A Cosmic Watershed: the WVF Void Detection Technique

TL;DR

This paper introduces the Watershed Void Finder (WVF), an objective and minimally assumption-dependent technique for identifying and analyzing cosmic voids within the large-scale structure of the Universe, effectively capturing their hierarchy and morphology.

Contribution

The WVF method applies the Watershed Transform to cosmic data, enabling accurate, parameter-light detection of voids and their hierarchical relationships, validated on Voronoi models.

Findings

Successfully identifies voids and their boundaries in simulated data.

Reproduces the full void size distribution function.

Accurately retrieves void shapes and sizes from noisy data.

Abstract

On megaparsec scales the Universe is permeated by an intricate filigree of clusters, filaments, sheets and voids, the Cosmic Web. For the understanding of its dynamical and hierarchical history it is crucial to identify objectively its complex morphological components. One of the most characteristic aspects is that of the dominant underdense Voids, the product of a hierarchical process driven by the collapse of minor voids in addition to the merging of large ones. In this study we present an objective void finder technique which involves a minimum of assumptions about the scale, structure and shape of voids. Our void finding method, the Watershed Void Finder (WVF), is based upon the Watershed Transform, a well-known technique for the segmentation of images. Importantly, the technique has the potential to trace the existing manifestations of a void hierarchy. The basic watershed transform is augmented by a variety of correction procedures to remove spurious structure resulting from sampling noise. This study contains a detailed description of the WVF. We demonstrate how it is able to trace and identify, relatively parameter free, voids and their surrounding (filamentary and planar) boundaries. We test the technique on a set of Kinematic Voronoi models, heuristic spatial models for a cellular distribution of matter. Comparison of the WVF segmentations of low noise and high noise Voronoi models with the quantitatively known spatial characteristics of the intrinsic Voronoi tessellation shows that the size and shape of the voids are succesfully retrieved. WVF manages to even reproduce the full void size distribution function.