NIRCam ERR measurements (from photutils) of same source are inconsistent depending on the pixel scale of the mosaic

[stscijgbot-jp]

Issue JP-3680 was created on JIRA by Maria Pena-Guerrero:

A user reported via Help Desk ticket, that NIRCam ERR measurements of the same source are inconsistent depending on the pixel scale of the mosaic. This was confirmed by the NIRCam team and they suspect a bug in the resample step.

The notebook the user provided can be found in the Help desk ticket: https://stsci.service-now.com/nav_to.do?uri=%2Fincident.do%3Fsys_id%3D9b3c5903934b8e50619dfe4c5cba103f%26sysparm_domain%3Dnull%26sysparm_domain_scope%3Dnull%26sysparm_view%3DJWST%26sysparm_view_forced%3Dtrue

 

[stscijgbot-jp]

Comment by Bryan Hilbert on JIRA:

The values in the SCI extension of the mosaics were consistent across changing output pixel scale values. It is the values in the ERR array that vary with the output pixel scale.

[stscijgbot-jp]

Comment by David Law on JIRA:

Maria Pena-Guerrero : Just to confirm, this bug is present in the B11.0 (jwst 1.15.0) code, right?

[stscijgbot-jp]

Comment by Maria Pena-Guerrero on JIRA:

correct, present in B11.0

[stscijgbot-jp]

Comment by Bryan Hilbert on JIRA:

Anton Koekemoer has had a chance to take a more detailed look at this issue. He reports:

"I still confirm the 2x inconsistency in errors, I also confirm that our test from the other day reproduces the photutils approach -- but photutils may be part of the issue, here’s why:

• since the ERR values are in surface brightness units (MJy/sr) their pixel values should be similar regardless of mosaic pixel scale. And we confirm this is the case when we look at the ERR arrays in these images (ie the ERR pixel values in the 30mas and 60mas mosaics are similar, so the ERR arrays are in fact consistent).
• instead, the photutils approach may be part of the issue: since your test reproduced the photutils results, then the photutils approach of calculating errors in these units may need to be modified to take into account the fact that ERR arrays are in surface brightness units, for non-default pixel scales."

So perhaps this is not a pipeline bug at all, but a photutils bug? Larry Bradley

[stscijgbot-jp]

Comment by Anton Koekemoer on JIRA:

To expand a bit on the above -- it all works fine, ie the pipeline ERR array and also the photutils measurements, when the mosaics are at the default pixel scale, so no issues there at all (ie this doesn't impact any of our default pipeline products).

But with a non-default pixel scale, the way in which the ERR array is defined (MJy/sr) means the interaction with photutils needs to be slightly adjusted to account for the relative change in pixel scale.

Attached here is the plot from the original helpdesk call (all public data), showing the 2x difference in the error values -- the photometry flux values are all consistent, just the error values are off by 2x, or by sqrt(relative pixel area)

!image-2024-07-05-11-53-01-534.png!

[stscijgbot-jp]

Comment by Larry Bradley on JIRA:

The very first thing the JWSTSourceCatalog step does when creating the source catalog is to convert the data and error array from MJy/sr to Jy. All of the photutils photometry calculations are done from data/error arrays in flux (Jy) units, so it's not a photutils bug.

The unit conversion from MJy/sr to Jy is performed in the source catalog step using the model.meta.photometry.pixelarea_steradians value of the input image. -Does the resample step update this value for non-native pixel scales? That could be the source of the bug.- Edit: that's not the issue. The issue is the input error array (see comment above).

[stscijgbot-jp]

Comment by Larry Bradley on JIRA:

I took a look at the images provided in the help desk ticket. The issue is definitely with the input data. The error arrays for 30 mas/pix and 60 mas/pix data are not equivalent. You can verify this by taking a single pixel in the 60 mas/pix image and comparing it with its corresponding 4 pixel (2x2 region) in the 30 mas/pix.

First convert the data and error arrays from surface brightness to flux density units (e.g., multiply by PIXAR_SR * 1e12 to convert to uJy).

Flux: Comparing the 60 mas/pix data value with the sum of the corresponding 4 pixels in 30 mas/pix data images agree.

Errors: The square root of the quadrature sum of the 4 pixels in the 30 mas/pix image is half the value of the same pixel in the 60mas/pix error image. They should give the same value. Therefore, one of the error arrays in the images provided by the user is wrong (scaled improperly).