Trans-Neptunian objects (TNOs) have historically proven challenging to discover partly because they appear faint. The upcoming Legacy Survey of Space and Time (LSST) is expected to discover some 30,000 TNOs over 10 years (Ivezíc et al., 2019), imaging the entire southern sky every three nights to a single exposure depth of . However, it is possible to probe far beyond the single exposure signal limit by employing “shift-and-stack,” a procedure that co-adds exposures from the same region of sky after offsetting each image to align with the orbit of a hypothetical moving object. Kernel-Based Moving Object Detection (KBMOD; Whidden et al. 2019; Smotherman et al. 2021) is a software package we developed that performs upwards of 109 such shift-and-stack searches per minute using GPUs. We demonstrated KBMOD's abilities by discovering >100 TNOs through the DECam Ecliptic Exploration Project (DEEP; Trilling et al., 2023; Trujillo et al., 2023; Bernardinelli et al., 2023; Smotherman et al., 2023; see also presentation by David Trilling). However, LSST will deliver orders of magnitude more data over a larger spatial and temporal baseline, so we set out to prepare for the forthcoming deluge.
We present a new strategy for discovering TNOs in astronomical image data with shift-and-stack techniques, especially applicable to large-scale datasets and enabling searches over longer timespans (i.e., >7 days). Our approach starts with a novel, tiered query method that efficiently identifies and assembles candidate images. Notably, our procedure accounts for Earth's motion around the Sun by transforming image coordinates to the solar system barycentric reference frame. This paradigm facilitates image selection beyond a few days. Secondly, we reproject images into the parallax-corrected reference frame, thereby better linearizing on-sky motion and simplifying the search parameter space. Lastly, we report on challenges in scalability, as well as approaches to alleviate some of these hurdles.