Upgrade to Pro — share decks privately, control downloads, hide ads and more …

Sourcefinding Strategy in MSSS, the first LOFAR survey

transientskp
January 10, 2014

Sourcefinding Strategy in MSSS, the first LOFAR survey

Dario Carbone

transientskp

January 10, 2014
Tweet

More Decks by transientskp

Other Decks in Science

Transcript

  1. Sourcefinding strategy in
    MSSS, the first LOFAR
    survey
    DARIO CARBONE
    TKP meeting, 10 January 2014

    View Slide

  2. Collaboration
    Job done in collaboration with:

    the LOFAR Sourcefinder WG (led by Chiara
    Ferrari)
    http://www.lofar.org/operations/doku.php?id=commi

    the MSSS team (led by George Heald)
    http://msss.astron.nl
    Dario Carbone

    View Slide

  3. Outlines

    PySE & PyBDSM
    – ASKAP Data Challenge
    – Performance
    – Parameters

    MSSS
    – MVF
    – First catalog
    Dario Carbone

    View Slide

  4. 4
    Sourcefinder tools in LOFAR

    Python Blob
    Detection and Source
    Measurement
    (PyBDSM)

    Designed by EoR
    group (Niruj Mohan
    Ramanujam), for an
    accurate detection of
    extended emission

    Get rid of all
    foreground signal

    Python Source
    Extractor (PySE)

    Designed by
    Transient group
    (Hanno Spreeuw) to
    be sensitive to point-
    like sources

    Detect radio transient

    View Slide

  5. 5
    PySE vs PyBDSM
    PySE is not able to detect extended
    sources
    Dario Carbone

    View Slide

  6. 6
    Focus on point-like sources

    Radio transients are
    point-like sources

    PySE was developed
    to find point-like
    sources

    Speed requirements

    Implemented in the
    TraP
    Dario Carbone

    View Slide

  7. 7
    PySE
    1. Fit and subtract a variable background across
    the image
    2. Fit the RMS noise across the image
    3. Identify islands of contiguous pixels which
    appear above some multiple of the RMS
    4. “Deblend” multi-component islands into the
    constituent parts
    5. Perform initial estimation of source parameters
    6. Fit each component with an elliptical Gaussian
    and return the result

    View Slide

  8. 8
    ASKAP Data Challenge
    Three simulated images including point sources and
    complex structures.
    14 different sourcefinder tools participated.

    View Slide

  9. 9
    Challenge 2 – point sources only
    σRA σDec
    PySE 0.39” 0.39”
    PyBDSM 0.44” 0.43”
    Credits to
    M. Whiting
    Dario Carbone

    View Slide

  10. 10
    Challenge 2 – PySE performance
    PySE completeness,
    reliability and flux recovery
    seems very good.

    Force-beam option was
    turned on.
    Credits to
    M. Whiting
    Dario Carbone

    View Slide

  11. 11
    Challenge 3 – extended sources
    PySE isn't detecting
    extended sources, but isn't
    detecting fake ones.

    Force-beam option was
    turned on.
    Credits to
    M. Whiting
    Dario Carbone

    View Slide

  12. 12
    Parameters for PySE
    Best parameters for source extraction with
    PySE:

    grid size = 50 pixels

    deblending turned on

    deblending threshold = 90

    detection threshold = 8 σ

    analysis threshold = 3 σ

    Force-beam option turned on Dario Carbone

    View Slide

  13. 13
    Force-beam option

    Avoid big ellipses
    collecting high flux

    Only point-sources
    Dario Carbone

    View Slide

  14. 14
    The extracted flux can change just because the
    ellipse shape changed
    Flux
    Time
    Flux
    Time
    No force-beam With force-beam
    Force-beam option
    Dario Carbone

    View Slide

  15. 15

    Very important for the transient pipeline

    Source could not to be associated

    Tests on 70 images of 10 fields, with a pixel
    size of 45 arcsec.
    Systematic position errors
    Dario Carbone

    View Slide

  16. 16
    Systematic position errors


    PyBDSM loses 13.3% PySE 3.4%
    R
    DR
    = 5.68
    Credits to A. Van der Horst
    Dario Carbone

    View Slide

  17. 17
    Systematic position errors

    Systematic position error of 20 arcsec

    PyBDSM loses 2.0% PySE 1.0%
    Credits to A. Van der Horst
    R
    DR
    = 5.68
    Dario Carbone

    View Slide

  18. 18
    MSSS intro

    LBA: 30 − 75 MHz;
    8x2 MHz band

    HBA: 119 − 158 MHz;
    8x2 MHz band

    Angular resolution ∼
    2′

    Sensitivity ∼ 10 mJy
    beam−1
    http://msss.astron.nl
    Dario Carbone

    View Slide

  19. 19
    MSSS recap
    Heald et al., in prep.
    Dario Carbone

    View Slide

  20. 20
    MSSS Verification Field (MVF)

    MSSS Verification Field (MVF): 100 square degree
    including no bright, complicated sources and distant
    from A-team sources.

    Coordinates of the center: RA = 15 h; dec = +69 deg
    Heald et al., in prep.
    Dario Carbone

    View Slide

  21. 21
    MSSS Verification Field (MVF)
    LBA combined image of MVF
    Heald et al., in prep.
    HBA combined image of MVF
    Dario Carbone

    View Slide

  22. 22
    Source extraction strategy

    Combined non-primary-beam-corrected image as
    detection image.

    Run PyBDSM and PySE separately on each band and
    then combine the results.

    For each sourcefinder, the detected sources are
    associated across the bands using the nearest neighbor
    criterion.

    Catalogues from PyBDSM and PySE are then cross-
    matched.

    Positions and fluxes are taken from PyBDSM outputs (if
    available).
    Dario Carbone

    View Slide

  23. 23
    Parameters for PySE

    grid size = 50 pixels

    deblending turned on

    deblending threshold = 90

    detection threshold = 10 σ

    analysis threshold = 5 σ

    Force-beam option turned on
    Parameter settings based on real maps and
    simulations and the requirement to be
    conservative for this first catalogue.
    Dario Carbone

    View Slide

  24. 24
    Comparison

    697 sources are detected in
    the area of interest.

    We compared HBA MSSS
    catalogue to the 7C survey
    @ 151 MHz.

    We compared LBA MSSS
    catalogue to the VLSS @
    74 MHz and the 8C
    catalogue @ 38 MHz.
    Heald et al., in prep.
    Dario Carbone

    View Slide

  25. 25
    Column in MSSS catalogue

    ID; Source ID, formed as Jhhmmss+ddmmss
    using the IAU convention.

    RA & eRA; Source J2000 Right Ascension, in
    decimal degrees & error on Right Ascension, in
    seconds of time.

    DEC & eDEC; Source J2000 Declination, in
    decimal degrees & error Declination, in
    arcseconds.

    SFFLAG000; Flag indicating how many
    sourcefinders detected the source (0 means it
    was detected in both; 1 only in PyBDSM; 2 only
    in PYSE.)

    Sint000 & eSint000; Source integrated flux at
    000 MHz, in Jy & error on integrated flux, in Jy.

    Spk000 & eSpk000; Source peak flux at
    000 MHz, in Jy & error on peak flux, in Jy.

    SLBA & eSLBA; LBA flux density (from
    combined map). Effective frequency is 000 MHz
    & error on LBA flux density.

    SHBA & eSHBA; HBA flux density (from
    combined map). Effective frequency is 000 MHz
    & error on HBA flux density.

    SPIX & eSPIX; Spectral index & error on
    spectral index.

    SCURV & eSCURV; Spectral curvature & error
    on spectral curvature.

    MAJAX000 & eMAJAX000; Major axis of fitted
    ellipse at 000 MHz & error on major axis.

    MINAX000 & eMINAX000; Minor axis of fitted
    ellipse at 000 MHz & error on minor axis.

    PA000 & ePA000; Position angle of fitted
    ellipse at 000 MHz, in degrees & error on
    position angle.
    Dario Carbone

    View Slide

  26. 26
    Column in MSSS catalogue

    ID; Source ID, formed as Jhhmmss+ddmmss
    using the IAU convention.

    RA & eRA; Source J2000 Right Ascension, in
    decimal degrees & error on Right Ascension, in
    seconds of time.

    DEC & eDEC; Source J2000 Declination, in
    decimal degrees & error Declination, in
    arcseconds.

    SFFLAG000; Flag indicating how many
    sourcefinders detected the source (0 means it
    was detected in both; 1 only in PyBDSM; 2 only
    in PYSE.)

    Sint000 & eSint000; Source integrated flux at
    000 MHz, in Jy & error on integrated flux, in Jy.

    Spk000 & eSpk000; Source peak flux at
    000 MHz, in Jy & error on peak flux, in Jy.

    SLBA & eSLBA; LBA flux density (from
    combined map). Effective frequency is 000 MHz
    & error on LBA flux density.

    SHBA & eSHBA; HBA flux density (from
    combined map). Effective frequency is 000 MHz
    & error on HBA flux density.

    SPIX & eSPIX; Spectral index & error on
    spectral index.

    SCURV & eSCURV; Spectral curvature & error
    on spectral curvature.

    MAJAX000 & eMAJAX000; Major axis of fitted
    ellipse at 000 MHz & error on major axis.

    MINAX000 & eMINAX000; Minor axis of fitted
    ellipse at 000 MHz & error on minor axis.

    PA000 & ePA000; Position angle of fitted
    ellipse at 000 MHz, in degrees & error on
    position angle.
    This catalogue has 93 columns for the point sources,
    and 93 + 96 = 189 columns for extended sources!

    View Slide

  27. 27
    Conclusions

    First catalogue of 100 squared degrees of MSSS is ready
    to be released.

    This will contain a lot of data on several hundreds of
    sources.

    Compared to other existing surveys at similar frequencies
    MSSS is behaving as expected.

    Sourcefinding was done following a slightly more
    conservative approach than usual.

    PySE is behaving very well on the Data Challenge maps
    as completeness, reliability, and flux recovery.

    The force-beam option should be turned on as we are
    looking for point sources only.

    View Slide