ERIS tutorial notes: Combining data from different arrays, 3C459 MERLIN + VLA

The science justification

3C459 has a very asymmetric radio structure, a high infrared luminosity and a young stellar population. The eastern component of the double-lobed structure is brighter, much closer to the nucleus and is significantly less polarized than the western one. This is consistent with the jet on the eastern side interacting with dense gas, which could be due to a merged companion or dense cloud of gas.

These data are published in Thomasson, P., Saikia, D. J. & Muxlow, T. W. B., 2003 MNRAS, 341, 91

The data used in this tutorial consist of the MERLIN 5GHz and VLA 5GHz data published in this paper. Alone both of these radio data sets provide a large amount of information regarding the radio structure of the double-lobed radio galaxy. However, only by combining these two data sets is it possible to image finer detail within the jets, seen in the MERLIN images, whilst also imaging the more diffuse radio emission that is observed in the VLA images.

The observations

These MERLIN 5GHz observations were made in July 1992 & July 1995 at 4 separate frequencies (4546, 4866, 4993 & 5186 MHz). These multi-frequency synthesis (MFS) data have been calibrated and combined to form a single calibrated u,v data-set.

The VLA A-array data were observed at 4885 MHz on September 8th 1983.

The calibration of both of these data-sets has followed the same methods as were outlined in earlier tutorials & lectures.

Further details of these observations can be found in Thomasson et al., (2003).

The data

The calibrated and split u,v data for this tutorial should already be loaded into the data area, alternatively they can be download from these links.

The MERLIN MFS data are 3C459_MER.UVTB.FITS.
The VLA A-array data are 3C459_VLA.UVTB.FITS.

Then load these two data files into AIPS using the AIPS task 'fitld'
Task 'fitld'
datain 'WORK:3C459_MER.UVTB.FITS'
inp $check all inputs
go fitld

Do this for the MERLIN and the VLA data.

Imaging the individual arrays

Initially both of the two data sets need to imaged individually in order that the astrometry of the two images can be aligned.

Initially imaging the MERLIN data
Using Imagr make a cleaned image of the MFS C-band MERLIN data.
Here are some suggested imagr inputs to make a naturally weighted image with a restoring beam of 120mas:
Task 'imagr'
cellsize 0.02
robust 7;uvwtfn '' $ Use natural weighting
rashift -2.71 0 $ So that the centre of the image between the two lobes.
decshift 0.0095, 0
imsize 1024 1024
niter 1500
bmaj 0.120;bmin 0.120
go imagr

A contour image of these data with a 120mas circular restoring beam is here.

Similarly make a naturally weighted VLA image, in this case use a beam size of 500mas.
Task 'imagr'
cellsize 0.08
robust 7;uvwtfn ''
rashift -2.71 0. $ so that the centre of the image between the two lobes
decshift 0.0095, 0
niter 1500 $ no boxes
imsize 1024 1024
bmaj 0.5;bmin 0.5
go imagr

A contour image of these data with a 500mas circular restoring beam is here.

Now re-make MERLIN image convolved with a beam-size matching the VLA image
Use the same inputs as used to create the first MERLIN image but this time set bmaj=0.5; bmin=0.5.

A contour image of these data with a 500mas circular restoring beam is here.

At this point we have made two images at the same resolution, one form the MERLIN data-set and one from the VLA data-set. As imagr run it will have outputed a value that aips calls the "sum of weights". This is written to the message window as imagr runs. Make a note of this number for both the VLA and MERLIN data-sets. For each data-set this value will differ depending on the observations. When combining two data sets with very different values "sum of weights" the final combined data-set will be dominanted by the data with the highest weight, in the case of MERLIN + VLA combinations this will normally the VLA data-set. Ultimately the purpose of combing two data-sets is to produce images that reflect the atributes of both of these data-set, thus it is usual to re-weight one of the date-sets so that both provide approximately equal weighting in the combinations. When making the final combination, these sum of weights can be used to calculate this re-weighting.

Checking alignment and core flux

From the VLA and MERLIN images check the position and peak fluxes of the core. Both of these observations have been made at essentially the same frequency and the core is compact on our highest resolution scales.

In order to do check the core position use tvwin to box the area around the core of each image and then type imstat;maxfit. This should give you these results similar to these:-

From VLA image @ 500mas beam size
Core position is RA 23h14m02.31443s; Dec 03d48'55.1528".
Peak flux from maxfit 381.367mJy/bm.

From MERLIN image @ 500mas beam size
core position is RA 23h14m02.31971s; Dec 03d48'55.0287"
Peak flux from maxfit 429.813mJy/bm.

Derive these numbers from the images you have created don't assume that the numbers above are correct.

As you can see at the moment the VLA and MERLIN images do not perferctly align, thus we need to shift the position of one of these images so it aligns with the other before combining the two data-sets.

Since the absolute astrometric accuracy of the VLA image is not as high as the MERLIN we will shift the position of the VLA pointing centre so that the cores of both images align.

Firstly, for cautions sake, use the AIPS task UVCOP to make a copy of the VLA u,v data.

Applying a shift in RA and Dec to the VLA uvcop data-set so that the core of the VLA aligns exactly with the core position as measured from the MERLIN data.

RA axis:
From the image header the pointing centre of the VLA data is 23h14m02.315s; Dec 03deg48'55.140".

Thus we need to shift the RA of the pointing centre of the VLA data by 0.00528s (difference between the two core positions in RA) to 23h14m02.3202801, so that when imaged the core will be at the correct RA. (calculate this position yourself, from your own images -- Do not trust the numbers given here make sure they what you believe they should be! There could easily be a typo in the tutorial!).

To re-define the RA axis use axdefine. Example inputs are as follows:
naxis 4,axtype 'ra';axref 1; axinc 0;axval 348.50 0.00967834
axdef;imh (check that the new pointing centre is where you expected it to be).

Note that the value of the RA as listed in the image header has now been changed. Is this position were you intended it to be? If not recalculate you axval.

DEC axis:
Doing the same exercise for the declination axis, the VLA image needs to be reduced by 0.1240997" to become Dec=03d48'55.0269"

Redefine the Dec axis using axdefine inputs as follows:
naxis 5;axtype 'dec';axval= 3.80999999 0.00528522; axdef; imh

Again check that this has done what you expected!

Flux scale
In addition to the miss-alignment of the position axis the flux scale also needs to be aligned.

Using uvmod rescale the fluxes of the modified VLA uv file so that they match thos in the merlin data set.
Within uvmod use a rescale by a multiplying factor of which is a will correct the VLA core flux relative to the MERLIN core flux.

Check the alignment

Now re-make the VLA image from the modified data and check the core position from the VLA image. Re-measure the core position and compare this with the position you derived from the MERLIN image. These should now be consitent with each other to better than a few mas.

Now combine & re-image

Weights of the two arrays:
As you re-call we checked the sum of weights (see above) of both of these data sets. If you check the two numbers from these two data-sets are quite similar. As such in this case there is no need to rewight these data sets, so we shall leave reweight =0.

Using the task dbcon combine the MERLIN data-set and the re-positioned VLA data-set.

Task 'dbcon'
getn i $ The MERLIN uv
get2n j $ The modified VLA data
This is the VLA data-set that you have just corrected.
dopos 1,-1
doarray =-1
go dbcon

A final dbconed u,v data set are 3C459_M+V.DBCON.FITS.

Dbcon will produce a combined u,v file, re-run imagr on this combined file.
Here are some suggested Imagr inputs, but please try your own to get the best image you can.
robust -2
bmaj 0.1 ;bmin 0.1

This is your basic combined image, although this can now be improved by careful imageing. For example, you may wish to use a combination of imagr (clean) and Vtess (mem) to image this data-set, this was how the figure 3 in Thomasson et al (2003) was created.

Image the polarisation

These data have been observed in full polarisation and calibrated. Hence at this stage you can also make a combined image of the polarisation.

Using imagr make a deep cleaned image in Q & U stokes from your combined VLA+MERLIN image.

Then combine the U & Q maps in comb
1)Task 'comb'; opcode 'poli'; getn q_map-file; get2n u_map-file;go
1)Task 'comb'; opcode 'pola'; getn q_map-file; get2n u_map-file;go

Use the task kntr to make a contour map fo this.

Here is an image of the combined data set with polarisation vectors (M+V_KNTR_POL.PS).

Making a better IPOL image

To improve the image we can carefully deconvolve the core with imagr and then use vtess in order to image the extended continuum. Or you can use try to use sdclean, within imagr.

These methods are simliar to those discussed in the lectures on Tuesday.

Here is an example of some inputs to create a restored image using clean and MEM.

Not that you do not have to use this method. Maybe you want to try to image the data using different weightings within imagr, for example. The choice is yours.

Use imagr to clean out the bright core components.
Task 'imagr'
cellsi 0.01
imsize 2048
bmaj -0.070;bmin -0.070
rashift -2.71;
decshift 0.0095
boxes 3
clboxes 540 908 733 1186 733 1000 756 1022 756 909 1552 1192 0
robust -3
phat 0.7
factor 0.3
imagprm(10) 1
niter 100
go imagr

Then vtess for the extended structures

Task 'vtess'
getn clean map
get2n dirty beam
Noise(1) 250e-5
bmaj 0.07;bmin 0.07
niter 50
blc 0;trc 0
dotv 1
go vtess

Restore the clean componets
task 'rstror'
getn vtc
get2n cleaned imagr
go rstor

Now kntr the final image. How does it look? How could it be improved further? Does it look like the images in the paper yet.

Do experiment with different cleaning and imaging methods for this data, as well experiment with presenting it.