The DESI Survey Validation: Results from Visual Inspection of Bright Galaxies, Luminous Red Galaxies, and Emission Line Galaxies

TL;DR

The study addresses validating DESI's target selections and redshift measurements during Survey Validation by constructing large VI catalogs for BGS, LRGs, and ELGs. It leverages visual redshift identifications to benchmark and tune the Redrock pipeline, particularly using Delta chi-squared criteria, to achieve high purity and substantial redshift recovery across tracers. The results show purity exceeding 99% for Main Survey samples, redshift precision of about 10 km/s for BGS and ELGs and ~40 km/s for LRGs, and cross-survey consistency in redshift systematics < 10 km/s. VI proves essential for identifying pipeline artifacts, guiding improvements, and uncovering rare objects like Lyα emitters and strong-lensing candidates, demonstrating its value for current and future spectroscopic surveys.

Abstract

The Dark Energy Spectroscopic Instrument (DESI) Survey has obtained a set of spectroscopic measurements of galaxies to validate the final survey design and target selections. To assist in these tasks, we visually inspect (VI) DESI spectra of approximately 2,500 bright galaxies, 3,500 luminous red galaxies (LRGs), and 10,000 emission line galaxies (ELGs), to obtain robust redshift identifications. We then utilize the VI redshift information to characterize the performance of the DESI operation. Based on the VI catalogs, our results show that the final survey design yields samples of bright galaxies, LRGs, and ELGs with purity greater than $99\%$. Moreover, we demonstrate that the precision of the redshift measurements is approximately 10 km/s for bright galaxies and ELGs and approximately 40 km/s for LRGs. The average redshift accuracy is within 10 km/s for the three types of galaxies. The VI process also helps improve the quality of the DESI data by identifying spurious spectral features introduced by the pipeline. Finally, we show examples of unexpected real astronomical objects, such as Ly$α$ emitters and strong lensing candidates, identified by VI. These results demonstrate the importance and utility of visually inspecting data from incoming and upcoming surveys, especially during their early operation phases.

Figures (14)

• Figure 1: Example of BGS spectra, ordered by their VI quality values from top with VI quality 4 (best quality) to bottom with VI quality 0 (poorest quality). The quality 4 spectrum has multiple spectral features including both emission and absorption lines. The quality 3 spectrum shows strong Ca II absorption lines with weak NaD and Mgb absorption lines. Spectra with quality 2 and below do not have spectral features that can be robustly identifiable. The right panels show the DESI Legacy Survey images of the BGS targets. The spectra in grey, black and orange colors are the original observed galaxy spectrum, the smooth spectrum with a Gaussian filter, and the error spectrum, respectively. The side length of the image is 18" and the white circle in the top panel image reflects the DESI fiber size.
• Figure 2: Example of LRG spectra, ordered by their VI quality values from top with VI quality 4 to bottom with VI quality 0. The quality 4 spectrum has multiple spectral lines, including CaII, G band, Mgb absorption lines and [O$\;$]$\lambda\lambda 3726,3729$ emission lines. The quality 3 spectrum shows Ca II and G band absorption lines. Spectra with quality 2 and below do not have spectral features that can be robustly identifiable. Sky residuals can be observed at the wavelength regions $>8000\, \rm \AA$. The right panels show the DESI Legacy Survey images of the LRG targets. The spectra in grey, black and orange colors are the original observed galaxy spectrum, the smooth spectrum with a Gaussian filter, and the error spectrum, respectively.
• Figure 3: Example of ELG spectra, ordered by their VI quality values from top with VI quality 4 to bottom with VI quality 0. The quality 4 spectrum shows a resolved [O$\;$]$\lambda\lambda 3726,3729$ doublet and interstellar absorption lines, including MgII and FeII. The quality 3 spectrum shows blended [O$\;$]$\lambda\lambda 3726,3729$ emission lines. The insets show the detected [O$\;$]$\lambda\lambda 3726,3729$ emission line profiles. The right two panels show the DESI Legacy Survey images and the Hyper Supreme-Cam images HSCDR2 of the ELG targets. The spectra in grey, black and orange colors are the original observed galaxy spectrum, the smooth spectrum with a Gaussian filter, and the error spectrum, respectively.
• Figure 4: Left: Distributions of VI quality values of BGS, LRGs, and ELGs from top to bottom respectively. Right: VI redshift distributions of BGS, LRGs, and ELGs with robust VI redshifts from top to bottom respectively. The faint color in each panel shows the distribution of sources selected in Survey Validation observations and the dark color shows the subset of those that are in the DESI Main Survey target selection.
• Figure 5: ELG SV and Main Survey g-r and r-z selections. Top: The fraction of ELGs with VI quality$\geq 2.5$. The color indicates the fraction and the size of the circles reflects the number of sources in the color-color bins. The dashed lines show the the LOP (blue) and VLO (green) color-color selection criteria used in the Main Survey. Bottom: The 75th-percentile redshifts of ELGs with VI quality$\geq 2.5$ in each color-color bin. We note that for ELGs with $r-z<0$, most of their spectra ($>50\%$) do not have features that can be identified visually as shown in the top panel. The remaining ELGs with VI quality$\geq 2.5$ tend to be at low redshifts as shown in the bottom panel.