Driving Discovery in Astronomy Using Scalable Computing and Fast Algorithms

Abstract

The Vera C. Rubin Observatory's Legacy Survey of Space and Time (LSST) will produce and publicly release an imaging and catalog dataset larger than any before it. At a total data volume of 500 PB, it will only be possible to realize the full scientific potential of this survey if tools and algorithms are available that can scale to the size of the dataset. This thesis contributes to this effort through three key advancements. First, I demonstrate how cloud-based science platforms can be used to access and analyze large catalogs using a distributed computing framework. The distributed computing tools utilized are user-friendly and accessible, allowing astronomers to scale their workflows to the entire LSST catalog. Second, I illustrate how the LSST Science Pipelines--a set of tools and algorithms for processing astronomical images--can be applied to a wide-field imaging survey, verifying their scalability and applicability in processing the LSST imaging dataset. I also provide tools that enable users and research groups to perform their own image processing campaigns. Finally, I introduce an optimized version of the shift-and-stack algorithm, enhancing its efficiency for detecting faint solar system objects in multi-epoch imaging surveys. This improved algorithm provides speedup relative to a naive implementation. This optimization lays the foundation for application of shift-and-stack to the entire LSST dataset, particularly in the search for faint inner solar system objects.