juliensimon/unified-radio-catalog

Unified Radio Catalog (SPECFIND v3)

Part of the Astronomy Datasets collection on Hugging Face.

The SPECFIND v3 unified radio source catalog, containing 1,658,207 cross-matched radio source measurements from 198 surveys spanning 16 to 31000 MHz. SPECFIND positionally cross-identifies radio sources across major surveys including NVSS, FIRST, SUMSS, TGSS, GLEAM, and dozens of others, then fits power-law radio spectra.

Dataset description

SPECFIND (Vollmer et al. 2005, updated Stein et al. 2024) is the largest positional cross-identification of radio continuum catalogs. Version 3 matches sources across 50+ radio surveys at frequencies from 16 to 31000 MHz, covering the entire sky. Each row represents a source detection at a specific frequency, grouped by a unique source identifier (source_id). For sources detected in multiple surveys, SPECFIND fits a power-law spectrum S(nu) = 10^b * nu^a, where a is the spectral index and b is the intercept.

The catalog contains 339,547 unique radio sources with measurements from surveys including NVSS (1.4 GHz), FIRST (1.4 GHz), SUMSS (843 MHz), TGSS (150 MHz), GLEAM (200 MHz), and many others.

Key columns

Column	Type	Description
source_id	Int32	Unique source identifier (groups cross-matched detections)
source_name	string	Survey-specific source designation
ra_deg	float64	Right ascension J2000 (degrees)
dec_deg	float64	Declination J2000 (degrees)
frequency_mhz	float64	Observation frequency (MHz)
flux_density_mjy	float64	Flux density at this frequency (mJy)
flux_density_error_mjy	float64	Flux density uncertainty (mJy)
spectral_index	float64	Fitted spectral index (a in S ~ nu^a)
spectral_intercept	float64	Fitted spectral intercept (b in log S = a*log(nu) + b)
n_frequencies	Int32	Number of frequency measurements for this source
flux_residual_pct	float64	Flux residual from spectral fit (%)
beam_arcsec	float64	Survey beam size (arcsec)
survey	string	Survey name extracted from source designation
frequency_band	category	Frequency band: VLF (<100), low (100-500), mid (500-2000), high (2-8 GHz), SHF (>8 GHz)

Full schema includes 16 columns with positional offsets and uncertainties.

Quick stats

• 1,658,207 total source measurements
• 339,547 unique radio sources
• 198 contributing surveys
• Frequency range: 16 to 31000 MHz
• Median flux density: 110.0 mJy
• Median spectral index: -0.75

Usage

from datasets import load_dataset

ds = load_dataset("juliensimon/unified-radio-catalog", split="train")
df = ds.to_pandas()

# Group by source to see multi-frequency data
source = df[df["source_id"] == df["source_id"].iloc[0]]
print(f"Source {source['source_name'].iloc[0]}: {len(source)} frequencies")

# Spectral index distribution
import matplotlib.pyplot as plt
si = df.drop_duplicates("source_id")["spectral_index"].dropna()
si.clip(-3, 3).hist(bins=200)
plt.xlabel("Spectral index")
plt.ylabel("Count")
plt.title("Radio Source Spectral Index Distribution")
plt.axvline(-0.7, color="red", linestyle="--", label="Typical synchrotron")
plt.legend()
plt.show()

# Sky coverage map
plt.hexbin(df["ra_deg"], df["dec_deg"], gridsize=100, mincnt=1)
plt.colorbar(label="Measurement count")
plt.xlabel("RA (deg)")
plt.ylabel("Dec (deg)")
plt.title("SPECFIND v3 Sky Coverage")
plt.show()

# Survey contribution
print(df["survey"].value_counts().head(10))

Data source

Stein, Y., Vollmer, B., Boch, T., et al. (2024), SPECFIND v3.0 — A catalog of radio continuum cross-identifications and spectra. VizieR catalog VIII/104. Based on Vollmer, B. et al. (2005, 2010). Via VizieR CDS.

Related datasets

• NVSS Radio Source Catalog — NVSS 1.4 GHz survey, 1.8M sources
• FIRST Radio Survey Catalog — FIRST 1.4 GHz survey
• VLASS Radio Sources — VLA Sky Survey 2-4 GHz

Pipeline

Source code: juliensimon/space-datasets

Citation

@dataset{unified_radio_catalog,
  author = {Simon, Julien},
  title = {Unified Radio Catalog (SPECFIND v3)},
  year = {2026},
  publisher = {Hugging Face},
  url = {https://huggingface.co/datasets/juliensimon/unified-radio-catalog},
  note = {Based on Stein, Vollmer et al. (2024) SPECFIND v3 via VizieR CDS}
}

License

CC-BY-4.0