Blind Source Extraction Catalog
Current version: 20171114
File name: cat_pybdsm_concatenated_141117.fits
Description of columns:
- RA — The right ascension (R.A.) coordinate in J2000 frame of the ALMA source.
- Dec — The declination (Dec.) coordinate in J2000 frame of the ALMA source.
- Total_flux — The integrated source flux density in units of Jy.
- E_Total_flux — The error in Total_flux, based on Condon (1997) method.
- Peak_flux — The peak source flux density in units of Jy/beam.
- E_Peak_flux — The error in Peak_flux, based on Condon (1997) method. Note that this is not the RMS noise of the image data.
- RMS — The RMS noise of the image data, measured by Gaussian fitting to the pixel value histograms.
- No source size information is given in this catalog...
- Beam_maj — The synthesized beam sizes of the ALMA data, measured as the full width at half maximum (FWHM) along the major axis of the synthesized beam, in units of arcsec.
Description of blind source extraction with the python code PyBDSM (from A3COSMOS paper in prep.):
- We use the Python-based software PyBDSM (http://www.astron.nl/citt/pybdsm) for blind source extraction. It is developed for the data analysis of Low Frequency Array (LOFAR) radio interferometer, and is also an ideal tool to blindly extract Gaussian-shape sources in ALMA images.
- PyBDSM first calculates the rms noise and the mean pixel value in the input image. The rms noise can be a constant value across the image or a map with variations, but can also be fixed to a given value as user input. In this work, since most of the observations are single pointings instead of mosaics, we measure the rms noise by ourselves via a Gaussian fitting to the pixel value histograms for each image.
- Then, PyBDSM searches for islands of contiguous emission, where two threshold parameters, \incode{thresh_pix} and \incode{thresh_isl}, mainly determine the outcome. \incode{PyBDSM} first searches for all pixels with values \incode{thresh_pix} times rms noise above the mean value, then these pixels are grouped into ``islands'' in which all pixel values should be \incode{thresh_isl} times rms noise above the mean value. \incode{PyBDSM} also properly accounts for overlaps of islands.
- Once islands were identified, \incode{PyBDSM} fits two-dimensional Gaussian(s) to each island. It iterates to find the best solution which achieves the least chi-square and meets a series of empirical flagging parameters as calibrated by the code developer. We verified that, for 95\% of our ALMA images, the default parameters work well, and tuning the parameters do not obviously improve the fitting results.
- After the source fitting, \incode{PyBDSM} obtains the peak and total flux densities and the FWHM of Gaussian major and minor axes and the position angle for each Gaussian component, then calculates the error of each quantity following the Monte Carlo simulation-calibrated, analytic error estimation equations by \cite{Condon1997}.
- Finally, \incode{PyBDSM} outputs the fitted Gaussian components and islands to catalogs. We choose the FITS format catalog as the output of \incode{PyBDSM}, which contains source peak and total fluxes, sizes, and their errors, as well as the rms noise for each island. We choose \incode{group_by_isl=True} to output islands as sources. Most islands/sources are fitted with only one Gaussian, except for only \%3.5 of islands/sources are fitted with multiple Gaussian components (e.g., blended sources, mergers, etc.). In the latter case, the island/source total flux is the weighted sum of all Gaussian components, and the errors are also properly propagated to the island values.
Statistics