Real-time and retrospective discovery, anomaly detection, classification, and similarity search of supernovae in time-domain surveys and data streams

Aleo, Patrick David

Permalink

https://hdl.handle.net/2142/124516 Copy

Description

Title

Real-time and retrospective discovery, anomaly detection, classification, and similarity search of supernovae in time-domain surveys and data streams

Author(s)

Aleo, Patrick David

Issue Date

2024-04-19

Director of Research (if dissertation) or Advisor (if thesis)

Narayan, Gautham

Doctoral Committee Chair(s)

Kemball, Athol

Committee Member(s)

• Liu, Xin
• Villar, Ashley

Department of Study

Astronomy

Discipline

Astronomy

Degree Granting Institution

University of Illinois at Urbana-Champaign

Degree Name

Ph.D.

Degree Level

Dissertation

Keyword(s)

• Supernovae
• Time-domain Astronomy: Anomaly Detection
• Classification
• Similarity Search
• Discovery

Language

eng

Abstract

Since the first "guest stars" appeared in the night sky thousands of years ago, discoveries of astronomical transients have been largely serendipitous, as have many significant advancements in understanding their phenomenology. Now, in the big data era of synoptic surveys, nightly streams of hundreds of thousands (and soon millions) of time-domain alerts require a new stratagem to maximize transient discovery and the study of rare or poorly understood transient phenomena: real-time algorithms for classification, anomaly detection, and similarity searches. This Dissertation presents such a solution through a diptych of works. First, we present a rigorous study into the development of a low-latency algorithm for classification and the construction of the largest low-redshift, multi-band, and multi-survey data release of supernovae (SNe) ever---the first data release of the Young Supernova Experiment (YSE DR1). This comprehensive multi-survey data release combines Pan-STARRS1 griz and Zwicky Transient Facility (ZTF) gr photometry of 1975 transients with host–galaxy associations, redshifts, spectroscopic and/or photometric classifications, and additional data products from 2019 November 24 to 2021 December 20. When validating our photometric classifier on 472 spectroscopically classified YSE DR1 SNe, we achieve 82% accuracy across three SN classes (SNe Ia, II, Ib/Ic) and 90% accuracy across two SN classes (SNe Ia, core-collapse SNe). Our real-time classifier performs particularly well on SNe Ia, with high (>90%) individual completeness and purity, which will help build an anchor photometric SNe Ia sample for cosmology. We then use our classifier to characterize our photometric sample of 1483 SNe, labeling 1048 (~71%) SNe Ia, 339 (~23%) SNe II, and 96 (~6%) SNe Ib/Ic, a necessity of population studies in the age of large, wide-field surveys where spectroscopic labels are increasingly rare (~10% of all transients today, ~1% in the imminent future). \par Next, we develop Lightcurve Anomaly Identification and Similarity Search (LAISS), a pipeline for "automating serendipity" within the nightly ZTF Alert Stream via the ANTARES alert broker. LAISS leverages statistical light curve and host galaxy features to flag interesting and rare astronomical transients in real-time to expand our samples of rare SN classes and transients uncommonly found in particular host galaxy environments. Tuned for high purity such that each transient flagged is a likely bonafide anomaly, the number of candidates tagged per night is low (~1--5 objects), enabling an expert to follow-up a manageable list of transients given limited spectroscopic follow-up resources. Lastly, we demonstrate the power of approximate similarity search methods to find analogs of transients with similar light curve evolution and host galaxy environments. In total, we have reported the discovery of ~1000 transients and have flagged ~100 bonafide anomalous transients. These methods will transition us into the era of the Vera C. Rubin Observatory's Legacy Survey of Space and Time (LSST), systematically processing large volume data streams for low-latency transient discovery, anomaly detection, classification, and similarity searches.

Graduation Semester

2024-05

Type of Resource

Text

Handle URL

https://hdl.handle.net/2142/124516

Copyright and License Information

Copyright 2024 Patrick David Aleo

Owning Collections