Background subtracted from co-adds

Dear all,
There is a problem I have been running into with the background estimation for co-added images. It is quite simple to find out the subtracted background for calexp images, but I am not able to do the same for the co-added ones.
If anyone has any pointers on how to do it I’ll be really grateful.

Thank you for your time

Hi Nandini, thanks for posting this question. Could you share an example of the code you tried and the error message that was returned when it failed?

Hello Melissa, thank you for your response. I am looking for some sort of documentation of the co-add images to retrieve the background estimation. So there is a tutorial for how to obtain the background for a calexp frame, as given in Notebook 5 on the RSP:
bkgd = butler.get(‘calexpBackground’, **dataId) [for example]
I am not sure what is the analogous command to get the background subtracted for the co-added images
(My apologies if I am doing something silly)

Try the deepCoadd_calexp_background dataset type.

1 Like

Thanks @price .

Since I didn’t know much about working with the subtracted backgrounds, I’ve drafted a notebook to explore them. It’s here in my branch of the delegate-contributions-dp01 repo:

I have a couple of questions about the deepCoadd_calexp_background – is there a place in the LSST Science Pipelines where I can learn more about it? I searched, but could not find descriptions.

In the notebook, I retrieve the background for a random coadd – it is uniform and very small values. Maybe this makes sense, since the backgrounds would be subtracted before coaddition, but then… was this tiny amount subtracted off the coadd after it was created? Is there something better to use that represents the backgrounds of the individual images, or should the user obtain the backgrounds for the individual calexps that went into the coadd? Suggestions of what else to include for users in this draft notebook would be very welcome.

1 Like

Backgrounds are subtracted off the individual images before coaddition, so there isn’t much background to subtract from the coadd.

Thank you for your reply @price
Are the values in the deepCoadd_calexp_background images weighted for the individual calexp images that have been stacked to create the Coadd frames?
And is there a place where it is possible to read more about the background dataset for such kind of information in general?

The coadd background is just measured from the coadd. There’s no direct weighting; maybe the variance plane is used, I don’t remember.

I don’t think there’s much LSST documentation on background estimation. I know there are Community posts about sky frames. There should be some information in the HSC SSP release and pipeline papers.

Thanks @price and @nandinihazra.

To the intro-to-backgrounds draft notebook I added links to Bosch et al. (2019) and to the Data Management Science Pipelines Design document which I think are currently the best two places to read more about background subtraction. @nandinihazra I think you might find answers to your questions in the latter.

I also added a bit more introductory information about the process, based on some of the comments above. We’ll get my branch of the delegate-contributions-dp01 repo merged to main soon.

Thanks again @nandinihazra for bringing this up, and for reviewing the notebook for me.

Any readers who are looking for this notebook, see: