The science goal of many extra-galactic SCUBA-2 observations is to detect unresolved point sources. In the examples below we work through the reduction of just such an extra-galactic field, A1835.
Most extra-galactic objects are on average only slightly brighter than the confusion limit. The fluctuations of the background sky brightness due to multiple super-imposed, unresolved sources within the telescope beam, below which individual sources cannot be detected. It is likely that any sources in the map will be at best, only a few standard deviations brighter than the noise in the map (caused by a combination of instrumental noise and source confusion).
The basic reduction method for maps like these follow two main steps: 1. running the data through the
map-maker using the dimmconfig_blank_field.lis
configuration file (see Section 3.7), and 2. applying the
Picard SCUBA2_MATCHED_FILTER
recipe (see Section 8.7 and Appendix D).
Step 1: Run the map-maker
In this example the raw data are stored locally in a directory called data
. We have three observations (#13, #18,
#21) of the field, which we will reduce independently.
Step 2: Combine the maps
These three maps are then combined using the Picard recipe MOSAIC_JCMT_IMAGES
. In this case we accept the
default of wcsmosaic mosaicking and nearest-neighbour pixel spreading and so do not supply a parameter
file.
The output map, cosmo3_mos.sdf
(automatically named for the last input file appended by _mos), is shown
in the left-hand panel of Figure 7.1. The advantage of using the Picard recipe over standalone
Kappa commands is that the exposure time is also propagated correctly to the output mosaic (it is stored in the
MORE.SMURF.EXP_TIME
extension).
Step 3: Apply the matched filter
In order to optimally find sources that are the size of the telescope beam, we apply the matched filter recipe,
namely SCUBA2_MATCHED_FILTER
. We create a simple parameter file called smooth.ini
containing the following
lines:
where SMOOTH_FWHM = 15
indicates that the background should be estimated by first smoothing the map and
PSF with a 15-arcsec FWHM Gaussian. Next, the recipe is executed as follows:
The output of this operation is a smoothed image that is named by appending _mf to the input file
(cosmo3_mos_mf.sdf
) and a cropped version is shown in the right-hand panel of Figure 7.1. You can
immediately see the contrast to the left-hand panel which is the output from the map-maker. A number of signal
peaks now emerge as possible sources.
Step 4: Crop the map
Next we shall crop the map to remove the noisy edges, in this case to a 900-arcsec radius circle. The parameter
file called crop.ini
contains the following lines:
The output file from the following command will be named cosmo2_mos_mf_crop.sdf
:
Step 5: Make an S/N map
Finally, we need to find sources. The filtered map contains a VARIANCE component, so it is easy to produce a
S/N map using the Kappa task makesnr:
The resulting map, cosmo2_mos_mf_snr
, is shown in Figure 7.2. Compared with the matched filter map the
edges no longer appear as noisy because they have been down-weighted by the larger noise values where there
were less data.
Step 6: Identify sources
The basic procedure for identifying sources would be to locate peaks above some threshold S/N. The S/N
image above shows peaks that are likely to be real sources. For a start, a source appears where expected at the
0,0 position.
But how can we check if these sources are real?
Although this method is considerably simpler to execute, the products have undergone more advanced processing than the manual method shown above. Due to these extra analysis steps, this pipeline method is particularly recommended.
Step 1: Create input file
Create a file with the full path names of all the files you wish to process, one per line (e.g. myfiles.lis
)
Step 2: Run the pipeline
The pipeline must first be initiated for the wavelength you are working on. In the case below this is
850 m.
Note that the date does not have to be specified when initialising the pipeline. The pipeline is run using the
REDUCE_SCAN_FAINT_POINT_SOURCES_JACKKNIFE
recipe; this uses dimmconfig_blank_field.lis
as the
configuration file. If you wish to provide an alternative file you will need to put the name of the new
configuration file in a recipe parameter file. See Chapter 4 for details.
You substitute the required date for YYYYMMDD
. The pipeline will write out a large number of files with the
following suffices.
sYYYYMMDD*_fmos | The map for each observation |
---|---|
sYYYYMMDD*_mappsf | The map for each observation with an artificial point source added at the map centre |
gsYYYYMMD*_wmos | The co-add of all the _fmos files
|
gsYYYYMMD*_whiten | The whitened version of _wmos
|
gsYYYYMMD*_cal | The calibrated version of _whiten
|
gsYYYYMMD*_mf | The matched-filtered version of _cal |
FAINT_POINT_SOURCES_JACKKNIFE
is a recipe designed to process blank field/extra-galactic data. The recipe
uses a jack-knife method to remove low-spatial frequency noise and generate a matched filter output
map.
The recipe processes each observation twice, a standard reduction first, then a re-run with a fake point source
added to the time series (see parameter “fakemap
”). This produces a co-added signal map (_wmos
) and a
co-added PSF map (_mappsf
).
FAINT_POINT_SOURCES_JACKKNIFE
can be used interchangably with
REDUCE_SCAN_FAINT_POINT_SOURCES_JACKKNIFE
.
After the map-maker has completed, the recipe will call SCUBA2_JACKKNIFE
. This routine divides the
observations into two groups (odd and even) which are co-added and then subtracted to create a jack-knife
map. This map contains only noise with no contribution from astronomical signal. The angular power spectrum
of this map is then used to estimate and remove the residual 1/f noise from the signal map and the PSF map;
this is the whitening step. The whitened jack-knife map is run through SCUBA2_MATCHED_FILTER
using the
whitened PSF map as the PSF input. It is this matched filter map which will be of most interest to
users.
See SUN/264 for more information on REDUCE_SCAN_FAINT_POINT_SOURCES_JACKKNIFE
and all other pipeline
recipes.
Step 3: (Optional) Re-run SCUBA2_JACKKNIFE
You may wish to run the SCUBA2_JACKKNIFE
step again independently from the pipeline. If your final map does
not look as expected you might first examine the individual mosaics from the pipeline (_fmos
), one of these
observations might show visible artefacts that you wish to exclude from the co-add. The size of the region in the
jack-knife image which is used to do the whitening step is determined automatically, but the method may fail if
the box is too small.
If you decide to re-run this step you first co-add all the _mappsf
files to create a co-added PSF using the
Picard recipe MOSAIC_JCMT_IMAGES
.
Next create a parameter file (recpars.lis
) for the jack-knife recipe (SCUBA2_JACKKNIFE
) containing the
following lines.
Another option for this parameter file is WHITEN_BOX
to set the size of the region used to calculate the angular
power spectrum. Finally run SCUBA2_JACKKNIFE
.
This will create files beginning with pgYYYMMDD
that should have the same suffices as above: _wmos
, _whiten
, _cal
, and _mf
.
This example is concerned with recovering bright extended emission. The signal from extended emission
varies slowly as seen by the array passing over it. It thus appears at lower frequencies in the power
spectrum and complicates the high-pass filter selection. Too harsh a filter will make flat maps but any
extended emission will have been removed in doing so (see Section 6.3 and Mairs et al. 2015 [15]).
Step 1: Running the map-maker
We run the map-maker using dimmconfig_bright_extended.lis
; we have also specified a couple of overrides on the
command line—maptol
= 0.04 is slightly more stringent than default and ast.zero_snr
= 3.5 constrains everything
below 3.5 to
zero (see also the ast.zero_snrlo
) parameter.
In this example we give the map-maker a file containing a list of the input files (filelist.txt
) and
dimmconfig_bright_extended.lis
is in the local directory.
The resulting map is shown in Figure 7.3.
Step 2: Generating an external mask
Next we create an external mask from the output of makemap. Here we follow the steps outlined in
Section 6.6.
This S/N map is thresholded to set everything below 3 to 0 and everything above to 1.
The thresholded map is shown in the left-hand panel of Figure 7.4. The next step is to smooth this map by convolving it with a Gaussian of 16 arcsec. For this we use a factor of 4 (pixels) for the FWHM parameter (since the default pixel size at 850 m is 4 arcseconds).
We threshold the map again to produce our mask. In this case all values below our threshold are set to ‘bad’. The smoothed map now has values scaled between 0 and 1, we set our threshold at 0.02 to include more of the emission beyond the 3 edge.
The final mask is shown in the right-panel of Figure 7.4. Note how it encompasses more emission and has softer
edges than the first threshold map.
Step 3: Re-running the map-maker with an external mask supplied
As a last step the map is re-made with this mask supplied as an external file. For this run we apply the
additional parameters in a personalised configuration file, mydimmconfig.lis
.
The configuration file, mydimmconfig.lis
, has the following format (note how it is based on dimmconfig
_bright_extended.lis
). It has decreased the convergence parameter to maptol
= 0.03 but increased the
number of iterations to compensate as 40 is unlikely to be sufficient.
Step 4: Cropping the map
We now crop the map to remove the noisy edges using the Picard recipe CROP_SCUBA2_IMAGES
. To determine
what to trim we can look at the exposure-time image with Gaia. See Figure 7.5.
The exposure-time map shows a sharp drop off at a radius of 30 arcmin. We can thus specify a parameter file
called crop1800.ini
like below:
and then run:
The final cropped map is shown in Figure 7.6. Compared with the first map out of the map-maker (Figure 7.3), slightly more of the faint extended emission is apparent.
One of the challenges facing this type of reduction is the need to account for both faint extended structure and very bright sources in the same map. You may find some degree of bowling remains around the brightest sources.
There are areas you may wish to experiment with. One is to adjust the filtering. Another option is to supply an external mask from a different dataset, e.g. a Herschel map. See Chapter 6 for further discussion.