RCAL-853: Apply velocity aberration correction to the WFI WCS

[stscieisenhamer]

Resolves RCAL-853

Closes #1192

This PR applies the velocity aberration correction to the WFI WCS.

Checklist

• added entry in CHANGES.rst under the corresponding subsection
• updated relevant tests
• updated relevant documentation
• updated relevant milestone(s)
• added relevant label(s)
• ran regression tests, post a link to the Jenkins job below. How to run regression tests on a PR

[stscieisenhamer]

@schlafly Current PR includes basically a duplicate of the JWST correction, that of applying the single scale as given by meta.velocity_aberration.scale_factor.

However, the questions in the issues relate more to whether scale_factor is sufficiently close to what ra_offset/dec_offset are, such that they are indistinguishable, or, more importantly, no systematic effect is introduced using the single value.

Nominally, all these values are being calculated in RSDP. Are we tasked to provide those calculations, and hence also deciding whether the single scale_factor or the individual offsets are to be used in the correction? Towards this, I am getting some sample data and seeing about making some empirical estimates.

[codecov]

Codecov Report

Attention: Patch coverage is 35.00000% with 13 lines in your changes missing coverage. Please review.

Project coverage is 78.14%. Comparing base (9331d16) to head (802d02a).

Files with missing lines	Patch %	Lines
romancal/assign_wcs/pointing.py	27.77%	13 Missing ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##             main    #1354      +/-   ##
==========================================
- Coverage   78.52%   78.14%   -0.39%     
==========================================
  Files         117      117              
  Lines        7833     7852      +19     
==========================================
- Hits         6151     6136      -15     
- Misses       1682     1716      +34     

Flag	Coverage Δ		*Carryforward flag
nightly	62.23% <ø> (-0.02%)	⬇️	Carriedforward from 9331d16

*This pull request uses carry forward flags. Click here to find out more.

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

[schlafly]

Hi Jonathan,

We will need to compute scale_factor. I think that's a simple function of the velocity of the spacecraft---some beta or gamma factor? I found the computation of that factor in the Webb code, and put the link in the issue.

It depends on the speed of the spacecraft, which is more annoying to estimate. I think the easiest thing to do would be to pretend that the spacecraft were exactly at L2, so that the spacecraft's velocity is exactly the same as the Earth's velocity times (l2_distance_from_sun / earth_distance_from_sun), where I guess I'm ignoring effects from the moon but that seems very small.

I had not appreciated until looking at your code that that the ra_ref and dec_ref correspond to the aberrated v2/v3 rather than the nominal v2/v3. Those ra/dec_ref come from romanisim at present and so I think it is straightforward to get that right. But I'm rather attached to the notion than ra_ref, dec_ref corresponds to v2_ref, v3_ref, which I claim this will break. Is that how things work for Webb?

[stscieisenhamer]

The previous comment generated much offline discussion. The summary and path forward is below; let me know if I got it wrong.

Concerning where the computation of va_scale occurs: For JWST, this is done in SDP using jwst cal code while the Level 1 file is built. For Roman, this can be done as JWST, or in the set_telescope_pointing equivalent.

For JWST, as described, the RA_REF/DEC_REF are the "semi-aberrated" versions. These values are determined in set_telescope_pointing and are calculated using the SIAF transformation from V1. V1 has been corrected. However, the SIAF transformation does not take into account the differential aberration, which is why the jwst wcs models add the shift component. These values are never updated in the headers. For Roman, these values can be updated in the assign_wcs step, or the set_telescope_pointing equivalent can take into account the DVA in the first place.

Just recently learned (27Aug24) about the plans on back-porting the Level3/GAIA corrected solution to the Level 2 files. Details of that does not matter. Note that this process will invalidate all the wcsinfo keywords.

For the moment, I believe the plan is to leave the assign_wcs code as is. I can add the updating of the wcsinfo to ensure that RA_REF/DEC_REF represents what V2_REF/V3_REF is pointing at. Note that this updating will use the WCS object which will also include distortion.

[schlafly]

It's fine to compute va_scale in romanisim (when the L1s are produced).

Let's follow Webb for ra/dec ref. i.e., radec_ref only match v1v2_ref exactly for the boresight; the others don't match up exactly due to differential aberration but they're within a pixel or so. This is tracked in the WCS but not in the wcsinfo. I'll add an issue to rad (spacetelescope/rad#437) bringing up whether we want to add v2a/v3a_ref which would allow these to match up better, though still not perfectly after tweakreg has been run since we don't update the wcsinfo with tweakreg information.

Re backporting the L2 tweakreg WCSes to the L1 files, we'll end up doing something like providing a pile of L2 WCS files separately from the rest of the L2s (which would include tweakreg, aberration tweaks) or just the tweakreg info plus code that turns the wcsinfo + tweakreg info into a new WCS (basically, repeating assign_wcs, aberration, ...). I don't think that invalidates the wcsinfo keywords any worse than they are already invalid in the L2s.

Let's not update ra_ref/dec_ref yet and instead just continue following Webb.