Processing CLS data with ORAC-DR

A dedicated recipe has been written to process data from the Cosmology Legacy Survey using a "jack-knife" method of estimating and removing residual large-scale noise. The recipe is based on a script written by Ed for S2SRO data (and used to produce the figures in the SMURF paper). In this article I will describe how to use the recipe, what steps it performs and what parameters may be used to control the processing. In the future, this technique may be made more widely available for applying to other data sets (not just blank field ones).

How do I use it?

To use this recipe, initialize the pipeline, create a text file with the names of all the files to process and a file with recipe parameters if necessary. Then add the recipe name REDUCE_CLS to the usual command line:

% oracdr -loop file -files filenames.lis -recpars params.ini 
  -log sf -nodisplay REDUCE_CLS

By default the recipe uses the "blank field" makemap config file (though this may be overridden with the MAKEMAP_CONFIG recipe parameter; see below). The recipe creates a slew of output files with the following suffices:

1. _fmos - signal map (1 for each observation)
2. _mappsf - map-filtered PSF (1 for each observation). This is the same as the above signal map but with an artifical gaussian added to the time-series (and located at the map centre).
3. _wmos - coadded signal map
4. _wpsf - coadded map-filtered PSF
5. _jkmap - jack-knife map
6. _whiten - whitened signal map
7. _whitepsf - whitened, map-filtered PSF map
8. _cal - calibrated, whitened signal map
9. _mf - above map processed with matched-filter (using the _whitepsf map as the PSF)
10. _snr - signal-to-noise map created from above map (_mf)

Note that the per-observation files begin with s, and the coadds (and subsequent products) begin with gs. The ones you are probably most interested in are the _mf and _snr files.

What does it do?

The recipe works as follows.

1. Each observation is processed separately to produce a signal map (_fmos). In addition, each observation is re-processed with an artificial gaussian source added at the map centre (this will be used to create the "map-filtered PSF" image, _mappsf).
2. The signal maps are combined using inverse-variance weighting to create a coadded signal map (_wmos). The images with the artificial gaussians added in are also coadded to produce the "map-filtered PSF" (_wpsf).
3. The data are split into two groups made from alternating observations. Each group is coadded, and the two coadds are subtracted to produce the jack-knife map (_jkmap).
4. The SMURF command sc2filtermap is used to estimate the radial angular power spectrum (circular symmetry is assumed) within a region defined by twice the minimum noise in the coadded signal map. (The size of this region is written to the FITS header of the output map under the keyword WHITEBOX.)
5. The inverse of this angular power spectrum is applied to the coadded signal and PSF images (_whiten and _whitepsf) to remove residual low spatial-frequency noise.
6. The amplitude of the fake source in _whitepsf is compared with the input value to derive a corrected FCF. This new FCF is used to calibrate the whitened signal map to create the _cal map.
7. A matched filter is applied to _cal using _whitepsf as the PSF to create _mf.
8. Finally, a signal-to-noise ratio image is created (_snr).

What parameters are available?

The following recipe parameters can be used to control the processing:

• FAKEMAP_SCALE - Amplitude of the fake source (in Jy) added to the timeseries to assess the map-making response to a point source.
• MAKEMAP_CONFIG - Name of a config file for use with the SMURF makemap task. The file must exist in the current working directory, $MAKEMAP_CONFIG_DIR, $ORAC_DATA_OUT, $ORAC_DATA_CAL or $STARLINK_DIR/share/smurf.
• MAKEMAP_PIXSIZE - Pixel size in arcsec for the output map. Default is wavelength dependent (4 arcsec at 850 um, 2 arcsec at 450 um).
• WHITEN_BOX - Size of the region used to calculate the angular power spectrum for removing residual low-frequency noise in the data. Default is a square region bounded by the noise being less than twice the minimum value.

What if I want to run it again?

In addition to the full pipeline recipe, there is a PICARD recipe called SCUBA2_JACKKNIFE which performs all of the post-map-making steps. This can be used to examine the influence of including or omitting individual observations (say ones with visible artefacts that the pipeline is not able to trap), or investigate the effect of varying the size of the whitening region, or trim the images to a specific size before re-running the jack-knife steps.

Note that the pipeline must have been run once to produce all the necessary files which go in to this recipe. And note that the full pipeline recipe must be run again if a different config file or input gaussian amplitude is to be used.

The majority of the flexibility in controlling the processing occurs after the individual signal maps have been created. As a refresher, running this recipe would mean typing:

% picard -log sf -recpars params.ini SCUBA2_JACKKNIFE myfiles*.sdf

All of the control is through the recipe parameters in the file params.ini. The most important item to note is that you must provide a map-filtered PSF (otherwise a default will be used). However, you already have one of these: the file ending _mappsf.sdf listed above.

• PSF_MATCHFILTER - the name of the map-filtered PSF file
• WHITEN_BOX - Size of the region used to calculate the angular power spectrum for removing residual low-frequency noise in the data. Default is a square region bounded by the noise being less than twice the minimum value.

This is only the first version of the recipe, and its limitations have yet to be determined. However, one case I have found where the recipe fails is in a large, multi-field map with highly-variable noise. In this case it might be better to process each field separately before creating the mosaic (though this has the disadvantage that different filtering may be applied to each field). Try it out and see what you get. All feedback welcome.