Science Imaging

Note:

  1. Our selfcal and science imaging workflow generally assumes a fixed image size, and may perform sub-optimally (e.g. sources missing from the CLEAN mask) if the image size changes at any point
  2. Imaging at higher spectral resolution (i.e. with more frequency channels) requires higher computational power / runtime, and may require increasing the time limit of your imaging jobs. In general we recommend self-calbrating at 1k (1024 channels). If your raw data is at higher spectral resolution, consider first averaging in frequency with the width or chanbin config parameters (also see spectral-line pre-processing here)

processMeerKAT implements science imaging using the tclean task in CASA to generate the images, and the katbeam package to perform primary beam correction on the final image.

The science imaging functionality exposes more of the advanced imaging capabilities of tclean and is intended to be run after all the self-calibration loops. By default the final imaging mask generated by self-calibration is passed in to science imaging, as well as the final set of outlier fields (if outlier imaging is enabled).

Similarly, the final rmsmap is passed in to science imaging, to enable S/N-based thresholding (see below). It is possible to run science imaging immediately after cross-calibration, without running selfcal, although these three fields will not be populated, but would need to be manually set based on a pre-existing set of custom output (e.g. a mask, RMS map and outlier file).

A primary beam (PB) corrected image is also produced, using the katbeam package. The threshold for the PB cutoff is specified via the pbthreshold config parameter, and the pbband parameter is used to select the coefficients for generating the primary beam (see below). However, in the event that images with different PB thresholds are required (after the completion of science imaging), the CASA task impbcor may be used to PB correct the flat-noise image. Please make sure to select the katbeam PB and not the default CASA .pb image while performing PB correction to ensure accurate fluxes. An example call for doing this (e.g. in addition to the PB-corrected image already output as my-amazing.science_image.katbeam_pbcor.image.tt0 at the default cutoff of 0.1) is the following:

impbcor(imagename='my-amazing.science_image.image.tt0', pbimage='my-amazing.science_image.katbeam.pb.tt0', outfile='my-amazing.science_image.katbeam_pbcor.0.3.image.tt0', cutoff=0.3)

Science imaging is expected to be run once, preferably at the end of self-calibration, and unlike self-calibration it does not accept tuples as arguments.

A note on PB correction

If the CASA gridders standard or wproject are used the default .pb image generated is incorrect for MeerKAT. We therefore use the package katbeam to perform the PB correction. The final data products are the flat noise image (produced by CASA), the katbeam PB image as well as the PB corrected image.

The katbeam package describes the MeerKAT PB as an axis-symmetric cos^2 function. In a future release of the pipeline we intend to support the AW projection algorithm, which will produce a correct .pb image for MeerKAT from within tclean, in full Stokes.

Science Imaging Config Options

The default science imaging section in the config file is shown below, and the details of each option are explained further down.

[image]
cell = '1.5arcsec'
robust = -0.5
imsize = [6144, 6144]
wprojplanes = 512
niter = 50000
threshold = 10                    # S/N value if >= 1.0 and rmsmap != '', otherwise Jy
multiscale = [0, 5, 10, 15]
nterms = 2                        # Number of taylor terms
gridder = 'wproject'
deconvolver = 'mtmfs'
restoringbeam = ''
stokes = 'I'
pbthreshold = 0.1                 # Threshold below which to mask the PB for PB correction
mask = ''
rmsmap = ''
outlierfile = ''
  • cell: The cell size of the pixels in the image plane, specified as a CASA compatible string (with units).

  • robust: The Briggs’ robust parameter used during gridding. Specified as a floating point between -2 (uniform weighting) and +2 (natural weighting).

  • imsize: The size of the image in pixels. The image dimensions can be specified in one or two dimensions, hence 6144 is treated as [6144,6144].

  • wprojplanes : The number of W-projection planes to use (in case the wproject gridder is specified). Must be specified as an integer.

  • niter: The number of iterations to perform during deconvolution. Imaging will continue until either the stopping threshold (described below) is reached, or the iteration limit is reached.

  • threshold: The stopping threshold for deconvolution. If the threshold value is larger than 1 and the rmsmap parameter is not blank, the value is treated as a signal-to-noise ratio. If the threshold value is less than 1, or if it is greater than 1 and the rmsmap parameter is blank, it is treated as an absolute value in Jy. If science imaging is requested after self-calibration, the rmsmap parameter will be auto-populated by the pipeline.

  • multiscale; The number of individual scales to use if multi-scale imaging is desired (and the appropriate deconvolver is chosen - e.g. mtmfs or multiscale) specified as a list of integers such as [0, 5, 10] etc. To switch off multiscale imaging, specify a blank list (i.e., multiscale=[]).

  • nterms : The number of Taylor terms to use for broadband imaging.

  • gridder : The gridder to use for imaging, typically one of either standard or wproject. Any CASA gridder is a valid entry for this parameter, although all the necessary configuration options for different gridders will not be exposed through the configuration file.

  • deconvolver : The deconvolver to use during imaging. Use clark or hogbom when the fractional bandwidth (bandwidth divided by band centre) is < ~15%.

  • restoringbeam : Specified as a string with units, such as '20arcsec'. If this is left blank, the default restoring beam is used. Specifying this parameter can be used to force a specific restoring beam during imaging.

  • stokes : The Stokes planes to image. Please note that generating multi-Stokes spectral cubes or multi-Stokes cubes with outliers, is not currently possible within CASA

  • pbthreshold : The PB gain threshold below which to mask the PB and the PB corrected image. The PB is generated using the katbeam package.

  • pbband : One of LBand, UHF, or SBand. The coefficients from the selected band are used to generate the primary beam. This is necessary because the different bands have overlapping frequencies, and the PB characteristics are different across the different bands, even at overlapping frequencies.

  • mask : Path to a CLEAN mask or a CASA region file. This is normally auto-populated by the pipeline if science imaging is queued up after self-calibration. However, if science imaging is being launched manually, this will have to be specified if masking is desired. If the parameter is left blank, no masking will be performed.

  • rmsmap : Path to an RMS image, used to determine local thresholds during deconvolution. This is normally auto-populated by the pipeline if science imaging is queued up after self-calibration. If this is left blank, the threshold parameter is assumed to be an absolute (global) threshold in Jy.

  • outlierfile : Path to a valid outlier fields file. This is normally auto-populated by the pipeline if science imaging is queued up after self-calibration. If this is left blank, no outlier imaging is performed.