A help desk user's suggestions to improve output from source finding

[stscijgbot-jp]

Issue JP-3889 was created on JIRA by Mihai Cara:

This came from help desk ticket INC0208380. User has several improvement suggestions related to documentation and a feature request:

1 - please include positional uncertainties in the output catalog. They are fundamental for astrometric measurement.

2 - please modify the description of "semimajor_sigma" and "semiminor_sigma" in the online documentation, because the way they are currently written is quite misleading and gives the strong impression that those quantities are, indeed, 1-sigma standard deviations on the source position.

CC: Tyler Pauly Nadia Dencheva

[stscijgbot-jp]

Comment by David Law on JIRA:

Hm, positional uncertainties could be nontrivial in a way that isn't misleading given the combination of stochastic uncertainties on the measurement (probably very good) and systematic uncertainties on the pointing (potentially quite bad for some exposures even if generally fairly small).

Tagging Anton Koekemoer Karl Gordon Greg Sloan Bryan Hilbert Misty Cracraft to see if they have any thoughts.

[stscijgbot-jp]

Comment by Anton Koekemoer on JIRA:

In terms of stochastic uncertainties, it might be useful to consider some of what the community currently is doing, eg SExtractor has a few different ways to present positional uncertainties:

https://sextractor.readthedocs.io/en/latest/Param.html

https://sextractor.readthedocs.io/en/latest/Position.html#poserr-iso-def

(it provides other uncertainties too, the above are just an initial example)

[stscijgbot-jp]

Comment by Karl Gordon on JIRA:

If there is a way to give the uncertainties and clearly state that these are just the measurement uncertainties, then that would be good. Maybe the uncertainties on the absolute pointing due to guide star positional uncertainties could be added to explicitly alert the users to the systematics. Do we know what this systematic uncertainty is? 0.3", more/less?

[stscijgbot-jp]

Comment by Tyler Pauly on JIRA:

It may also be useful to discuss the expected value and use of the source_catalog step outputs - I have discussed issues and possible improvements with INS in the past, and the response fell more into "source catalog outputs are not science-ready", "use the catalog output as a 'quicklook'-like product" than "we should improve our source finding algorithm and validate its output". It's not clear to me that everyone is on the same page.

[stscijgbot-jp]

Comment by David Law on JIRA:

Just adding a note from my point of view that we should do what we can to provide reliable source catalogs.  They won't be ideal for all use cases, and many astronomers will want to do their own analysis, but if there are any major problems with the step we should open tickets and address them.

With regard to absolute uncertainty, I imagine it will vary between instruments (larger for MIRI than NIRCam), whether or not tweakreg has been run successfully, etc.  Most of the time my understanding is that NIRCam absolute pointing uncertainty should be 0.1" or so (I think), but that's a long way from the mas precision that astrometry work would want.  It's probably fair to say that the automatic source catalog step may never be sufficient for folks who want to push astrometry to extreme precisions, and that this would require out-of-pipeline analysis.  Being clear about the limitations is good.

[stscijgbot-jp]

Comment by Kevin Volk on JIRA:

This NIRISS imaging people have discussed this ticket. First, concerning the position uncertainties: as far as we know the positions in the catalogue are derived from the segmentation map signal centroid values, and this is a complex quantity for extended sources. Hence, in many cases it would be very difficult to estimate the uncertainties. Even for point sources the centroid position uncertainties are not necessarily simple to estimate because it depends on the contrast between the PSF peak and the background level which affects the size of the segmentation area. For a bright source one may get the diffraction spikes in the segmentation area, and then it might be that this can affect the centroid values. There is also the question of whether this means the absolute astrometric positional uncertainty or the internal relative astrometric uncertainty. These are rather different things. So we thought that this was not easy to do, and possibly not very useful. Second, in terms of the definitions of the "semimajor_sigma" and "semiminor_sigma" values, the definition of these is copied from the photutils documentation, since the pipeline is using the photutils routine and taking the output values. Hence it does not seem to us that there is any reason to change the documentation of these qualities. If the documentation is misleading that is a photutils issue not something for the JWST pipeline. It would not make sense for us to be using these values from the photutils routine but using a very different description of what the values are.