Hi,
I am trying to make coadd of difference images (e.g. stacking tens of difference images taken by LSST in one night) and run DiaSource detection with the DP0.2 data on the Rubin Science Platform (RSP). Our purpose is to identify transients fainter than the detection limit of standard 30s exposures.
I have checked the nice and detailed tutorial notebooks on the RSP, but it seems that making coadded difference images is not covered yet. May someone help me please? Any advice would be greatly appreciated!
Hi @hbz, you are correct that there is no tutorial on how to stack DP0.2 difference images.
Also, the tutorials on how to create custom coadds (notebooks 9a and 9b) with DP0.2 data have recently been decommissioned. The LSST Science Pipelines have evolved significantly since DP0.2 was processed, and these tutorials could only run with a very old version from 2022. They were no longer an appropriate learning tool and have been removed from the tutorial set. We are now working towards Data Preview 1, and will provide guidance for custom coadds for the DP1 data.
I’m going to mark this reply post as the solution, because the answer is that the advice is not to try and stack DP0.2 difference images (but rather to wait for guidance based on DP1).
However, I’d like to hear more about your science use case for detecting sub-threshold transients (e.g., this would be only for Deep Drilling fields?), and more about doing this by stacking difference images (instead of, e.g., using forced photometry or stacking direct images). This would help us to put together guidance for DP1.
It is nice to hear that the DP1 data and guidance are coming. We greatly appreciate the LSST team’s continuous efforts and will wait for the new guidance.
Regarding our science use case, indeed we plan to detect sub-threshold transients only in Deep Drilling fields and then carry out deep spectroscopy for these faint transients.
Because the difference images for single exposures are already available on the RSP, we thought that stacking difference images might be the easiest way to detect sub-threshold transients. On the other hand, stacking direct images and then subtracting the reference image should also work for our science, although we are not sure which is the better and recommended way for LSST data (difference->stacking or stacking->difference).
Thanks @hbz, that additional information about your use case is really helpful. We’ll look towards providing guidance on stacking differences vs. differencing stacks for DP1, in the context of deep drilling field analysis and faint transient detection.
A side note, because you mention follow-up spectroscopy, and perhaps you’re already aware but – one thing to keep in mind is that during LSST operations there will be an 80-hour embargo on all newly acquired images.