I’ve been reprocessing a portion of the DECam-based “DECaLS” survey with the LSST pipelines. Starting from the raw DECam data, we’ve been able to generate calibrated CCD-level and coadd-level products, which has been great.
One element included in the published DECaLS processing but currently missing from our version built with the LSST pipelines is infrared (WISE) forced photometry. So I am curious to know about the status of cross-instrument forced photometry with the LSST pipelines. To be specific, I want to photometer WISE images using the LSST pipelines, at the sky positions of DECam sources that we’ve extracted/cataloged with the LSST pipelines (also adopting the LSST pipelines’ source morphologies). Is this currently possible?
I was pointed to a DESC paper (h/t @sfu) that possibly did something like this using simulated Rubin+Roman data, though with v19 of the LSST pipelines, which I understand to now be quite outdated. I am using a Gen3 version of the LSST pipelines.
Also, if I were to attempt this WISE forced photometry with the LSST pipelines, I would want to start with science-ready coadded WISE sky maps, not the raw WISE exposures…would this be a problem? The documentation about making a new obs package seems to emphasize “raw data”, but maybe that’s just a matter of wording. Thanks very much.
It all depends on how hard it is to make the coadded WISE images look like coadds produced by the LSST pipelines, in terms of mask planes, PSF models, WCSs, etc. You may also need a custom lsst.skymap.BaseSkyMap implementation to define the regions of sky you actually have coadds for. That’s all together typically a pretty tall order, and it’s why we usually recommend starting from raws. But if you can get that all working, you should be able to ingest them into a butler data repository and then the coadd processing pipelines should mostly just work.
I can’t offer very much tangible help in the very near term – you know the WISE data model far better than I – but in the longer run this question is very interesting indeed to me. We are using a Butler-based pipeline for SPHEREx, at IPAC, and will be deploying forced photometry services based on that. If there were a way to build WISE forced photometry around the same framework, that would be most useful.
Thanks for your note, @gpdf . The mention of WISE’s data model also brought to mind the fact that (to my knowledge) the truly raw WISE/NEOWISE single-exposure images are not publicly available.
Given that both the available single-exposure and coadded WISE images already have at least some detrending applied, this makes me wonder: could one do something like systematically leaving various calibration steps as stubs when creating a WISE obs package, and/or make it so that e.g., the master bias is a frame full of 0’s and the master flat is a frame full of 1’s? Or is this a worse road to go down than @jbosch 's suggestion? Thanks again…
Instead of stubbing out IsrTask, I’d recommend creating a wholly different PipelineTask that takes the external detrended images (which you’d ingest directly into the data repository as something other than raw) and writes postISRCCD datasets that look like the ones produced by IsrTask. You’ll still need to come up with mask and variance planes and an initial WCS, but you won’t need a PSF model (that’s fit later) or a particularly good WCS, and that sounds pretty feasible to me.