Hawaii Supernova Flows (HSF) Dataset

Updated 21 September 2025

The Hawaii Supernova Flows Dataset is a comprehensive collection of near-infrared and optical observations of over 1,200 transients, serving as a cornerstone for accurate distance measurements and local velocity field analysis.
It employs advanced multi-method light curve fitting and rigorous calibration techniques, achieving sub-0.1 mag systematics and robust quality control across multiple model frameworks.
The dataset underpins cosmological research by enabling cross-survey analyses and providing a critical local anchor for future space-based NIR supernova surveys.

The Hawaii Supernova Flows Dataset is a large, systematics-limited compilation of near-infrared (NIR) and optical photometry and host galaxy redshifts for Type Ia supernovae and other astronomical transients, created to facilitate high-precision distance measurements and peculiar velocity studies in the local universe. As of its first data release, the dataset comprises 1,217 unique transients, including 668 spectroscopically classified SNe Ia, with extensive NIR photometry predominantly in the J band and supporting host spectroscopy from Maunakea-based facilities. Its methodology leverages advanced multi-method fitting, real-time triggering, and state-of-the-art calibration and reduction pipelines, aligning the dataset as a critical anchor for both cosmological analysis and future space-based NIR supernova surveys.

1. Composition and Structure of the Dataset

The Hawaii Supernova Flows (HSF) dataset consists of 1,217 astronomical transients, of which 668 are spectroscopically confirmed Type Ia supernovae. The dataset includes:

NIR photometry, chiefly in the J band, acquired via extensive follow-up (almost 5,000 observations) using UKIRT/WFCAM.
Optical photometry contributed from all-sky surveys such as ATLAS, ASAS–SN, and ZTF.
Host galaxy spectroscopic redshifts, obtained either through literature (HyperLEDA, NED) or from targeted campaigns using SNIFS (UH 2.2m) and FOCAS (Subaru).

After rigorous quality cuts—based on photometric phase coverage, light curve S/N, and model fit acceptability—the cosmology-ready samples consist of approximately 350 SNe Ia for each primary fitting method. The final dataset supports three parallel model fits per supernova (SNooPy EBV_model2, SNooPy max_model, and SALT3–NIR), yielding slightly different event subsets due to individual selection criteria.

2. Observation, Reduction, and Calibration Methodology

HSF employs a real-time triggering system that identifies transient candidates via network alerts from optical surveys; NIR imaging is then executed to sample the SN near peak light. Photometric extraction utilizes:

Forward-modeling algorithms that disentangle SN flux from host galaxy contributions, refined using late-time reference images for galaxy subtraction.
Host-galaxy redshift measurements, with typical uncertainties of 45 km s⁻¹, gleaned from either existing catalogs or new spectral acquisition.
Calibration cross-checks using standard star observations (see practices similar to those of SweetSpot (Weyant et al., 2017)), aiming for sub-0.1 mag systematics in NIR.

Quality requirements for inclusion in final samples involve constraints on shape parameter values (e.g., sBV for SNooPy fits), phase coverage, low-fit uncertainties, and reduced χ² thresholds.

3. Distance Estimation Techniques

Three model-based fitting frameworks are applied:

SNooPy (EBV_model2 and max_model): Fits the SN light curves for time of maximum, stretch (sBV), and dust/colour parameters (E(B−V)).
SALT3–NIR, implemented via SNCosmo: Fits for time of max, shape (x₁), and colour (c), using NIR-extended spectral templates.

Each approach incorporates K- and S-corrections for effective rest-frame standardization and applies Bayesian mixture models (UNITY) for outlier suppression. The statistical spread in Hubble residuals is RMS ≈ 0.165 mag and NMAD ≈ 0.123 mag for the EBV_model2 fit—consistent with or marginally lower than contemporary NIR SN Ia studies.

HSF builds upon and is directly cross-calibrated with datasets such as:

Cosmicflows-2 (Tully et al., 2013): Utilizes SNe Ia distances as high-precision indicators, anchoring the large-scale velocity field at depths up to z ≈ 0.1 and supporting absolute scale calibration alongside Tully–Fisher and TRGB distances.
SweetSpot (Weyant et al., 2017): Shares calibration methodology and NIR focus, with similar treatment of photometric standards and host galaxy subtraction.
SNLS (Perrett et al., 2012): Provides context for SN Ia rates and delay time distribution models, which underpin population synthesis in the local volume and inform systematic uncertainties tied to star formation histories.

The dataset's host redshift and photometry curation create synergies with the Open Supernova Catalog (Guillochon et al., 2016), facilitating advanced cross-survey analyses.

5. Sources of Systematic Error and Quality Control

HSF aims for systematics-limited precision. Control strategies include:

Multiple independent model fits (SNooPy, SALT3–NIR) with cross-validation.
Use of several extinction law parameterizations (O’Donnell 1994, Fitzpatrick 1999/2019) to assess dust model dependence; fits are run with each for comparison.
Outlier rejection using Mahalanobis distances and Bayesian mixture models; iterative exclusion of events with poor fit metrics or parameter space excursions.
Explicit reporting and correction for S-corrections, distinguishing filter systematics between observer and rest frame.
Careful photometric calibration against standard stars and late-time templates for improved galaxy subtraction.

Comparisons indicate the achieved RMS scatter in Hubble residuals is on par with or better than those from other NIR surveys; differences in methodology (e.g., variable versus fixed light-curve parameters in distance estimation) are noted as sources of variation in reported dispersions.

6. Scientific Utility and Prospective Applications

The principal aims and applications include:

Mapping the local peculiar velocity field (dark matter distribution, testing cosmic gravitation models).
Investigating SN Ia uniformity and dust extinction in the NIR regime; reduced sensitivity to extinction yields more robust cosmological distance anchors.
Differential cosmological studies leveraging the combined optical+NIR data for chromatic systematics tests (e.g., constraining extinction curves and host dust models).
Calibration of SNe Ia as standard candles for upcoming NIR-focused space-based surveys (Roman Space Telescope); HSF offers a local anchor at z ≲ 0.1 for comparison to future high-z samples.
Enabling cross-survey analyses and meta-studies, as data and fits are publicly available—allowing consistent external model testing and systematic effect probes.

7. Data Accessibility and Community Engagement

HSF data products are openly available at https://www.github.com/ado8/hsf_DR1, including:

J-band light curves for all transients (full set of 1,217 objects).
Forced photometry outputs and host–galaxy spectroscopy results.
Model fit parameters (and uncertainties) for all spectroscopically confirmed SNe Ia, including optical only, optical+NIR, and multiple model permutations.
Associated analysis code resources.

Raw data may be accessible upon request, supporting reproducibility and facilitating extended analysis by the astronomical community. The dataset's openness aligns with contemporary standards and is designed for extensibility as further releases are made (e.g., with increased phase coverage, additional calibration stars, and host galaxy references).

In summary, the Hawaii Supernova Flows Dataset represents a systematics-limited, open-access, near-infrared–centric resource for supernova cosmology and local velocity field studies. Its rigorous methodology, calibration strategies, and multi-faceted model fitting ensure robust distance measurements critical for both local anchor analyses and next-generation extragalactic surveys. Its ongoing release protocol and integration with related datasets position it as a cornerstone for transient astronomy in the coming decade.