1. The low probability phase - takes effect with any object detected
that has an impact probability below 0 on the Palermo Scale. Most
objects listed on the "impact risk pages" managed by NEODyS and JPL fall
into this category. No specific public information is required for such
an event.

2. The moderate probability phase - gets underway with the detection of
an object that has an impact probability above 0 on the Palermo Scale.
It would be sensible to post clarifying information on the various
internet "risk pages." However, at this stage it would be important to
emphasise that the impact risk will be removed, in all likelihood, as a
result of additional observational data. Accordingly, there is no need
for any official press release during this phase.


Benny Peiser, Liverpool John Moores University, Faculty of Science,
Liverpool L3 2ET, United Kingdom,

Paper presented at the International Workshop On Managing Global-Scale
Disasters, 12 April 2002, Irvine, California


Disaster warning systems have become essential social mechanisms in the
forecast, detection and mitigation of natural disasters. People exposed
to natural hazards are increasingly relying on the effectiveness of
warning systems. This dependence has been acknowledged by the United
Nations which has called for improved early warning systems (Parker

Disaster warning systems are social and technological organisations that
detect impending disasters, estimate and analyse their potential
effects, convey that information to those in danger and those that can
mitigate the catastrophe. Warning systems are complex structures because
they connect many organisations, hubs and institutions, ranging from
research and science labs, to engineering, technology, government
agencies, news media and the public (Sorensen, 2000). "People at risk
from disasters, whether natural or human in origin, can take actions
that save lives, reduce losses, speed response, and reduce human
suffering when they receive accurate warnings in a timely manner"
(Effective Disaster Warnings, 2000).

Disaster warning systems are most effective for natural catastrophes
that develop gradually and relatively slowly, such as floods or tropical
cyclones. In 1991, for example, 600,000 people in the Indian state of
Andhra Pradesh were evacuated in advance of a tropical cyclone, thus
minimising the number of fatalities to just over a thousand. 13 years
earlier, in comparison, over 10,000 people were killed in a similar
cyclone that transpired without any warning (Parker 1999). Also in 1991,
350,000 people were moved from the vicinity around Mount Pinatubo
(Philippines) in anticipation of a volcanic eruption, because of
appropriate warning (Davis et al. 1998). The decline in tornado-related
deaths since 1950 has been attributed to improvements in tornado-warning
systems (Noji 1997). Yet, hazard warnings are effective only if they are
coherent, correct and result in appropriate action.

The main principle that has emerged in warning research is the
realisation that an "integrated warning system" will maximise public
protection. "Integration refers to a melding of scientific monitoring
and detection with an emergency organisation that utilizes warning
technologies coupled with social design factors to rapidly issue an
alert and notification of the public at risk. Thus, warning systems must
be considered as having scientific, managerial, technological, and
social components that are linked by a variety of communication
processes" (Sorensen, 2000). In this paper, I examine the strengths and
weaknesses of the current impact warning system. In addition, a number
of impact emergency simulations are explored in order to identify the
different categories of impact warnings and alerts most accommodating
and effective within a coherent impact warning system.

The NEO Impact Warning System

During the last twenty years, technological and computational advances
have significantly improved the rate of detection of potentially
hazardous near Earth objects (NEO). This progress can be measured
against an intensification of NEO discoveries, improved telescopic
capability and incessant evolution of computer technologies. Equally
important has been the expansion of advanced computational systems at
the Minor Planet Center (MPC) and the establishment of impact
probability programmes at Pisa University and NASA's NEO Program Office
at the Jet Propulsion Laboratory (JPL). These developments have greatly
augmented predictive capabilities and thus expanded impact warning times

NASA's current "Spaceguard" objective is to discover 90% of the near
Earth asteroids (NEAs) that are larger than 1 km within the next six
years. It is thought that about half of the estimated 1000-1200 NEAs in
this category have been discovered to date. In line with incessant
technological progress during the last 20 years, current advances in the
discovery rate are due to constant expansion in the detection limits of
the major US search programmes. Both telescopic and computational
amplification have resulted in much improved detection capability which
allow the discovery of NEAs with fainter magnitude. As the detection
rate of NEAs improves, so does the impact warning system which has
increased the potential length of warning times accordingly.

The advances of the impact warning system are in line with similar
progress in other monitoring systems where, as a result of improved
sensoring and advanced technologies, natural hazard warning times have
been lengthened greatly (Sorensen, 2000). However, in contrast to most
other, more frequent natural disasters (such as earthquakes, volcanic
eruptions, tropical storms, tsunami, etc.), we have very little
empirical knowledge, let alone experience of NEO impacts and their
secondary environmental and societal effects. It thus remains difficult
to determine and test the effectiveness of the embryonic impact warning
system. At some unknown time in the future the day will come that a NEO
is found to be on a collision course with our planet. In the absence of
any preceding experience, we will be confronted with a unique and
unprecedented crisis situation. In order to be best prepared for such an
emergency, disaster management strategies and arrangements need to be
carefully planned and developed well in advance. At the same time, the
quandary of how to inform and communicate with an apprehensive public
has to be addressed in order to avoid misunderstandings or blunders.

There has been an ongoing debate about the problem of how best to
publish or announce the discovery of potentially hazardous asteroids
(PHAs). Ever since the debatable announcement of a small but non-zero
impact possibility of asteroid 1997 XF11 back in March 1998, the problem
of how to handle these exceptional asteroids that are, temporarily, on a
possible collision course with Earth has troubled the NEO research
community (Marsden, 1999; Binzel 2000; Chapman et al., 2001). The rapid
increase in NEO discoveries in recent years has inevitably led to
significant numbers of PHAs with non-zero impact probabilities. The vast
majority of these "potential impactors", however, is no longer
exceptional due to their minuscule chances of actual impact (in the next
century or so). The calculations of their orbital dynamics are now made
routinely by researchers in Europe and the US, and particularly
interesting objects are regularly monitored.
Despite many technical and computational advances in the impact risk
assessment process, no such progress has been made pertaining to
proposals for improved impact disaster management procedures such as
"notifying the public and relevant officials/agencies about an impact
prediction, and putting in place (in advance of such predictions)
procedures for coordination among relevant agencies and countries"
(Chapman et al. 2001).

As long as we are dealing with temporary low-probability impact alerts
that will most likely be eliminated by future observations, the absence
of a coherent impact management system does not pose an immediate
problem. However, once faced with a more significant impact risk, the
accustomed scientifically formulated announcements by NEO astronomers
may no longer be appropriate as a warning message for the general
public. For a start, it regularly includes technical and probabilistic
information which is known to be confusing to non-scientists. Since NEO
astronomers are generally inexperienced and untrained in disaster
management (in particular on the subject of secondary environmental,
social and economic effects of impacts), their scientific impact threat
announcement most likely will not include plans for adaptive actions,
either for the general public or for civil defence organisations. "This
raises a fundamental question concerning the formulation of the public
warning: To whom should scientists transmit their scientific warnings?
While this may appear to be a simple question, it may not be easy to
answer. An adequate answer would have to include both jurisdictional
considerations as well as technical skills" (Nigg, 1995).

Notwithstanding these shortcomings, it is guaranteed that one day
astronomers will detect a near Earth object that is on a collision
course with Earth. This could happen tomorrow or it could occur in 100
years time. It could be a small asteroid or medium-sized comet, or it
could be a larger object. Happily, chances are extremely small that this
will happen soon. Nonetheless, such an event will transpire one day. And
when it happens, it will be unprecedented. Since we have no experience
with such an emergency, the first predicted impact disaster crisis will
confront us with considerable social, technological and managerial
challenges. Even though the first predicted impact event will intrude
upon us all of a sudden, it is prudent to anticipate such an emergency
so that it will not find us unprepared or ill-equipped.

With the exception of science fiction, there have been very few attempts
to ponder the management and public handling of an impact emergency in
the event of a predicted and confirmed NEO impact. I deliberately use
both terms, prediction and confirmation, in the context of impact
disaster management because the words denote two entirely different
phases in an impact crisis situation. The current lack of a clear
understanding of how an impact crisis may eventually build up and travel
as well as ambiguous linguistic terminology may contribute to
misinterpretations of and incoherent operations during an impact crisis
situation. Ultimately, malfunctions in the warning process can manifest
themselves as deficient alerts, ineffective responses, failure of timely
information of people at risk or the inability to mitigate avoidable
consequences of impact (Alexander 1993). In order to improve the current
impact warning system and to prepare the dissemination of appropriate
impact warnings, it is essential to understand the complexities of how
an impact crisis situation would unfold.

Impact Warnings & Impact Crisis in Science Fiction

In the event of a confirmed impact risk, we will be faced with an
unprecedented situation that lacks any historical prototype. Since we
are short of any experience of an impact emergency situation, some
people may resort to the only impact disaster scenarios they know, i.e.
those described by science fiction. Even as most science fiction impact
fantasies conclude with a moderately happy ending, the science and
astronomy they portray is often flawed and thus misleading if not
adverse to attempts at reassuring an anxious public.

Take, for example, the impact disaster movie Judgement Day. In Judgement
Day, a large asteroid is expected to crash into Earth in a matter of
days - an impact that would cause global devastation. US government
officials are desperately searching for David Corbett, the architect of
a missile defence system, the only defence technology that could
notionally prevent the disaster. However, the government is not the only
group trying to locate Corbett however. A militant apocalyptic cult
leader believes that Judgment Day is about to happen and tries to do
anything to keep Corbett from stopping the asteroid. While this may look
like a good Hollywood plot, it is highly unlikely that an apocalyptic
cult or movement would have any significant effect on planetary
protection actions taken.

In Fire in the Sky, a US astronomer spots a small comet and computes
that it will hit Arizona in less than a month. The president of the USA
instructs the Air Force to launch a space mission, armed with nuclear
explosives, with the purpose of deflecting the comet shortly before
impact. In view of the fact that the Governor of Arizona decides to keep
the menace to his populace secret, the astronomer goes public and
announces the impending danger. "Not until 72 hours before impact, when
both of the nuclear weapons miss (firing a second too late) does the
city (and the Governor) finally realize the magnitude of their problem.
The National Guard is called in, martial law declared, and civil defense
authorities try to direct an evacuation. Strangely, they cannot seem to
pull this off in 72 hours, as massive gridlock blocks the highways. The
comet arrives right on schedule, completely destroying Phoenix and (by
accident) killing both the hero astronomer and the evil governor"
(Morrison 1998).

In Deep Impact, the issue of governmental conspiracy and cover-up also
plays an important role. Here, the world is kept in the dark about an
approaching comet for almost a year not just by the American and Russian
administrations - but also by the scientific community. One only wonders
why the news is broken in the first place (other than for dramatic
reasons) before the Messiah mission lifts off for it ineffective
mitigation task.

In Asteroids, astronomers discover that two asteroids are on collision
course with Earth and will impact in less than a week. One of the
asteroids is calculated to hit Kansas City, which is evacuated in the 72
hours before impact. Fortunately, the first asteroid, Helios, breaks up
during atmospheric entry, and only one fragment (apparently of meter
size) hits the surface. However, this fragment strikes the face of a
high concrete dam outside Kansas City, the dam ruptures, and the city is
flooded. The astronomers point out that Eros, coming behind, is large
enough to cause a global disaster. Accordingly, the military is called
upon to prevent it. Under command of the Air Force Space Command at
Cheyenne Mountain, Colorado, two experimental high-power
aircraft-mounted lasers are used to strike Eros, disrupting it. Some
hours later, however, the asteroid reappears as a swarm, with hundreds
of fragments still on collision course, including one large  (hundred
meter, perhaps) fragment headed for Dallas.

Two main issues can be identified here as potential problems: Firstly,
the question of scientific credibility and openness in the event of an
emergency; and secondly, the problem of optimism vs. despair is
approximating the odds that mitigation will actually work.

The concern regarding open information policies will, in all likelihood,
be the easiest to handle since scientists from around the world will
largely confirm the impact date calculations that are based on generally
available data. However, there may be less consensus in scientific
estimates about the secondary effects of the impact in question.

More importantly, though, are potential problems arising from
disagreeing estimates about the chances for a successful impact
mitigation mission. Here, the lack of any experimental data as well as
the still rather feeble results from much less complex ballistic missile
defence tests currently do not allow any credible public assurances by
the authorities. Morrison et al. (2002) argue that policies that
reassure the population and prevent panic should be considered. "A
relatively low-cost generic interception system could be built and
tested, then shelved. In the event of emergency, the system would enter
surge production, with industrial capacity commandeered from other
programs. In this manner, there would be a tailored response to the
threat, and operational flexibility would be enhanced. Salvoes could be
launched, and in many cases, shoot-look-shoot would be possible.
Uncertainty in threat composition could be countered by adding an
additional production margin." Whether or not such a generic
interception system will be adequate to successfully neutralize an
impending impact threat will ultimately depend on the extent of the
warning time and the technological progress civilisation will have made
whenever we face the first major impact crisis.

Impact Disaster Management

Garshnek et al (2000) address the issue of impact disaster management by
differentiating between three distinct disaster periods: 1.) a
pre-disaster phase, 2.) an acute disaster phase, and 3.) a post-disaster
(or rehabilitative) phase. The first of these three phases is split once
more into two sub-phases. Pre-disaster phase A is defined as "the
situation now, with no impact predicted", while pre-disaster phase B is
characterized as "the period from the prediction up to the impact." The
acute disaster phase and the post-disaster (rehabilitative) phase, on
the other hand, proceed an impact disaster and deal primarily with the
management of secondary social and environmental effects.

The characterization of the two pre-disaster phases is to some extent
mistaken. The first sub-phase (pre-disaster phase A = "no impact
predicted") no longer exists and should consequently be discarded. As of
April 8, 2002, there are almost 40 near Earth asteroids listed on JPL's
impact risk page ( which are predicted
(albeit with an extremely remote probability) to collide with earth in
the next 100 years.

The categorization of pre-disaster phase B (i.e. "the period from the
prediction up to the impact") is also rather deceptive. Predictions of
impact events (usually with minute impact probabilities) are made
regularly and are listed on a number of professional impact risk
websites (operated by the MPC, University of Pisa and JPL). None of
these predictions, however, have yet resulted in the confirmation of an
impact or an impact warning. Even so, one of the predicted impacts may,
one day, evolve into a confirmed impact. Evidently, there is a need for
clarity and accurateness in the impact warning system if we wish to
understand how an impact crisis may evolve one day.

In order to refine the current impact warning system, it would be
prudent to integrate certain terminology from other hazard warning
systems. In disaster management, a forecast is normally used as a
relatively vague estimate about the likelihood of a future event
(Alexander 1993). With reference to the impact warning system, the
estimated rate of terrestrial impacts can be regarded as impact
forecast. A prediction, on the other hand, is more specific and
indicates, on the basis of observed data, what is probable to happen. In
our case, the currently known 40 NEAs with non-zero impact probabilities
which are catalogued on professional impact risk websites serve as
impact predictions. A warning, conversely, is different again and
generally denotes a recommendation, an advice or order deriving from a
specific prediction. An impact risk warning, for example, may be issued
if the probability of a predicted impact is both substantial, e.g. above
0 on the Palermo Technical Scale (for details see Chesley et al., 2001)
and warranted due to particular circumstances such as a relatively short
warning period.

As far as an explicit impact confirmation is concerned, I suggest
dividing the impact warning systems into the two separate elements. The
first component entails the detection of a specific impact threat in
what I call the pre-confirmation phase. The second, or acute impact
phase, involves the preparations for avoiding the disaster or, at least,
minimising its secondary environmental knock-on effects during the final
impact confirmation phase. In contrast to many other warning systems,
however, the impact warning system requires the incorporation of a
third, critical disaster management component. It comprises of the
crucial period between the detection of an object that has a significant
probability of colliding with the Earth (a NEO that scores above 0 on
the Palermo Scale) and the ultimate verification - or elimination - of a
specific impact threat. This "wait-and-see period" will be a time of
uncertainty and nervousness because it will require careful
observations, refined calculations and a prolonged waiting period before
the impact threat can be confirmed or eliminated. This can be called the
pre-confirmation phase, or the period from the discovery of a Palermo
Scale 0+ object up to the confirmation that the object has either a 0%
or a 100% probability of impact. Depending on the specific time before
predicted impact as well as the opportunities for further observations,
the duration of the pre-confirmation phase can vary from days to years.
In the case of asteroid 1950 DA, the first object scoring higher than 0
on the Palermo Scale, the pre-confirmation phase may last for a number
of decades before additional, physical observations will either
eliminate or confirm the impact threat for the year 2880.

Given these uncertainties, the pre-confirmation phase will have to be
broken up in a number of additional disaster management segments. The
most crucial questions during the pre-confirmation phase deal with the
timing, decision processes and responsibilities of initiating expensive
and perhaps controversial mitigation activities. These actions may range
from the design and launch of a spacecraft mission to a potential
impactor with the task of refining its orbit and studying its
composition, to kick-starting the design and construction of a costly
asteroid deflection system, to local and regional preparations for mass
evacuation. While it is highly likely that the final impact confirmation
phase will provide us with an adequate amount of warning time for
disaster mitigation measures, it would be wise to consider unlikely
scenarios with much shorter warning periods which would put more
pressure on the decision-making processes and disaster management during
the pre-confirmation phase.

As far as the issue of public warnings during these phases are
concerned, I will propose to divide the impact warning system into five
separate warning phases. Once it is understood that any predicted impact
will advance in this way, although not necessarily going through all the
phases, it will become obvious that each of these phases require a
completely different approach to publicity and information policies.

For any predicted impact, we can expect the developing impact emergency
crisis to travel through a number of pre-disaster phases. I have listed
the main segments below but should draw attention to the fact that an
identified impact threat can easily get underway at phase 2 or 3, thus
skipping the earlier "low" or "moderate" probability phases. The impact
simulations further below are good illustrations of this point.
Nevertheless, it is equally possible that the evolution of an impact may
travel through all the 6 phases here suggested. Obviously, each phase
calls for different actions and announcements of the impact threat:

1. The low probability phase - takes effect with any object detected
that has an impact probability below 0 on the Palermo Scale. Most
objects listed on the "impact risk pages" managed by NEODyS and JPL fall
into this category. No specific public information is required for such
an event.

2. The moderate probability phase - gets underway with the detection of
an object that has an impact probability above 0 on the Palermo Scale.
It would be sensible to post clarifying information on the various
internet "risk pages." However, at this stage it would be important to
emphasise that the impact risk will be removed, in all likelihood, as a
result of additional observational data. Accordingly, there is no need
for any official press release during this phase.

3. The high probability phase - takes effect the moment an object
attains an impact probability above 1%. Depending on the length of the
warning period, ongoing public information, updates and clarifications
will be indispensable. Despite a significant impact probability, it is
prudent to stress that the chances remain very good that the threat will
go away with further data. Once again, the matter of warning time and
NEO size will be of crucial importance in the decision making process.

4. The very high probability phase - sets off with a very high impact
probability of somewhere between 30 and 50%, depending on the warning
interval before probable impact, the size of the object and the length
of time covered by the observations. During this phase, the emphasis
would certainly switch to mitigation potentials and the necessity for
mitigation projects to commence in earnest.

5. The confirmation phase - will provide both researchers and the public
with almost complete certainty about the veracity of the impact threat:
this is less of a phase and more of a moment at which (as a result of
additional data) the impact probability either drops to 0% or goes up to

6. The mitigation phase - sets off with the confirmation of an
approaching impact and would start the clock of the remaining warning
time until impact (or impact prevention). Obviously, the public would
have to be constantly updated, informed and reassured about the
considerations of the disaster management teams, and the scientific,
political and perhaps the military communities. The computation of
impact location and secondary impact effects as well as specific local
and regional warnings will be essential elements for cases where impact
prevention either is impossible or proves unsuccessful.

Up to now, we have had some fickle experience with objects reaching the
"low probability phase". Only recently, the exceptional case of asteroid
1950 DA with its 1:300 chance of impact in 2880 has introduced us to a
first encounter with the "moderate probability phase". The biggest
dilemma we are most likely to face in the event of a phase 3 or phase 4
incident, is the problem of highly erratic and unpredictable impact
probabilities which would set off a prolonged period of uncertainty and
anxiety that can last for many years. The impact simulations presented
below are very instructive in this respect.

Simulating Impact Emergency Crises

Tunguska-type impacts without warning

Given the rudimentary structure of the current impact warning system,
the impact event most likely to occur during our life-times is the
undetected and thus unpredicted explosion of a Tunguska-sized NEO in the
atmosphere. Of the 29 documented smallish impacts that occurred in the
decade between 1990 and 2000, more than 90% happened in or over
uninhabited parts of the world (Atkinson et al. 2000).  Since the
largest parts of the world are uninhabited or sparsely populated areas,
the odds are high that such an impact will take place over or in a
remote region. As long as NEO search programmes are primarily
terrestrial rather than space-based, almost all of the small,
Tunguska-type impacts will likely transpire "out of the blue", i.e.
without prior detection or anticipation. Accordingly, it will be
impossible to issue any public warning. No one needs to agonize over the
risk of Tunguska-class impacts. Nevertheless, it would be prudent not to
underestimate the social and psychological disturbances a sudden,
unpredicted asteroid impact may set off.

A frightful experience such as an unexpected impact disaster will most
likely produce emotional distress and anxiety - regardless of where it
may occur on Earth. Consternation, anger and apprehension are some of
the sensations that may result from a relatively small, Tunguska-size
(let alone a medium-size) impact. Uneasiness is a common response to a
hazardous situation such as the impact hazard. However, this would
change quite significantly if we were to witness a 20-100 MT impact
disaster. The view of many people about their sense of safety would
change dramatically. Some individuals may become more anxious about the
risk of future impacts given that increased media coverage may intensify
emotional arousal. Many people distressed by a minor impact disaster
will start to see the world as filled with cosmic danger.

Some people may experience problems in dealing with even a small impact
due to its random and 'terrorising' nature. It will stir up anxieties
not least because the impact is likely to be blown out of all
proportions by the mass media. Some people will feel very angry,
irritable and scared. They will blame their governments, space agencies
and astronomers for failing to protect them from cosmic disaster.
Evidently, by contemplating what may happen in the event of a small
impact, we need to recognize the psychological and social implications
of traumatic events and the emotional and irrational reactions they can
activate. Sometimes, it will not be sufficient to issue the mantra of
statistical risk estimates.

Tunguska-type object (or smaller) detected on final trajectory shortly
before impact

While considering disaster management of impacts that might threaten
civilization (Garshnek et al. 2000), we must not lose sight of the very
much more frequent impacts by much smaller objects that might be largely
harmless to people or the environment but which could be very
fear-provoking. According to Morrison et al (2002), "a short lead time
for an NEA is extremely unlikely - we can expect either decades of
warning or none at all." It is indeed more likely that we will have many
years of warning time in the unlikely event of a local or regional, not
to mention a global impact disaster. Nevertheless, it is possible that
NEO search programmes may one day detect a Tunguska-sized (or smaller)
object just days or weeks before it enters the atmosphere. With
incessantly growing telescopic aptitude and the future prospect of
space-based NEO search programmes, the probability of detecting a small
NEO on its final trajectory before impact is slowly but gradually
increasingly. As Chapman at al. (2001) point out, "even the impact of a
very small body, if the time and location are predicted with an hour's
warning or more, might merit evacuation of people from ground zero."

Given NASA's current Spaceguard policy which focuses on the detection of
large NEAs, it is not surprising to observe that "to date, very little
attention has been given to the demands that would be placed on
governmental and private disaster-response systems by even a small
(Tunguska-class) impact in a populated region" (Morrison et al. 2002).
As a matter of fact, research on impact disaster management almost
exclusively addresses the difficulties that would arise from the threat
or impact of a very large asteroid (Garshnek 2000).

What then would be the main problems in handling such an impact
emergency crisis in the weeks, months or years before impact? Here, I
will deal with a number of issues that need to be anticipated in the
event of the detection of a small asteroid on its final trajectory, a
week before impact.

Discovery and initial announcement

The first time we detect a definite impactor before it enters the
atmosphere, even if the object is too small to do any physical damage,
we can expect to encounter a sever problem of (potential) disaster
management. For that reason, it seems far-sighted to anticipate the
potential problems we may face. 

A 50-100m object found by LINEAR, NEAT or Spacewatch data today would be
announced by the Minor Planet Center on The NEO Confirmation Page
because of apparent sky motion faster than for main-belt asteroids. At
that point, it would remain unknown even roughly how big the object is.
Under normal circumstances, the MPC would receive some follow-up
observations during the following 24 hours. At that point, the MPC would
revise the initial prediction published on The NEO Confirmation Page
(even though the orbit would still be somewhat speculative). This would
help other search programmes and amateurs as they frequently follow up
the NEOCP objects, sending their measurements to the MPC the next
morning. With now the three groups of observations, the MPC can expect
reasonably to calculate an initial orbit. The MPC would run the orbit
forward for the MPEC announcement of the observations, orbit and
ephemeris. The first MPEC is usually issued when the MPC have three
groups of observations. The orbit is therefore known rather minimally,
and if a group is incorrectly timed (for example), there will be some
unexpected uncertainty. After another couple of days, however, there
would  be many additional observations. Any inconsistencies would be
uncovered and the orbit should be known much better. At that point, it
should be pretty obvious there is a good chance (and a significant
impact probability) that the object would hit in a few more days.  

Pre-confirmation phase

Given that the earth seems to be in the range of impact solutions, all
orbit calculation and impact risk programmes at Pisa University and JPL
and other interested observers would immediately catch on, even if the
impact probability at this stage is "only" 10 percent (maybe even if
only 1 percent). The question everyone will then want answered is
whether the impact probability will go to zero - or to 100 percent, for
the earth as a whole.

After the first two days, the impact probability will substantially
increase, though not necessarily all the way to 100 percent. It is
important to emphasise that every case is quite different and thus it is
difficult to generalize. While the impact probability will rise
significantly and the object will thus trigger international anxiety, it
would be possible at this stage to announce that large parts of the
earth (the southern hemisphere, say) would be quite safe from the
potential impact.

While the international public would anticipate an imminent impact to
occur, it would be possible at this stage to announce that some large
areas of the earth are safe. However, there remain some uncertainties in
connection with the estimated size of the impactor. The NEO community
will have calculated by now that the object is perhaps some 100 meters
across. Given the notorious difficulties to work out the actual mass of
such an object (Chesley et al. 2001), any size estimate based on its
intrinsic brightness means that its actual size may vary between 50 and
200 meters, the latter possibility being somewhat troubling for those
residing near the coasts surrounding oceans that have not been ruled out
for direct impact.

In order to analyse the structure and to refine the orbit of the
impactor, radar observations will be particularly effective during the
last four days (in fact, it may take the first two days after discovery
to arrange for them. These data together with additional observations
should elevate the impact probability to 99% or 100% within a day or

Impact mitigation phase

Once the impact probability has reached 90-100%, the most important task
will be narrow down the impact location and to estimate the probable
secondary impact effects most likely to result from an impact in the
calculated region. The last three to two days before impact would be
spent  narrowing down the region where the impact will occur. It can be
assumed that the impact point will be known for a radius of 500 km
accuracy some 72-48 hours before impact, while the radius will gradually
become more reduced as more observations become available.

Consequently, the last two to three days could be used to evacuate
populations out of an impact area - in the unlikely event that it were
to hit a populated area. Of course, the bigger the object and the larger
the effected areas, the more difficult a rapid mass evacuation would be.
Nevertheless, the last 48 hours will be crucial to reduce the potential
cost of life in such a disaster.

The idea of calculating the impact probability for certain areas of the
globe is important because it introduces a new aspect to the question of
impact risk. It would also help to reassure and apprehensive population
around the globe. In the event of a confirmed impact of a 50-200m
object, the probability of any given person being affected by the impact
will be minimal. This knowledge should be part of the main message
conveyed to the international public. Only a relatively small
geographical area of the globe will be affected, and the chances are
extremely high that it will actually be a remote part of the world. In a
real impact emergency, it would be crucial to narrow down the impact
location and to warn those most likely affected. While this will create
stress and anxiety for people in or around the affected impact areas, it
would at the same time provide reassurance to all those who will not be
physically affected.

Nonetheless, many questions regarding the organisation and structure of
disaster management during the final days of the impact emergency are
still unidentified. No disaster plan currently exists that can identify
all hazards likely to threaten communities or that specifies who will
organise, pay and lead the mitigation efforts. 

During the impact mitigation phase, the most important policy group
would not be the IAU, but preferably a U.N. NEO Panel, with its
geographical involvement and disaster-management expertise accompanying
that of the astronomers and military experts. If such a scenario would
occur tomorrow, we would not be well prepared - neither regarding the
way we would handle the PR nor with regards international management
structures that are simply non-existent at the present. We would have to
set everything in place, ad hoc, and would almost inevitably commit a
number of mistakes given our complete lack of impact disaster

Impact Simulation of Asteroid 1997 BR

The following impact simulations are based on research conducted by Paul
Chodas. Chodas modified the orbit of an existing asteroid (1997 BR) so
that it would impact the Earth instead of just passing close by. This
was done by slightly altering two of the six orbital elements. Both the
discovery date and the observation times remained unchanged. In the
first simulation, the discovery was dated 180 days before impact. Since
observation errors, both random and systematic, have a major effect on
our ability to predict impact (as Chodas learned from his predictions of
the Shoemaker-Levy 9 impacts), he preserved the actual errors and used
them in his simulation. He computed the best possible orbit for the real
1997 BR, (now assigned the number 13651) using both optical and radar
observations, then predicted what the observations of the real asteroid
would have been, had they been perfect, and finally differenced these
values from the actual observations.  He treated the resulting set of
'residuals' as errors to add to his simulated observations of the
impacting object. The resulting set of observations was a fairly
accurate simulation of observations of an impacting object (except that
there may well be many more observations of a real object on an impact

Discovery 180 days before impact

In this simulated impact, the asteroid is discovered 180 days before
impact, at a distance of 0.46 AU. This is quite a long time for an
object impacting on the discovery apparition. The estimated diameter of
1997 BR is 1.5 km, so it was bright enough at this large distance. After
one week of observations, the object had an impact probability of 1% and
thus scored higher than 0 on the Palermo Scale. After the IAU technical
review team has confirmed that a significant impact risk exists, the
pre-confirmation phase kicks in. With only a week of observations, a
public announcement about the potential impact threat will be released
by the IAU together with the customary declaration that the threat will
most likely be eliminated by further observations in the next few weeks
and months. Yet, contrary to official assurances the impact probability
does not decrease, let alone drop to zero. Instead, the impact
probability starts to fluctuate wildly over the next four months.

Intuition would suggest that the impact probability increases
progressively with every additional set of data. But this is not the
case mainly due to systematic observational errors. The evolution of
impact probability during the first 100 days shows a rather erratic
pattern: In this simulated scenario for 1997 BR, the asteroid was
discovered on 1997 Jan. 20, and the impact probability after one week
(ie, on Jan. 27) was about 1%. After two weeks (ie, on Feb. 4) it was
15%, after week 3 it was 30% (week 4: 12%; week 5: 2%; week 6: 12%; week
7: 15%; week 8: 37%; week 9: 36%; week 10: 35%; week 11: 33%; week 12:
30%; week 13: 6%; week 14: 7%, and then jumped to 86% on May 2, which is
78 days before impact.

The main reason for the quick jump is that we received 10 good
observations on May 2. In his case study, Chodas was using actual
observation times. He makes the point (private communication) that in an
actual impact emergency, we may have many more observations than he
simulated (because of the great interest in the object) which might
diminish the extreme fluctuations in the impact probability somewhat.

What is crucial to recognize is the length of the pre-confirmation
phase. The probability of impact did not rise above 50% until 78 days
before impact (I-78d), but it rose past 99% only 8 days later. Part of
the reason that the impact probability remained low until I-78d was that
the systematic observation errors held the nominal solution 'off' the
Earth until this time. If the systematic errors had been smaller (as
they will be with more modern star catalogues), the 50% impact threshold
may have been reached at about I-90d. The impact point was determined to
within 500 km about 45 days before impact. 

The fluctuations in the probability are caused by systematic errors in
the observations pulling the so-called uncertainty ellipse away from the
Earth. According to Chodas, this effect was one of the most interesting
results of his study. It in fact explained some of the behaviour he saw
when predicting the impact times and locations for the fragments of
comet Shoemaker-Levy 9 before they smashed into Jupiter in 1994.

According to Chodas: "One could argue that this wasn't a realistic case
because an object on a near-impact trajectory would be observed more
frequently than was 1997 BR, and possibly with bigger-aperture
telescopes and better detectors.  This is true to some extent, but the
systematic error problem would remain, as these are errors in the star
catalogs used to reduce the observations.  The situation has improved
somewhat since 1997, however, since modern star catalogs based on
Hipparcos results have smaller systematic errors: the 50% probability
might be reached 4-5 weeks earlier nowadays" (private communication).

Once radar data is used for the refinement of the impact orbit, one can
pinpoint the impact area rather well. However, radar cannot be used long
in advance of impact because an object of this size has to be within 0.1
AU of the Earth for the echo to be detected from Goldstone. 
One of the biggest disaster management problems for the decision making
processes in these hypothetical set of circumstances concerns the
question of planetary protection. After all, the impact probability rose
above 50% only after 100 days of observations and only passed the 99%
mark 70 days before impact. The length of the pre-confirmation phase is
a critical dilemma because the prolonged period of uncertainty could
result in dithering and hesitancy, thus further shortening the time for
successful mitigation operations. While space missionsare currently
impossible in the event of a short warning period, the dilemma could
become more awkward in a case with a medium warning period limited to a
few years.

Chodas has used a 50% impact probability threshold to define the start
of the "warning time." While he believes that plans for direct action
should begin at a lower probability level, he does not spell out at what
level of impact probability (or after what time) action should kick in.
Given a short warning time of only 180 days, it is reasonable to assume
that mitigation procedures would be underway without any delay.

Unlike the discovery of a long-period comet that may be detected only
three or four months before a possible impact, it is very unlikely to
find a large asteroid such as 1997 BR with only a few months warning. On
the other hand, if we ever had to face such an unlikely impact emergency
with a warning time of less than half a year, we would have to make a
decision about possible planetary defence initiatives at some crucial
point. That raises the question: how high does the impact probability
have to go in order to force a political decision that would initiative
highly expensive and controversial planetary protection action? In other
words, at what point during the crucial pre-confirmation phase would
researchers advise their governments and military experts: "Now is the
time to prepare for the worst"? Given the unique features of every
feasible impact scenario, and in particular our inexperience with such
an unprecedented situation, there seems to be no obvious answer. One can
only hope that we will never be confronted with such a short-term

Discovery 20 days before impact

Next, Chodas considered a 'short arc' case in which the asteroid was discovered
only 20 days before impact. The impact probability passed the 50% level
after 4 days of observations (16 days before impact), and the 99% level
was reached 2 days later. The impact point could be determined to within
a 500 km radius 6 days before impact. Since 1997 BR is a rather large
asteroid, the warning time is obviously far too short for any deflection
mission. By the same token, a week of warning would be of little benefit
to evacuate millions of  people threatened by a 1.5 km asteroid. It is
important, however, to stress the improbable circumstances of this
simulation. Should we be faced with a similarly short warning period of
only 20 days, it is a good deal more likely to involve a much smaller
object. And in such an event, even 6 days of warning would seem
sufficient for mass evacuations.

Discovery 54-year before impact

The impact probability remained below 1% for the first 60 days, and
never rose above 7% during the 375-day apparition. 600 days later, when
the second apparition began, the probability rose from 4% to 75% over
the course of the first 25 days. As with the other case, Chodas was
using realistic observation errors, and the probability was bouncing
around during the early stages. With the more modern Hipparcos-based
star catalogues we have now, this effect would be less but still
present. According to Chodas, this effect needs to be re-examined using
these newer catalogues, and also using the newer techniques we have now
to estimate impact probability (private communication).

With 54 years of warning time, the pre-confirmation phase of uncertainty
spans almost 2 years after detection. While impact probabilities
fluctuating between 1% and 7%, the object would be classed as a category
7 object on the Torino Scale (causing high anxiety levels around the
world). During this lengthy period of time, the question on every ones
mind would be whether the impact probability will rise to 100% or drop
to zero once we get additional data during the second apparition.

Calculating Impact Location & Radar problems

The above simulations used optical observations only. Chodas performed
additional simulations which used radar data as well. Radar can lengthen
the warning time somewhat. Crucially, radar observations, almost as soon
as they are made, assist the efforts to narrow the impact point
prediction. Current radar observations, however, are possible only
within the last two weeks or so. (Another good way of boosting the
accuracy of the impact prediction is to find old observations, although
these won't help so much at a very late stage.)

Determining the impact point is a simple extension of determining the
close approach circumstances (Chodas, private communication). The best
estimate for the impactor's orbit is used in order to determine the
location where the orbit intersects the planet. These calculations can
provide both a time and a position of the asteroid when it touches the
Earth's surface. Given a mathematical model for the rotation of the
planet, researchers can determine the latitude and longitude coordinates
of the impact point. Radar observations, however, are not possible until
the object comes within range, where the precise range limit depends on
the object's size.
What this means is that the refining of the actual impact area depends
crucially on radar observations, and the very availability of these
depends on a number of things. For one thing, there is no radar in the
southern hemisphere. Even in the north, one should note that Arecibo can
observe only over a relatively small range of declination. While these
limitations currently restrict the inception of precise impact point
calculations, it is anticipated that this technology too will advance
significantly over time, thus further expanding the time-span of impact
point calculations.

Perhaps the most important lessons to come out of the simulation of
impact emergency crises is that as the impact probability for the earth
as a whole goes to 100 percent, the probability for most specific areas
on the surface actually decreases.

Fluctuating Impact probabilities: When to take action?

The big question is: how high does it have to go before we take an
interest? In assessing that question, one needs to be watching how the
probability is changing with time, particularly in comparison with the
time remaining until the assumed impact. It's questions like these that
are important here while the Torino Scale, which pays no attention to
the time factor, would not prove very effective during such an impact

Unlike most other natural hazards warnings that are usually delivered to
just the people at risk from a local or regional disaster, an initial
impact warning will be delivered most likely to a global audience. In
order to minimise public anxiety in response to an impact warning, it is
imperative to calculate and announce (as early as possible) impact
location probabilities for certain areas of the globe. The successful
predictions of where and when the fragments of comet Shoemaker-Levy 9
would hit Jupiter in 1994 demonstrate the accuracy of such calculations
given enough data. Even if the impact probability for a small or
medium-sized NEO reaches 100%, the personal risk probability (i.e. the
chance of being physically effected as a result of the impact and its
secondary destructive effects) would be minute. As a matter of fact, the
computation of and public communication regarding the probable impact
point will be one of the most important messages of any impact warning.
After all, only a relatively small geographical area of the globe will
be affected by a small or medium-sized NEO impact. Besides, the chances
are very high that it might not even be an inhabited part of the Earth.
Naturally, the bigger the size of the impactor, the higher the personal
risk probability will be - until the NEO reaches the 1-2km threshold
that would more or less affect everyone.

In a genuine impact emergency it will be essential to narrow down the
exact impact point as early as technically possible in an attempt to
reduce personal risk both for the public that is largely safe as well as
for those people living close to the impact location. While this might
create stress and anxiety for people in the vicinity of the predicted
impact, it would provide vital warning time to evacuate areas should
this be necessary. It would certainly provide re-assurances to all those
around the world that will not be effected whatever.

But at what impact probability should we actually initiate an expensive,
controversial and perhaps even dangerous mitigation scheme? Ultimately,
all depends on two questions: 1) how much warning time do we have and 2)
what are the expected consequences of the impact, if it should occur?

One of the highest impact probabilities experienced so far was the
1-in-500 chance of 2000 SG344 colliding with Earth in 2030. In this
case, we had 30 years, it was a small object - and we knew there were
more data that might change the result anyway. Even with the
theoretically possible 1-in-100 impact probability we might have had for
the much larger 1997 XF11 in 2028, it would have been appropriate to
look for old data before trying anything impulsive. But what if the 1990
observations perhaps accompanied by 1998 radar data had increased the
2028 impact probability to 50 percent? What would we have done in such a
case? Since we had exhausted our earth-bound possibilities, it would
have been time to plan a visit to the asteroid and examine it
composition and place a transponder on the surface. This would certainly
be a very expensive mission. Yet, we would soon be able to gather
crucial information about whether or not XF11 would collide with the
Earth in 30 years time. In any case, such a mission would provide
researchers with valuable scientific information. But what if we still
lack absolute certainty about the object's orbital evolution? Do we then
try a relatively easy and decisive deflection with 10 years to go at 90
percent impact probability or a more difficult one with 2 years to go at
99 percent - or somewhere in between? That's where things become more
difficult (Marsden, private communication).


We are essentially unprepared for the mitigation and management of an
impending impact crisis at present. Indeed, the media hype and
hullabaloo over the re-entry of Skylab and MIR does not bode well, given
that there was even some control over where MIR would fall. In short, we
are currently not well prepared for an impact emergency crisis - neither
regarding current disaster management structures nor in the way we would
handle public risk communication and impact warnings. If truth be told,
we would need to establish most of these disaster management structures
form scratch, ad hoc, and thus run the risk of committing a number of
blunders given our inexperience. In consideration of the rudimentary
structure of the NEO impact warning system as well as in the expectation
that we will have sufficient time for significant progress, I hope that
this paper will help to improve the effectiveness of the budding impact
warning system.


I would like to thank Brian Marsden and Paul Chodas for critical
comments and essential information about their research. The simulation
of impact emergency management is largely based on the astronomical
impact simulations by Paul Chodas who kindly provided me with the
results of his unpublished findings.


David Alexander (1993) Natural Disasters. London: UCL Press

Atkinson, H., Tickell, C., Williams, D. (2000) Report of the Task Force
on Potentially hazardous Near Earth Objects. London: HMSO

Atwod, L.E. and Major, A.M. (1998) Exploring the "Cry Wolf" Hypothesis.
International Journal of Mass Emergencies and Disasters, 16(3), 279-302.

Binzel, R. (2000) The Torino Impact Hazard Scale. Planetary and Space
Science, vol. 48, 297-303

Bowell, E. and Muinonen, K. (1992) The End of the World: An orbital
uncertainty analysis of a close asteroid encounter. Bulletin of the
American Astronomical Society 24, p.965.

Bekey, I. (2001) An international approach to detecting
Earth-threatening asteroids and comets and responding to the threat they
pose. In: 6th International Space Cooperation Workshop Report, (Brennan,
H., ed.), pp. 15-21.

Canavan, G.H., Solem, J.C., Rather, J.D.R. (1994) Near-Earth Object
Interception Workshop. In: Hazards due to Comets and Asteroids.
(Gehrels, T., ed.), Tucson & London: University of Arizona Press, pp.

Chapman, C.R. (2000) The asteroid/comet impact hazard: Homo sapiens as
dinosaur? In: Prediction: Science, Decision Making, and the future of
Nature, (Sarewitz, D., Pielke, R.A., and Byerly, R., eds.) Washington:
Island Press, pp. 107-134.

Chapman, C.R., Durda, D.D. and Gold, R.E. (2001) The comet/asteroid
impact hazard: a systems approach

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

Chesley, S.R, Chodas, P.W., Milani, A., Valsecchi, G.B., Yeomans, D.K.
(2001) Quantifying the risk posed by potential Earth impacts. Icarus [in

P.W. Chodas and D.K. Yeomans (1997) Impact Warning Times for Earth
Crossing Asteroids, Bulletin of the American Astronomical Society 29,

P.W. Chodas and D.K. Yeomans (1996) The Orbital Motion and Impact
Circumstances of Comet Shoemaker-Levy 9. In: The Collision of Comet
Shoemaker-Levy 9 and Jupiter, (K.S. Noll, P.D. Feldman, and H.A. Weaver,
eds.), Cambridge University Press, pp. 1-30, 1996.

De Marchi, B. (1995) Risk communication to reduce vulnerability. In: In:
Natural Risk and Civil Protection, (Horlick-Jones, T, Amendola, A. and
Casale, R., eds.), London: E&FN Spon, pp. 381-393 

Effective Disaster Warnings (2000). Report by the Working Group on
Natural Disaster Information Systems. Subcommittee on Natural Disaster
Reduction. Washington: National Science and Technology Council

Fitzpatrick, C. and Mileti, D.S.(1994) Public Risk Communication. In:
Disasters, Collective Behavior, and Social Organisation, (Dynes, R.R.
and Tierney, K.J., eds),  London & Toronto: Associated University

Friedman, G.J. (1997) Risk Management applied to planetary defense. IEEE
Transactions on Aerospace and Electronic Systems 33(2), 721-733

Garshnek, V., Morrison, D., Burkle, F.M. (2000) The mitigation,
management, and survivability of asteroid/comet impact with Earth, Space
Policy 16, 213-222

Hunt, J.C.R. (1994) Forecasts and warnings of natural disasters and the
roles of national and international agencies, in: Wadge, G. (ed.)
Natural Hazards and Remote Sensing. London: Royal Society

Lindell, M.K. and Perry, R.W. (1987) Warning mechanisms in emergency
response systems. International Journal of Mass Emergencies and
Disasters 5:2, 137-153

Marsden, B (1999) A discourse on 1997 XF11". Journal of the British
Interplanetary Society 52, pp. 195-202

Morrison, D, Chapman, C.R. and Sommer, G. (2002) Managing the Impact
Hazard. In: Asteroids III (Binzel, R., ed.) in press

Nigg, J. (1995) Risk Communication and warning systems. In: Natural Risk
and Civil Protection, (Horlick-Jones, T, Amendola, A. and Casale, R.,
eds.), London: E&FN Spon, pp. 367-379

Noji, E.K. (1997) The Public Health Consequences of Disaster. Oxford
University Press, New York

Dennis Parker (1999) Criteria for Evaluating the Condition of a Tropical
Cyclone Warning System, Disasters, 23:3, 193-216

Sorensen, J.H. (2000) Hazard Warning Systems: Review of 20 Years of
Progress. Natural Hazards Review, 1(7), 199-125

CCCMENU CCC for 2002