skip to main content


Title: Fitting HI Spectra with Neural Networks
The number of extragalactic sources of HI detected in radio surveys is growing exponentially. It will soon no longer be feasible for human researchers to individually fit spectra. We present algorithms for automatically extracting the typical parameters of interest for the 21 cm HI line—recessional velocity, velocity width, and integrated flux—using neural networks. Features are produced by convolving spectra with templates generated with the Busy Function. We present the results of fitting hundreds of spectra with many different shapes, and at a wide range of signal to noise ratio. Additionally, we compare with prior methods of automated source extraction. This work has been supported by NSF grants AST-1211005 and AST-1637339.  more » « less
Award ID(s):
1637339
NSF-PAR ID:
10097652
Author(s) / Creator(s):
; ;
Date Published:
Journal Name:
American Astronomical Society, AAS Meeting
Volume:
233
Page Range / eLocation ID:
245,23
Format(s):
Medium: X
Sponsoring Org:
National Science Foundation
More Like this
  1. The Arecibo Pisces-Perseus Supercluster Survey (APPSS) will provide strong observational constraints on the mass-infall rate onto the main filament of the Pisces-Perseus Supercluster. The survey data consist of HI emission-line spectra of cluster galaxy candidates, obtained primarily at the Arecibo Observatory (with ALFA as part of the ALFALFA Survey and with the L-Band Wide receiver as part of APPSS observations). Here we present the details of the data reduction process and spectral-analysis techniques used to determine if a galaxy candidate is at a velocity consistent with the Supercluster, as well as the detected HI-flux and rotational velocity of the galaxy, which will be used to estimate the corresponding HI-mass. We discuss the results of a preliminary analysis on a subset of the APPSS sample, corresponding to 98 galaxies located within ~1.5° of DEC = +35.0°, with 65 possible detections. We also highlight several interesting emission-line features and galaxies discovered during the reduction and analysis process and layout the future of the APPSS project. This work has been supported by NSF grants AST-1211005 and AST-1637339. 
    more » « less
  2. The Arecibo Pisces-Perseus Supercluster Survey (APPSS) is an observing project undertaken by the Undergraduate ALFALFA Team that aims to detect HI in galaxies in the Pisces-Perseus neighborhood and analyze the dynamics and the properties of the galaxies. The galaxies targeted in APPSS are suspected from their optical properties (color, morphology, surface brightness) to lie in the Pisces-Perseus Supercluster (PPS) but are below the detection threshold of the ALFALFA blind HI survey. Here we present results for galaxies targeted in a strip across the PPS region in declination from 30o to 32o. This region is along the main filament of the supercluster and includes objects such as the Pisces Cluster. The data was recorded by the L-Band Wide receiver of the Arecibo Observatory. Data reduction was done using routines derived for the APPSS in IDL. After baselining the spectra and sifting out radio interference, we fit either a gaussian or two-horned profile to their 21-centimeter line to measure the HI line flux density, velocity, and velocity width. From these parameters we calculate distances, hydrogen gas mass, and rotational velocities. As expected, the galaxies analyzed in this slice of declination have consistently lower mass than the ALFALFA detections thus extending the sampling of galaxies within the PPS. The combined ALFALFA and APPSS HI line detections will be used for future applications of the Baryonic Tully-Fisher Relation in this region. This research has been supported by NSF grant NSF/AST-1714828 to M.P. Haynes and by the Brinson Foundation for the Arecibo Pisces-Perseus Supercluster Survey (APPSS). 
    more » « less
  3. Ultra-diffuse galaxies (UDGs) are galaxies with a very low optical surface brightness; they have very few stars for their given radius. Since UDGs are thus difficult to study in visible light, we observe radio emission from neutral hydrogen gas (HI) in these galaxies. Here we present observations of the HI gas in the UDGs AGC 749290 and AGC 238764. Initially selected from a sample of Ultra-Diffuse Galaxies detected in the ALFALFA survey, these sources were imaged as a part of a follow up program using the Jansky Very Large Array (VLA) in both C and D configurations. We reduce the data using the CASA software suite, removing radio interference, applying calibration, and creating images. From these data we obtain spectra and maps of the galaxies' HI distribution and radial velocities. We find that both sources show ordered gas distributions and rotation, and that the HI gas extends well beyond the already extended optical emission. Further, we estimate inclinations and plot these sources on the Baryonic Tully-Fisher relation, providing tentative evidence that these sources are rotating too slowly for their given mass. This work has been supported by NSF grant AST-1637339. 
    more » « less
  4. Manual flagging of RFI is extremely time-consuming and error-prone. We present a machine learning algorithm which automatically identifies radio frequency interference (RFI) in HI spectra. Our algorithm uses the features of polarization asymmetry (defined as |polA - polB|/[polA + polB] ) along with the skew and standard deviation of each channel over time to evaluate the presence of RFI. The algorithm was tested on hundreds of spectra taken by the Undergraduate ALFALFA Team (UAT) as part of the APPSS survey. It outperforms humans not only in speed, but in visually identifying RFI when it is weak or mimics properties of signals. This work has been supported by NSF grants AST-1211005 and AST-1637339. 
    more » « less
  5. null (Ed.)
    We present our work on constructing a template Baryonic Tully-Fisher Relation (BTFR) from galaxies in the local universe that have primary distances. We utilize HI 21 cm line data from the complete Arecibo Legacy Fast ALFA (ALFALFA) survey and the digital HI archive from Springob et al. 2005; we also use photometry from the Sloan Digital Sky Survey (SDSS) and the NASA Sloan Atlas (NSA) MANGA v1_0_2 database; lastly, we have also made use of the Extragalactic Distance Database (EDD) for identifying galaxies with primary distances. After cross-matching the galaxies in these catalogues, we identify some 144 galaxies which meet our requirements for having all the necessary HI and photometry data, having primary distances, residing within 30 Mpc, and having low enough uncertainties to be considered reliable data points. An important trait of this data set is the prominence of low-mass, low-luminosity dwarves. Notably, we find the values for the slope, intercept and intrinsic scatter of the relation to be around 2.3, 4.8, and 0.4, respectively. Further, while unresolved velocity widths have historically produced shallower slopes, and while the BTFR has been shown to have a higher intrinsic scatter for low-mass galaxies, these precedents are not enough to explain the deviation of our data from the “standard” values of the BTFR. This work therefore raises several questions about why this discrepancy exists, how it can be resolved, and what we can learn from it. The authors would like to acknowledge the support of NSF/AST-1714828 and the Brinson Foundation. 
    more » « less