SPLASH: A Rapid Host-Based Supernova Classifier for Wide-Field Time-Domain Surveys

Fuente: arXiv
Salvato in:
Dettagli Bibliografici
Autori principali: Boesky, Adam, Villar, V. Ashley, Gagliano, Alexander, Hsu, Brian
Natura: Preprint
Pubblicazione: 2025
Soggetti:
Accesso online:
Tags: Aggiungi Tag
Nessun Tag, puoi essere il primo ad aggiungerne!!
_version_ 1866910977670250496
author Boesky, Adam
Villar, V. Ashley
Gagliano, Alexander
Hsu, Brian
author_facet Boesky, Adam
Villar, V. Ashley
Gagliano, Alexander
Hsu, Brian
contents The upcoming Legacy Survey of Space and Time (LSST) conducted by the Vera C. Rubin Observatory will detect millions of supernovae (SNe) and generate millions of nightly alerts, far outpacing available spectroscopic resources. Rapid, scalable photometric classification methods are therefore essential for identifying young SNe for follow-up and enabling large-scale population studies. We present SPLASH, a host-based classification pipeline that infers supernova classes using only host galaxy photometry. SPLASH first associates SNe with their hosts (yielding a redshift estimate), then infers host galaxy stellar mass and star formation rate using deep learning, and finally classifies SNe using a random forest trained on these inferred properties, along with host-SN angular separation and redshift. SPLASH achieves a binary (Type Ia vs. core-collapse) classification accuracy of $76\%$ and an F1-score of $69\%$, comparable to other state-of-the-art methods. By selecting only the most confident predictions, SPLASH can return highly pure subsets of all major SN types, making it well-suited for targeted follow-up. Its efficient design allows classification of $\sim 500$ SNe per second, making it ideal for next-generation surveys. Moreover, its intermediate inference step enables selection of transients by host environment, providing a tool not only for classification but also for probing the demographics of stellar death.
format Preprint
id arxiv_https___arxiv_org_abs_2506_00121
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle SPLASH: A Rapid Host-Based Supernova Classifier for Wide-Field Time-Domain Surveys
Boesky, Adam
Villar, V. Ashley
Gagliano, Alexander
Hsu, Brian
High Energy Astrophysical Phenomena
Astrophysics of Galaxies
Instrumentation and Methods for Astrophysics
The upcoming Legacy Survey of Space and Time (LSST) conducted by the Vera C. Rubin Observatory will detect millions of supernovae (SNe) and generate millions of nightly alerts, far outpacing available spectroscopic resources. Rapid, scalable photometric classification methods are therefore essential for identifying young SNe for follow-up and enabling large-scale population studies. We present SPLASH, a host-based classification pipeline that infers supernova classes using only host galaxy photometry. SPLASH first associates SNe with their hosts (yielding a redshift estimate), then infers host galaxy stellar mass and star formation rate using deep learning, and finally classifies SNe using a random forest trained on these inferred properties, along with host-SN angular separation and redshift. SPLASH achieves a binary (Type Ia vs. core-collapse) classification accuracy of $76\%$ and an F1-score of $69\%$, comparable to other state-of-the-art methods. By selecting only the most confident predictions, SPLASH can return highly pure subsets of all major SN types, making it well-suited for targeted follow-up. Its efficient design allows classification of $\sim 500$ SNe per second, making it ideal for next-generation surveys. Moreover, its intermediate inference step enables selection of transients by host environment, providing a tool not only for classification but also for probing the demographics of stellar death.
title SPLASH: A Rapid Host-Based Supernova Classifier for Wide-Field Time-Domain Surveys
topic High Energy Astrophysical Phenomena
Astrophysics of Galaxies
Instrumentation and Methods for Astrophysics
url https://arxiv.org/abs/2506.00121