An automated probabilistic asteroid prediscovery pipeline

TL;DR

This work tackles the challenge of prediscovering near-Earth asteroids by automating an end-to-end, probabilistic pipeline that propagates the orbital covariance from post-discovery MPC data back to archival survey epochs to generate sky regions for search. It combines robust orbit fitting, isometric sampling of the uncertainty region, low-threshold source catalogs, and a likelihood-ratio based linking framework to identify self-consistent prediscovery detections across multiple images, achieving significant arc extensions in real archival data. The authors demonstrate the approach with ZTF, recovering prediscoveries for 2021 DG1 and 2025 FU24 and showing potential for thousands of objects across the SBDB, with arc extensions up to factors of ~78 and substantial reductions in future sky-plane uncertainties. The method is survey-agnostic and scalable, enabling rapid orbit refinement for discoveries from Rubin, NEO Surveyor, and NEOMIR, thereby strengthening planetary defense capabilities by leveraging archival imaging for long-term orbital accuracy.

Abstract

We present an automated and probabilistic method to make prediscovery detections of near-Earth asteroids (NEAs) in archival survey images, with the goal of reducing orbital uncertainty immediately after discovery. We refit Minor Planet Center astrometry and propagate the full six-parameter covariance to survey epochs to define search regions. We build low-threshold source catalogs for viable images and evaluate every detected source in a search region as a candidate prediscovery. We eliminate false positives by refitting a new orbit to each candidate and probabilistically linking detections across images using a likelihood ratio. Applied to Zwicky Transient Facility (ZTF) imaging, we identify approximately 3000 recently discovered NEAs with prediscovery potential, including a doubling of the observational arc for about 500. We use archival ZTF imaging to make prediscovery detections of the potentially hazardous asteroid 2021 DG1, extending its arc by 2.5 years and reducing future apparition sky-plane uncertainty from many degrees to arcseconds. We also recover 2025 FU24 nearly 7 years before its first known observation, when its sky-plane uncertainty covers hundreds of square degrees across thousands of ZTF images. The method is survey-agnostic and scalable, enabling rapid orbit refinement for new discoveries from Rubin, NEO Surveyor, and NEOMIR.

Figures (6)

• Figure 1: A visualization of the prediscovery algorithm. The orbital uncertainty from post-discovery observations is propagated back in time to the survey to identify images and search regions within those images (black ellipses). Sources detected in the search regions are candidate prediscoveries. Each candidate source corresponds to a trial orbit, which is propagated to the other images to look for coincident detections. A true prediscovery will result in multiple detections along a single physical orbit.
• Figure 2: Prediscovery of the near-Earth asteroid 2021 DG1 in a ZTF image from August 2018, 2.5 years before its discovery date. Black points are locations of sample orbits consistent with the post-discovery observations. They are used to construct a convex hull (gray region) which forms the search region for this image. Image sources detected above $4\sigma$ significance are shown as green circles (filled if they are in the search region, empty if outside it). A detected source in another image is considered as a candidate prediscovery and the resulting trial orbit is propagated to this image. The inset shows the predicted location ($+$) and uncertainty (black contours) for this updated orbit. The trial orbit is consistent with the location of a source (green point) detected in this image, contributing greatly to this particular orbit's likelihood ratio.
• Figure 3: Potential arc extension ratios of 2,676 NEAs queried from JPL SBDB with at least 5 candidate images in ZTF. Arc extension ratio is defined as the potential increase in arc divided by the current arc.
• Figure 4: Predicted light curves of selected NEAs based on their post-discovery orbit fits. Gray points show all ZTF images that should contain the asteroid. Candidate intersections (blue) are ZTF images in which the asteroid is predicted to be brighter than the image's $3\sigma$ detection threshold. The presence of candidate intersections before the discovery date allows for significant arc extension. The points circled in red in the top two panels mark successful prediscovery detections in ZTF. They correspond to candidate sources with a likelihood significance greater than $10\sigma$ (see Fig. \ref{['fig:dg1_hist']}). The bottom two panels demonstrate null detections: we artificially brighten the reported magnitudes for 2022 DB4 and 2022 ED1 by 5.3 mag but they are in fact undetectable in ZTF.
• Figure 5: Prediscovery significance of candidate sources obtained from our pipeline for selected NEOs. Each panel shows a histogram of the standardized log-likelihood ratio (in units of standard deviations above the null hypothesis expectation) for orbits corresponding to trial sources detected in search regions. There is clear evidence of prediscovery for 2021 DG1 and 2025 FU24, where orbits successfully predict the asteroid's position across multiple images. The bulk of orbits with standardized log-likelihoods near 0 reflect the null distribution corresponding to spurious sources unrelated to the asteroids (orbits with significance less than $-5$ are not shown). The asteroids 2022 DB4 and 2022 ED4 are too faint to be detected in ZTF.