|Title:||Projected Near-Earth Object Discovery Performance of the Large Synoptic Survey Telescope|
|Author(s):||Steven R. Chesley and Peter Veres|
|Reference:||JPL Publication 16-11 (April 2017)|
|Source File:||Download (PDF)|
This report describes the methodology and results of an assessment study of the performance of the Large Synoptic Survey Telescope (LSST) in its planned efforts to detect and catalog near-Earth objects (NEOs).
LSST is a major, joint effort of the US National Science Foundation and the Department of Energy, with significant support from private donors. The project has a number of key science goals, and among them is the objective of cataloging the solar system, including NEOs. LSST is designed for rapid, wide-field, faint surveying of the night sky, and thus has an 8.4m primary mirror, with 3.2 Gigapixels covering a 9.5 deg2 field of view. The system is projected to reach a faint limit of V ≃ 25 in a 30-second exposure visit to a given field and perform nearly 2.5 million visits in its 10-year survey.
The baseline LSST survey approach is designed to make two visits to a given field in a given night, leading to two possible NEO detections per night. These nightly pairs must be linked across nights to derive orbits of moving objects. However, the presence of false detections in the data stream leads to the possibility of high rates of false tracklets, and the ensuing risk that the resulting orbit catalog may be contaminated by false orbits. NEO surveys to date have successfully eliminated this risk by making 3–5 visits per night to obtain confirming detections so that the single-night string of detections has a high reliability. The traditional approach is robust, at the expense of reduced sky coverage and a diminished discovery rate. The baseline LSST approach, in contrast, is potentially fragile to large numbers of false detections, but maximizes the survey performance.
One of our key objectives was to investigate this fragility by conducting high-fidelity linkage tests on a full-density simulated LSST detection stream. We also sought to quantify the overall performance of LSST as an NEO discovery system, under the hypothesis that the NEO detections arising from the baseline LSST survey observing cadence can be successfully linked.
We used the latest instantiation of the LSST baseline survey and the most current NEO population model to derive the fraction of NEOs detected and cataloged by LSST from among the source population. As a part of this we developed a high-fidelity detection model that accurately represented the LSST focal plane and implemented a smooth degradation in detection efficiency near the limiting magnitude, rather than the usual step function. The study carefully modeled losses from trailed detections associated with fast moving objects, and we investigated other minor effects, such as telescope vignetting, and asteroid colors and light curves.
For the linking tests, we included all major sources of detections for a single selected observing cycle (full moon to full moon), leading to 66 million detections, of which 77% were false detections, 23% were main-belt asteroids, and only 0.14% were NEOs. Using only a single 8-core workstation we were able to successfully link 94% of potential NEO discoveries in the detection stream. We are confident that with appropriately-sized computation resources, some algorithmic improvements and careful tuning of the linking algorithms the linking efficiency can be significantly improved. In this simulation, 96% of objects in the NEO catalog were correctly linked. Of the 4% that involved erroneous linkages, almost all comprised detections of two distinct main-belt asteroids. This situation, known as “main-belt confusion” is fundamental to the NEO search problem and is readily resolved as the main-belt asteroid catalog becomes filled in over time. Despite the 77% rate of false detections, less than 0.1% of derived NEOs in this simulation included false detections. From these results, within the study hypotheses, we conclude that the two visits-per-night observation cadence can be successful in cataloging NEOs. This conclusion does assume a certain rate of false positives, but is unlikely to be sensitive to increases by factors of a few in the false detection rate, given the significant computational resources allocated by LSST for the problem.
Our simulations revealed that in 10 years LSST would catalog ∼ 60% of NEOs with absolute magnitude H < 22, which is a proxy for 140m and larger objects. This results neglects linking losses and the contribution of any other NEO surveys. Including our worst-case linking efficiency we reach a overall performance assessment of 55% completeness of NEOs with H < 22. We estimate that survey mis-modeling could account for systematic errors of up to 5%. We find that restricting the evaluation metric to so-called Potentially Hazardous Asteroids (PHAs) increases the completeness by 3–4%, and that including the benefits of past and expected future NEO survey activity increases completeness at the end of the baseline LSST survey by 15-20%. Assembling these results leads to a projection that by the end of the baseline LSST survey the NEO catalog will be 80±5% complete for PHAs with H < 22.
As described in detail in the report, these results are largely consistent with other results obtained independently; indeed the small 1–2% variation among independent estimates of NEO completeness is remarkable and reassuring. The results above require pairs of observations in three distinct nights over no more than 12 days. A maximum linking interval of 20 days, for which high linking efficiency has not been demonstrated, leads to a 2–3% improvement in completeness. We also tested a special-purpose LSST cadence designed to enhance the NEO discovery rate, but our results show little improvement over the baseline for a ten-year survey. Surveying longer does provide an increase in completeness, by roughly 2% per year for a lone LSST and 1% per year when including contributions from other surveys. Thus, in our judgement, the H < 22 PHA catalog can be expected to approach ∼ 85% completeness, but not 90%, after 12 years of LSST operation.