Académique Documents
Professionnel Documents
Culture Documents
1
University of Southampton, Engineering and the Environment, Southampton, UK
2
Lancaster University, Faculty of Science and Technology , Lancaster, UK
3
University of Southampton, Geography and Environment, Southampton, UK
4
Queen's University Belfast, School of Geography, Archaeology and Palaeoecology, Belfast,
UK
*
Corresponding Author. Email c.rumpf@soton.ac.uk
Key Points:
Human vulnerability models for asteroid impacts for seven impact effects (e.g.
tsunami, wind blast, etc)
Introduction of risk and applicability to asteroid impact hazard.
Protective function of continental shelf against tsunamis by impactors on shelf.
1
Abstract
An asteroid impact is a low probability event with potentially devastating consequences. The
Asteroid Risk Mitigation Optimization and Research (ARMOR) software tool calculates
whether a colliding asteroid experiences an airburst or surface impact and calculates effect
severity as well as reach on the global map. To calculate the consequences of an impact in
terms of loss of human life, new vulnerability models are derived that connect the severity of
seven impact effects (strong winds, overpressure shockwave, thermal radiation, seismic
shaking, ejecta deposition, cratering and tsunamis) with lethality to human populations. With
the new vulnerability models ARMOR estimates casualties of an impact under consideration
of the local population and geography. The presented algorithms and models are employed in
two case studies to estimate total casualties as well as the damage contribution of each impact
effect. The case studies highlight that aerothermal effects are most harmful except for deep
water impacts, where tsunamis are the dominant hazard. Continental shelves serve a protective
function against the tsunami hazard caused by impactors on the shelf. Furthermore, the
calculation of impact consequences facilitates asteroid risk estimation to better characterize a
given threat and the concept of risk as well as its applicability to the asteroid impact scenario
are presented.
2
characteristics of the asteroid and the Risk
surface. For example, an asteroid that enters
the atmosphere at a shallow angle is more The concept of risk is applicable to a
likely to experience an airburst than a wide variety of subjects (e.g. finance,
surface impact with very different impact insurance, politics and decision making).
effects (e.g. lack of seismic shaking and Risk, defined as the expected loss, is the
cratering). Similarly, a large or dense product of three factors: the probability that
asteroid is more likely to reach the surface an event occurs, exposure, the value that is
than a small or highly porous one. at stake (or exposed), and vulnerability - the
Furthermore, the hazard of a large impact portion of the exposed value that is affected
near a densely populated metropolitan area if the event occurs. Specifically, for the
will yield more damage than a similar asteroid impact hazard, this relation can be
impact in an unpopulated desert. Therefore, stated in mathematical terms as:
it is crucial to consider the impact situation
of each asteroid taking into account impact = () (1)
location as well as the physical
circumstances (e.g. impact angle and speed) where is asteroid hazard risk, is the
of the event. asteroid impact probability, is the
At the University of Southampton, population (exposure) and () is the
the Asteroid Risk Mitigation Optimization vulnerability which is a function of the
and Research (ARMOR) tool is under severity of harmful effects generated by
development to analyse the threat posed by an asteroid impact. The information needed
discovered asteroids. ARMOR calculates to assess the asteroid impact probability is
impact effects and determines the lethally provided by ARMOR as the asteroids
affected population in the impact zone spatially distributed impact probability
considering their vulnerability. (shown in (Rumpf et al. 2016b)) which not
Consequently, ARMOR allows for the risk only allows identification of the possible
of known asteroids to be calculated in terms impact locations but also provides
of expected casualties. This paper presents information about localized impact
a new method for calculating asteroid risk probability. The global population map
including and emphasising the derivation of feeds the exposure term and provides
the necessary vulnerability models. The exposure values as well as its spatial
applicability of the vulnerability models is distribution. Here, the global population
demonstrated by few examples that present map for the year 2015 (CIESIN et al. 2005)
estimation of total casualty numbers as well with a grid resolution of 4.64.6 km2
as the contribution of each impact effect in (Figure 1) is used. Vulnerability describes
three scenarios. what portion of the exposed population is
lethally affected by the asteroid impact and
RISK AND VULNERABILITY this term depends on the severity of the
MODELS impact generated effects. The process of
impact effect modelling and vulnerability
In this section, the concept of risk is estimation is described in the following
defined, the percentage of the outside (of sections along with the derivation of
buildings) population is derived and vulnerability models.
vulnerability models are presented.
3
Figure 1: World population density map for the year 2015 based on data by (CIESIN et al.
2005). Data represents population density as people per square km and the scale represents
powers of ten. Note that maximum population density is higher than represented by this scale
in certain regions.
4
/3 corresponding to an estimated
0.13 2.5 + 0.26 (7.3 2.5)
= 0.22 (2) yield strength of 381315 Pascal (Pa)
7.3
(Collins et al. 2005). Ordinary Chondrites
For further analysis, it was assumed that account for about 90% of all known
22% of the global population is meteorites (Britt 2014).
unsheltered at any given time. The process that was used to
determine if an asteroid experiences an
Impact effect and vulnerability airburst or surface impact is visualized in
modelling Figure 2 and it is based on (Collins et al.
2005). This process employs analytical
Upon colliding with the Earth, an models to calculate the outcomes of
asteroid deposits most of its energy either physical processes that occur during
in the atmosphere, during an airburst, or on atmospheric passage (e.g. break-up altitude,
the surface after it passes the atmosphere airburst altitude, impact velocity, etc.) as
mostly intact. Whether a surface impact or well as the severity of subsequent impact
airburst occurs depends on the entry effects. Once the asteroid airbursts or
conditions of the asteroid: impact angle, impacts the surface, its energy is released in
impact speed, size of the asteroid, and a variety of impact effects and in this
material. In this analysis, impact angle and analysis, seven impact effects are modelled:
speed are provided by ARMORs orbit High winds, overpressure, thermal
dynamic impact simulation. Furthermore, radiation, cratering, seismic shaking, ejecta
size values are published by ESA and blanket deposition and tsunami. The first
NASA on their NEO webpages and the three of these may occur in both, airburst or
sizes were estimated based on the surface impact, while the latter four occur
asteroids brightness. Finally, the asteroid only in a surface impact.
body was assumed to be similar to
ordinary chondrites with a density of 3100
5
Asteroid Asteroid Asteroid
Break Up Above Airburst Above
enters Yes breaks up Yes experiences
altitude ground? altitude ground?
atmosphere in mid air Airburst
No
Asteroid
Asteroid fragments Airburst
reaches No
impact Effects
surface ground
Asteroid
impacts Ocean Impact
solid Effects
ground
Figure 2: Impact effect flow diagram showing how an airburst or surface impact is
determined and the corresponding impact effects.
6
impact effects, the term = 1 The calculation of energy scaled
, conversely, describes the chance airburst altitude 1 is described in (Collins
that this individual survives the impact et al. 2005) as scaling the result of equation
(18) using equation (57) of that reference.
effect. After all effect vulnerabilities, and,
The pressure shockwave is reflected from
similarly, all effect survivability chances,
the surface and it interacts constructively
are computed, the chance of an individual with the original shockwave at sufficient
to survive all effects in sequence may be distance from the airburst. In fact, this
calculated as: condition is already described in equation
(5) and the switching distance 1 in
= = (4) meters between equations (6) and (5) is:
7
sense of the sensitivity of the impact effect not included in the present vulnerability
models. The resulting overpressure model.
vulnerability models are dependent on Strong winds accompany the
overpressure (at a given distance) and overpressure shockwave and the severity of
the best fit values for the coefficients , strong winds is expressed by equation (10).
and for the best, expected and worst cases In fact, overpressure shockwave and strong
are: winds occur together and depend on each
other. However, they are treated separately
Table 1: Overpressure vulnerability in terms of vulnerability models because
coefficients. their mechanism of harming humans differs
(overpressure: internal organs, wind:
Case dislocation of bodies or objects). A wind
vulnerability model was derived based on
Expected 1.0 2.424 105 4.404 105
the severity making use of the similarity
Best 1.0 1.899 105 5.428 105 between strong wind gusts and the criteria
in the Enhanced Fujita (EF) scale, which is
Worst 1.0 2.847 105 3.529 105 used to classify tornado strength (Wind
Science and Engineering Center 2006) in
the United States of America (USA). In the
EF scale, tornados are classified based on
Thus, the vulnerability function to the damage that they cause during the peak
overpressure is: 3 seconds wind gust and Table 2 provides
an overview of the EF category, wind speed
1
( ) = (11) and expected damage.
1+ (+)
The vulnerability function is plotted in
Figure 3 along with the experimental data
points.
8
Table 2: Enhanced Fujita scale. population that is outdoors is hit by missiles
Categories, wind speeds and damage or thrown against objects (affected
population) and that 2% of these
Category 3s Typical Damage individuals die as a direct result of the
Wind-
Gust injury. These assumptions provide a
[m/s] vulnerability of 0.000044 for strong winds
corresponding to a category EF0 tornado.
EF0 29-38 Large tree branches broken;
Trees may be uprooted; Strip
mall roofs begin to uplift.
EF1
With increasing wind speeds, a larger
EF1 38 - 49 Tree trunks snap; Windows portion of the outside population will be
in Institutional buildings affected; more people will be thrown
break; Facade begins to tear
off. against solid objects because the strong
wind will be able to lift up more people. The
EF2 49 - 60 Trees debark; Wooden wind will also generate more missiles that
transmission line towers could hit victims. Furthermore, the lethality
break; Family residence
buildings severely damaged for each person also increases because the
and shift off foundation. impact speed of the body or the missile will
be higher. Reference (Glasstone and Dolan
EF3 60 - 74 Metal truss transmission 1977) estimates that 50% lethality is
towers collapse; Outside and
most inside walls of family reached when a body contacts a solid object
residence buildings collapse. with a speed of 16.5 m/s and 100% lethality
is reached at 42 m/s. It seems plausible that
EF4 74 - 89 Severe damage to
institutional building
a body could be accelerated to speeds of
structures; All family 16.5 - 42 m/s in an EF1 tornado. However,
residence walls collapse. it can be assumed that some of the outside
population finds sufficient shelter. Hence, it
EF5 >89 Severe general destruction.
was assumed that 10% of the outside
population is affected and that 5% of those
EF0
affected die. Housing still provides good
According to the EF scale, EF0
protection against EF1 level winds but it
corresponds to wind speeds between 29-38
was assumed that 1% of the inside
m/s. Humans can be harmed in this
population can be affected and that 5% of
condition by being thrown against objects
those affected die. Vulnerability for winds
or objects being hurled at them. In
corresponding to an EF1 tornado was, thus,
(Glasstone and Dolan 1977) lethality
set to 0.0015.
estimates are provided for objects turned
missiles that hit the body. According to this The above assumed increase in
source, a 5 kg object entails a near 100% vulnerability agrees well with the increase
rate of fracturing a skull when hitting the of lethality of recorded tornados between
head with a velocity exceeding 7 m/s. 2000 and 2004 (NOAA 2015). During that
Furthermore, lethality may occur when the time period 4284 EF0 tornados killed 2
body is thrown against solid objects with people resulting in a casualty rate of
velocities in excess of 6 m/s. It is 0.00047 per EF0 tornado. In the same time
conceivable that these events may be 1633 EF1 tornados killed 20 people
produced in a category EF0 tornado and, yielding a casualty rate of 0.012 per EF1
indeed, category EF0 tornados have been tornado which is a 26 fold increase.
lethal in the past (NOAA 2015) but the Similarly, assumed vulnerability for strong
casualty rate is low (3 people were killed by winds increased by a factor of 33.
EF0 tornados between 1997 and 2005).
Here, it was assumed that 1% of the
9
EF2 and Sutter 2005) and (French et al. 1983)
Increasing wind speed renders that the presence of a warning decreases
shelters less effective as houses start to mortality by a factor of about three. In
exhibit significant damage. It is assumed addition, housing standards in the USA
that, in addition to 40% of the outside ensure that protection of the population
population, 5% of the housed population is against windblast is better than the global
affected yielding a total of 12.7% affected average by an assumed 50% (factor 1.5).
population. Lethality for the affected Taking into account the influence of
population increases to 10% as wind speeds warning and better protection, the observed
are capable of accelerating bodies beyond vulnerability of 2.5% can be expected to
the 42 m/s body impact speed assumed for increase to a global average of over 9%
100% lethality (Glasstone and Dolan 1977) matching well with the windblast
and objects turned missiles have higher vulnerability found here.
damage potential. Consequently,
vulnerability is equal to 0.013. EF4
The increase in vulnerability from Persons who are sheltered in very
EF1 to EF2-like wind speeds of a factor of well constructed concrete buildings will be
8.5 matches the casualty rate increase from protected against these winds. It was
EF1 to EF2 tornados. Between 2000 and assumed that 40% of the inside and 90% of
2004, 439 EF2 tornados killed 51 persons the outside population would be affected
yielding a casualty rate of 0.12 per EF2 with corresponding lethality rates of 30%
tornado corresponding to a 9.6 fold and 40%, respectively. Thus, vulnerability
increase. is 0.17.
EF3 EF5
Tornados of this category destroy The great majority of structures
most housing shelter leaving basements and collapse in these winds offering
well-constructed concrete buildings as diminishing protection. Consequently, it is
viable shelter options. It was assumed that assumed that 95% of the outside and 50%
80% of the outside and 30% of the inside of the inside population is affected with a
population would be affected by winds of lethality rate of 50% and 40%, respectively.
this strength. For those affected outside and The resulting population vulnerability is
inside, lethality increased to 30% and 20%, 0.26 at 89 m/s wind speed.
respectively, due to hitting missiles or by Evidently, (Wurman et al. 2007)
being thrown against fixed structures. The modelled EF5 tornados in an urban setting
vulnerability thus increases 8 fold to 0.10. and assumed that 10% of the inside
The record shows that 116 persons were population would be affected lethally.
killed by 127 EF3 tornados yielding a Taking into account the criticism that this is
casualty rate of 0.913 that corresponds to an likely an overestimation for a setting in the
8 fold increase from EF2 to EF3 tornados. USA, 5% lethality seems more likely.
In fact, (Paul and Stimers 2014) show Considering the influence of warning
that the vulnerability inside a zone affected (factor 3) and extrapolating to a global
by an average of EF3 tornado winds setting (factor 1.5), a value of 22.5% was
(Kuligowski et al. 2013), was 2.1%. In obtained which correlates closely to the
contrast to a tornado, the windblast in the value found previously.
case of an unforeseen asteroid impact Based on these data, three
would arrive without prior warning by the vulnerability models were derived: One
government or by meteorological cues that model that describes the expected case
the population could be expected to and two for a worst and best
correctly interpret. It is shown in (Simmons case. The expected case uses the
10
median wind speed for each EF category generates a plume with very high pressure
with the corresponding vulnerability value, (>100GPa) and temperature (10000K)
while the worst and best case models use that rapidly expands. This is called the
the wind speeds of one category lower or fireball. As a result of the high temperature,
higher, respectively. The model function is: the gas is ionized and appears opaque to
thermal radiation due to the plasmas
1 radiation absorption characteristics.
( ) = (12)
1 + (+) Consequently, the plume expands
and the corresponding coefficients are: adiabatically and only starts to radiate
outwards when the plasma cools to the
Table 3: Wind vulnerability coefficients. transparency temperature (Zeldovich
and Raizer 1966). (Collins et al. 2005)
Case report an empirical relationship for the
fireball radius when it reaches
Expected 1.0 5.483 102 1.124 102
transparency temperature as a function of
Best 1.0 5.036 102 1.293 102 impact energy :
1
Worst 1.0 5.549 102 9.898 101 = 0.002 3 (13)
Figure 4 shows the vulnerability models Only a fraction of the kinetic energy
plotted over the relevant range of wind gust released during impact is transformed into
speeds. thermal radiation (Nemtchinov et al. 1998).
This fraction is called the luminous
efficiency and (Ortiz et al. 2000)
determined that it is on the order of 104 to
102. The received thermal energy per area
unit (assuming a hemispheric dissipation of
heat radiation) is given by (Collins et al.
2005) as:
= (14)
22
where, is the fraction of the fireball that is
visible over the horizon at distance which
Figure 4: Wind vulnerability models with
is also a function of and the
data points.
corresponding geometric relationship is
Thermal Radiation given in (Collins et al. 2005) (equation
(36)). The effect severity of thermal
Surface impacts as well as airburst radiation is subject to additional
produce thermal radiation but the two considerations such as local weather (e.g.
events require separate modelling as fog will reduce severity) and the
presented in the following. topography providing shadowing
opportunities. Additionally, luminous
Ground Impact efficiency is currently not well constrained
and these considerations require a
If the impacting meteoroid travels in cautionary note that thermal radiation
excess of 15 km/s, enough energy is severity could vary significantly from
released to evaporate the asteroid and some estimates provided by the present model.
of the ground material. This violent event This comment is also valid for the severity
11
estimation of other impact effects as
uncertainty in severity estimation (Collins =40 = 700000 1 1010 (17)
et al. 2005) is not considered in the best and
worst case scenarios here which are Finally, a linear interpolation scheme
designed to capture uncertainty in estimates for any given airburst altitude
vulnerability modelling only, based on the calculated values for
=25 and =40.
Airburst 25 40
= = (40000 ) + 40 (18)
15000
Besides the air blast, some of the
kinetic energy carried by the meteoroid that The distance was estimated using
is released during airburst dissipates as Pythagoras relationship with airburst
thermal radiation. (Nemtchinov et al. 1994) altitude and surface distance as
investigates the radiation emitted by parameters:
meteors and the following airburst thermal
radiation model was derived here based on
= 2 + 2 (19)
this research. Equation (11) of the reference
provides an expression for thermal energy
flux density based on airburst intensity: With these relations, equation (15) can be
solved and a thermal energy flux density
0 2 may be obtained for any airburst event.
= [ ] 5 (15) Note that the unit of equation (15) is
2
[W/m ] and that for subsequent analysis the
where is the energy flux density in thermal radiation energy density [J/m2] was
[W/m2] at the target distance, (the needed. Based on visual observations of the
reference uses ) is the energy flux Chelyabinsk (Popova et al. 2013) and other
density of the meteoroid at a given altitude, meteors (Borovika and Kalenda 2003) it
0 is the asteroid diameter and is the was determined that a break-up occurs
line of sight distance from the airburst to the within a time span on the order of one
target. Table 1 of the reference provides second. Therefore, one second was
values for as a function of speed for the assumed as the default break-up duration
two altitudes of 25 km and 40 km. Here, an for airbursts and the unit [W/m2] is
interpolation function was built that equivalent to the energy density [J/m2]
produces values for any given speed, when integrated for this timespan because
altitude pair based on table 1 in the energy is the integral of energy flux
reference. To this end, a six degree [W]=[J/s] over time [s]. This relation is
polynomial was least square fitted to the expressed by the following example
data describing =25 at 25 km altitude as equation assuming that energy flux is
a function of meteoroid speed : constant over time:
12
for the thermal radiation vulnerability expected when exposed to a certain radiant
model. It should be noted that the spectral energy.
intensities in the burn relevant portion of
asteroid and nuclear explosion generated
radiation spectra will differ from each
other. This could lead to non-identical
efficiencies in translating radiation energy
into burn injury. However, given the sparse
evidence basis of asteroid explosions and
few literature sources, the approach
presented here represents a best effort to
treat asteroid caused radiation vulnerability.
The burn probability as a function of radiant
exposure and explosion yield is given in
Figure 12.65 of the reference. While the
dependency of burn probability to radiant Figure 5: Burn degree distribution as a
exposure [J/m2] is obvious, its dependency function of radiation intensity based on
on explosion yield should be explained. data in (Glasstone and Dolan 1977)
The dependency on explosion yield is assuming the explosion signature of a
rooted in the observation that the process of 1Mton TNT yield nuclear device.
small yield explosions takes less time to
Aside from burn degree, the total
unfold than large yield explosions resulting
body surface area (TBSA) of a human that
in different energy flux rates. For smaller
is burned determines the expected
explosions, a given amount of radiant
mortality. In (American Burn Association
energy is delivered in a shorter time
2012), statistical analysis of 143199 burn
compared to a larger explosion and, thus,
victims in the United States were analysed
the radiation intensity differs with
for their mortality rate based on burned
explosion yield. Higher radiation intensity
TBSA. The reported numbers apply to
causes injuries more readily than low
persons who have been treated in medical
intensity radiation even though the same
facilities after the burn injury. This means
cumulative energy might be delivered in
that the burn injury itself could be treated
both cases. The reason for this behaviour is
adequately but also that possible
that the heated surface has more time to
subsequent medical complications
dissipate the incident radiation energy in a
(pneumonia, infection) that are linked
low intensity radiation case. Unlike nuclear
directly to the burn injury could be
explosive devices, meteoroids are not
addressed. Here, it shall be assumed that
optimized for explosion and it is thus
mortality rates are twice as high because
assumed that their explosion signature is
proper and timely treatment of burn injuries
more comparable to that of a large nuclear
is unlikely in the event of an asteroid impact
device because the explosion process takes
that will potentially affect a large region
a relatively long time. The data used to
and its medical infrastructure. Figure 6
build the vulnerability model correspond to
visualizes the data in Table 9 of (American
the results produced by a 1Mton TNT
Burn Association 2012) and shows the
equivalent yield nuclear device as shown in
mortality rate as a function of burned TBSA
Figure 12.65 in (Glasstone and Dolan
for treated and untreated victims.
1977).
The burn severity distribution is a
function of radiant exposure and the data in
the reference forms the basis for Figure 5
which shows the burn degree that can be
13
against a short lived energy burst of thermal
radiation and it is therefore assumed that
only one third of TBSA can be burned for
people standing outside. Figure 7 visualizes
the resulting TBSA curve as a function of
radiant exposure.
14
Figure 8: Mortality rate as a function of Figure 9: Mortality rate as a function of the
radiant exposure. full, applicable radiant exposure range.
1
() = (23)
1 + (+)
For the expected case, a maximum of
47% of the population is exposed.
Additionally, in the worst case scenario it
was assumed that the entire population is
outdoors (exposed) while in the best case
15
scenario the entire population is sheltered generally an unstable structure and is
and the corresponding coefficients are: similar to the crown-like shape that forms
in a water surface immediately after a
Table 4: Thermal radiation vulnerability droplet falls into it. The crown ring
coefficients. surrounds the impact point that forms a
bowl shaped depression and represents the
Case
crater bottom. A transient crater is not self-
Expected 0.47 5.623 106 7.316 105
supporting and collapses under the
influence of gravity to form the final crater
Best 0.25 5.623 106 7.316 105 shape. The transient crater diameter is
given in (Collins et al. 2005) with:
Worst 1.0 5.623 106 7.316 105
=
Figure 10 visualizes these 1
3
0.44 0.22 1 (24)
vulnerability models. 1.161 ( ) 0.78
0 0 sin 3
16
km2. Cratering vulnerability in a given grid given distance from the impact point and
cell was determined by calculating the determines vulnerability.
fraction of the crater area that covers this A literature review was conducted to
specific grid cell with respect to the grid cell find suitable data to support a seismic
area. Note that the impact point grid cell vulnerability model. Specifically, data were
might be covered completely by the crater needed to relate seismic shaking magnitude
but that cells that are located on the rim of at a given location to the mortality rate at
the crater are only partially covered and this location. However, typical earthquake
ARMORs algorithm accounts for such records report only peak intensity and total
situations. To this end, the final crater area losses and these data are too convoluted for
was assumed to be circular: usage here because they depend on
population density and affected area in the
2 location of the earthquake which are
= ( ) (26) typically not reported in the respective
2
publications (Norlund et al. 2009). The data
Seismic Shaking reports fatalities that occur in an area that
encompasses the entire earthquake region
The seismic shock is expressed in and relates this casualty figure to the peak
terms of the Gutenberg-Richter scale intensity shaking. However, not all
magnitude. It is assumed that a fraction of fatalities occur at the location of peak
104 of the impacting kinetic energy is shaking intensity (the epicentre) and some
transformed into seismic shaking (Schultz casualties are found at a distance away from
and Gault 1975). The Gutenberg-Richter the epicentre. Thus, it would be wrong to
magnitude energy relation provided the use these data because they attribute the
magnitude of the expected shock as: casualties of the entire earthquake region to
the peak shaking intensity and would
= 0.67 log10 5.87 (27) produce an overestimation for a given
seismic intensity.
where is the impacting kinetic energy in A function is needed that provides the
Joules, and is the magnitude on the mortality rate with respect to local shaking
Richter scale. With increasing distance magnitude because mortality varies with
from the impact site, the force of the shocks distance from the epicentre. In other words,
decreases and (Collins et al. 2005) present an earthquake produces a high mortality
an empirical law that describes the effective rate close to the epicentre where seismic
magnitude at a distance from the shaking is severe and a lower mortality rate
impact site: at a distance from the epicentre because
seismic shaking attenuates with greater
5 |
distance. In (Wu et al. 2015), mortality rates
2.38 10 < 60 (28)
= { 4.8 106 1.1644 | 60 < 700 are provided as logistic functions with
1.66 log10 6.399 | 700
respect to seismic intensity based on
earthquake records in China and these
It should be noted that seismic shaking functions were validated against four severe
severity estimation (as is true for other earthquake events. It should be noted that
effect severity estimation) carries the reported mortality rates are equivalent
uncertainty in itself and could produce to the vulnerability rates that are of interest
casualty estimation errors beyond what is here because the mortality rates describe
captured in the best and worst cases for the the observed number of casualties for a
vulnerability models. given seismic shaking intensity. The
The effective magnitude can be vulnerability logistic function that best
related to the expected destruction at the fitted the validation data (mean estimation
17
error of 12%) with seismic intensity
(Modified Mercalli Intensity (US
Geological Survey 2015)) as free parameter
is:
1
= (29)
0.01 + 2.691 106 0.170
18
= 0 (33)
19
where is the ocean depth at the impact numerical methods and an analytical
site. In accordance with (Korycansky and solution is presented instead.
Lynett 2007), deep water has been defined The magnification of a tsunami wave
as any depth >800 m. This equation adopts as it approaches the shore is called run-up
a 1 wave distance attenuation and this parameter ultimately determines
relationship which seems to match how much of the coastal area is inundated
observations made for the 2004 Sumatra and, thus, threatened. Much uncertainty in
tsunami (Weiss et al. 2006; Fritz et al. the field of tsunami modelling arises around
2007). The transient cavity diameter if tsunami waves break at the continental
may be calculated with equation (24) using shelf in what is coined the "Van Dorn"
the constant factor 1.365 instead of 1.161 effect (Van Dorn et al. 1968). If the Van
(Collins et al. 2005). These relationships Dorn effect is real, much of the tsunami
were used in ARMOR to propagate the energy is dissipated at the continental shelf
initial wave height through deep water to and the run-up of the wave would be greatly
the point where the bathymetry profile rose diminished (Melosh 2003). To address this
to less than 800 m water depth. If the impact issue, run-up is treated in (Korycansky and
point was already located in shallow water, Lynett 2007) utilizing the concept of the
equation (37) was used to calculate initial "Irribaren number" . The Irribaren number
wave height outside the transient crater for concept is promising because it shows good
subsequent run-up estimation outlined agreement with various modelling
below. conditions, such as varying bathymetry
To determine the coastline affected profiles and wave heights as demonstrated
by an impact generated tsunami, a fast ray by comparison with numerical results and
tracing algorithm was employed which wave-tank experiments (Korycansky and
connects each coastline point inside the Lynett 2007). The Irribaren number is
reach of the tsunami wave (maximum reach defined as (Hunt 1959; Battjes 1974):
set to roughly 13,000 km) with the impact
1
point. The algorithm traces the coastline 2800 2 (38)
map along the connecting rays to detect if = ( )
an island or another coastline does not
obstruct the targeted coastline point. The where is bathymetry slope from deep
algorithm has a priori knowledge of how far water to the coast and is wavelength. The
the closest coastline is away from the wave amplitude at a sea depth of 800 m is
impact point and only starts the 800 and this value is obtained using
computationally intensive ray tracing from Equation (37). The Irribaren number
this distance outwards until it intersects a approach suggests that run-up scales as
coastline. Consequently, the algorithm is the product of wave amplitude 800 and
able to extract such coastlines that are Irribaren number :
directly "visible" from the impact site and
also the corresponding points along the = 2 800 (39)
connecting rays where deep water
transitions to shallow water (800 m). In the numerical asteroid ocean impact
Actual wave propagation across the analysis (Gisler et al. 2011) the
ocean is more complex because refraction approximate, empirical relationship, that
of waves around land features, and wavelength, , is double the transient
subsequent interference of the wave with crater diameter, , can be found:
itself may occur. However, the aim of this
work was to provide fast models, which did = 2 (40)
not allow for computationally expensive
20
Combining equations (38), (39) and (40)
yields:
1
800 2 (41)
= 2800 ( )
Figure 13: Tsunami run-in as it relates to
Run-up calculation requires slope run-up U and beach slope s. Red terrain in
estimation of the bathymetry profile leading one pixel is inundated and green terrain is
up to the exposed shoreline. Suitable, safe. Note, that tsunami wave height and
global bathymetry (or, more correctly, run-up are not the same.
hypsography when it also considers land
Assuming a sufficiently large wave,
topography) data are available from
more than one map pixel may be inundated.
(Patterson and US National Park Service
In contrast to the linear model portrayed in
2015). Slope is calculated using the simple
Figure 13, multi-pixel inundation cannot be
"rise over run" definition given by:
calculated linearly, because, depending on
|800 | the local topography, the terrain slope
= (42) changes from pixel to pixel. ARMOR's
800 code accounts for this fact by recalculating
slope between map pixels and comparing
where 800 is the sea depth where deep the topography height with the run-up
water stops and 800 is the distance height to determine wave run-in and local
between the 800 point and the exposed pixel inundation. Local inundation is
shoreline in the direction of the equivalent with local run-up. Further, the
approaching wave. Note that a shallower code calculates the mean run-up height
depth than 800 m was used in case the for each map pixel, whether fully or
impact occurs in shallow water. The partially inundated, for subsequent
calculated run-up height was used to vulnerability estimation. In other words, if
estimate population vulnerability and > , then the entire pixel is inundated
inundation range. and mean run-up in this pixel is =
2. In the next map pixel, run-up is
Vulnerability Model
A tsunami can inundate coastal reduced by of the previous pixel and the
regions and, thus, harm the population new local medium run-up is, thus, =
living there. How much of the coastline is ( 1 ) 2. This formulation
inundated depends on the run-up height can further be expressed as:
1
and terrain slope . A steeper terrain slope
limits the extent to which water of a given = (43)
2
=0
run-up height can reach inland. Figure 13
visualizes this concept and in this figure, where counts the pixels from the shore
the red portion of one pixel (pixel length is (pixel 1) to the local pixel and 0 = 0.
) is inundated by a wave with run-up The above equation holds true for
while the green portion is beyond the reach
completely inundated pixels. The last pixel
of the wave because of terrain elevation
will generally only be partially inundated,
over the length of one pixel.
as shown in Figure 13, and, thus, the local
run-up is equal to the original run-up less
the terrain height of the previous pixels.
Finally, the mean local run-up of the last
21
pixel is half of the run-up height at the last with the case parameters in Table 6:
pixel:
Table 6: Tsunami vulnerability model case
1
=1
coefficients.
= (44)
2 Case
22
Table 7: Casualty estimates for airburst (50 m) and cratering (200 m) events over Berlin and
London. Impact angle is 45, impact speed is 20 km/s and asteroid density is 3100 kg/m3.
Berlin London
Size [m] 50 200 50 200
Type Airburst Cratering Airburst Cratering
Wind [%] 85.5 48.6 84.6 49.1
Pressure [%] 0.0 24.9 0.0 23.4
Thermal [%] 14.5 23.6 15.4 24.3
Seismic [%] 0.0 0.1 0.0 0.1
Cratering [%] 0.0 0.4 0.0 0.8
Ejecta [%] 0.0 2.4 0.0 2.2
Tsunami [%] 0.0 0.0 0.0 0.0
Total expected [-] 1,180,450 3,511,397 2,818,507 8,761,812
Best [% variation] -34.1 -4.0 -31.6 -6.2
Worst [% variation] 35.5 4.8 32.3 7.7
The second case study analysed the contributions are colour coded to help
offshore region of Rio de Janeiro. A 200 m interpretation of the results. In addition, the
sized object was simulated to impact water depth at the impact locations is given.
onshore (Latitude: 22.98S, Longitude: Rio de Janeiro was chosen because it is
43.22W) and at eight distances offshore exposed to tsunamis and because the
moving directly South from the onshore bathymetry profile is relatively benign with
location. The same impact parameters were a continental shelf reaching about 120 km
used as in case study one. The total casualty offshore and a subsequent transition into
estimates as well as the damage deep ocean.
distributions per impact effect are provided
in Table 8. Similarly to Table 7, the damage
23
Table 8: Impact scenarios near Rio de Janeiro providing casualty estimates (expected) as well
as damage contribution per impact effect in terms of casualty numbers and percentage of total
casualties (percentages are colour coded). Asteroid size is 200 m, impact angle is 45, impact
speed is 20 km/s and asteroid density is 3100 kg/m3.
24
percentages as well as overall casualty ARMORs tsunami code is sensitive
numbers, and this indirect mechanism has to sea depth for initial wave height
not been considered for pressure effects calculation and the coastal region of South
(but it has been for wind which occurs Americas East exhibits a continental shelf
together with pressure). Thermal radiation that reaches about 120 km offshore close to
causes significant harm and exhibits a Rio de Janeiro. A continental shelf typically
tendency to contribute larger proportions of features a gentle, constant bathymetry slope
the total casualties for higher energetic of about 0.1 ( 0.0017 rad) (The Editors
events. The relative contribution of of Encyclopdia Britannica 2016) such that
aerodynamic and thermal effects for the the sea depth at a given distance from
airbursts agree well with the observations the shore can be calculated as:
made after the comparatively sized
Tunguska event in 1908. There, over 2,000 = (46)
km2 of forest were flattened due to
aerodynamic forces and an area of 300 km2 where is given in radians. Substituting this
was charred (Boslough and Crawford 2008; relationship into equation (37) yields that
Nemtchinov et al. 1994). The two airburst the coastal height of a wave which is
events showed higher variability in terms of generated by an impact on the shallow
total casualty outcomes than the continental shelf is independent of distance
corresponding surface impacting events. to the shore and only depends on initial
Airburst events are generally less severe impact energy (reflected in transient crater
than surface impacts and this places diameter ) and slope:
airbursts mainly in the transitioning regime
(where vulnerability function slope is
= (47)
steepest) of the vulnerability models. Best 2
and worst case vulnerability models show
For a 200 m impactor such as was utilized
highest variability in this regime and this
here, a transient water cavity of about 5 km
fact is reflected in the variability of the
in diameter was produced resulting in a
results.
wave height of only 4.3 m at the shore
The second case study (Table 8)
according to equation (47). Impacts
focused on the contribution of tsunamis. As
elsewhere on the shelf will produce similar
reference, a cratering land impact was
wave heights at the shore as long as the
produced at the shore and the impact
assumptions about constant slope and
location was subsequently moved offshore
shallow water impacts (defined by <
resulting in varying impact effect
0.14 in equation (37) corresponding to
contributions. The land and near-coastal
about 750m for a 5km transient water
impactors (at 10 km) showed a similar
cavity) hold true. The underlying
damage distribution as the land impactors
mechanism is that while sea depth increases
in case study one. However, as the impact
seawards, which allows for the generation
point recedes farther from the inhabited
of larger waves, these waves have to travel
land, the varying effect ranges become
farther to the shore attenuating with
apparent. As already observed in case study
distance and these opposing mechanisms
one, the vulnerability to pressure decreases
result in a constant wave height at the shore.
quicker relative to wind vulnerability.
Consequently, the continental shelf,
Similarly, thermal radiation has a shorter
through its shallow water and gentle,
reach compared to the wind blast which
constant slope, serves a protective function
only loses lethality at a distance of 130-150
which limits the tsunami wave height that
km. Perhaps surprising is the small
can reach the shore from impacts that occur
contribution of tsunamis to loss in the near-
anywhere on the shelf. This protective
coastal region.
effect is different from the Van Dorn effect
25
which describes wave breaking at the edge coastline. By reaching farther land inwards,
of the continental shelf (Korycansky and aerothermal effects are, thus, able to cover
Lynett 2005; Van Dorn et al. 1968). It larger population numbers overall. In
should be noted that ARMORs code addition to the limited initial wave height
calculates wave height at the shore due to shallow sea depth for near coastal
according to the Irribaren approach which impactors, this observation contributes to
also accounts for run-up wave height due to the finding of low tsunami casualty
shoaling. However, the analytical example numbers relative to those of aerothermal
(using equation (37)) illustrated more effects.
suitably the underlying mechanisms of
wave height attenuation over distance and CONCLUSIONS
sea depth limited initial wave height to
explain the relatively small contribution of This work derives and presents new
tsunami damage in Table 8 at distances vulnerability models that can be used by the
corresponding to the continental shelf. The wider community to assess the asteroid
relative damage contribution of tsunamis impact hazard in relation to human
only started to increase significantly as the populations.
impacts reached the edge of the continental The vulnerability models facilitate
shelf and the correspondingly deeper water. the calculation of asteroid impact
This was the point where aerodynamic and consequences in terms of loss of human life
tsunami lethality reached equal magnitude under consideration of the environmental
(at about 120-125 km) and tsunamis effects produced in an impact event. The
became most harmful in deeper sea. algorithm that is employed by ARMOR to
Notably, the possibility for larger waves in calculate whether an asteroid experiences
deeper water even allowed for an increase an airburst or surface impact was presented.
in total casualties for impacts at distances of Additionally, the algorithm determines
120-130 km from the shore despite the effect severity for seven impact effects:
continuous weakening of aerodynamic strong winds, overpressure shockwave,
effects at those distances. thermal radiation, seismic shaking, ejecta
The results in Table 8 illustrate the deposition, cratering and tsunamis. A
significantly farther reach of tsunamis comprehensive, analytical model for
compared to those of other impact effects. tsunami propagation was presented that is
Casualty numbers for an impactor at 100 amenable to varying bathymetry profiles.
km distance to the shore were 99.6% To enable casualty calculation,
smaller compared to those produced by a vulnerability models for all impact effects
shore impactor because most effects were derived that connect effect severity to
(except tsunami) did not reach the shore lethality for human populations.
from this distance. By comparison, casualty Two case studies examined casualty
numbers only decreased by about 63% over estimation utilizing the new vulnerability
the next 200 km (between 100 km and 300 models. In these case studies, it was found
km) where tsunamis were the only relevant that aerothermal impact effects were most
hazard. The results illustrate why tsunamis harmful with the exception of deep water
pose a significant threat by deep water impacts, where tsunamis became the
impacts through their far reach. dominant hazard. Analysing casualty
While tsunamis generally affect far outcomes for ocean impactors in the Rio de
longer coastlines than could be covered by Janeiro area revealed the protective
the other impact effects (especially function of continental shelfs against the
aerothermal), they usually do not penetrate danger of tsunamis by impactors on the
land inwards as far as other effects if the shelf. It also illustrated that aerothermal
impact occurs suitably close to the effects, through their farther reach land
26
inwards, can harm a larger population than cretaceous-tertiary extinction. Science 208:
tsunamis in near-coastal impacts even 10951108.
http://www.sciencemag.org/content/208/4448
though tsunamis may cover longer /1095.
coastlines. American Burn Association. 2012. National Burn
Furthermore, the case studies found Repository Report of Data From 2002-2011,
seismic shaking, cratering and ejecta Chicago, IL.
deposition to contribute only negligibly to Battjes J. A. 1974. Surf similarity. In 14th
international conference on coastal
overall casualty numbers. engineering. pp. 466480.
Casualty numbers of an on-shore Berryman K. 2005. Review of Tsunami Hazard and
impact decreased significantly when the Risk in New Zealand Review of Tsunami
impact location moved few tens of Hazard and Risk in New Zealand, Lower
kilometres off-shore due to the fast effect Hutt, New Zealand.
Binzel R. P. 2000. The Torino Impact Hazard
severity attenuation of the dominant Scale. Planetary and Space Science 48: 297
aerothermal impact effects with distance. 303.
The calculation of impact Borovika J., and Kalenda P. 2003. The Morvka
consequences facilitates asteroid risk meteorite fall: 4. Meteoroid dynamics and
estimation and the concept of risk as well as fragmentation in the atmosphere. Meteoritics
& Planetary Science 38: 10231043.
its applicability to the asteroid impact http://onlinelibrary.wiley.com/doi/10.1111/j.1
hazard were introduced. The results 945-
produced in the two case studies represent 5100.2003.tb00296.x/abstract%5Cnhttp://doi.
casualty estimates for a given impact threat wiley.com/10.1111/j.1945-
and location. Casualty estimates correspond 5100.2003.tb00296.x.
Boslough M. 2013. Airburst warning and response.
to the product of exposure and Acta Astronautica 103: 370375.
vulnerability () in risk equation (1). http://linkinghub.elsevier.com/retrieve/pii/S0
Consequently, the new vulnerability 094576513003500 (Accessed June 9, 2014).
models enable asteroid risk estimation Boslough M., and Crawford D. a. 2008. Low-
when the impact probability (third factor altitude airbursts and the impact threat.
International Journal of Impact Engineering
in Equation (1)) is taken into account for 35: 14411448.
example in the form of the spatial impact http://linkinghub.elsevier.com/retrieve/pii/S0
probability distribution (Rumpf et al. 734743X08001784 (Accessed October 30,
2016b). 2013).
Britt D. 2014. The Physical Properties of Near
Earth Asteroids. In Asteroid Grand
ACKNOWLEDGMENTS Challenge: Virtual Seminar Series. pp. 130.
Chesley S. R., Chodas P. W., Milani A., Valsecchi
The work is supported by the Marie G. B., and Yeomans D. K. 2002. Quantifying
Curie Initial Training Network Stardust, the Risk Posed by Potential Earth Impacts.
FP7-PEOPLE-2012-ITN, Grant Agreement Icarus 159: 423432.
317185. The authors acknowledge the use http://linkinghub.elsevier.com/retrieve/doi/10
.1006/icar.2002.6910 (Accessed October 30,
of the IRIDIS High Performance 2013).
Computing Facility at the University of Chyba C. F., Thomas P. J., and Zahnle K. J. 1993.
Southampton. James G. LaDue from the The 1908 Tunguska explosion: atmospheric
National Oceanic and Atmospheric disruption of a stony asteroid. Nature 361.
Administration (NOAA) supported this http://www.nature.com/nature/journal/v361/n
6407/pdf/361040a0.pdf.
work with advice on tornado vulnerability. CIESIN, Columbia University, United Nations
We extend our sincere thanks to Gareth Food and Agriculture Programme - FAO, and
Collins who has served as thorough CIAT. 2005. Gridded Population of the
reviewer for this paper. World, Version 3 (GPWv3): Population
Count Grid, Future Estimates, Palisades, NY:
REFERENCES NASA Socioeconomic Data and Applications
Alvarez L. W., Alvarez W., Asaro F., and Michel Center (SEDAC)
H. V. 1980. Extraterrestrial cause for the http://sedac.ciesin.columbia.edu/data/set/gpw
27
-v3-population-density-future-estimates. effect revisited. Geophysical Research
Collins G. S., Lynch E., McAdam R., and Davison Letters 32: L10608.
T. 2016. A numerical assessment of simple http://www.agu.org/pubs/crossref/2005/2004
airblast models of impact airbursts. GL021918.shtml.
Meteoritics and Planetary Science submitted. Korycansky D. G., and Lynett P. J. 2007. Run-up
Collins G. S., Melosh H. J., and Marcus R. a. 2005. from impact tsunami. Geophysical Journal
Earth Impact Effects Program: A Web-based International 170: 10761088.
computer program for calculating the Kuligowski E. D., Lombardo F. T., Phan L. T.,
regional environmental consequences of a Levitan M. L., and Jorgensen D. P. 2013.
meteoroid impact on Earth. Meteoritics & Draft final report, National Institute of
Planetary Science 40: 817840. Standards and Technology (NIST) Technical
http://doi.wiley.com/10.1111/j.1945- investigation of the May 22, 2011 tornado in
5100.2005.tb00157.x. Joplin, Missouri, USA
Diffey B. L. 2011. An overview analysis of the http://www.nist.gov/customcf/get_pdf.cfm?p
time people spend outdoors. British Journal ub_id=915628%5Cnhttp://nvlpubs.nist.gov/n
of Dermatology 164: 848854. istpubs/NCSTAR/NIST.NCSTAR.3.pdf.
http://www.ncbi.nlm.nih.gov/pubmed/211289 Melosh H. J. 2003. IMPACT-GENERATED
11. TSUNAMIS: AN OVER-RATED HAZARD.
Van Dorn W. G., Le Mehaute B., and Hwang L.-S. In Lunar and Planetary Science.
1968. Handbook of Explosion-Generated NASA. 2014. NASA NEO Impact Risk.
Water Waves, Washington D.C.: Office of http://neo.jpl.nasa.gov/risk/.
Naval Research National Research Council et al. 2010. Defending
http://oai.dtic.mil/oai/oai?verb=getRecord&a Planet Earth: Near-Earth Object Surveys
mp;metadataPrefix=html&identifier=A and Hazard Mitigation Strategies,
DB041508. Washington D.C.
French J., Ing R., Von Allmen S., and Wood R. Nemtchinov I. V., Popova O. P., Shuvalov V. V.,
1983. Mortality from flash floods: a review and Svetsov V. V. 1994. Radiation emitted
of national weather service reports, 1969-81. during the flight of asteroids and comets
Public health reports 98: 584588. through the atmosphere. Planetary and Space
http://www.pubmedcentral.nih.gov/articleren Science 42: 491506.
der.fcgi?artid=1424497&tool=pmcentrez&re Nemtchinov I. V., Shuvalov V. V., Artemieva N.
ndertype=abstract. a., Ivanov B. A., Kosarev I. B., and
Fritz H. M. et al. 2007. Extreme runup from the 17 Trubetskaya I. a. 1998. Light flashes caused
July 2006 Java tsunami. Geophysical by meteroid impacts on the lunar surface.
Research Letters 34: 15. Solar System Research 99114.
Gisler G., Weaver R., and Gittings M. 2011. NOAA. 2015. NOAAs National Weather Service,
Calculations of asteroid impacts into deep Storm Prediction Center, Killer Tornados by
and shallow water. Pure and Applied Year. Data on Webpage.
Geophysics 168: 11871198. http://www.spc.noaa.gov/climo/historical.htm
Glasstone S., and Dolan P. J. 1977. The Effects of l (Accessed April 28, 2015).
Nuclear Weapons, Third Edit. USA: United Norlund C. 2013. Evacuation Simulation
States Department of Defense & Energy Modelling in the event of a Near Earth Object
Research and Development Administration impact. University of Southampton.
https://www.fourmilab.ch/etexts/www/effects Norlund C., Lewis H. G., and Atkinson P. M. 2009.
/. Evacuation Simulation Modelling in the
Hunt I. A. J. 1959. Design of Seawalls and Event of a Near Earth Object impact. In First
Breakwaters. Journal of the Waterways and International Conference on Evacuation
Harbors Division 85: 123152. Modeling and Management. Delft, The
Klepeis N. E., Nelson W. C., Ott W. R., Robinson Netherlands. p. 10.
J. P., Tsang A. M., Switzer P., Behar J. V., Ortiz J., Sada P., Bellot Rubio L. R., Aceituno F.,
Hern S. C., and Engelmann W. H. 2001. The Aceituno J., Gutierrez P., and Thiele U. 2000.
National Human Activity Pattern Survey Optical detection of meteoroidal impacts on
(NHAPS): a resource for assessing exposure the Moon. Nature 405: 9213.
to environmental pollutants. Journal of http://www.ncbi.nlm.nih.gov/pubmed/108795
exposure analysis and environmental 26.
epidemiology 11: 231252. Patterson T., and US National Park Service. 2015.
http://www.nature.com/jes/journal/v11/n3/pd CleanTOPO2. CleanTOPO2 - 3D version.
f/7500165a.pdf. http://www.shadedrelief.com/cleantopo2/
Korycansky D. G., and Lynett P. J. 2005. Offshore (Accessed October 26, 2015).
breaking of impact tsunami: The Van Dorn Paul B. K., and Stimers M. 2014. Spatial analyses
28
of the 2011 Joplin tornado mortality: Deaths Earthquake. Website.
by interpolated damage zones and location of http://pubs.usgs.gov/gip/earthq4/severitygip.h
victims. Weather Climate and Society 6: tml (Accessed August 24, 2015).
161174. Ward S. N., and Asphaug E. 2000. Asteroid Impact
Pomonis A., Spence R., and Baxter P. 1999. Risk Tsunami: A Probabilistic Hazard
assessment of residential buildings for an Assessment. Icarus 145: 6478.
eruption of Furnas Volcano, Sao Miguel, the Weiss R., Wnnemann K., and Bahlburg H. 2006.
Azores. Journal of Volcanology and Numerical modelling of generation,
Geothermal Research 92: 107131. propagation and run-up of tsunamis caused
Popova O. P. et al. 2013. Chelyabinsk airburst, by oceanic impacts: Model strategy and
damage assessment, meteorite recovery, and technical solutions. Geophysical Journal
characterization. Science (New York, N.Y.) International 167: 7788.
342: 106973. Wetherill G. W. 1990. Formation of the Earth.
http://www.ncbi.nlm.nih.gov/pubmed/242008 Annual Review of Earth and Planetary
13 (Accessed June 23, 2014). Sciences 18: 205256.
Rumpf C. M., Lewis H. G., and Atkinson P. M. http://adsabs.harvard.edu/full/1990AREPS..1
2016a. On the influence of impact effect 8..205W.
modelling for global asteroid impact risk Wind Science and Engineering Center. 2006. A
distribution. Acta Astronautica 123: 165170. Recommendation for an Enhanced Fujita
http://linkinghub.elsevier.com/retrieve/pii/S0 Scale (EF-Scale), Lubbock, Texas
094576515302988. http://www.depts.ttu.edu/nwi/Pubs/FScale/EF
Rumpf C. M., Lewis H. G., and Atkinson P. M. Scale.pdf.
2016b. The global impact distribution of Wu S., Jin J., and Pan T. 2015. Empirical seismic
Near-Earth objects. Icarus 265: 209217. vulnerability curve for mortality: case study
http://linkinghub.elsevier.com/retrieve/pii/S0 of China. Natural Hazards 77: 645662.
019103515004996. http://link.springer.com/10.1007/s11069-015-
Ryder G. 2002. Mass flux in the ancient Earth- 1613-1.
Moon system and benign implications for the Wnnemann K., Collins G. S., and Weiss R. 2010.
origin of life on Earth. Journal of Impact of a cosmic body into EarthS Ocean
Geophysical Research 107. and the generation of large tsunami waves:
http://onlinelibrary.wiley.com/doi/10.1029/20 Insight from numerical modeling. Reviews of
01JE001583/full. Geophysics 48: 126.
Schultz P. H., and Gault D. E. 1975. Seismic Wurman J., Alexander C., Robinson P., and
Effects From Major Basin Formations on the Richardson Y. 2007. Low-level winds in
Moon and Mercury. The Moon 12: 159177. tornadoes and potential catastrophic tornado
Simmons K. M., and Sutter D. 2005. WSR-88D impacts in urban areas. Bulletin of the
Radar, Tornado Warnings, and Tornado American Meteorological Society 88: 3146.
Casualties. American Meteorological Society Zeldovich Y. B., and Raizer Y. P. 1966. Physics of
20: 301311. Shock Waves and High-Temperature
The Editors of Encyclopdia Britannica. 2016. Hydrodynamic Phenomena, New York:
Continental shelf. Encyclopaedia Britannica. Academic Press.
https://www.britannica.com/science/continent
al-shelf (Accessed August 1, 2016).
Trades Union Congress. 2012. Average
Commuting Times in UK. Labour Force
Survey. https://www.tuc.org.uk/workplace-
issues/work-life-balance/men-their-early-40s-
have-longest-commutes (Accessed July 21,
2015).
U.S. Census Bureau. 2011. Average Commute
Times in the United States. U.S. Census
American Community Survey 5-year
estimates. http://project.wnyc.org/commute-
times-us/embed.html#5.00/36.114/-92.526
(Accessed July 21, 2015).
Universita Di Pisa, and European Space Agency.
2014. NEODyS-2 Risk Page.
http://newton.dm.unipi.it/neodys/index.php?p
c=4.1.
US Geological Survey. 2015. The Severity of an
29