Skykatana: a scalable framework to construct sky masks for the Vera Rubin Observatory and large astronomical surveys

TL;DR

Skykatana offers a scalable, open-source framework to build, combine, and visualize sky masks at survey scale using HEALPix/HEALSparse and MOC representations. It integrates on-demand star masking via Gaia queries through HATS/LSDB, supported by empirically calibrated radius–magnitude relations from Rubin DP1, with memory-efficient, streaming I/O and a modular pipeline of stages. Key contributions include geometry-, property-, and catalog-based mask builders, breadth-first MOC chunking, and on-demand, high-order masking that can generate multi-billion-pixel masks within modest resources for a target order $o_{ m max}$, enabling reproducible angular selection functions. Demonstrations on an HSC–WISE composite mask and Rubin bright-star masks show practical applicability for Rubin/LSST-scale surveys and point toward extensions to weighted maps and probabilistic analyses.

Abstract

Modern wide-field surveys require robust spatial masks to excise bright-star halos, bleed trails, poor-quality regions, and user-defined geometry at scale. We present Skykatana, an open source pipeline that builds and combines boolean HEALPix/HEALSparse maps into science-ready masks and engineered for low-memory operation. Skykatana can efficiently construct, visualize multi-order coverage maps and generate random points in high-resolution masks over half of the celestial sphere with very limited resources and leveraging the hierarchical partition of data the HATS/LSDB framework. We demonstrate two end-to-end applications: (1) a Subaru HSC-WISE composite mask; and (2) Rubin star masks generated on demand in the Rubin Science Platform by querying HATS/LSDB Gaia data and assigning radii from empirical fits to Rubin DP1 data. We release full bright-star masks for various regions of the Rubin footprint and describe performance and scaling. The code, documentation, and examples are publicly available at https://github.com/samotracio/skykatana, and the LSST masks can be obtained from https://osf.io/r5vw6

Figures (18)

• Figure 1: Schematic view of pipelines and stages in Skykatana.
• Figure 2: Mean radial density profiles of DP1 sources around Gaia stars in the same magnitude bin ($10.3 < G_{\mathrm{Gaia}} < 11.6$). The top panel shows the profile obtained when all detections are included, displaying the characteristic features of bright-star contamination: a spurious central overdensity from core fragmentation, a depletion region caused by incompleteness within the stellar halo, and a secondary excess associated with the fragmented outer halo.The bottom panel shows the corresponding profile after removing halo-related artifacts using blendedness_flag == False, yielding a smooth, monotonic increase toward the field density. Shaded regions in both panels represent the $1\,\sigma$ dispersion across stars in the bin.
• Figure 3: Fitted radius–magnitude relations for all LSST bands in the halo and no-halo cases. For the halo case, three power-law segments are fitted per band, with boundaries at $G_{\mathrm{Gaia}} = 13.6$ and $G_{\mathrm{Gaia}} = 15.0$, capturing the transition from strong halo contamination to the weak-halo regime. The no-halo case is well described by a single power law per band. Best-fit parameters are listed in Tables \ref{['tab:halo_band_parameters']} and \ref{['tab:no_halo_and_parameters']}.
• Figure 4: Best-fitting radius–magnitude relations in the $r$ band for three completeness thresholds (90%, 85%, and 80%), computed after removing halo artifacts. The corresponding fitted parameters are listed in Table \ref{['tab:r_completeness']}
• Figure 5: Radius–magnitude relations in the $r$ band for different limiting magnitudes ($r_{\mathrm{lim}} = 27.00, 26.75, 26.50, 26.25, 26.00$) used when computing background densities, all adopting the 85% completeness threshold. The fitted parameters for each limit are given in Table \ref{['tab:limit_mag_parameters']}