The Earth is subjected to a continuing "celestial rain" of cosmic debris. Nearly all of it is fine dust or larger objects that mostly burn up in the atmosphere and are of little consequence to our environment. Occasional pieces of larger, stronger objects survive as meteorites. Every few years a meteorite causes damage, penetrates a roof, strikes a car, or is otherwise a nuisance; such events are newsworthy because of their rarity and unusual nature. Yet the power-law that approximately describes the size distribution of cosmic debris has no large-size cut-off (comets of unlimited size could conceivably approach the Earth), unlike other terrestrial natural hazards. Very rarely (every few 100,000 years or so, or 1 chance in several thousand during a human lifetime) a comet or asteroid more than a mile wide strikes the Earth with serious global environmental consequences that could well threaten the future of civilization as we know it (Chapman & Morrison 1994). Even more rarely (every 100 million years or so, but it could happen this decade) a cosmic projectile 5 to 10 miles across strikes, with consequences so terrible that most species (like the dinosaurs 65 million years ago) are threatened with extinction. It is plausible that an even larger object, perhaps larger than 25 miles across will strike the Earth during our Sun's lifetime, possibly virtually sterilizing the surface of our planet. Technical background on the impact hazard is found in two recent books: Gehrels (1994) and Remo (1997). The impact hazard has often been discussed in the context of establishing an international telescopic survey to find threatening asteroids, especially those larger than 1 km in diameter that dominate the hazard. This "Spaceguard Survey" was first proposed by a NASA committee chaired by Morrison (1992) and updated by one chaired by the late Eugene Shoemaker (Shoemaker et al. 1995). The current status of such surveys is reviewed by Harris (1999) and Binzel et al. (1999).

This "impact hazard" by near-Earth objects (NEOs) is strikingly different from most natural hazards in two ways: (1) the potential consequences of a major impact exceed any other known natural or man-made hazard (including nuclear war), and (2) the probability of a major impact occurring in a politically relevant timescale (e.g. during our lifetimes) is extremely low. It is interesting, however, that multiplying the probability of a major impact by its consequences yields an annualized death rate similar to that of some of the traditional natural hazards discussed in this volume (hundreds of deaths per year worldwide, primarily due to near-Earth asteroids (NEAs) 1 to 3 km in diameter, which have a 1-in-a-thousand chance of striking during any century, producing a global climatic catastrophe that would threaten civilization: see Morrison et al. 1994). Cosmic impacts, therefore, are the ultimate high-consequence, low-probability hazard.

A feature of this hazard, in the context of "prediction", is that scientific awareness has evolved from nearly zero two decades ago (when a serious impact would certainly not have been predicted) to one in which three-quarters of potential impacts could, within another decade, be predicted so exactly that effective, technologically feasible mitigation measures could be applied. For the remaining one-quarter, the prediction would either be too late to mount more than partial countermeasures or might even occur as a wholly unforecast "act of God". The public, including civilian and military officials, were once wholly unaware of the impact hazard. Now it is a common cultural touchstone, largely due to an "impact scare" in March 1998 followed by release of two major impact-disaster movies.

This is an immature hazard, in that there has been little analysis of its policy implications and minimal development of methodologies or an infrastructure for dealing with predictions of impending impacts. Simple recognition by policy-makers and federal and international governmental agencies that this hazard exists is only just dawning. I believe that a national and international dialog should commence, informed by an active research program, to assess the appropriate level of response to a threat that is very real and serious, if also very unlikely to be realized in our lifetimes. An informed political decision may be that little priority should be given to the impact hazard at this time. A decision, instead, that something must be done should then trigger development, from scratch, of ways to evaluate the hazard, of protocols for issuing impact predictions, and of potential approaches to mitigation. Governmental inaction so far should not be construed as a tacit decision not to act but rather as expected from the lack of serious engagement, outside of the specialty of asteroid astronomy, with this "new" hazard.

This volume's theme -- problems of prediction -- might seem to be of little relevance to the impact hazard. For asteroids that have been discovered, current technology can, in principle, provide very precise predictions about when and where they might impact. (Astronomers have been applauded for centuries for accurate predictions of sunrise, sunset, and eclipses and engineers routinely guide spacecraft to distant planets.) Technical approaches to mitigation, while preliminary in development and very expensive, are inherently simpler than for many hazards (e.g. deflection of the oncoming object or evacuation of ground zero); much more likely, lead-times before impact will be measured in decades, rather than months or years. And for those potential impactors that haven't been discovered (so far the great majority, although that may soon change if the proposed Spaceguard Survey is fully mounted), the probability of impact is a simple cosmic lottery: our inherent chances-of-losing are readily calculated and well known (Chapman & Morrison 1994). There are, indeed, uncertainties in the precise environmental consequences of impact (Toon et al. 1997) -- akin to those of other meteorological phenomena, like storms, global warming, the ozone hole, and El Niño -- but the sudden, exceptional drama of an impact event itself would minimize the perceived importance of such variables. (Even the 1979 re-entry and impact of the relatively tiny Skylab caused public anxiety.)

However, the impact hazard actually does have major predictive issues. They concern how individuals, society, and political institutions may prepare for and respond to predictions of such a horrific but unlikely disaster as the impact of a body hundreds of meters to a few kilometers in size. Given the difficulty people have in conceptualizing such an event, never witnessed during modern times, and human irrationalities in responding to very low probabilities, uncertain or faulty impact predictions (or misperceptions of actual innocuous impacts) may dominate policy makers' concerns. Predictions themselves will be the consequential events, not actual impacts.

The erroneous prediction of a possible impact issued on March 11, 1998, which made banner headlines around the world, illustrates the problem and forms the core of this Case Study. The error -- public assertions by supposedly credible astronomers that there was a "small" (e.g. >0.1%) chance that the Earth would be hit in the year 2028 by a mile-wide asteroid (1997 XF₁₁, hereafter XF11) -- was corrected a day later; revised predictions were reported by a press somewhat disenchanted with the astronomers' credibility. Yet, such temporary errors, occasioned by an impatient press looking over the shoulders of naive astronomers (who should not have alerted the press until they knew what they were doing), was hardly as serious as forecasts of earthquakes that never happened (after counties had spent millions in preparation) or failures to forecast hurricanes that did happen. Impact odds only changed from "very unlikely" to "zero", and no public funds were spent on mitigating an event 30 years in the future. Nevertheless, astronomers appeared to act like Chicken Little and lost some credibility in an arena in which they may someday have to forecast an event deserving to be taken seriously at the highest governmental levels.

Important policy issues must be addressed. One concerns mechanisms whereby astronomers evaluate data about a newly discovered, potentially hazardous asteroid before it can be proved that the object cannot hit; how they should validate calculations of impact probabilities during that interval; and under what protocols they should advise officials and the public about such possibilities. This problem, articulated in 1997 drafts of this Case Study, played out dramatically during the early days of the March 1998 media frenzy over XF11. More sober treatment of two other cases, the predicted non-zero impact probability by 1999 AN₁₀ (hereafter AN10) and another object, provides a counterpoint and suggests that lessons are being learned from the XF11 fiasco.

Other policy issues include clarification of how a plethora of national and international laws (e.g. the National Environmental Protection Act and the Outer Space Treaty) might apply to development of a defensive infrastructure to deal with asteroids or, in the unlikely event that an object were found to be on an imminent collision course with Earth, to emergency countermeasures (cf. Gerrard & Barber 1997; Kunich 1998). Secondly, there is the question of whether responses by political entities to the impact hazard will be driven primarily by legitimate public concerns, or instead be torqued by the self-interested motives of those who would stand most to benefit from mounting large scientific or military responses to the threat. For instance, some commentary (e.g. Park 1992), following a Los Alamos National Laboratory conference on asteroid interception (Rather et al. 1992), questioned the objectivity of ex-Cold-War technologists' sudden interest in planetary defense against asteroid and comets. Finally, the impact hazard presents a stark example of society's dilemma in responsibly deciding the relative allocation of funds to address, mundane, everyday problems that kill millions (e.g. diseases and accidents) versus focussing instead on rare but horrifying disasters (e.g. biologi- cal/chemical warfare, terrorism, airliner crashes, or nuclear power plant meltdowns), epitomized by the extreme case of asteroid impacts. As an end-member in the spectrum of natural hazards, cosmic impacts are more likely to provide us with a mirror for assessing other rare but important hazards than to become real-life disasters.

The impact hazard made unprecedented inroads in public consciousness worldwide during the second week of March 1998. The first apparently "official" prediction of a significant chance of impact during our lifetimes by a dangerous asteroid dominated news around the globe. A day later, headlines expressed relief as newly found observations were reported to yield a miss-distance twenty times farther away and an absolutely zero chance of impact. Follow-up coverage was often non-critical toward the astronomers who made the predictions, yet a cynical undercurrent persisted about the initial hasty and erroneous warning. National Public Radio commentator Daniel Schorr asked on March 22, discussing the asteroid event as well as another premature announcement about terrorists, "couldn't they have waited a few days before scaring us half to death?" Behind the scenes, a complex, often bitter exchange of e-mail ensued between asteroid researchers trying to assess data while dealing with a media frenzy.

During the subsequent year, a technically fallacious myth developed among media commentators about the causes and resolution of the XF11 affair, which had the potential of misleading policy makers about how to deal with future impact scares. Meanwhile, astronomers researched impact predictions and by spring 1999, when two other cases analogous to XF11 occurred, there was a constructive attempt to devise a generic approach for responsibly dealing with such predictions in the future.

This Case Study describes and evaluates this critical transitional period in which a hazard outside of the mainstream hazard community, in the purview of a specialty naive about dealing with matters of practical societal consequence, initially confronted the public policy arena. Space scientists are neither trained nor organized like other predictive scientists to evaluate predictions in the context of how they may be perceived by users in society. It remains to be seen whether fledgling efforts to develop an infrastructure and protocols for handling impact predictions will mature and be implemented responsibly and successfully.

Discovery of 1997 XF11. 1997 XF₁₁ was discovered on December 6, 1997, by Jim Scotti, using the Spacewatch telescope on Kitt Peak, Arizona. It was the brightest NEO discovered by Spacewatch -- a program focussed on smaller and fainter NEOs -- in more than a year. Follow-up observations by Japanese amateur astronomers permitted calculation of an approximate orbit for the asteroid with a MOID (minimum orbital intersection distance with Earth) of close to zero, making it the 108th known PHA (potentially hazardous asteroid: those asteroids for which MOID <0.05 Astronomical Units, or 7.5 million km [one AU = 149,600,000 km, the mean distance of the Earth from the Sun]). In general, an asteroid's orbit has a different size, ellipticity, and orientation than the Earth's orbit; most such orbits do not intersect the Earth's orbit, so that a collision could never occur, barring an exceptional and predictable change in the asteroid's orbit (e.g. by passage by another planet). If an orbit comes as close as 0.05 AU to intersecting the Earth's orbit, then small changes in the asteroid's path -- perhaps induced by the Earth's gravity itself during close passages -- could make the orbits truly intersect. Even then, the chance of a collision is very small, because both the asteroid and the Earth would have to be at the intersection point at the same time. Because of the consequences of any collision, even such small probabilities must be taken seriously. So once an asteroid is discovered, the astronomer's first task is to measure sufficient positions to calculate its orbit and determine if it has a small MOID and whether that MOID might decrease to zero in the next century. If so, then more observations may be needed to be sure that the position of the asteroid as it moves around its orbit is never located at the intersection point during the fraction of an hour (each year) that it takes the Earth to pass through the intersection point.

Brian Marsden, who directs the International Astronomical Union's Minor Planet Center (MPC) in Cambridge, Massachusetts, maintains a list of PHAs on his web site. His classification of XF11 as a PHA, a few weeks after discovery, placed it higher on observers' priority lists; but as new measurements trickled in, astronomers remained generally unaware of any near-term close approach of XF11 to Earth, for reasons discussed below. By February, Marsden's private orbital calculations and extrapolations indicated to him that XF11 would pass by the Earth very roughly twice as far away as the Moon (0.005 AU) in October 2028.

As new observations are added, orbital calculations can be run again, resulting in changes, generally improvements with smaller uncertainties. In early March, Peter Shelus of the Univ. of Texas submitted new positions of XF11 to Marsden. In accordance with Marsden's usual procedure (one criticized by his orbit-calculating colleagues), Marsden did not post Shelus' positions on the web -- his practice was to embargo them for a month -- but he did use them in a new calculation of XF11's potentially close pass in 2028. Marsden was amazed that the nominal miss distance jumped down to just 46,000 km (29,000 miles, or 25,000 miles above the Earth's surface); from Marsden's decades of experience, this seemed to him to be a remarkably close approach.

Announcement of impact possibility. Without checking with any outside scientists, Marsden published an International Astronomical Union Circular (IAUC) (#6837) during the early afternoon of March 11th, announcing that XF11 would pass just 0.00031 AU from the Earth on Oct. 26, 2028, thirty years hence. The usually restrained Marsden added an exclamation mark. As an indication of the estimate's uncertainty, Marsden wrote that an approach within 0.002 AU -- a distance closer than the Moon -- was "virtually certain". (On April 18th in IAU Circular 6879, Marsden explained that the "virtually certain" statement was in error and that miss distances could have been 10 to 15 times larger.)

The IAU Circulars (IAUC) -- distributed by e-mail, posted on the Internet, and mailed in postcard format to observatories worldwide -- are the chief way that astronomers learn about rapidly changing events in the heavens (e.g. supernovae, X-ray bursters, and comets), requiring observational follow-up that can't wait for later publication in scientific journals. The IAUCs, which began as telegrams many decades ago, are an official service of the International Astronomical Union (IAU); in addition to his MPC responsibilities, Marsden edits and publishes the IAUCs. Marsden's stated reason for mentioning the close-approach on the IAUC was to motivate new observations of XF11, including searches for any unreported past observations, in order to refine its orbit and the prediction.

Footnote 2: The International Astronomical Union's leadership criticized Marsden for issuing the PIS under its name. Marsden says that he did not indicate his IAU affiliation on the PIS but that Maran added it in the AAS press release. There is no practical distinction since Marsden's credibility in such matters derives from his IAU mandate, whether he is issuing a Circular or a press statement about the Circular.

The Swift-Tuttle precursor. An earlier 1992 event (cf. Steel 1995) provides context for appreciating how Marsden's XF11 announcement was received by some of his colleagues. As astronomers searched for the returning long-period comet Swift-Tuttle, Marsden worked on observers' positional data, just as for XF11. While talking with Boston Globe science reporter David Chandler, he made a back-of-the-envelope calculation that the comet would have a 1-in-10,000 chance of colliding with Earth during a close approach early in the 22nd century. Marsden repeated his estimate to the New York Times, whose reporter phoned me to ask what would happen if the comet hit. My remarks, made assuming that Marsden's calculations were correct, accompanied Marsden's prediction in the Times news service story. I felt embarrassed when Don Yeomans, of the Jet Propulsion Laboratory (JPL), showed that the Earth was actually safe from the comet. A public argument between Marsden and Yeomans was highlighted by Newsweek in a cover article entitled "Doomsday Science." The experts' consensus is that Marsden's estimate was so simplified as to be wrong, but he has never fully admitted his error and his relationship with Yeomans has since been cool.

Initial reactions to XF11. Although Marsden never stated a quantitative impact probability for XF11, his words had a straightforward meaning for astronomers. From the quoted miss-distance and estimated error of many times the miss-distance, a simple calculation indicates a probability of impact that is at least 0.1% (ratio of the cross-section of the Earth to the area of the circle of uncertainty; as we will see, the uncertainty was actually nothing like a circle). Alan Harris, of JPL, was among those to draw this simple conclusion, and he e-mailed it to nine colleagues within 2 hours of the IAUC's posting. Undoubtedly, many others made the same elementary calculation (cf. Goldman 1998).

Since XF11's observed brightness meant that its size was about 1 or 2 km (roughly that of a civilization-threatening impactor), a 1-in-a-thousand chance of civilization's demise occurring on a specific date just 30 years hence is awesome, deserving the public attention it received. Richard Binzel, of M.I.T., considered the threat to measure 3.5 on his proposed 5-point impact hazard index (Binzel 1997; see below) -- higher than he had imagined would ever be registered on the scale. Still, the chance of one of the ~1000 so-far-undiscovered asteroids as large as XF11 colliding with Earth in the next 30 years is about 0.01%, only ten times less than the (incorrect) apparent collision probability of XF11. So the predicted impact was not all that unlikely...what was remarkable was that we had found such a threatening object when the search for such objects -- the Spaceguard Survey -- was barely underway!

Don Yeomans and Paul Chodas at JPL approached Marsden's announcement with understandable skepticism, given the Swift-Tuttle history. They immediately sought Shelus's original XF11 data from Marsden, in order to make an independent assessment. While waiting about two hours for Marsden to comply, they calculated an effectively zero probability of impact from just the data available through early February. Within 15 minutes of receiving the later data from Marsden, Chodas reaffirmed his previous calculation that the chances for collision in 2028 were zero ("that's zero folks," Yeomans emphasized in e-mail sent to numerous colleagues that evening, 5 hours after the IAUC posting). The Chodas calculation placed the nominal miss-distance farther from Earth than Marsden's, but still unusually close. More significantly, Chodas' calculated uncertainty was 4 times larger than Marsden's. (What we now know to be the true miss distance, far beyond the Moon, is consistent with the JPL error bars but not with Marsden's.)

Footnote 3: It was unclear just what Marsden's quoted error meant. Usually, scientists in this specialty quote "3-sigma" errors, meaning that there is a 99.7% chance that the correct answer will fall within the error bar. Marsden used the words "virtually certain" to refer to his original estimate that the asteroid would come within 80% of the distance of the Moon, but he later suggested that those confident-sounding words referred to a "1-sigma" error, meaning only 68% chance. Even as a 1-sigma error, it appears inconsistent with what was eventually determined to be the miss-distance, 2.5 times the Moon's distance. Five weeks later, Marsden agreed that calculations allowed miss-distances up to 10 to 15 times as large as the estimate of 80% of lunar distance.

Two of Marsden's colleagues called on him late on March 11th to distribute, via a new IAU Circular, the Yeomans/Chodas probability = zero calculation so as to forestall the developing media frenzy. But Marsden rejected the suggestion on the flimsy basis that an IAUC shouldn't be used to correct text in a PIS. Reporters phoning me early on the 12th, were unaware of the JPL result so I suggested that they talk with Yeomans, which they did. But upon checking back with Marsden, they found him sticking by his original estimate that the asteroid might impact, despite the Chodas result. Mid-day on the 12th, I telephoned Marsden and implored him to compare notes with Yeomans, but he refused.

"No Impact" aftermath. The following morning's newspaper headlines told the story of the Earth's escape from cosmic doom. The New York Post was emphatic: "Kiss Your Asteroid Goodbye!" Another Post story headlined "NASA Needs a 'Crash' Course in Math," and continued that the "Doomsday figures were way off [the] mark." Domestic and international news reports tended to report that "NASA scientists" (meaning JPL researchers Yeomans and Chodas, using data from JPL observer Helin) had corrected the work of "the International Astronomical Union" (meaning Marsden), while actually all are funded by NASA.

Headlines in the March 23rd issues of Time and Newsweek illustrate the tone of media reaction to what one labelled a "Cosmic False Alarm": "Oops, Never Mind!" and "For a day, it looked like we could all be toast as an asteroid hurled through space. Then astronomers double-checked their figures." Many reports noted the fortuitous promotion of the Hollywood movies "Deep Impact" and "Armageddon", due to premier within the next few months. At least one conspiracy-minded reporter persistently attempted to uncover illicit connections, which never existed, between Marsden and the Dreamworks producers of "Deep Impact." While the media generally handled the coverage accurately and benignly, astronomers surely lost some of their vaunted reputation for unassailable predictions thanks to the XF11 affair. Let's examine some elements of the impact hazard illuminated by this episode.

Despite Marsden's IAUC 6879, issued in mid-April which can be parsed into an obscure acknowledgement that the March prediction was mistaken, he continued to rationalize his behavior in March. In essays contributed to the Internet newsletter CCNet Digest, Marsden vehemently objected to assertions that there never was a chance of impact by XF11 based on the data he used for making his prediction; at one time, he even asserted that a 2028 impact remained possible after inclusion of the 1990 data. Marsden continues to discount the relevance of impact probability calculations and claims his pursuit of prediscovery observations is what saved the day. Adopting this view, several analyses of the interactions between science and science journalists have cited the XF11 affair as a case where science essentially "worked," rather than as a case of a wholly erroneous prediction that never should have been made in the first place (cf. Gladstone 1998). The July/August 1998 Skeptical Inquirer (Gardner 1998) and the July 1998 Astronomy (Gordon 1998) exemplify the myth that "science worked" in the XF11 case and that, as originally reported just after the scare (Browne 1998), "old photos" of XF11 saved the day.

In fact, subsequent work has confirmed the near-zero impact probability calculations initially made within a few hours of the March IAUC. Karri Muinonen, of the Univ. of Helsinki, using different computational techniques, took several days to firmly reach the same conclusion. Muinonen (1998a, 1998b) later reported that data obtained as early as December 22, 1997 and made available somewhat later were sufficient to calculate a 2028 impact probability of less than 10^-42. Much smaller probabilities are derived when data through February are properly analyzed. Obviously, such nearly zero chances of impact have no practical relevance and would hardly be newsworthy. Incorporation of the 1990 data greatly reduced the length of the error ellipse; but they were found many hours after the correct, effectively zero impact chance in 2028 had already been demonstrated. The unchecked March 11th prediction was simply erroneous and should not have been announced; moreover, Marsden should have withdrawn it, or qualified it, immediately that evening when he learned of Chodas' calculation, without letting another day of media frenzy persist.

Given that adequate (if incomplete) data were published on Marsden's web site so that correct calculations of zero impact probability could have been made anytime during the more-than-two months before Marsden's IAUC, why weren't they? Especially given Marsden's colleagues' belief that he lacks the programs to calculate impact probabilities, why didn't others with the appropriate software monitor PHA's like XF11 and demonstrate the safety of XF11's passage in 2028 long before March, when Marsden took an interest in it? A simple answer is that they weren't organized or funded to do so. NASA officials assumed, incorrectly, that Marsden was adept at impact probability calculations; his chief expertise is actually calculating ephemerides, which are predictions of where moving objects will appear in the Earth's skies. It is really at JPL where computational techniques for encounters have been developed in the context of spacecraft missions to distant bodies, but the JPL experts were not funded to analyze future impacts of newly discovered NEOs. Moreover, despite long talk about coordinating the Spaceguard Survey, neither the IAU nor any other entity had mandated procedures to track NEO future paths. NEO analysis remained a cottage industry at the MPC, supplemented by sporadic research elsewhere.

Tiny numbers, chaos, and keyholes. Scientists who work with infinitesimal probabilities actually care about differences between, say, 10^-10 and 10^-42, despite lacking practical relevance. So arcane arguments ensued -- some of it confusingly and inappropriately carried out in the press -- about whether Yeomans had been correct in reporting, "that's zero, folks." After all, in the world of tiny probabilities, extraordinary things might happen that could deflect XF11 into the Earth in 2028. Marsden, taking comfort in the moralistic view of some commentators that any chance of impact with a large asteroid was too big to accept (akin to "risking one life is unacceptable"), spent the summer of 1998 trying to prove that XF11 could have struck the Earth, after all, based on the data available to him on March 11th -- if not in 2028, then in 2037 or 2040.

Indeed, there are a few tiny places along the immense length of the 2028 error ellipse (prior to its shortening by the 1990 data) where, if XF11 were to pass right through such a "keyhole" in 2028, then it would hit the Earth in a later year. Chodas (1999) concurs with Marsden (1999) that there was about 1 chance in a hundred thousand that XF11 could have passed through such keyholes -- that is, until the 1990 observations eliminated such possibilities. However, this revisionist discovery of tiny chances of impact in 2037 and 2040 hardly justifies the original scare about the always-safe 2028. And the probabilities that XF11 would impact Earth anytime in the next 50 years (including the 2037 and 2040 possibilities) were always much smaller than the background probability, which is always with us, that some other, undiscovered asteroid larger than a kilometer would hit during the next 50 years: 1 chance in a hundred thousand is roughly the chance that such an asteroid will hit next year! It would be reasonable to check, sometime before 2028, whether XF11 were headed for a dangerous keyhole, but there was hardly any urgency to do so in March 1998. Hence, there never was a legitimate need to hurriedly issue an IAUC or PIS, in order to motivate new observations, without double-checking the accuracy of the released information.

Marsden's emphasis on keyholes has inspired more thorough analysis of keyholes, special resonances, and other phenomena that describe the attributes of the celestial dynamical "chaos" that governs the motions of small bodies that pass through the gravity fields of large bodies like Earth. After a close pass, the trajectory of the small body becomes genuinely chaotic and inherently unpredictable over even rather short periods of time. Subsequent encounters can no longer be described by neat "error ellipses," yet Monte Carlo statistical simulations can still be run to estimate impact probabilities. In essence, after a future close approach with the Earth, an asteroid begins to rejoin the "background" of undiscovered objects...until precise observations are made once again to determine its new trajectory. In that sense it once more becomes no more, nor less, dangerous than any other similar asteroids which have not yet been discovered.

1999 AN10 and 1998 OX4. On January 13, 1999, the LINEAR program (a major element of Spaceguard, operated by M.I.T. Lincoln Laboratory with U.S. Air Force and NASA funding) discovered a mile-wide Earth-approaching asteroid designated 1999 AN10. Andrea Milani, of the Univ. of Pisa, and his colleagues showed that there was approximately a 1-in-a-billion chance of AN10 passing through a keyhole during a 2027 close approach so that it would impact the Earth several decades later. Rather than publicizing this unremarkable result (a thousand times less likely to hit than XF11's might-have-been scenario), Milani sought a peer review. In late March 1999, he sent his manuscript, drafted for a professional journal, to several international colleagues (mostly members of the IAU Working Group on Near Earth Objects), requesting a reply within two weeks. Upon receiving no complaints, he posted his paper (Milani et al. 1999a) -- but with no fanfare -- on his web site.

Milani intended his actions to demonstrate a responsible way to handle impact calculations, as a counterpoint to Marsden's handling of XF11. Indeed, officials of the IAU -- including IAU General Secretary Johannes Andersen -- were kept informed by Milani and they considered his effort as a forerunner of procedures that might be routinely adopted by the IAU for such matters. Unfortunately, in mid-April, Benny Peiser, the moderator of the CCNet Digest Internet forum, discovered the unheralded paper on Milani's web site and charged "cover-up." Needless to say, conspiratorial minded journalists, and others, picked up the story of AN10. So once again, although without XF11's page-one hype, newspapers around the world reported the slight possibility of an impact with yet another asteroid. A chance-in-a-billion of an impact would hardly seem to be news in a world fraught by war, disease, and accidents, but such is the result of mankind's inability to deal rationally with small probabilities (witness playing the lottery).

Milani arranged a more focussed challenge for attendees of the international IMPACT workshop (sponsored by the IAU, NASA, the European Space Agency, and other entities) held in Turin, Italy, the first week of June 1999. On the meeting's first day, Milani displayed a vu-graph covered by an opaque sheet; he claimed that it described yet another asteroid with a non-zero chance of impacting on a particular date in the next century. He asked that the group devise clear recommended guidelines, which he promised to follow. Should he just release the information, or should he submit his analysis for peer review? What specific criteria should he use to decide? The challenge helped inspire draft recommended procedures; on the final day, Milani applied them by immediately uncovering the vu-graph: by the recommended criteria, the object -- a recently discovered PHA named 1998 OX4 -- was both too small (a couple hundred meters in diameter) and too unlikely to hit (less than 1-in-a-million) to be worthy of peer review (or news media interest, for that matter). In essence, it scored as a zero in Binzel's new 10-point impact hazard scale, which I describe below.

Astronomers studying the chaotic and resonant behavior of Earth-approaching asteroids increasingly realize that the so-called background impact probability is in no way constant. In reality, it is represented by numerous "spikes" in probability for individual asteroids, which occur on specific future dates when their MOIDs are nearly zero and the asteroids may be near an orbital intersection point with the Earth. Our knowledge, today, of whether a particular asteroid has a chance of hitting on such a date is rendered uncertain by two factors. First, there are observational errors in the asteroid's orbit and position along its orbit, which can be improved by future (or past) telescopic positional measurements. Second, should there be a near-term close pass by the Earth, then the inherent chaos introduced may make it impossible to calculate future encounters deterministically, and the object will tend to rejoin the background.

In fact, 1998 OX4 is already "lost". While its orbit is well determined, its position in its orbit is very poorly known: it could be at virtually any orbital longitude, and hence would seem to have rejoined the background. However, Milani et al. (1999b) have developed a way to learn for certain whether OX4 will hit during the next century, without requiring an impractical search around 360 degrees of sky to recover OX4. They have calculated the very few places where OX4 could be in its orbit that would result in collision. So if astronomers look in just those few places and don't find OX4 there, then we can know that we are safe from its minuscule threat.

The studies of 1999 AN10 and 1998 OX4 have greatly helped astronomers' understanding of the nature of NEO encounters and of how to assess impact probabilities. At the same time, Milani's pedagogic use of these cases to explore ways of handling public release of impact predictions is helping entities like the IAU and NASA to develop protocols for assimilating and reporting such results.

The 1997 XF11 impact scare raises issues involving the affected communities -- asteroid astronomers, institutional and public officials, the news media, and the public -- concerning both their internal workings and their mutual interactions. This section evaluates several such issues, focussing on the astronomical community, which I know best.

Urgency vs. Peer Review. There is a fundamental dilemma within the predictive sciences concerning the balance between being careful and being timely. One wishes not to issue a hazard warning without good basis; yet, if one waits too long, it will be too late. While astronomers are academics where the only pressure to conclude a research project is "publish-or-perish," the IAU Circulars were created to handle matters of comparative urgency, though generally lacking practical importance. Some things change in the sky, and require observational follow-up quickly compared with the year it takes to publish a paper. Brian Marsden was operating within this tradition when he called for observations of XF11 in his March 11th IAUC. Although the IAUCs are a form of publication, they are issued quickly and generally without peer-review. Of course, they rarely deal with things of such popular and practical importance as a catastrophic impact.

As in the Swift-Tuttle case, Marsden was criticized because of the public announcement of a sensational, unchecked result that proved wrong. A lesson for astronomers surely is that the normal, out-of-the-limelight, ivory-tower protocols that work for informal communications among themselves have very different effects if the topic is inherently sensational. One could readily forgive Marsden for his haste had the prediction demanded an immediate response. But, with 30 years to go, there simply was no public urgency. In an astronomical context, it is always desirable to try to follow-up interesting NEOs quickly, since they may become much fainter within hours or days. (The problem is compounded by the inflexible tradition of assigning observing time on telescopes typically half-a-year in advance.) However, there was less urgency for XF11 because any new observations of it during spring 1998 (before it became unobservable near the Sun) would reduce errors little; moreover, as a relatively bright asteroid, it could readily be reobserved in later months or years.

More careful procedures should be adopted for communications among astronomers about inherently sensational topics. For example, Marsden could have omitted mentioning the 2028 encounter in the IAUC and simply sought additional observations for this "very interesting" asteroid. Or he could have just telephoned the several most likely observers who could make additional observations. Marsden might answer that there are many potentially interesting asteroids and he needed to highlight the reason for raising XF11's priority. Clearly, in the future such motivations should at least be weighed against the downsides of issuing a hasty, and inevitably public, announce- ment.

Technical Confusion and Meta-Errors. Specialists who calculate orbits normally have weeks or even years to do their work. The perceived urgency of the XF11 case resulted in unusually hasty, error-prone work, done under a media spotlight. It is human to err and scientists operating outside of their normal routines are bound to err as well, and they did. In such situations, researchers should realize that potential errors of interpretation or judgement might yield errors much larger than what are known as formal estimates of uncertainty. Until requirements of "checking one's math" and peer-review are met, there must be restraint in believing one's preliminary results, let alone broadcasting them to the world. Headlines in the New York Times should be reserved for dispassionate and well-checked conclusions, barring true urgency.

Consider Marsden's original report ("virtually certain" to be closer than the Moon), which was seriously erroneous. Marsden first responded that his calculations were technically correct, while admitting that his words had been "unfortunate". Later he confessed to initial misunderstandings of his technical calculations. The new Shelus data could have resulted in the nominal pass anywhere along the skinny error ellipse; what triggered Marsden's excitement -- that the center of the error ellipse happened to fall near the Earth -- was actually accidental and insignificant. Various uncertainties that could have affected anybody's calculations were not fully evaluated. The data sets available for analysis are not uniform, coming from various amateur and professional observers, subject to various systematic errors (e.g. reference star positions), and distributed peculiarly in time (e.g. due to the difficulties of observing during the two weeks when the Moon is bright). It can be misleading to apply formal statistics to evaluate uncertainties in orbits calculated from such disparate data. Yet scientists routinely report "formal" estimates of errors, occasionally multiplied by an arbitrary factor (like 3) to account for unknown errors. In the case of XF11, the errors reported by several calculators (in addition to Marsden) were sometimes too small.

There were technical reasons to quibble about exactly how close to zero the impact probability might be. Muninonen, for example, believe they highlighted technical differences between analytical calculations and his own Monte Carlo simulations. But such arguments confuse the public, and Marsden took advantage of the Yeomans/Chodas "that's zero, folks" 10^-300 impact probability. At such ridiculously small levels of probability, a host of what would be wholly negligible factors in the practical world become possible (Marsden pointed to passage of XF11 near another asteroid before 2028, or comet-like "non-gravitational" accelerations). Such arcane debates shifted the focus of the public (tuning in via the Internet and the Boston Globe) toward philosophical questions about the meaning of "zero" and away from whether a catastrophe might really occur 30 years hence.

Asteroid astronomers need to appreciate what many more practical scientists have long known: they must report uncertainties that take into account factors beyond just the formal uncertainties (which they often calculate within restricted contexts) that they normally discuss with each other. A public official needs to know what the real chances are of the predicted event happening, and that includes the chance, which is uncomfortable for a researcher to deal with, that he or she has made an error. For the impact hazard, where probabilities of only 1-in-a-million must be taken seriously because of the potential consequences, one must jump into an unfamiliar frame-of-mind about uncertainties. Human beings are familiar with estimating odds involving flips-of-a-coin, or even 1%, but they are demonstrably irrational when confronting the small odds that typify impact probabilities. The chances for computer errors, illness of the researcher, or a host of usually minor technical problems to happen are far larger than the infinitesimal chances of impacts; so it is irresponsible for a researcher to insist, for example, on the validity of calculated odds that are smaller than, say, 1-in-a-billion. Meta-errors must replace formal errors.

Centralized vs. Distributed Evaluation. Marsden has been the czar of NEO data, empowered by the IAU to run the MPC as the clearinghouse for all data. Yet he sifts through and validates the data, and sometimes even uses it for his own scientific purposes, before releasing it to the wider community. Marsden's scientific competitors have long argued that he should make all data immediately available, so that his work may be checked and so that the data can be rapidly used for other purposes. NASA, which has recently funded part of the MPC's operations, has tried to mandate such an open policy. The fact that the critical data that motivated Marsden's March 11th announcement were not yet public brought this matter to a head.

Several NEO researchers, primarily observers who provide Marsden with positions, have defended Marsden's March 11th IAUC and his unique role as guardian of the data. In the context of the themes of this volume, (e.g. "who becomes empowered when a prediction is made?"), it would note that these observers depend on Marsden to perform his traditional role. Through him, they have seen their observations integrated into something important (asteroid and comet orbits). Indeed, a chief reward for observers is having a comet named after its discoverers, and it is Marsden who judges and applies whose names (lately Marsden has de-emphasized such rewards). Several of these observers have prominently advocated on the Internet that Marsden should be praised, rather than criticized, for the XF11 announcement because the higher public visibility may yield increased funding of the observers' programs. (Despite such obvious self-interest, I believe that most such observers sincerely believe that the impact hazard is a serious threat and that it is in the public interest to augment the surveys.)

Clearly, if the data were available to all experts, independent verifications of potential impacts would be easier. Marsden's colleagues have different, independent, computer programs for calculating orbits and impact probabilities. As exemplified by Chodas' quick work upon receiving the full XF11 data, analytic calculations are fast and the more thorough Monte Carlo simulations only take a matter of hours to a day-or-so. Whether such checks and balances would actually have been done, had the data been available, is another matter. Thanks to a combination of lack of funding and organization, such work was not being done with the partial/delayed data Marsden was already posting. For example, Chodas' calculations made a few hours before receiving the full XF11 data, which were more than sufficient to rule out an impact, could have been done weeks earlier but simply weren't.

As a result of pressure since the XF11 affair, Marsden is making a larger fraction of NEO data available much more rapidly than before. Also, other researchers are now making routine calculations of future close encounters. But it remains to be determined whether formal (and funded) responsibility will stay with a single person or instead be distributed within the international community, as is now feasible, thanks to the Internet.

A process has developed within the IAU, assisted by Milani's handling of the AN10 and OX4 cases, to formalize some voluntary procedures for vetting a potentially sensational prediction in a timely peer review before announcement. While voluntary, there is little doubt that there would be strong peer-pressure and insistence by funding agencies to adhere to such policies.

Secrecy versus Peer Review. In Internet chats and newspaper editorials in the aftermath of the XF11 affair, some commentators worried about "secrecy" and urged that scientists immediately make impact calculations public. This view must be taken seriously for it resonates with conspiracy-oriented people who are unfamiliar with the scientific process. Moreover, traditional methods of ensuring credibility of scientific results are undergoing revision in the face of the realities of the Information Age.

The world has long been exposed to the uncensored "babble" of anyone who wants to say something and has access to a printing press. Yet society, and the scientific community, have developed ways to deal with it, to sort out the wheat from the chaff. With the advent of the Internet, the babble has become a roar and new methods must be developed. Traditionally, the scientific community has "certified" reputable results by having them checked, peer-reviewed, and published in technical journals. Also, by self-regulation (or imposed by editorial policies of journals like Nature), scientists don't go to the media until after the publication date of the peer-reviewed technical article.

Of course, such procedures don't prohibit lay people, amateurs, fortune-tellers, pseudo-scientists, and even professional scientists who don't care about their reputations from publishing their hasty results immediately -- on the Internet, or anywhere -- and such is the grist for supermarket tabloids. But serious people, opinion leaders, and policy makers rely on those media that attempt to abide by society's self-regulating procedures in order to report more reliably. Just days before the 1997 XF11 affair, there was Internet chat of a claim by unknown foreign scientists that the asteroid Icarus might impact Earth in the year 2006, but the false story generated few headlines around the world, presumably because the source was suspect and because such "official" sources as the MPC did not endorse the prediction.

The Minor Planet Center represents itself to be, should be, and was taken (by the media) to be a reliable source of information, representing the astronomical community. In that role, it must help formulate and adhere to modern procedures to ensure that centuries of traditional peer-review procedures are maintained in the Information Age. Conspiracy theorists will complain. But that is a small price to pay for helping scientists retain their current high level of public credibility. There is no possible public benefit from releasing incomplete, premature, unchecked scientific results prior to cross-checking and review.

Media relationships. A fractious mutual disrespect is common in relationships between scientists and journalists (Hartz & Chappell, 1998). Thus many astronomers, in the wake of the XF11 media frenzy, had the knee-jerk reaction that journalists were responsible for the sensationalism. While media accounts of XF11 were replete with the usual errors of the reporters' own making, the fundamental fault lay with the scientists. Given the statements in Marsden's PIS, and given that Marsden was as reputable and official a source on this topic as anyone in the world, the subsequent headline treatment was fully warranted. Even at small odds, the threatened impact in 2028 was clearly sensational on its face. The error was not only that the prediction was wrong but that simple, normal procedures (checking with colleagues) would have demonstrated the error before the announcement was made.

It is a common ivory-tower view that it is not the responsibility of scientists if the media listens in on scientists' traditionally-open, technical discussions and blows them out-of-proportion. Such views were expressed by several astronomers defending Marsden's actions in the XF11 affair. (This view would have a purer basis if Marsden had released only the IAU Circular and not the accompanying PIS.) This is not, however, the accepted view of ethicists who study the interaction of scientists with the media and the public. Scientists are obliged to be aware of the social context in which their work is done, and it is their responsibility to help minimize the chances of sensational misreporting of their work.

Astronomers working on topics of public interest must try to meet science journalists half way in order to help them present complex issues understandably. Yet journalists must also recognize that few astronomers are culturally attuned to effective interactions with the media, the way so many other public players (including scientists in more practical disciplines, like weather forecasting) are. As Hartz and Chappell (1998) recommend, professional societies (like the American Astronomical Society) could do much more to foster improved interactions with the media, including maintenance of useful web sites.

Uncertain Orbits and the Hazard Scale. Several asteroid experts have been concerned about the process that evolves from first discovery of an NEO to its eventually certain prediction that it will not (or unluckily will) hit the Earth, and how to evaluate and report such results. Bowell and Muinonen (1992) first presented a might-be scenario. This was amplified by Chodas & Yeomans (1997) and by myself in the first draft (July 1997) of this Case Study. Initial observations of an NEO permit only crude estimates to be made of its orbit, which might admit of a tiny chance of Earth impact at some moment in the future. Subsequent observations will shrink the size of the error ellipse (unless there is a close approach, generating chaos); the smaller ellipse will, almost always, exclude the Earth. In a highly unlikely case, the ellipse could shrink and yet include the Earth, which would raise the unexpected scenario of a very close approach or actual impact.

1997 XF11 presents an excellent example, of such an evolving situation, beginning with Marsden's late December 1997 listing of XF11 as a PHA and concluding with the March 12th evaluation of 1990 observations that ruled out any impact in 2028. Had anyone checked the data and made proper calculations, it would have been known from the outset that the Earth was never within the error ellipse in 2028. According to Chodas & Yeomans (1998), the highest impact probability their algorithm ever would have calculated subsequent to December 24th, as data trickled in, was 2x10^-30, based on the 65 observations available as of January 10th. Muinonen (1998a, 1998b), using the Monte Carlo technique, finds collision probabilities even closer to zero based on the early data.

Marsden and some NEO observers argue that the answer to any uncertainty in the potential close approach by an NEO is acquisition of new data, which never hurt. They argue that debates about what analysis of pre-March data on XF11 might have shown are moot: the correct thing to do, they say, is always to get more data and make sure. Unfortunately, there are unintended but inevitable consequences of using a prediction to motivate astronomers to obtain more data: e.g., scaring the world half to death. In future cases, it may be difficult to obtain timely data -- lacking prediscovery observations, the vagaries of weather and rapid motion of an NEO may delay successful observing for months. So it is vital to assess uncertainties of calculations based on partial data sets as the orbit of a new PHA is gradually improved. A lesson that should be learned from the XF11 affair -- obscured by the common media story that the prediscovery data were responsible for the corrections -- is that if the choice is to pay observers to make follow-up observations or to pay for better orbital calculations using existing data, the latter might often be more effective.

How do we best report to the public discovery of a new NEO that can't yet be ruled out as potentially impacting the Earth in the foreseeable future? While, as for XF11, errors may often be reduced and impact ruled out quickly, there will be cases (1998 OX4 is an example, although with very low probabilities -- less than 1-in-a-million -- of hits on 20 Jan. 2044 and again on 20 Jan. 2046) where ambiguities may persist for months or even years, depending on asteroid observability. Impact probabilities may be low during such a period of ambiguity, but so long as it isn't essentially zero (as it actually was for XF11), the situation will have a different tone from the purely statistical argument that an as-yet-undiscovered object might impact next year with some small probability. As with OX4, we would have a real, known body that might strike -- with some probability -- at an exact date in the foreseeable future. When does such a prediction merit the interest of the public, and how should such a prediction be presented?

Richard Binzel, of M.I.T., has attempted to condense the detailed complexities facing astronomers by employing a simple numerical scale, like the Richter scale used to measure earthquakes or the scale used to characterize hurricanes. He proposed (Binzel 1997; see also Verschuur 1998) a 5-point scale, and associated evaluation of concern, that is based on two vital measures of a specific impact threat: (a) probability of impact and (b) size, hence impact energy and resulting damage, of the impactor. Clearly, we are more concerned if the impact probability is high or if the threatening object is large -- a few km across as distinct from a meteoroid that will burn up harmlessly in the upper atmosphere. Another feature of a potential impact, not included in Binzel's scale, is how soon the predicted impact will occur. Obviously, the seriousness and urgency of responding to a predicted impact depends on whether it will occur in a year, a decade, or a millenni- um.

Some controversy about the scale and its implementation has delayed its adoption. For instance, should the scale involve the integers, or should decimal numbers be allowed? Some want to normalize the scale to the mean background probability of impact, so that it registers above nominal only if the chance of a particular impact exceeds that for a random impact by an unknown body. Others worry that it would be confusing to have a specific asteroid change values on the scale, as it quickly evolves toward 0 or (conceivably) 5, although hurricanes do just that.

More recently, in consultation with colleagues and science journalists, Binzel has modified his scale and made it explicitly applicable to evaluating predictions of specific impacts that might occur during just the next century, not beyond. It is now a ten-point scale (see Fig. 2) and emphasizes the definitions of 0 and 1, the only numbers likely to be used, which are defined with respect to the background annual chances of impact; the boundary between 0 and 1 is 1/10th the annual chance of impact by a kilometer or larger asteroid, or 1 chance in a million (a common rule-of-thumb threshold for significance in the field of risk assessment; cf. Okrent, 1987). The higher numbers are defined primarily to serve a pedagogic purpose, as users learn about the impact hazard by trying out "what if" scenarios. The scale was endorsed by the Turin IMPACT workshop in June 1999; accordingly, the scale is named the Torino Scale. Whether it will be adopted by official bodies and actually used by astronomers and science journalists remains to be seen. Furthermore, just when and how it should be used remains problematical, as illustrated by Binzel's judgement that XF11 was an astonishingly high 3.5 on his original 5-point scale, due to unquestioning application of the erroneous estimate of 0.1% chance of impact.

NASA and Scientific Community Reactions. NASA has walked a tightrope with regard to the impact hazard. Initially NASA, perhaps scared of the "giggle factor" that suffused news reports of the impact hazard in the early 1990s, responded minimally to calls by Congress that it study the hazard and propose a program to deal with it. A previous Associate Administrator for Space Science regarded the impact hazard as lacking scientific importance; other parts of NASA were happy to leave "planetary defense" to the Defense Department. After the XF11 scare, NASA's stance shifted as asteroid impacts became the second most common subject of public communications directed to NASA Administrator Dan Goldin (after the "Face on Mars"). In his May 1998 testimony to the House Subcommittee on Space and Aeronautics, Carl Pilcher, director of NASA's Solar System Exploration program, committed NASA to achieving the goals of the Spaceguard Survey -- discovery of 90% of Earth-approaching asteroids larger than 1 km by 2009. NASA claims that it has doubled its previous ~$1.5 million annual funding of NEO search programs; that amount probably needs to be doubled again if Spaceguard goals are to be met (Chapman 1998). In fact, there continues to be a mismatch -- complained about by several Congressmen in the 1998 hearings -- between NASA's promised level of funding and that actually spent: only about $700K of new money was said to be available for NEO research and search programs for FY 2000. Harris (1999) believes that the current discovery rate lags what is needed by a factor of ~8; an effective acceleration, which probably requires constructing one or more larger, 2-meter class wide-field telescopes, is unlikely to be accommodated within NASA's present spending profile.

NASA likes to say that it is fostering NEO research in the form of spacecraft missions to asteroids and comets, although the scientific goals that justified those missions have little to do with the impact hazard. Pilcher attended the June 1999 IMPACT workshop, obviously hoping to deflect pressure from NASA and encourage broader international funding contributions toward what is, obviously, a global issue. Meanwhile, the U.S. Air Force -- which already operates the element of Spaceguard, LINEAR, that is discovering more NEAs than all other worldwide search programs combined -- may choose to undertake the Spaceguard Program on its own. However NASA might respond to such a move, there would be misgivings by civilian astronomers in the U.S. and consternation on the part of non-American scientists.

Pilcher has expressed more concern about smaller impactors than the kilometer-scale objects being sought by Spaceguard. Those, of course, are more likely to occur "on the watch" of politicians and public officials, despite the fact that once-per-century, several megaton impacts constitute only about 1% to the total expected mortality of all natural hazards (floods, earthquakes, etc.) of similar potency (Morrison et al. 1994). In fact, NASA is even more concerned about impact predictions, since the XF11 event caught NASA officials pretty much off-guard. And predictions, erroneous and otherwise, will occur with much greater frequency than the 1-in-a-hundred-thousand-years between actual impacts with kilometer-scale bodies.

The impact hazard raises issues involving the interface between scientists and public policy. Some are similar to issues presented by other natural hazards, although the specifics often differ because of the unusual character of the impact hazard. Because the impact hazard has only been recognized recently, and chiefly within a scientific discipline outside of the existing natural hazards community, there has been very little integration of this topic into the existing private and governmental structures that deal with natural hazards and their mitigation.

I have had very few inquiries from entities involved in hazard mitigation and analysis. In the early 1990s, I received an inquiry about the impact hazard from the German insurance consortium, Münchener Rückversigherungs-Gesselshaft, which analyzes natural hazards. I have had one telephone call from a FEMA official. Even within the earth-and-space science community, the impact hazard is often overlooked; in late 1996, I learned that a committee of the American Geophysical Union was drafting recommendations about hazards and I was able, at the last minute, to get language into the adopted resolution that included the impact hazard.

Inevitably, the hazards community will be confronted by the impact hazard, especially as the now mostly-unknown potential impactors become discovered. About 290 NEAs, perhaps 15% or 20% of the total population larger than 1 km diameter, have been discovered (the total is roughly 1600; Harris 1999) plus 10 Earth-approaching short-period comets. That means that the vast majority of asteroids large enough to potentially threaten civilization have not yet been discovered and virtually none of the comets. Should the Spaceguard Survey be undertaken at full strength, that situation will change over the next decade to one in which most of the potential impactors will have been surveyed and it will have been reliably determined whether or not (almost certainly not, but if so, then when and where) any of them will impact Earth during the next century. Even if the Spaceguard Survey is not formally undertaken, rapid advances in astronomical instrumentation -- including that available to amateur astronomers -- combined with the augmented interest in this topic is likely to accomplish the Survey's objectives in the next several decades.

During this time of augmented discovery, there could be many "false alarms," especially because of the period of uncertainty between first discovery and eventual determination that an impact is impossible. Even though the odds are strongly (1-in-1000) against the possibility that even one such large object will be found to be in an orbit that will actually strike Earth within the next century, some will be found that will appear to have, for a period of days (conceivably years), a non-zero (even if very low) probability of collision with the Earth. Such a scenario is even more likely for the many smaller objects that Spaceguard will inevitably discover in the course of searching for kilometer-sized bodies. Indeed, the size of such objects may be uncertain for some time and several factors (observational biases as well as psychological factors) are likely to cause small objects to be reported as being much bigger, hence more dangerous, than they really are.

In addition to rare, real threats, the Swift-Tuttle and 1997 XF11 cases illustrate that well-intentioned experts may erroneously announce a potential danger even when it does not actually exist. Less reliable people, who nevertheless are -- or represent themselves to be -- genuine scientists, frequently make bad predictions. Several years ago, a well known French astronomer was quoted in normally reliable news media as predicting that the NEA Toutatis might collide with the Earth. Competition among different observing programs and among different orbit-calculators also inspires poorly qualified press releases. And then the media often mis-report even accurately qualified scientific announcements (a recent example was a CNN report in July 1997 that an asteroid 3/5ths of a mile across would "destroy the Earth" -- it resulted from oversimplification of a longer, more accurate CNN story based on a scientifically factual press release by E. Helin).

Preliminary attempts to establish regular communications links and coordination between the potential discoverers of NEOs (professional and amateur astronomers in the U.S., Japan, France, Russia, China, and other countries, as well as Air Force observers) and those who calculate orbits and collision probabilities are in their infancy. In the current, uncoordinated environment, competitive pressures actually work against efficient attainment of Spaceguard goals. Observers feel rewarded for the most discoveries, which is accomplished by "skimming the cream" of asteroids near the opposition point (directly opposite the Sun in the sky, where asteroids are brightest) and ignoring the sky along the Earth's orbit (where similar-sized asteroids are fainter, but which includes a larger fraction of bodies that might actually hit the Earth). Furthermore, the diverse telescopic resources being applied to Spaceguard cannot be used optimally if the leading program -- currently LINEAR -- doesn't leave some of the sky for them to search in. Analysis of the total system, sharing information about observing plans and actual sky coverage, and active coordination are among the first steps toward rectifying this situation, and those are goals of NASA's newly formed NEO Program Office at JPL, headed by Donald Yeomans. Perhaps the rewards of participation in the Spaceguard Survey can be restructured so that competitive motivations can be channeled more constructively toward meeting the project's goals.

Communications and coordination between such astronomers and those public/military officials who might be called on to respond to potentially threatening impacts remains wholly ad hoc. Fortunately, it is much more likely that a threatening object will be found decades, rather than hours or weeks, before impact (as exemplified by XF11's 30-year advance warning). But, especially if sizes or overestimated or if exaggerated attention is given to small objects, imminent impacts could well find themselves thrust into the laps of unprepared high government officials. The fall of Skylab and the more recent (1996) crash of a Russian would-be Mars probe in the southern hemisphere exemplify how even minuscule sprace debris can create a sensation. It is reliably known that the White House was alerted in 1994 when a modest atmospheric impact (10's to 100's of kilotons) was detected by downward-looking surveillance satellites (the event is described technically by McCord et al., 1995). Exploding meteors, with yields approaching a megaton, enter the upper atmosphere occasionally, and are spectacular even if they do little or no damage on the ground. Indeed, with widening media interest in the topic, even modest impacts -- of rocks that yield meteorites on the ground -- sometimes generate major media reports (such common fireball or bolide events are dramatic, but are widely misinterpreted by non-professional observers). Bolides or actual ground impacts by meteorites could even be misinterpreted as aggressive military activities (Chyba et al. 1998).

We should plan in advance for how public and military officials should act in the face of a genuine near miss, with attendant media and public reactions. For instance, there is roughly a 5% chance that an object at least 1 km in diameter really will, sometime in the next century, pass as close to the Earth as Marsden originally reported for XF11. And a couple times a decade, an object large enough to cause a Tunguska-like event (the 1908 15-megaton asteroid explosion over Siberia) passes within the same distance. While astronomers might consider these to be comfortably safe encounters, political and military leaders might not take such a dispassionate view. Efforts might be mounted to "do something" (expensively or even dangerously) about such close approaches, despite astronomers' assurances.

There may already have been significant tsunamis in human history caused by impacts into the ocean, but whose causes went unrecognized. From data given by Toon et al. (1997), it seems plausible that every thousand years, or so, an impact (probably by an iron projectile with kinetic energy of about a megaton) would cause a tsunami equivalent to the biggest recorded during the 20th century. Warnings of impact-caused tsunamis could be prepared, but the infrastructure for doing so is now lacking. (Of course, impacts contribute minimally to generating normal tidal waves, but the very biggest ones on a time scale of centuries will be disproportionately caused by impacts.)

In summary, the impact hazard has failed to become incorporated into the natural hazard policy arena because of: the newness with which the hazard has been recognized; its unusual high consequence, low probability character; the wide gap between astronomers and the natural hazards community; the reluctance, in an era of budget-cutting, of federal agencies to begin new research programs; controversial issues related to mitigation (i.e. diverting an incoming object with nuclear explosives; cf. Sagan & Ostro 1994, Harris et al. 1994, Morrison & Teller 1994); and self-interested motives of both astronomers and ex-Cold Warrior technologists that have distorted projects proposed for funding.

If policy officials and the public become sufficiently educated about the impact hazard to make an informed decision about what to do, then what, if anything, should be done? Several years ago, Fortune magazine published a cost/benefit analysis demonstrating that it was worth spending hundreds of millions of dollars a year on planetary defense. Some advocates of planetary defense within the DoE community have argued, in effect, that billions should be spent annually. Gerrard and Barber (1997) show how one could even defend an annual expenditure of $16 to $32 billion, thousands of times what is now being spent. However, others argue that the probability of impact is so low (often incorrectly thought of as being equivalent to a long waiting time until the next impact: see below) that few public funds should be expended on this hazard but instead on other hazards that certainly will happen (even if with more modest consequences) during the next years.

It is unusually difficult to think rationally about how to deal with this unusual hazard. A common mistake is to think that there is no reason to deal with it during our lifetimes (or for a politician to think it is irrelevant to his or her "watch") because there is a long waiting time until the next impact. The chances of a civilization-threatening impact are extremely low during a politician's term, but they are the same that it will occur on "her watch" as on her successor's...or during a few-year period several hundred thousand years from now. People build in floodplains saying, "the hundred year flood occurred here a few years ago, so I need not worry." It is a faulty argument. A valid argument is that the chances of a flood (or an impact) are so low that we choose to do nothing (or little) to avoid the consequences and, instead, spend our funds to avoid lesser but more certain consequences (e.g. investing in automobile safety).

Another choice is to let future generations, with their no-doubt superior knowledge and technology, deal with the problem. But that choice avoids responsibility for consequences that could happen on our watch. Future generations cannot deal with an impact that happens during the next 30 years. If we have failed to search for the impactors that may hit during our lifetimes, then not only we, but future generations will suffer if civilization is knocked back into the Dark Ages (or, heaven forbid, our species goes the way of the dinosaurs) because we failed to discover the object that actually will hit in 2028. To be sure, the odds are so extremely small that such a thing will happen during our lifetimes that it may well be justifiable to do little or nothing about it. The choice is necessarily subjective, but the issues need to be engaged.

Because of little research on fundamental questions about the impact hazard, it is understandable that there has been no formal analysis by an advisory body (like the National Academy of Sciences) or by governmental policy-makers about the priority of dealing with this threat. (An NAS/NRC report [COMPLEX, 1998] on Near Earth Objects, recently released, explicitly chose not to address measures motivated by the impact hazard, as distinct from scientific research on the origin and nature of NEOs.) It seems to me that, at the very least, a conscious decision should be made by society about whether to deal with the impact hazard or not.

Astronomy has, traditionally, been the quintessential predictive science of unexcelled precision. For millennia, astronomers have been famously successful in predicting eclipses, the Sun and Moon unerringly rise and set on time, and planetary spacecraft reach their distant targets within seconds of the times predicted. Of course, other astronomical predictions are much less secure (e.g. predictions of the date of the next "great" Leonid meteor shower, of solar flares, or of how bright a comet may become). Still other aspects of astronomy are so uncertain as to preclude prediction. The impact hazard, since it is based on the same kind of physics that successfully gets spacecraft to planetary encounters, would seem to be fairly foolproof, but the 1997 XF11 affair illustrates the kinds of complexities and errors that can make it seem unreliable as well.

Public officials rarely have to deal with astronomical predictions. Powerful solar flares are the chief exception. The impact hazard may be another. A general lesson from this Case Study is that whenever the public is introduced to a "new" hazard or a new type of prediction, one must expect a rocky road as scientists, who have been unfamiliar with communicating their predictions in ways that are useful to the public or to public officials, learn how to do it. On the other side of the coin, the particular journalists and public officials called on to deal with a new arena may be those (in this case, astronomy journalists and NASA officials) who have not previously been engaged in dealing with such issues. They need to learn, as well.

There is a spectrum of responses to various hazards on the part of the public and policy officials that ranges from great concern to minimal concern. The impact hazard generates especially strongly held but opposite views ranging from "this is the most important issue affecting the human species because it is the only one that could actually result in the near-term eradication of our species" to the view that "it is ridiculous to waste a moment thinking about such an unlikely catastrophe." As with mad-cow disease, the danger of nuclear plant melt-downs, or alleged terrorist sabotage of pills or grapes, the impact hazard Case Study demonstrates that public perceptions and media treatment of technical issues tend to overwhelm the "objective" technical facts and details. Astronomers have a responsibility to learn how to frame the technical facts in ways that are useful and understandable to policy officials, so that the officials remain grounded in the technical realities while they also respond to public opinion, no matter how "irrational" it may seem.

Many of the "lessons learned" from the predictive sciences project reported on in this volume appear to be applicable to the impact hazard. A prediction itself (as in the case of 1997 XF11) may dominantly affect public perception of the entire issue. There are issues of multiple ways of calculating the predictions, assessing uncertainties (precision and accuracy), reporting predictions and uncertainties, being aware of societal and international contexts, potential conflicts of interest (and noting what people and agendas are promoted by predictions), and so on. While attempts by scientists to educate the public about the objective aspects of hazards should proceed, the narrowness of scientists' awareness of the societal and political context of their work means that the goal of greater understanding is often not met. The ever more rapid and superficial exchange of information fostered by popular culture, educational institutions, and the Internet is increasingly inimicable to the kind of in-depth understanding of complex technical issues on which society's future depends.

It is disappointing that some technical experts in NEOs as well as many of the best science writers in the world, largely failed to understand the essential realities of how the false prediction of XF11 was made and retracted. A year after the XF11 frenzy, the predominant view remains that the matter was resolved by finding prediscovery observations of the asteroid on existing plates. One might conclude from that false perception that more attention should be given to procedures that will increase the efficiency of finding images in plate archives. Actually, the predominant failings were (a) failure for 2½ months of the NEO community to apply existing software to data that were publicly available, which would have demonstrated that the asteroid presented no hazard, and (b) failure of individuals and institutions to adopt and adhere to procedures that would increase the likelihood that public announcements would be valid instead of erroneous. We may hope that there are few other arenas in which public policy officials, doing the best they can, nevertheless adopt "solutions" that are only tangential to the technical realities of the associated problems.

I hope that the kind of interdisciplinary interchanges developed in the Predictive Sciences project will foster greater understanding between those who make predictions and those who use them.

I thank David Morrison for his direct and indirect assistance. Discussions with Alan Harris, Ted Bowell, Rick Binzel, Paul Chodas, Don Yeomans, Hal Levison, and numerous other colleagues in astronomy, as well as participants in the two Prediction Workshops, have helped me to formulate the issues that face us.

References

Binzel, R.P. 1997. A Near-Earth Object hazard index. In Near-Earth Objects: The United Nations International Conference (ed. J. Remo, Annals N.Y. Acad. Sci. 822) 545-551.

Binzel, R.P., M. A'Hearn, C.R. Chapman, C. Chyba, A.L. Cochran, A.W. Harris, D.K. Yeomans, B.G. Marsden, R.L. Millis & D. Morrison 1999. From the pragmatic to the fundamental: the scientific case for near-Earth object surveys. Report prepared for the Astronomy and Astrophysics Survey Committee (AASC), 10 May 1999.

Bowell, E. & K. Muinonen 1992. The end of the world: An orbital uncertainty analysis of a close asteroid encounter. Bull. Am. Astron. Soc. 24 965.

Browne, M. 1998. Old photos helped refine progress of asteroid. New York Times, 14 March.

Chapman, C.R. 1998. The threat of impact by near-Earth asteroids (www.boulder.swri.edu/clark/hr.html) and Action plan statement to: House Subcommittee on Space & Aeronautics (www.boulder.swri.edu/clark/actnea.html).

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

Chodas, P.W. 1999. Presentation at AAS Division on Dynamical Astronomy meeting, Estes Park CO, April.

Chodas, P.W. & D. Yeomans 1997. Impact warning times for Earth Crossing Asteroids. Bull. Am. Astron. Soc. 29 960.

Chodas, P.W. & D. Yeomans 1998. Could asteroid 1997 XF11 collide with Earth? Bull. Am. Astron. Soc. 30, 1029-1030.

Chodas, P.W. & D.K. Yeomans 1999. Orbit determination and estimation of impact probability for near-Earth objects. Presented at "21st Annual American Astronautical Society Guidance and Control Conference", Breckenridge CO, 3-7 February.

Chyba, C.F., G.E. van der Wink, C.B. Hennet 1998. Monitoring the Comprehensive Test Ban Treaty: possible ambiguities due to meteorite impacts. Geophys. Res. Lett. 25 191-194.

COMPLEX 1998. Exploration of Near Earth Objects (Committee on Planetary and Lunar Exploration, National Research Council, National Academy Press, Washington D.C.) 44 pp.

Gardner, M. 1998. Near-Earth objects: monsters of doom? Skeptical Inquirer, July/August issue.

Gehrels, T. (Ed.) 1994. Hazards due to Comets & Asteroids (Univ. Ariz. Press, Tucson), 1300 pp.

Gerrard, M.B. and A. W. Barber 1997. Asteroids and comets: U.S. and international law and the lowest-probability, highest consequence risk. N.Y.U. Environmental Law Journal 6 (1), 4-49.

Gladstone, B. 1998. Report on Morning Edition, National Public Radio, 4 June 1998.

Goldman, S.J. 1998. The most dangerous rocks in space. http://impact.skypub.com/rocks.html (from Sky & Tel., June 1998, 33).

Gordon, B.B. 1998. The asteroid caper, editorial in Astronomy, July issue, pg. 6.

Harris, A.W. 1999. Near-Earth asteroid surveys. In Collisional Processes in the Solar System (H. Rickman and M. Marov, eds., Kluwer ASSL series, Dordrecht) in press.

Harris, A.W., G.H. Canavan, C. Sagan & S. Ostro 1994. The deflection dilemma: use versus misuse of technologies for avoiding interplanetary collision hazards. In Hazards due to Comets & Asteroids (ed. T. Gehrels, Univ. Ariz. Press) 1145-1156.

Hartz, J. and R. Chappell 1998. Worlds Apart: How the Distance Between Science and Journalism Threatens America's Future (First Amendment Center, Freedom Forum, Nashville TN), 177 pp.

Kunich, J.C. 1998. Posted on World Wide Web, August 1998: sac.saic.com/space_warfare/planlrev.htm.

Marsden, B.G. 1999. Presentation at AAS Division on Dynamical Astronomy meeting, Estes Park CO, April.

McCord, T.B. et al. 1995. Detection of a meteoroid entry into the Earth's atmosphere on February 1, 1994. J. Geophys. Res. 100 3245-3249.

Milani, A., S.R. Chesley & G.B. Valsecchi 1999a. Close approaches of asteroid 1999 AN10: resonant and non-resonant returns. Astron. Astrophys. 346, L65-L68.

Milani, A., S.R. Chesley, A. Boattini & G.B. Valsecchi 1999b. Virtual impactors: search and destroy. Preprint, submitted for publication.

Morrison, D. (chair) 1992. The Spaceguard Survey: Report of the NASA International Near-Earth-Object Detection Workshop (JPL/NASA, Jan. 25, 1992).

Morrison, D., C.R. Chapman & P. Slovic 1994. The impact hazard. In Hazards due to Comets & Asteroids (ed. T. Gehrels, Univ. Ariz. Press) 59-91.

Morrison, D. and E. Teller 1994. The impact hazard: issues for the future. In Hazards due to Comets & Asteroids (ed. T. Gehrels, Univ. Ariz. Press) 1135-1144.

Muinonen, K. 1998a. Asteroid and comet encounters with the Earth: Impact hazard and collision probability. In Proceedings of The Dynamics of Small Bodies in the Solar System: A Major Key to Solar System Studies (eds. A.E. Roy and B.A.Steves, NATO Advanced Study Institute, Maratea, Italy) in press.

Muinonen, K. 1998b. Upper bounds for the Earth-XF₁₁ collision probability. Bull. Am. Astron. Soc., in press.

Okrent, D. 1987. The safety goals of the U.S. Nuclear Regulatory Commission. Science 236, 296-300.

Park, R.L. 1992. Star warriors on sky patrol. New York Times (March 25, 1992) A19.

Rather, J.D.G., J. H. Rahe, and G. Canavan (organizers) 1992. Summary Report of the Near-Earth-Object Interception Workshop (JPL/NASA, Aug. 31, 1992).

Remo, J.L. (Ed.) 1997. Near-Earth Objects: The United Nations International Conference (Annals of the N.Y. Acad. of Sci. 822), 632 pp.

Sagan, C. and S. Ostro 1994. Dangers of asteroid deflection. Nature 369 501.

Shoemaker, E.M. et al. 1995. Report of the Near-Earth Object Survey Working Group (NASA).

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

Toon, O.B. et al. 1997. Environmental perturbations caused by the impacts of asteroids and comets. Reviews of Geophys. 35 41-78.

Verschuur, G.L. 1998. Impact hazards: truth and consequences. http://impact.skypub.com/page1.html (from Sky & Tel., June 1998, 26-34).

Fig. 1. (a) Near-Earth section of uncertainty ellipse for 1997 XF11's 2028 passage, calculated from positions obtained Dec. 1997 - Mar. 1998 (magnification of next panel). The asteroid cannot impact Earth. (b) Full uncertainty ellipse based on positions obtained Dec. 1997 - Mar. 1998. Asteroid could pass well beyond the Moon's orbit. (c) Full uncertainty ellipse calculated with older 1990 positions included. It falls within the uncertainty ellipse shown in previous panel, but it is now clear that the asteroid will pass far beyond the Moon's orbit. From Chodas & Yeomans (1999).

Fig. 2. Revised Torino Scale for rating the seriousness of concern for impact predictions that may be made for specific dates in the 21st century. Associated with each number, 0 through 10, is a one or two sentence description of the category. Courtesy R. Binzel.