200m or more the deepwater wave amplitude h is estimated by:
        h = 7.8 * [(D/406)^3 * (V/20)^2 * (M/3)]^0.54 (metres, at a distance of 1,000km)
       D is stony asteroid diameter in metres (note that Hills uses asteroid radius. Also, in the figures,  wave height is double amplitude)
       V is velocity in km/s (range 11 to 70, typical 20km/s)
      M is asteroid density in grams per cubic centimetre (range 1 to 6, typical 3g/cc)

      Results from this formula are shown in Table 2.

      Results of recent work by LANL will be presented at the Tsunami Symposium. The following is an extract from the abstract of a paper by Hills & Goda:
      "The critical factor in the third part of the study is to accurately determine the dispersion in the waves produced by the smaller impactors. Dispersion may greatly reduce the effectiveness of the smaller impactors at large distances from the impact point. We wish to understand this effect thoroughly before going to the Monte Carlo study. We have modeled mid-Atlantic impacts with craters 150 and 300 km in diameter. We are proceeding to Pacific impacts. The code has been progressively improved to eliminate problems at the domain boundaries, so it now runs until the tsunami inundation is finished. We find that the tsunami generated by such impacts will travel to the Appalachian mountains in the Eastern USA. We find that the larger of these two impacts would engulf the entire Florida Peninsula. The smaller one would cover the Eastern third of the Peninsula while a wave passing through the Gulf of Cuba would cause the inundation of the west coast of Florida."

      Wave dispersion

      According to Ward & Asphaug (1999) the deepwater wave height reduces approximately in proportion to distance travelled: H is proportional to 1/R. In the absence of dispersion, H can be expected H to be proportional to 1/R0.5 since energy is proportional to the square of wave height (this relationship also exists when considering the work done in depressing a water surface against hydrostatic pressure).

      For large distances from the impact the above log-log plot based on Crawford and Mader agrees, roughly, with the 1/R relationship (the actual relationship is about 1/R0.85). The main differences between methods appear to result from differences in estimates of the initial wave size and wave dispersion over the first 100km or so.

      Calculation of risk to Earth's inhabited regions

      This section attempts to estimate the risk to regions important for population, agriculture and resources. The total land area of these regions is estimated in the following table.

      Table A1. Estimate of inhabited land areas

      Continent Land Area 
      (millions sq km) Assumed % inhabited Area inhabited 
      (millions sq km) Africa 30 30% 9 Antarctica 14 0% 0 Asia 45 30% 14 Europe 10 90% 9 N.America 24 30% 7 Oceania 9 20% 2 S.America 18 20% 4

      45 = 9% Earth

      For a given inhabited area the risk from direct devastation by an asteroid impact is related to the total area of the region plus a boundary representing the radius of destruction of the impact event - the larger the impactor the larger the boundary.
      For the purpose of estimating risk it is necessary to assume the typical size of an inhabited area. In this analysis this is assumed to be 500km by 500km*. Around this area will be a boundary which varies according to the size of the asteroid. The method is illustrated in the diagram, where B is the radius of devastation from a given impact.

        * Caution: this analysis is quite sensitive to the value chosen. For example, in the table below the "target area" for a 500m asteroid impact varies from 15% for a 1,000x1,000km area to 50% for a 200x200km area.

      Based on the above assumptions (9% of Earth's surface "inhabited" and typical inhabited region is 500km by 500km), an estimate can be made of the total "target area" which represents a risk to inhabited regions for each size of asteroid. Note that the estimate of target area for 1km asteroids is probably high because the resulting "boundaries" will overlap adjacent inhabited regions. However, the significant indirect effects of these impacts have not been taken into account and could more than compensate for this problem. Note also that boundary areas can include seas and oceans therefore the total "target area" can exceed the land area of Earth (as in the case of 1km impacts in the following table).

      Table A2. Estimate of the risk of an inhabited region being within an area of direct devastation

      Asteroid Diameter 
      (m) Width of "Boundary"* 
      (km) Target Area (% of Earth) Annual Probability Average Interval (Years) Chance in 50 years  1 in ... 50 24 10.8% 1.1E-3 900 18 100 48 12.7% 1.3E-4 8000 160 200 96 17% 3.4E-5 30 000 600 500 149 22% 5.6E-6 180 000 3600 1000 252 34% 3.4E-6 290 000 5800 All 1.3E-3 800 16 * radius of direct devastation due to impact. There is a high risk of death within this radius

      These values are used in Table 1 and they represent the risk of an inhabited region being within an area of direct devastation - the consequences of that impact depend on the size of impactor, population density and numerous other factors. In general, the consequences of a large impact are much graver than those of a smaller asteroid and indirect effects, such as global starvation, could lead to greater loss of life than the initial impact.

      Estimate of fatalities
      Building on a method of analysis presented by Steel (1995), an estimate can be made of the likely fatalities from a particular type of impact.

      The most violent explosion in historical times was the Indonesian Tambora volcano eruption in 1815 which resulted in a 6km diameter "crater" (see image). This is recorded as having caused 10,000 deaths immediately due to blast and ash and a further 80,000 deaths in the region over subsequent weeks, due mainly to starvation. The eruption also pushed an estimated 80 cubic kilometres of ejecta into the atmosphere and is a possible cause of the "year without summer" (1816) in the Northern Hemisphere, when freezing weather hit the USA during June and there were widespread crop failures. This suggests that indirect deaths from such a major disruptive event can exceed eight times the direct death toll. This ratio has been used in the following calculations for asteroids 500m diameter and more. A lower value has been used for the smaller asteroids due to the increased likelihood of airburst explosions.

      Assuming that 9% of the Earth's surface is inhabited by 6 billion humans then the average population density of inhabited regions is about 130 persons per square kilometre. The "area devastated" in Table 1 can be combined with the estimated risk of an inhabited region being devastated to derive a (very rough) estimate of potential fatalities: This takes into account the reduced population density in "boundary areas". Also the consequences of an impact by a 2km asteroid are included, based on assumption that one quarter of the human population would perish - mainly from indirect effects.

      Table A3. Estimate of death toll from various types of  impact

      Asteroid Diameter 
      (m) Area devastated 
      (sq km) "Typical" 
      Direct Fatalities Ratio of 
      fatalities Total fatalities Annual 
      chance for inhabited regions 
      1 in ... Equivalent annual death toll 50 1900 200 000 4 1 million 900 1100 100 7200 650 000 4 3 million 8000 400 200 29 000 2 000 000 6 14 million 30 000 500 500 70 000 4 000 000 8 35 million 180 000 200 1 km 200 000 7 000 000 8 63 million 290 000 200 2 km - - - 1.5 billion 1 million 1500 All 800 3900

      For comparison, the average annual death toll from earthquakes is about 10,000 per year. That of commercial airliner crashes is about 700 per year!

      Comparison with risk estmates by John Lewis
      Email from John S Lewis, University of Arizona, May 1999:
      Thought you might be interested in seeing the results of a very elaborate Monte Carlo simulation (repeated calculations using random input parameters) of impact hazards on a time scale of 10^4 years and less.  The calculations use the best available orbital and taxonomic data on NEOs, laboratory chemical and physical properties of impactor materials, realisric strength-vs.-size models, 3-D entry geometry, detailed modeling of ablation, luminosity, fragmentation, airburst blast waves, S injection, NOx production, Ir signatures, etc.
      A popular account of the simulations appeared in my book "Rain of Iron and Ice", Addison-Wesley (1996), and a detailed technical account of the modeling will appear in "Comet and Asteroid Impact Hazards on a Populated Earth", due out this year from Academic Press.
      Using this model, I found the same basic importance of Tunguska-type airbursts on normal human (1-100 years) and societal (100-10,000 years) time scales.  The majority of the fatalities, however, are caused by the largest single lethal event in the simulation.
      In "Rain of Iron and Ice" (1997) Dr Lewis describes the results of ten simulations of 10,000 year 'runs'. They include tsunami effects and 30% of the deaths from 1gigaton+ events were due to tsunami. The largest impact was an 8.5gigaton event (e.g. a 3km asteroid). Over the ten runs the equivalent annual fatalities range from 720 to 6,170, with an average of 2,450 deaths per year. Impacts of 20Mt (e.g. a 60m asteroid) are violent enough to kill 100,000 people - equivalent to 360 deaths per year.
      The values in Table A3 are therefore in the right ballpark.
      Dr Lewis also refers to an event in China in 1490 when "stones fell like rain" and over 10,000 people were killed.

      5 Dec 1999: Just received my copy of a new book by Planetary Scientist John Lewis "Comet and asteroid impact hazard on a populated Earth". It includes a diskette with a Monte Carlo program to run simulations of Earth impacts over time. The book is basically a handbook for the software with a wide range of physical information about NEOs, impacts and effects on the human population. An excellent resource covering physics, chemistry and environment . My own rough estimates of human fatalities may prove too optimistic.

      The paper "Meteorite falls in China and some related human casualty events" by Yau, Weisman and Yeomans (downloadable PDF from NASA Astrophysics Data System ADS Abstract Service) also refers to the 1490AD event. This paper estimates the worldwide fatalities from meteorite impacts (mainly due to collaspsed buildings) at around one fatality every four years. It does not cover larger impact events.

      Tsunami risk
      A similar estimate of risk can be derived for coastlines vulnerable to asteroid-generated tsunami.

      Table A4. Estimate of the length of inhabited coastline

      Ocean Estimated inhabited coastline
      (km) Indian (Africa, Asia, Australia) 16 000 Pacific East (Americas) 11 000 Pacific West (Asia, Oceania) 15 000 Atlantic West (Americas) 17 000 Atlantic East (Africa, Europe) 12 000 Southern (Australia) 3 000 Total 74 000 Applying the "danger radius" values from Table 5, and assuming a typical run-up factor of 5, a target area of ocean can be derived for each size of asteroid. Only asteroids 200m diameter and larger are considered because, with smaller asteroids,  the area of direct devastation is likely to be similar to that the tsunami threat.

      Table A5. Estimate of risk of an asteroid-generated tsunami (run-up height 10m or greater) striking an inhabited coastline (tsunami run-up factor 5)

      Asteroid Diameter
      (m) Tsunami Danger Radius (km) Tsunami Danger Area (% of Earth) Annual Probability Average Interval between events (years) Chance in 50 years
      1 in ... 200 140 2% 4.1E-6 250 000 5000 500 800 12% 2.9E-6 350 000 7000 1000 2800 41% 4.1E-6 250 000 5000 All 1.1E-5 90 000 1800 Note that, due to the method of calculation, these risks are not independent of the risk of direct devastation.
      Conclusions about the risk to inhabited areas
      Overall, this tentative analysis suggests:
      • Asteroids 1km diameter and larger pose the greatest threat to humankind, in terms of the number of fatalities - the death toll from a 1km impact would probably exceed 63 million (somewhere between 1km and 2km diameter the event becomes a global catastrophe, with over 1 billion deaths). It is estimated that there is a 1 in 2,000 chance of the Earth being struck by a 1km asteroid in the next 50 years.
      • The most likely type of impact for an inhabited area of the Earth is the Tunguska-size event (asteroid diameter about 50m). The effects would probably be quite localized (unless it triggers a nuclear war). It is estimated that there is a 1 in 18 chance of an inhabited region somewhere on Earth being devastated by such an impact in the next 50 years and the total fatalities could be around 1 million. However, for a given location, the chance of devastation by a Tunguska-size impact in the next 50 years is about 1 in 600,000.
      • Due to tsunami, vulnerable coastal locations are at increased risk from impacts by asteroids 200m or larger, compared with "inland" locations. It is estimated that there is a 1 in 1,800 chance of an inhabited coastal location somewhere on Earth being inundated by an asteroid tsunami in the next 50 years (assuming a typical run-up factor of 5). The chance of a given coastal location being inundated by an asteroid tsunami depends on the run-up factor and other factors. It is estimated that a "high-risk" location with a run-up factor of 10 has a 1 in 20,000 chance of being inundated in the next 50 years. This is less risk than that from the global effects of a 1km asteroid or larger striking somewhere on Earth but considerably more than the risk of direct devastation by a Tunguska style impact.
      See books for further reading on this subject. Also a paper "Damage from the impacts of small asteroids" by Hills & Goda is available in PDF format.

      Are you insured?

      My home insurance policy covers me for "impact by space debris or debris from a rocket, satellite or aircraft" but not "the action of the sea, tidal wave, high water or tsunami". Interestingly, Tsunami is defined as "An unusually high wave or series of waves caused by an earthquake or volcanic eruption". Tsunami generated by asteroid impacts or underwater landslides would not meet this definition, but are probably still excluded as "action of the sea". Anyhow, I guess insurance would be the least of my worries - I live 150 metres above sea level. Most of Sydney would be less than 50 metres above sea level!

      Tsunami Links

      Several links no longer work. Suggestions for correct URLs are welcome. Frame from USGS animation. Greatly magnified vertical scale.

      Australian sites


      Tsunami resources

      More NEO links


      Note: The tsunami sizes in some of these books are based on earlier work and are much larger than those described in Table 3 above.
      • "Rogue Asteroids and Doomsday Comets" by Duncan Steel (John Wiley & Sons, 1995) has information about tsunami and NEO impacts.
      • "Impact! The threat of comets and asteroids" by Gerrit Verschuur, Oxford University Press, 1996.
      • "Rain of Iron and Ice" by John S Lewis, Addison-Wesley (1996) includes a Monte-Carlo simulation of impact hazards
      • "Hazards due to comets and asteroids", edited by Tom Gehrels from Spacewatch, has dozens of scientific papers about impact hazards. It includes a paper "Tsunami generated by small asteroid impacts" by Hills, Nemchinov, Popov and Teterau.
      • Comet and Asteroid Impact Hazards John Lewis
        • by Edward A. Byrant, School of Geosciences, University of Wollongong, Australia
          0 521 77244 3   Hardback                £55.00/$74.95
          0 521 77599 4   Paperback       £19.95/$27.95
          Publication c. July 2001
          For more details and how to order, please visit the (UK) website
          It can also be ordered from Cambridge University Press Melbourne.
            In the past decade over ten major tsunami events have impacted on the world's coastlines, causing devastation and loss of life. Evidence for past great tsunami, or 'mega-tsunami', has also recently been discovered along apparently aseismic and protected coastlines. With a large proportion of the world's population living on the coastline, the threat from tsunami can not be ignored. This book comprehensively describes the nature and process of tsunami, outlines field evidence for detecting the presence of past events, and describes particular events linked to earthquakes, volcanoes, submarine landslides and meteorite impacts. While technical aspects are covered, much of the text can be read by anyone with a high school education. The book will appeal to students and researchers in geomorphology, earth and environmental science, and emergency planning, and will also be attractive for the general public interested in natural hazards and new developments in science.
            (from CCNet)