1 Introduction

The collision between the Earth and an asteroid can be described as an extreme event. On one hand, it is extremely rare: although tons of material enter the Earth’s atmosphere on a daily basis, humankind has not yet witnessed the impact of a body belonging to the asteroid class. On the other hand, the effects of such an impact could be extremely catastrophic: the energy released in the process ranges from tens of Megatons (for 50-m bodies) to millions of Megatons (for bodies with several Kilometers in diameter), reaching anyway global consequences for asteroids with diameter equal or greater than 1 km.

The extreme character of the impacts between the Earth and the asteroids makes it difficult for the general public to understand the true nature of the problem we are facing. The fact that this type of collisions are rare does not mean that they are impossible. In fact, there are evidences both indirect (craters on the surfaces of the rocky celestial bodies, including the Earth) and direct (collision of the comet Shomaker–Levy 9 with Jupiter in 1994) showing that this kind of episodes have occurred in the past and will take place again in the future.

Among the population of asteroids only those with orbits close to that of the Earth, which are known as Near Earth Asteroids (NEAs), represent a real risk. There is a number of different measures that can be taken to mitigate this hazard. The most basic measures have a preventative nature and aim at cataloging the whole NEA population. A complete catalog of accurate orbits would allow us to know, well in advance, the asteroids that are on a collision course. The second type of measures, more selective and targeted for a particular threatening asteroid, include different deflection techniques to avoid its collision with the Earth. Finally, if the previous actions fail, it is still possible to implement a third kind of measures, such as population evacuations before the impact or construction of refuges near the shock area, to mitigate the effects of a certain collision. It is worth noting that the success of the measures of the second and third kind heavily relies on the time interval ranging from the publication of a certain impact to the impact itself. Obviously, those predictions can only be made when there is enough information available on the asteroids, hence the importance of the cataloging.

This paper is focused on the real case provided by the asteroid initially known as 2004 MN4 and later numbered (99942) and named Apophis. According to recent measurements, this object has a diameter around 250 m and an estimated mass of 2.1  ×  1010 kg, which is derived from an assumed density. The collision problem posed by this asteroid has not been fully solved yet. However, the largest threat raised by Apophis, a possible collision in the year 2029, was completely removed by using the first preventative method mentioned above. The collaborative work to rule out that impact provides a great example of the good practices needed to handle a delicate situation in which the information that becomes public must be carefully presented and commented to avoid unnecessary alarms.

The paper is organized as follows: in Sect. 2 we provide an introductory description of the NEODyS/CLOMON2 system. The risks scales used in assessing the hazard posed by an asteroid potentially colliding with the Earth are presented in Sect. 3. In the next Section we review the collision scenario for Apophis during the Christmas holidays in 2004 from the perspective of the NEODyS team. The last Section collects some concluding remarks that can be extracted from this experience.

2 The NEODyS/CLOMON2 System

Let us start by presenting a brief tour through the NEODyS/CLOMON2 system to help understand the context in which the computations for 2004 MN4 were done.

NEODyS (Chesley and Milani 1999), which is an acronym for Near Earth Objects Dynamic Site, is a database containing all the orbital information concerning the currently known NEAs and was created in 1999, almost simultaneously with the NASA NEO Program Office at the Jet Propulsion Laboratory (JPL). It is accessible on the Internet at the following web addresses:

The first corresponds to a server physically located in Italy, at the University of Pisa, while the second one is a duplicate maintained at the University of Valladolid in Spain. Both together make NEODyS a very reliable service, almost completely independent from possible computer failures, network disruptions or other technical events, due to its location in different parts of the world.

The system input is taken from the Minor Planet Electronic Circulars (MPECs), sent via e-mail by the Minor Planet Center (MPC). The data is processed by means of different perl and shell scripts and fortran programs (OrbFit Footnote 1). Its main tasks are scheduled as follows:

  • 10:00 am (CET): Update of the data, orbits and priority list,

  • 12:00 am (CET): Update of the ephemerides,

  • 13:00 am (CET): Impact monitoring.

Most of the system runs automatically, although some parts still require human intervention.

Through the web interface of the NEODyS system it is possible to access a large amount of information such as the list of all known NEAs, the list of observatory codes and their statistics, as well as a selected collection of related sites. Concerning a particular NEA, you will find all the orbital information including orbital elements with uncertainty, information about the orbital fit and a list of close approaches to the terrestrial planets and Jupiter, as well as to Ceres, Pallas and Vesta, from 1950 to 2100. It is important to bear in mind that all this information refers to the nominal solution, that is the best fit in the sense of leasts squares. On the other hand, NEODyS offers interesting services from the observer point of view: prediction of ephemerides and observations or tools for searching the database. However, what makes NEODyS different from other systems is what is known as Risk Page Footnote 2. This service lists all the objects for which it has not yet been possible to discard the possibility of impacting the Earth until 2080. In reality, its content is not but the final result of the impact monitoring program CLOMON2, which was the last feature added to NEODyS at the end of 1999.

The CLOMON2 (Milani et al. 2005a) program consists of three stages:

  1. 1.

    Generation of multiple Virtual Asteroids (VAs) (Milani et al. 2005b), in our case 2401, that sample the confidence region.

  2. 2.

    Propagation of these VAs until 2080.

  3. 3.

    Analysis of the close approaches of all these propagated orbits.

On average, the run of CLOMON2 takes about 2–3 h per object, although for some complicated cases it can even last more than one day. The risk table posted on the web is, in fact, the output of the third stage and the meaning of the different columns can be found in the help available in the same page. Currently, CLOMON2 and Sentry Footnote 3 are the only two impact monitoring systems that operate in a systematic and automatic way on a daily basis. Regarding automation, it is important to note that the NEODyS/CLOMON2 system does not post any risk table automatically, but there is always human intervention. On the contrary, JPL/Sentry posts automatically the risk tables containing low risk Virtual Impactors (VIs), that is with Palermo Scale (PS) below  −2 and zero Torino Scale (TS), while human intervention is required for the others. This is so because, according to the IAU standard, we have to undergo through the Technical Verification Procedure for every event with PS above  −2 and we also consult for whatever is TS  >  0. Note that this verification is always slowed down by the local time differences.

3 Risk Scales

The risk scales just mentioned try to quantify in a simple way the risk associated with the possible collision of an asteroid with the Earth and its purpose is to serve as a mean of communication for the astronomers and the general public when evaluating the seriousness of the potential collisions.

In general terms, we can state that when evaluating the risk of an impact at least three factors have to be considered:

  • on one hand it is obvious that the risk has to be an increasing function of the probability of the impact, in such a way that less probable impacts lead to low risks and viceversa.

  • On the other hand, a risk scale must be sensitive to the energy released in the collision. This quantity is directly related to the size of the asteroid and its relative velocity with respect to our planet.

  • Finally, we cannot neglect the time left till the predicted possible impact since, for instance, a well in advance prediction would allow us to adopt measures aimed at the mitigation of the possible effects of the threat.

The first risk scale to be introduced was the Torino Scale (Binzel 2000). It uses numbers from 0 to 10 in combination with colors and words to classify the impact risks (Fig. 1). This scale is discrete and the fact that it takes into account the third factor only in a binary way Footnote 4 makes it a bit inconvenient, so that a new scale was proposed: the Palermo Scale (Chesley et al. 2002). This scale is continuous, includes the three forementioned factors and is mainly used among professionals. Basically, the PS compares the destructive effect of a certain impact with that of the whole population of asteroids, both known and unknown.

Fig. 1
figure 1

A simple graphical tool to compute the number in the Torino Scale and hence the associated color and word to classify the risk from white (no risk) to red (certain impact) passing through green, yellow and orange

Full size image

Most of the asteroids posted in the Risk Pages of NEODyS and JPL are TS = 0 and PS < −2. The first notorious case was 2002 NT7, since although it was TS = 1—but very close to the TS = 2 region, on July 23, 2002—it was the first asteroid to reach a positive PS value. 2002 NT7 held this PS record until our protagonist 2004 MN4 came into scene beating any previous record, both in the Torino and the Palermo scales.

4 The Apophis Case

On December 20, 2004 the MPC issued MPEC-Y25 (Gilmore et al. 2004) containing observations over three consecutive nights of a new Potentially Hazardous Asteroid (PHA). Figure 2 shows the relative position of this asteroid and the Earth when Gordon Garradd observed it from Siding Spring on December 18. However, it was not the first time this object had been observed. Indeed, it had been discovered on June 19, 2004 by Roy Tucker, David Tholen and Fabrizio Bernardi from Kitt Peak (see also Fig. 2), who observed it over two consecutive nights by using an instrument never previously used to find asteroids.

Fig. 2
figure 2

Position of the Earth (dark sphere) and Apophis (light sphere) on their respective orbits when the asteroid was discovered on June 19, 2004 (right) and rediscovered on December 18, 2004 (left). In addition to the orbits, the plots also show the ascending (white) and descending (black) nodes of Apophis joined by the nodal line

Full size image

As a result, the processing of the astrometry and photometry had some problems related to distortions in the large CCD mosaic and clock errors. Then, when the asteroid was serendipitously recovered by G. Garradd it was soon found to be the same as 2004 MN4, but the fit was very poor for the June data as shown in Fig. 3.

Fig. 3
figure 3

Depiction of the residuals in right ascension (RA), declination (DEC) and magnitude (MAG) corresponding to the data available on December 20. In the case of RA and DEC the grey strip has a width of 2 arcsec, while for MAG the width is 2 magnitude units

Full size image

When CLOMON2 ran on this data it found a VI in 2029 with a few 10−4 Impact Probability (IP) and TS = 1. In the afternoon of December 20 CET, we discussed the results with the JPL people. Both monitors were showing the VI in 2029 in a reasonable, though not perfect, agreement. In view of the poor fit (Fig. 3), we considered the possibility of impact in 2029 a dubious result and decided not to post it in our Risk Pages, waiting for new and/or corrected data, which we expected soon, since Steve Chesley, Sentry’s lead member, had contacted D. Tholen concerning the problematic June data and asked for new observations to a group of observers.

This call for observations had no response on December 21, but the following day four new observations from E12 were published and we received the remeasurements of the June 19 data. The improvement for these observations was clear (Fig. 4) but the fit of the “uncorrected” June 20 data became even worse. When we processed all the available data, CLOMON2 got even more worrying results than the previous day, the 2029 VI reaching TS = 2 and PS = +0.72. After consultation with the Sentry team, we decided to wait for the second half of Tholen’s work to be completed.

Fig. 4
figure 4

Comparison of the residuals of the discovery observations of Apophis. The fit on December 20 is at the top. The middle plot represents the fit after correcting the observations of the first night. The fit at the bottom includes also the correction in the data of the second night. Note that the scale measuring the residuals has been shrunk in the second case to show the markers correspoding to outliers

Full size image

On December 23, Tholen obtained accurate remeasurements for all the June Kitt Peak observations (Fig. 4), sent them to us and we processed them. The results of CLOMON2 were a little less bad than the previous day, but still at the TS = 2 and PS  >  0 level. The agreement between CLOMON2 and Sentry was good to the point that there was no way to cast into doubt the existence of the 2029 VI, although the quality and time distribution of the data was not at all what we would have liked. At that point it was decided to post the results in the Risk Pages simultaneously and including a note in each of the two web sites stressing that the situation was bound to change as new data became available.

At the end of the day, a few dozen additional observations came directly to us from Robert McNaught. They were incorporated to the orbital solution and, in the evening (CET), we posted an update of the results.

On Christmas’ Eve, the new available observations gave rise to new outputs of CLOMON2 with unprecedented TS = 4 and PS > +1. We posted the results near 20:00 CET and continued to make a call for observations.

During the following two days (25–26) new observations started to flow in. The MPC issued four special MPECs (three on the 25th and one on the 26th) for 2004 MN4. Due to the seriousness of the case, we realized we needed to apply some weighting (Carpino et al. 2003) to the observations due to the large number of observations coming from some observatories. M. E. Sansaturio carried out those discussions with Chesley through e-mail, despite the holidays and the local time difference.

There is an important aspect to comment at this point: during those days the impact monitoring tasks involved not just running the existing computer programs and doing simple discussions on the residuals in order to get the best orbital solution. On the contrary, the NEODyS team was furiously changing all parts of the software which were not performing as expected because of the new features of the case. Certainly, 2004 MN4 has proven to be a challenge for both monitoring systems.

CLOMON2 had a mistake in the definition of the Target Plane (TP) coordinates, irrelevant but for exceptionally low velocity at infinity and exceptionally high IP. Sentry also had some problems, but they appeared in late January 2005, when the first radar measurements were got, also because of some new features of the 2004 MN4 encounter in 2029. Thus, the software was being changed on the fly and under a great preasure.

Finally, it is also relevant here to remember that, on December 26, a dramatic tsunami took place in Asia. This and the fact that we were immerse in the core of the Christmas holidays help to understand why this outstanding case hardly had media coverage.

On December 27, we got the maximum IP ever: the infamous 1 in 38 chances of impact. The high IP can be easily explained by looking at the situation in the Target Plane (Fig. 5).

Fig. 5
figure 5

The picture on the left shows the trail left on the TP by the set of VAs that generates the 2029 VI. The picture on the right shows a detail of the trail in the neighborhood of the Earth. In both plots the circle represents the impact cross section of the Earth once the gravitational focusing has been taken into account

Full size image

On one hand, there is almost no dispersion of the Line Of Variations prior to the impact. The close approach shows all the 2401 VAs almost perfectly aligned. On the other hand, there are many VAs within a distance of 0.7 Earth radii.

This was the situation at 14:00 CET when CLOMON2 finished its run. However, the MPC issued 4 new special MPECs. In particular, MPEC-Y70 (Gleason et al. 2004) contained pre-discovery observations for 2004 MN4 from observatory code 691 (Spacewatch), which extended the arc back to March 2004. In addition, almost at the same time, Tholen sent time corrections to the June 19 observations. We processed this new data and started a new run of CLOMON2 near 23:00 CET. The run finished at 1:37 December 28, the most outstanding result being that the 2029 VI had been ruled out, which brought an end to the “2004 MN4 crisis”.

Nevertheless, there is something to be noted about the March Spacewatch observations. In late January 2005, when the first radar data points were got for Apophis, it was apparent that these observations were not compatible with the radar data, the miss distance being adjusted by a factor two closer to Earth. According to the information provided by the Spacewatch team at the time, these images were found even though the ephemeris prediction plus trailing losses placed the asteroid well below the detection threshold for the frames. Chesley contacted the Spacewatch team to ask them to revisit the measurements, which they did and even had an external astrometric reduction (by the MPC staff Tim Spahr) done for verification. The corrected March observations provided by Spahr were then properly weighted and incorporated in the orbital solution.

However, it is important to point out that the problems exhibited by the March observations were only detected a posteriori when the orbital solution included the radar data points. Contrary to what happened with the discovery observations from Kitt Peak, for which the poor fit was evident from the very first day, the March observations showed no clear indication of their low quality in the fit. Thus, no action to “improve” them was taken when they were made available by the MPC. In any case, the corresponding computations lead to the conclusion that the radar observations obtained at the end of January 2005 would have ruled out the 2029 VI, irrespective of the March data weighting.

Finally, apart from the already discarded 2029 VI, there were some other low risk VIs out of which the one in 2036 has persisted till nowadays with a current impact probability of one in 45,000. If not before, it is expected (Chesley 2006) that new radar measurements in 2013 will exclude the whole set of VIs associated to this asteroid.

5 Conclusions

Table 1 provides a brief summary of the story of the 2029 VI of Apophis, showing the evolution of the risk published on the NEODyS Risk Page from December 20 to December 27, 2004. It exhibits the usual pattern followed by a large VI that is not a real impactor: tipically, obtaining the very first new observations reduces the size of the uncertainty region, although the VI is still there and therefore the IP grows. When the number (and the accuracy) of the observations is sufficiently large the confidence region becomes so small that the VI is excluded, finally dropping the IP to zero.

Table 1 Evolution of the 2029 VI of Apophis
Full size table

From a scientific point of view the Apophis case was really extraordinary: in only one week it generated more than 200 observations, more than 20 runs of CLOMON2 and hundreds of technical e-mails. To discard the possibility of the 2029 impact it was necesary to coordinate the specialized work of a lot of people, including both professionals and amateurs, distributed all around the globe. This is an excelent example of the good results that can be achieved with an efficient collaboration among the scientific community.

Finally, we want to stress that a great work has been done to assess the hazard posed by the NEA population and nowadays we know that it is really possible to put it under control. Also the public opinion is starting to be aware of the problem. It is now the responsibility of the governments to do the economic efforts to provide the scientific community with the necessary material means to complete the task.