Same old tale of an asteroid: 2001 PM9
(or the worst holiday of my life)
by Andrea Milani *- Copyright Tumbling Stone 2001
The story begins on the morning of Saturday 11 August 2001, Central European Time (CET), but in Arizona it is full night, and the new asteroid observing station on Mount Palomar, where the NEAT camera has started working just a few months ago, is doing its job very efficiently. One of the discoveries of this night moves faster than allowed for a main belt asteroid, thus the data are immediately forwarded to the Minor Planet Center, where it is posted on the Near Earth Objects Confirmation Page.
In the evening (CET) of the same day, 11 August, in the island of Elba, I am going home on a Suzuki motorbike when the front wheel slips underneath the bike. Do not ask how it happened, because I am the first one who would like to know: maybe the road was abnormally slippery, maybe I am too old for motorbiking. By the time my wife Anna arrives, I am lying flat on the road, and the ambulance has already been called for: I am in rather bad shape.
|The same evening, but later, around midnight (CET), in the Czech Republic, Jana Ticha, Milos Tichy, and a coworker (J. Jelinek) are at their 57 cm telescope to recover the Near Earth Asteroid discovered few hours before by NEAT, and as usual they are successful. Little more than one hour later the same object is observed from the nearby Ondrejov, and the next day, when the night sets on the American continent, both NEAT telescopes in Palomar and in Hawaii are at it again. There are observations from an amateur observatory, Powell Observatory, Louisburg and in the evening of Sunday, 12 August Ticha and Tichy observe it again. There are now 22 individual data points, from 5 observatories, over a time span of 2 days. Gareth Williams, of the Minor Planet Center, decides that this is enough to announce the discovery of a new Near Earth Asteroid (NEA): only at this point the object receives its 'designation', 2001PM9, by which it will be known from now on (and until it changes again name, when it is numbered). The Minor Planet Electronic Circular (MPEC) to announce the discovery, prepared by Williams late in the evening of Sunday, is disseminated through internet and by the time it is received the morning of Monday has risen in Europe.||
At 10 AM (CET) the NEODyS site at the University of Pisa updates the list of NEA by processing all the MPEC received in the last 24 hours. In the process, it computes the Minimum Orbital Intersection Distance (MOID) and the orbital uncertainty, and as a function of these two quantities it updates a priority list for impact monitoring: with a MOID of the order of the distance to the Moon, and a very poor orbit resulting from an observation arc of 2 days only, 2001 PM9 gets immediately to the top of the list. At 13:00 another computer program, called CLOMON, following the indications of the priority list begins to scan for all possible impacts of 2001 PM9 with the Earth between now and the year 2080. At 17:46 CLOMON sends to all the coworkers of NEODyS the warning message, with:
|Subject: POSSIBLE COLLISION DETECTED:
Please see 2001PM9.new
You may think this is scary, but we get this kind of messages almost daily. Most of them are false alarms, that is cases where a close approach within the distance of the Moon is possible, but upon examination by a human operator the output file .new turns out to indicate that a collision is not compatible with the available observations. Only in a few cases another file, the .risk file, is generated with indication of possible encounters within 1 Earth radius, which means collisions. Unfortunately, in this case the 2001PM9.risk file not only exists, but contains the indication of a possible impact in the year 2003. The probability of such an event, as estimated on the basis of the knowledge available at that time, is minute, something like 1 in 10 millions, but the date is so close, and the diameter of the asteroid can be estimated to be between 500 and 1100 meters, that is large enough for catastrophic effects.
So far the story appears to contain a positive
message: the system has been working, the problem of this
"Virtual Impactor (VI)" (click here to know more about Virtual impactors
(dict.)) is being taken care of,
everybody involved has done his job in a very professional way,
why should you worry? And indeed nobody of the people involved so
far, at the observatories, at the MPC and in the NEODyS team, has
even thought of sounding some kind of alarm, because the problem
is very likely to be solved in a routine way: as new observations
come in, the scanning for impacts will be repeated with the
reduced orbital uncertainty, and in most cases the VI just
vanish. The devil is in these three words: "in most
On the morning of Tuesday, 14 August, new observations of 2001 PM9 received in another MPEC are processed by NEODyS at 10:00. As expected, the uncertainty decreases significantly, but has the VI vanished? The software robot CLOMON would rerun 2001 PM9 at 13:00, but in this special case somebody gets impatient and launches manually the impact scan at 10:38. You may wonder who did this, since I was still on the Elba island, laying in bed with ice on my knee. Actually, who are the people behind NEODyS ?
On 14 August, the day in which I should have been back from my motorbike weekend, Prof. Sansaturio, universally known as Genny, remotely connected from Valladolid (Spain), is running NEODyS. Under normal circumstances, she can always consult me if there is something she is not sure about. In particular, she has never taken care by herself of a Virtual Impactor crisis: with my continuing absence, on 14 August she is for the first time fully in charge. At 13:37 the scan of the future close approaches of 2001 PM9 is complete, and unfortunately the worst has happened: the possibility of impact in 2003 is also compatible with the observations over a 3 days arc, and the probability has increased to about one in 5 millions. This is the way it must be: if a different set of observations is used, the probability, must change; in most cases the probability decreases to zero, but in rarer cases it first increases, to decrease later (again, in most cases) when further observations are available. Genny consults Steve Chesley, who is following the situation from JPL.
A quick computation done by Steve shows that the Torino scale (dict.) of this possible event is still zero.
Thus the conclusion reached by Genny and Steve on 14 August is the following: since the case of 2001 PM9 is Torino scale zero, there is no need to inform the IAU officials and to start a formal IAU technical review before any announcement. Since the asteroid is still easily observable, there is no need to notify the observers that they must observe it: they will do this anyway. On the other hand, it could be extremely harmful to keep the information on the possibility of impact as a secret: this could be known later, and rated as cover up. as lack of respect for the right to know of the public, and could even fuel some of the conspiracy theories which are so popular among the lunatic fringe. Thus they decide to exhibit the 2001PM9.risk file on the 'risk page' of the NEODyS web site.
Of course Genny knows that, even if she is acting without being able to consult me, I will stand by her because it is my responsibility anyway. Still she is quite nervous about it, and she decides she has to do something more that just stand by and wait for new observations. The opportunity to do so has to do with another island: Mallorca, where she has met Jaime Nomen, who is in charge of an amateur observatory near Tarragona, on the Spanish coast: Ametlla del Mar (obs. code 946). He has only a 40 cm telescope, and 2001 PM9 is fainter than magnitude 19, but when he receives from Genny the request to observe, and the epehmerides, he is ready to try. Without repeating their story (see this issue of T.S. "Feelings of an amateur observer of NEAs" by Jaime Nomen), it is enough to say that the run of CLOMON concluded on 16 August shows that the observations from Ametlla del Mar are enough to contradict the possibility of an impact in 2003, but they leave open the possibility of impacts in either 2005 or 2007. There is also a possibilty of impact in the year 2042, with a probability of 1 in 1 million.
I am still out of reach, but Genny and Steve have been able to contact Giovanni Valsecchi, who is on holiday on still another Mediterranean beach. He supports the idea of exhibiting on the NEODyS risk page the output of the latest CLOMON run, but when requested of activating the Spaceguard Foundation Central Node to request more observations he is in difficulty, because everybody is on holiday. 15 August, or Ferragosto, is the most sacred holiday in the Italian calendar, really nobody works. Only the next day, 16 August, Valsecchi is able to get in touch with Andrea Boattini, who is in San Marcello Pistoiese (obs. code 104) to observe, and indeed will try to observe 2001 PM9, although he also has only a 40 cm telescope.
The JPL group, which of course was kept informed by Chesley, has also computed the possible close approaches of 2001 PM9, by using a Monte Carlo, a method very different from the one of CLOMON. On 16 August, Paul Chodas confirms that the possibility of impact in either 2003 or 2005 did exist, but he is uneasy with the new data: "With the Aug. 15 data, the probability has nearly gone down, but I have not computed a precise new value because Steve and I are not very happy about the large residuals on the new observations." Indeed, the observations from the amateur observatories, for an asteroid at the limit of detectability with such small telescopes, have quite large residuals, some even have to be discarded. Now this can appear a very technical point, but let me try to explain. The probabilities of impact we compute are anyway nothing but the transformation of the probabilities we attach to observation errors. E.g., if we believe that the astrometric observations cannot have residuals of more than 2 arcsec, and the orbit leading to a collision leaves residuals of 3 arcsec, then we can say that such an orbit is incompatible with the observations. But, if we are unsure about the quality of the observations, because they have been obtained under particularly difficult conditions by telescopes too small for the task, how can we claim that such a lower quality fit is impossible? We would like to have some professional, comparatively large telescope to provide more reliable data. This should not be understood as lack of appreciation for the efforts done by the amateurs, but we need to be sure before proclaiming that impacts can be excluded, for the very good reason that after such an announcement less attention would be dedicated to that asteroid, and if the "all clear" announcement had been premature, we would take the risk of losing a pontentially dangerous fellow.
On the evening of 16 August I finally limp back to Pisa (by train). As soon as I am again able to connect to internet from my home computer, I try to figure out what has happened from the messages and also by phone. The situation is still fluid, in that we still have Virtual Impactors which have not been contradicted by the observations, but other observations are coming in and they have to beprocessed. Normally this is done only once a day, not only because of the CPU time required for the computations, but because the MPC supplies observations of NEA only once per day. As an alternative, we can run NEODyS and CLOMON manually when we receive some observations directly from the observatories, but this involves more possibility of mistakes. Thus I take the heroic decision of doing nothing until we can look at the new computations based upon the data available from the MPC on the morning of 17 August, in the meantime consulting all the coworkers about what to do. The problem is always the same: should we try to do something more visible, to obtain more collaboration, also from professional observatories with larger telescopes, or should we keep quiet and wait for the observations to come in 'naturally'?
Maybe I should take this opportunity to reveal one of those 'secrets' which are not a secret at all for the experts. We can find out exactly who and when has accessed our risk page (or any other web page in our server) by consulting the log of the http daemon. Thus we were aware that no one, in the professional astronomical community, had taken any notice of our posting, although it had been out for more than 2 days. That is, although we were not keeping it secret, it was nevertheless not known at all. Is it our responsibility to wake up the members of the asteroid scientific community? Or is it their responsibility to follow up a site which is well known to have sometimes important announcements? I do not know the answer to these questions.
On 17 August the MPC provides additional observations of 2001 PM9 made on 15 August from Sormano (an amateur, but especially well organized, observatory in Italy), but the ones from San Marcello are not yet ready (because Boattini needs to use a tricky procedure, coadding the frames, to detect the faint target; the new ones from Ametlla del Mar are also not there, because of the bizarre problem encountered by Jaime (see this issue of T.S. "Feelings of an amateur observer of NEAs" by Jaime Nomen). When the CLOMON run ends at 13:42, the results are quite discouraging. The 2042 Virtual Impactor is gone, but the 2005 and 2007 ones are still there, and of course their probability has increased, to 1.7 and 1.3 per million, respectively.
I decide to spend a couple of hours analysing the results on the basis of the 'Palermo' scale (dict.). The results are quite bad: the best estimate of the impact energy for this asteroid, using 'average' albedo and density, is 11,700 MegaTons (MT). Thus the 'expected energy' for the VI of 2005 and 2007 are 20 KiloTons (KT) and 15 KT, respectively. The expected energy is a tricky concept, and is obtained by multiplying the energy of the impact, in case it takes place, by the probability that it indeed occurs. Then we can compare the specific risk of each Virtual Impactor to the risk for our planet to be hit by an object with this energy or larger within the time until the date of possible impact. In this case the conclusion is the following: the 2005 VI represents 2.5 % of the background risk (until 2005), while the 2007 one represents 1.3 % of the background risk (until 2007!). This is rated on the Palermo scale as -1.59 and -1.89 (0 would correspond to an event fully equivalent to the entire background risk until that time).
These numbers do not look much by themselves, but they are impressive if they are compared with the previous cases (click here to know more about XF11 and AN10).
In simple words, 2001 PM9 is a delicate case: the Torino scale 0 lets us make an announcement without going through the formal IAU procedure, but this does not allow us to take it lightly.
At this point I formally ask Valsecchi and Boattini to post the announcement of a new special observation campaign on the SGF web page; this they do in the same afternoon, by recalling from holidays G. D'Abramo. Then I write a message to Hans Rickman, General Secretary of the IAU, Dave Morrison, WGNEO chairman, and Rick Binzel, WGNEO secretary; copy also to Franco Pacini, IAU President. Beside explaining the 2001 PM9 situation to the IAU officials, I point out that this is not a formal IAU review procedure, but as a matter of common sense, if we agree that the proximity of the impact date is relevant in the perception of risk by the public (and by ourselves), this is a case of which they should be informed. This does not imply that some special action is immediately required by the IAU, although IAU support may be requested later if we do not manage to have enough response from the observers.
Saturday 18 August. CLOMON is running the new computations, including the observations by Boattini, Nomen and also from Garradd, and I am at my computer running a close approach analysis for 2001 PM9 until 2007, by using another program, this one an interactive, menu driven version which I normally use for development and verification of the output of the automated computations. The results are as bad as they can be. Not in the sense that the new observations have been useless: on the contrary, they have significantly improved the orbit, and indeed when the CLOMON run is complete at 12:28 the 2001PM9.risk file has disappeared: that is, our automatic impact monitoring system cannot find anymore Virtual Impactors.
Unfortunately, a close scrutiny of the output, as well as my interactive runs, indicate that this is not a reliable result. The problem of the data quality, of which we were already aware, has caught up with us. In simple words, we base our statements that some collision has been contradicted by additional observations on the fact that, to allow the orbit with collision to be the true solution for a given asteroid, we have to admit values of the residuals (observations minus prediction) larger than allowed by our understanding of the observation procedures. To do this we normally set the limit at a value of the conventional parameter sigma=3 (which sets the dimension of the region of uncertainty (dict.)). The collision in 2005 can occur at a value of sigma between 3 and 3.5. That is, residuals only slightly larger than those we normally allow would be compatible with an impact. But we do know that the observations of this object are of somewhat lower quality than usual, because after the first two nights it has been observed only with telescopes smaller than the ones which should be used for such a faint target. Thus the data from the amateur observatories are very useful, but they cannot entirely replace the ones from the observatories with bigger telescopes, which are still missing. Under these conditions, I cannot send out an all clear message: I am fully convinced that the probability of an impact in 2005 has decreased with respect to the computations of the previous day, but I am not in a position to assign a new value to the probability. Moreover, for 2007 the output of our automatic programs contains an error message, indicating that the program has failed to compute the minimum possible approach distance. This has a completely different cause, called interrupted resonant return. (click here to know more about resonance (dict.))
At 13:09 I send to all the
people involved in this discussion so far the following message:
"There is nothing we can do now, apart from 1) making more noise, in particular asking for professional observers with bigger telescopes to get more accurate observations [...] 2) developing in a rush a new theory, so that our monitoring, advertised as providing safety to our planet until 2080, can at least work for 2007. Maybe this kind of adrenalina-driven development of new theories and software, which was so successful in 1999, is what we need now. I am working on CLOMON2 this afternoon and tomorrow, Giovanni Valsecchi is going to be here on Monday. The development of the theory or resonant deflection maybe can wait until the end of next week." The last obscure phrase refers to the most pessimistic hypothesis: what if the worst case happens, that is if we were to end up by concluding that an impact in either 2005 or 2007 is possible, actually if such a possibility was confirmed by later observations until it becomes likely? The only way to protect our planet would be to exploit the 2003 close approach to apply a deflection impulse to the asteroid; but this requires to develop a suitable theory of such manoeuvers for a resonant orbit. This gives an idea of how bad was my mood. For my good luck, I tend to react to these crises in a positive way, by setting down to work. Maybe the work is made even more intense by the fact that I can hardly walk, so to sit in front of the computer all the weekend is the best choice to allow my knee some time to recover.
On Sunday, 19 August at 16:37 I send another, more optimistic message, for two reasons. First, new data, arrived with the morning MPEC, have further restricted the possibilities for close approaches in 2005 and 2007. Second, I have succeded in applying the unpublished theory of interrupted resonant returns to the 2007 close approach, and found that there is indeed a minimum possible distance even in that strange case. This minimum distance, also thanks to the new data, is quite large, more than 2000 Earth radii. Thus we can claim the crisis is over, at least for the years until 2008.
This apparent happy ending is, however, not the real end. To be able to remove the risk page from the NEODyS web site we need to be able to claim that all the impact possibilities until the year 2080 can now be dismissed. Although the Virtual Impactors for later years have lower probability, and of course a much lower rating in the Palermo scale, still they have been detected and we cannot stop our computations until we are able to dismiss them too. This requires to run our impact scan programs up to that time: but, should we run the old CLOMON, which did not work for 2007, or should we run the new CLOMON2 which is just beginning to work and has not been properly debugged? And can we draw these conclusions on the basis of the amateur observations only, knowing as we know that they are of lower accuracy than usual?
I understand that on this point there have been, and there will be again in the future, many polemics. But, from my point of view to declare that a Virtual Impactor crisis is over is a much more delicate step than opening such a crisis. When we have a non-empty .risk file, there is no question that we have to post it on the web and take the necessary actions to ensure that additional observations are obtained. To remove a .risk file from our web server we need to know that all the possibilities of impact listed in there have been reliably contradicted by observations. What are we supposed to do when 'some' of the impact possibilities have been contradicted, but for others the results are inconclusive, or even not yet available because our computations are not finished? Should we edit by hand the file and add some comments to some of the lines of the table? Are we supposed to make some announcement of the form "the worst possibilities of impact, the ones for the next 6 years, have been eliminated; for the others we are still investigating"? Who would feel reassured by this?
After some discussion with my coworkers, I decide to keep the 'old' .risk file on the web server. The point is that the crisis is not completely over, observations are still needed, and this is the message which we are sending by leaving some risk file there; the specific content of the file is not up to date, but this is stated in the file itself by a clear indication of the observations used in the scan. The SGF also decides that the special observation campaign for 2001 PM9 should not be declared closed at this time. Indeed the first observations of 2001 PM9 from a professional observatory (after the ones of the discovery and the first two days) have ben made in the night between Saturday and Sunday (by the Tichy), but they have not yet been disseminated by the MPC.
The impact scan computer programs, including the new CLOMON2 with its supposed capability to handle 'interrupted resonant returns', are running all the night between Sunday and Monday (understandably, the new program has not been optimized properly and is much less efficient). On Monday, 20 August, G. Valsecchi is with me in Pisa, and we analyse the output of the old and the new program. Although this is all new, and there are many points both in the theory and in the software which need to be clarified, in the afternoon we get to the conclusion that, even taking into account the incomplete theory of 'interrupted', even allowing some margins to account for the lower than usual accuracy of the observations, there are no possibilities of impact until 2080.
|Some comments by the
Andrea Milani * - Director of NEODyS
The images and the animations of this issue are courtesy of Jaime Nomen.