More formally, the impact probabilities computed by the software behind this visualization tool are based on observation error statistics, assigned by means of an astrometric observations error model based upon Gaussian (normal) distributions. Because the number of individual observations for each object is very small, the law of large numbers does not apply. Thus the actual errors of the observations included in a single tracklet, normally between 3 and 5, may not be a representative sample of the corresponding random variable, which is normally distributed.
In simple words, the probability for a single tracklet to have large errors is small, but not as small as detecting an imminent impactor (less than 1 in a million tracklets). Thus some apparent detections of imminent impactors can be spurious, and there is no way to avoid this, short of abandoning the statistical error model and resorting to a careful human inspection of the image to reveal possible causes of degradation of the data.
NEOScan is a system dedicated to the scan of the Minor Planet Center NEO Confirmation Page (NEOCP). The goal is to identify asteroids as NEOs, MBOs or distant objects to confirm or remove from the NEOCP, and to give early warning of imminent impactors, to trigger immediately follow-up observations.
The service NEOScan is based on the algorithm presented in Spoto et al. (2018), which we are going to summarize.
- Scanning of the NEOCP every 2 minutes. New cases or old cases just updated are immediately run
- Computation and sampling of the Admissible Region using a 2-dimensional representation in the range/range-rate plane with a grid or a spider web (used when there exists a reliable nominal solution, it samples a neighborhood of the nominal range/range-rate by following the level curves of the quadratic approximation of the target function)
- Computation of the Manifold Of Variations, obtaining a set of virtual asteroids
- Propagation of the virtual asteroids in the future (currently for 30 days)
- Projection on the Modified Target Plane, searching for virtual impactors
- If virtual impactors exist, computation of the impact probability
We assign to each NEOCP object an integer flag related to
the computation of its impact probability. It is
called impact flag, and it depends on the impact
probability value and on the arc curvature, as shown in the
table below. We say that an arc has significant curvature if
χ2 > 10, where χ is the chi-value of the geodesic
curvature and the acceleration. The impact flag can take the
integer values from 0 to 4: a 0 value indicates a negligible
chance of collision with the Earth, whereas the maximum
value 4 express an elevated impact risk. It is conceived as
a simple and direct communication tool to assess the
importance of collision predictions, and to give the
priority for the follow-up activities.
|0||IP ≤ 10-6||Negligible|
|1||10-6 < IP ≤ 10-3||Very small|
|2||10-3 < IP ≤ 10-2||Small|
|3||IP > 10-2 and no significant curvature||Moderate|
|4||IP > 10-2 and significant curvature||Elevated|
When we perform the sampling of the Admissible Region, we also compute a score for the object. The score gives us a first insight into the nature of the object, even though the asteroid were not a potential impactor. The score is a rigorous probability of the object to belong to different classes (NEO, MBO, DO, and SO).
The systematic ranging allows one to express the orbital elements as a function of the range and the range-rate. Therefore, each orbit on the Manifold Of Variations can be matched to one of the four classes. Their definitions are listed below.
- Near Earth Object (NEO): q < 1.3 au
- Main Belt Object (MBO): (1.7 au < a < 4.5 au and e < 0.4) or (4.5 au < a < 5.5 au and e < 0.3)
- Distant Object (DO): q > 28 au
- Scattered Object (SO): anything else