Cosmic Covers: A Geometry-First Pipeline for 3D Comoving Astronomical Catalogs

Janos Horvath, Istvan Horvath

Keywords: astronomical catalogs, 3D point clouds, comoving coordinates, geometric covers, K-center, K-means, density estimation, noise-contrastive estimation, selection-preserving null models, time-domain astronomy

TL;DR: We unify multiple astronomical catalogs into a reproducible 3D comoving point-cloud and summarize their spatial structure using density baselines and geometric covers that provide interpretable representative events with error radii.

Abstract: We present a fully reproducible pipeline for turning heterogeneous public astronomical catalogs into a unified 3D comoving point-cloud representation and for summarizing these distributions with geometric covers. Starting from the ZTF Bright Transient Survey (BTS) supernova sample, the Swift/BAT GRB catalog (with redshifts), and the redMaPPer cluster catalogs, we (i) harmonize metadata and categories across surveys, (ii) convert sky coordinates and redshift to Cartesian comoving coordinates using a standard cosmology, (iii) fit density baselines (Gaussian kernel density estimation and noise-contrastive density ratio models), and (iv) build compact cover sets (K-means and K-center) that approximate the spatial support and yield interpretable “representative” events with error radii. The same codebase generates publication-ready summary tables and a large suite of plots, including sky footprints, distance distributions, and cover allocation visualizations (static and interactive). On the processed releases used in this workshop paper, the pipeline produces N = 11,514 SNe, N = 522 GRBs, and N = 26,898 clusters; KDE baselines and density-ratio AUCs provide a reference point for future method development.

Submission Number: 4