Correcting for Fibre Assignment Incompleteness in the DESI Bright Galaxy Survey

TL;DR

This paper addresses fibre assignment incompleteness in the DESI Bright Galaxy Survey by applying inverse probability weighting and dithering techniques to correct clustering measurements, ensuring unbiased results even with limited passes.

Contribution

It introduces a novel correction method combining pair inverse probability weighting and angular upweighting for fibre assignment incompleteness in galaxy surveys.

Findings

Completeness depends on galaxy surface density.

Inverse pair weighting corrects clustering biases.

Method effective even after a single pass.

Abstract

The Dark Energy Spectroscopic Instrument (DESI) Bright Galaxy Survey (BGS) will be a survey of bright, low redshift galaxies, which is planned to cover an area of ~14,000 sq deg in 3 passes. Each pass will cover the survey area with ~2000 pointings, each of area ~8 sq deg. The BGS is currently proposed to consist of a bright high priority sample to an r-band magnitude limit r ~ 19.5, with a fainter low priority sample to r ~ 20. The geometry of the DESI fibre positioners in the focal plane of the telescope affects the completeness of the survey, and has a non-trivial impact on clustering measurements. Using a BGS mock catalogue, we show that completeness due to fibre assignment primarily depends on the surface density of galaxies. Completeness is high (>95%) in low density regions, but very low (<10%) in the centre of massive clusters. We apply the pair inverse probability (PIP) weighting correction to clustering measurements from a BGS mock which has been through the fibre assignment algorithm. This method is only unbiased if it is possible to observe every galaxy pair. To facilitate this, we randomly promote a small fraction of the fainter sample to be high priority, and dither the set of tile positions by a small angle. We show that inverse pair weighting combined with angular upweighting provides an unbiased correction to galaxy clustering measurements for the complete 3 pass survey, and also after 1 pass, which is highly incomplete.