Pyse.py position errors in AARTFAAC

[gijzelaerr]

We need to have a way to deal with position errors in AARTFAAC when using pyse.py/TraP. There are two options for this:

• convert extremes of the positions to position error
• assume zero error to begin with, and then add in a systematic error (pixels?)

[dariocarbone6]

We might want to add a systematic position error (on top of the one calculated by PySE) in order to have the brightest sources being correctly associated. This is due to the fact that the de Ruiter radius depends on the position error and this is smaller for brighter sources (they are better identified). To estimate if we need to add this systematic error and, if so, quantify it we need a dataset with bright sources with well known positions; I think this would be better done with simulated images. When Alexander and I did this for LOFAR images we used both simulated and real images.

[gijzelaerr]

ok @dariocarbone6 do you actually have some good simulated images with bright sources so we can do this?

[dariocarbone6]

Hi! The tests we performed to include a systematic errors were performed using real LOFAR maps. I can point you to the ones I used, but this was mainly done by Alexander, he can tell you more about it. We could create simulated maps to test this, but the main contribution that is causing the missing association is mainly due to ionospheric effects that are not so easy to simulate.

[gijzelaerr]

@dariocarbone6 can you ask alexander what came out of these tests and report this here?

[ajvdhorst]

Hi guys, we indeed used real LOFAR maps for this. If you want to do the same exercise for AARTFAAC, what you'll have to do is make AARTFAAC images, determine the positions of sources with PySE, and them compare those source positions with the known positions of those sources. If you do this for several different images and many different sources, you'll get an idea of the magnitude of the difference between the measured position and the real position (the latter you can get from the skymodel); which will then give you an idea of what the systematic error is (since the statistical error is already reported by PySE; and total error = statistical + systematic). I think it would really be good to do this for AARTFAAC again, because there may be different systematics that you're dealing with, compared to 'regular' LOFAR images.

[AntoniaR]

This is an ongoing issue to check if still useful with new AARTFAAC data but not essential for R7. Moving to the long term milestone.

[HannoSpreeuw]

How does this discussion tie in with PySE issue #71, in the broader context of being able to add calibration uncertainties and biases to a config file which pyse.py taps into?

Would accommodating for such a config file solve this problem?

That config file would probably be a simple Python dataclass, with attributes such as ew_sys_err, ns_sys_err, eps_ra, eps_dec, frac_flux_cal_error and clean_bias_error.

[AntoniaR]

I think this discussion does indeed tie in with issue PySE issue 71. In principle, that and the current systematic position uncertainties should work for AARTFAAC data just as well as for other radio images. @HannoSpreeuw you have the most recent experience working with AARTFAAC images, have you noticed any anomalous behaviour for AARTFAAC sources?

[HannoSpreeuw]

I am frequently running PySE on an AARTFAAC image from Transient Buffer Board data (credit: Mattia Mancini) and that works fine, but noticing any anomalous behaviour would require me to cross-check with e.g. LOFAR images.

I guess this issue is not about anomalous behaviour as such but about being able to take account of e.g. large positional uncertainties due to poor calibration solutions from an unstable ionosphere.

The need for providing a convenient way to propagate (large) positional errors and biases into TraP has been acknowledged and will probably be fulfilled this year.