What is the sky subtraction algorithm being used currently by the pipeline? It is using background matching from the 2020 algorithm workshop or are there any updates? Thanks.
The current background estimation and subtraction algorithms in use for processing LSST data are largely an evolution of those described in Bosch et al. 2018 (Section 4.6).
In brief, the background is estimated per-detector in bins of 128x128 pixels, masking pixels identified as either belonging to a detected source or otherwise unsuitable for science. We calculate the sigma-clipped mean of the remaining pixels within each background mesh element, and then fit a 6th order Chebyshev polynomial to these data in order to determine a background flux level for every pixel in the image. This is our final visit-level (single frame measurement) background solution.
To support processing coadded data, we make use of the sky correction algorithm (Aihara et al. 2019, Section 4.1). The goal of the sky correction task is to fit a background model to all detectors from an entire visit, rather than handling each detector separately. The visit-level backgrounds are added back on to the science data, and the background is estimated in superpixel meshes of 8k x 8k pixels (approximately 2x2 LSSTCam detectors). We then fit and subtract a sky frame (see Aihara et al. 2019, Figure 4) which captures features of the camera which remain static as we move around the sky. We then perform a final round of background estimation in superpixel meshes of 4k x 4k pixels to account for any residual small-scale structure in the background.
We don’t currently employ any form of background matching as part of standard LSST data reduction efforts, however, a good amount of exploratory work and prototypes have been pursued. Any developments will be announced in the Community Forum when we have news to share.