Unable to get WCS information from fits files

[lsmercadante]

I have downloaded the difference images and read in the fits files to a jupyter notebook using astropy. The problem I am having is I cannot convert the pixels to WCS coordinates. Whenever I attempt this I get an error. Here is how the image is read in:

And here is the error I encounter:

I'm not sure how to fix this problem.

[genghisken]

Hi Lilah

Thanks for raising this. It helps us keep track of any solutions and time spent on them.

As you may have seen there is quite strong feeling about astropy standards compliance! I'm not a WCS expert, but one solution I've found that kind-of works is below. I'm not sure what effect deleting the corner cards have, so I'd need to double check with local experts. But try that and see what happens. The solution will still whine about SIP, but should only produce a warning, not an error this time.

from astropy.io import fits
from astropy.wcs import WCS
h = fits.open('01a57801o0580o_diff.fits')
header = h[0].header

w = WCS(header)
# whines about SIP but then will also barf out with an error.

# Delete the corner info

del header['CNPIX1']
del header['CNPIX2']

# Now try again - still whines about SIP but should't crash this time.

w = WCS(header)

[lsmercadante]

This seems to have solved the problem, thank you! As a follow up question, I noticed that when the pixel position changes from say (200,200) to (199,200), instead of just the RA value changing, both the RA and DEC values change.

This seems unusual to me and I was wondering if there is a reason for this?

[genghisken]

Nothing unusual I don't think - let us know if you see something particularly unusual. Is the dec change particularly large? Although it's an equatorial mount there will be slight variations in the orientation of the camera, plus the source image is a very large field of view (29 sq degrees), so astrometric corrections will have been applied to account for distortion, etc.

[djbrout]

Excellent. Thank you Ken!

[rsiverd]

In addition to using a few keywords in ways we shouldn't, the reduced ATLAS images actually contain two different WCS solutions. A first pass is performed using astrometry.net that installs SIP terms. A second solve adjusts the solution and saves its result using the PV distortion convention (TPV CTYPE). Unless you modify the headers, you will be getting the TPV solution.

I use the following code to clean things up on-the-fly and take the SIP solution:

import astropy.io.fits as pf
import astropy.wcs as awcs

## Load image header:
ihdr = pf.getheader(context.input_file, extname='image')

## Modify header for compatibility:
whdr = ihdr.copy(strip=True)                            # make a working copy
toss = ['RP_SCHIN', 'CDANAM', 'CDSKEW', 'RADECSYS']     # these prompt warnings
toss += ['CNPIX1', 'CNPIX2']
toss += [x for x in whdr if x.startswith('PV')]         # these are PV terms
for kk in toss:
    whdr.remove(kk, ignore_missing=True, remove_all=True)

## Update CTYPEx to prompt SIP interpretation:
whdr['CTYPE1'] = 'RA---TAN-SIP'                         # overwrite RA---TPV
whdr['CTYPE2'] = 'DEC--TAN-SIP'                         # overwrite DEC--TPV

## This should work correctly and without tons of warnings:
twcs = awcs.WCS(whdr)
```

[genghisken]

Rob's solution above is the definitive one. Please use it, rather than my hack!