BOSS & DESI Spectroscopic Surveys

• BOSS and DESI are large-scale spectroscopic surveys delivering millions of galaxy, quasar, and stellar spectra, optimized for BAO and cosmic structure studies.
• They employ state-of-the-art instrumentation and rigorous pipelines that ensure high signal-to-noise, precise redshift measurements, and robust spectral calibrations.
• The surveys provide extensive value-added products and mock catalogs that support statistical analyses, cross-correlation studies, and tests of cosmological models.

The Baryon Oscillation Spectroscopic Survey (BOSS) and the Dark Energy Spectroscopic Instrument (DESI) represent successive generations of massively multiplexed, wide-area galaxy and quasar redshift surveys optimized for large-scale structure, baryon acoustic oscillation (BAO), cosmology, and stellar population studies. Each offers public spectroscopic datasets comprising millions to tens of millions of spectra, accompanied by rigorously validated reduction pipelines, instrument models, and value-added products suitable for both statistical cosmology and detailed astrophysical analysis.

1. Survey Design, Samples, and Scope

BOSS (SDSS-III; 2009–2014) utilized the upgraded SDSS 2.5-m telescope, delivering $\sim$1.5 million galaxy spectra, $\sim$150,000 Lyα-forest quasars, and over 100,000 stellar spectra over $\sim$10,000 deg$^2$ at red optical wavelengths, with a sky-limited exposure scheme targeting $z<0.7$ galaxies for BAO and high-redshift ($z>2.15$) quasars for Lyα forest-based cosmology (Dawson et al., 2012).

DESI (KPNO 4-m Mayall, operational 2020 onward) extends this legacy with a 5000-fiber robotically actuated focal plane over a 7–8 deg$^2$ field, targeting $\sim$14,000 deg$^2$ for $\sim$40 million galaxy/quasar redshifts and $\sim$10 million stellar spectra, probing redshifts $0$R\sim2000$–5000 (Guy et al., 2022, Fulmer et al., 2023). DESI classes include Luminous Red Galaxies (LRG) to $z\sim1.0$, Emission Line Galaxies (ELG) to $z\sim1.6$, and quasars extending the survey’s redshift reach.

Survey	Telescope, Fibers	Sky Area	Main Tracers	$z$-Range	$N_\mathrm{spec}$ (goal)
BOSS	SDSS 2.5m, 1000	$\sim$10,000 deg$^2$	LRG, QSO, stars	0.1–0.7 (gal), 2.1–3.5 (QSO)	$\sim$1.5M (gal), 150k (QSO)
DESI	Mayall 4m, 5000	14,000 deg$^2$	LRG, ELG, QSO, stars	0–1.7 (gal), 0–3.5 (QSO)	$>40$M (gal, QSO), 10M (stars)

DR9 of BOSS provides 535,995 new galaxy spectra (plus previous SDSS samples), 102,100 new quasar spectra, 90,897 new stellar spectra (Collaboration et al., 2012). DESI’s DR1/2 are projected to exceed 10 million unique spectra, with 98.9% redshift success for LRGs (Zhou et al., 2022).

2. Instrumentation, Resolution, and Data Products

BOSS features twin spectrographs with 1000 fibers per exposure, 2″ fibers, wavelength coverage 3600–10,400 Å, resolution $R=1500$–2500, and end-to-end throughput up to 31% in both channels due to volume-phase holographic gratings and advanced CCDs (Smee et al., 2012). Signal-to-noise per Å is $S/N\approx 3$–4 at $i=19.9$ in 1 hr exposures.

DESI advances on all hardware fronts: 5000 actuated 1.5″ fibers per tile, 10 spectrographs (3 cameras each), simultaneous 3600–9800 Å coverage, $R=2000$–5000 (blue to NIR), median S/N and redshift performance on all primary target classes exceeding 98% (Guy et al., 2022). Key performance: 0.03 Å wavelength accuracy, star-based flux calibration to 2% (3700–9800 Å), fiber-to-fiber throughput reproducibility to 0.2–2%, sky subtraction matching variance (continuum), and broadband photometric repeatability to 6% in $r$.

BOSS FITS products include per-exposure “spPlate” files (flux, inverse variance, resolution matrix, bitmask arrays), per-object “spSpec” files, and value-added catalogs for LSS, stellar parameters, and emission lines (Fulmer et al., 2023). DESI adopts a more modular file hierarchy, centered on the “spectro redux” model, with per-tile, per-HEALPix coadd files storing flux, variance, mask, pixel-wise resolution matrices, and merged target, fiber, and redshift tables (Guy et al., 2022).

3. Reduction Pipelines, Calibration, and Quality

The BOSS “idlspec2d” pipeline performs flat-fielding, bias correction, optimal extraction, arc-lamp wavelength calibration, sky subtraction using 80–150 sky fibers/plate, and F-star-based flux calibration, with repeat-visit velocity accuracy $\sim$38 km/s and typical systematic redshift shifts $<10\%$ (Collaboration et al., 2012, Dawson et al., 2012). Redshift measurements employ cross-correlation with empirical galaxy, QSO, and stellar templates.

DESI’s reduction pipeline, described in detail by Guy et al. (Guy et al., 2022), generalizes the “spectro-perfectionism” paradigm with 2D PSF Gauss–Hermite modeling per fiber and pixel, decorrelated flux vector extraction, and full error propagation including resolution matrices. Automated redshift determination utilizes principal-component eigenspectra, machine-learning-based “QuasarNET” for QSO identification, and per-exposure DeltaChi2 quality metrics (e.g., ZWARN flags). Quality assurance is monitored nightly with full end-to-end, open-source pipelines and GUI visualization tools (Fulmer et al., 2023).

Both pipelines implement sophisticated cosmic ray, bad pixel, and artifact masking, along with per-pixel sky-residual quality flags and continuum-correction mechanisms, essential for Lyα-forest and emission-line analyses (Lee et al., 2012). DESI’s adoption of fully empirical PSF models and pixel-level forward modeling has improved calibration accuracy ($<0.03$ Å wavelength rms, relative flux stability $<0.6\%$/night) and sky subtraction (continuum-level, line residuals 1–3%).

4. Value-Added Products and Mock Catalogs

BOSS released specialized value-added data products for statistical Lyα forest analysis, including the “speclya” DR9 catalog: 54,468 quasar spectra with $z_\text{VI} > 2.15$, supporting transmission studies over $2.0 < z_\alpha < 5.7$ and covering $\sim$20 $h^{-3}$ Gpc$^3$ (Lee et al., 2012). Per-spectrum products include masks for sky lines and DLAs, empirical corrections for pixel-by-pixel pipeline noise bias, DLA damping wing subtraction, MF-PCA continuum estimates, and full masks/flags for problematic data regions. Data are provided in FITS format with documentation and machine-readable catalogs.

BOSS also distributed DR11/DR12 mock Lyα forest datasets—100 realizations of $\sim$150,000 quasar sightlines each—incorporating cosmological, astrophysical (HCD, metals), and instrumental systematics, with per-skewer FITS storage and public code for conversion to pipeline-compatible formats (Bautista et al., 2014). These mocks have enabled covariance estimation, systematics validation, and end-to-end BAO pipeline testing.

DESI analogously provides mock catalogs and expanded cross-survey simulation suites. File formats and metadata fields reflect the increased multiplexing and data volume, supporting efficient ingestion into both legacy BOSS and next-generation analysis pipelines (Fulmer et al., 2023).

5. Major Scientific Applications and Cosmological Impact

BOSS and DESI datasets underpin a wide range of cosmological and astrophysical investigations:

• BAO and Cosmological Distance Ladder: BOSS’s primary deliverable is percent-level determination of the BAO scale via galaxy clustering and Lyα-forest cross-correlations, anchoring $D_A(z)$, $H(z)$, and $D_V(z)$ at $z\sim0.3, 0.57, 2.34$ and forming a “distance ladder” when combined with higher-redshift DESI BAO points at $z>1$ (Feng et al., 27 Oct 2025). DESI’s n($z$), sky coverage, and superior $R, S/N$ allow robust BAO/radial modes out to $z\sim2.3$ with fractional errors $<1\%$ per bin.
• Structure Growth and Modified Gravity: Full-shape analyses of the BOSS and DESI galaxy power spectrum (EFT-of-LSS modeling) constrain the matter clustering amplitude and test extensions of $\Lambda$CDM (EFT of dark energy, modified gravity, massive neutrinos), with current results limited by degeneracies in the large-scale shape and broadband systematics (Taule et al., 2024, Noriega et al., 2024).
• Intergalactic and Circumgalactic Medium: BOSS Lyα forest and DLA products provide a fiducial data set for mapping the large-scale distribution of neutral hydrogen, with MF-PCA continuum-corrected transmitted flux fields suitable for 3D clustering, void analysis, and astrophysical absorber studies (Lee et al., 2012).
• Stellar Populations and Galactic Archaeology: Both surveys deliver high-fidelity Milky Way stellar spectra for empirical template library construction, atmospheric and chemical abundance analysis, and calibration of synthetic stellar population models (Kesseli et al., 2017). The precision and uniformity enable studies of velocity dispersions, mass functions, and chemical tagging across the Galaxy.
• Cross-correlation Science: The high density and area of the DESI sample enable cross-correlation with imaging surveys (e.g., Euclid, LSST), lensing analyses (E$_G$ tests), and photometric redshift calibration at the $\sigma(\langle z\rangle)\lesssim0.002(1+z)$ level (Naidoo et al., 2022, Rauhut et al., 21 Jul 2025).

6. Data Access, Query, and Visualization Infrastructure

Both BOSS and DESI provide open-access data distribution through Science Archive Servers and web interfaces. For BOSS, this includes CasJobs SQL, SkyServer browser, and direct FITSfile download tools. DESI expands this, enabling ADQL/TAP service, RESTful API endpoints, bulk data staging into VOSpace, and full-programmatic integration via Python packages (“dltools”, “astroquery.adl”) (Fulmer et al., 2023). Interactive multiwavelength spectral viewers with caching, batch navigation ($\leq 0.75$ s per update), and on-the-fly template overlay are now standard, enabling rapid inspection of thousands of spectra with real-time pipelines for masking, smoothing, and line identification.

7. Comparative Advances and Future Prospects

DESI builds directly on the BOSS legacy but exceeds it in survey scale, sampling density (LRGs: 605 deg$^{-2}$ vs. BOSS CMASS 110 deg$^{-2}$), redshift reach ($z\sim1.6$ vs. $z\sim0.7$), instrumental resolution, and systematic mitigation (Zhou et al., 2022). The increase in fiber multiplexing, improved throughput, and enhanced data model facilitate scientific programs previously unreachable due to sample variance limitations (e.g., cross-correlation lensing, high-$z$ BAO, evolution of galaxy quenching) (Ditrani et al., 14 Jan 2026). DESI’s redshift success (98.9% for LRGs) and purity ($<0.5\%$ stellar contamination) set a new benchmark for wide-area surveys.

The increasing synergy between high-density spectroscopic grids (DESI), deep imaging (LSST, Euclid), and advanced simulation/mocking frameworks is enabling high-fidelity end-to-end cosmological constraints, robust systematics calibration, and an expanding scientific reach to high precision tests of gravity and the neutrino sector (Feng et al., 27 Oct 2025). Both BOSS and DESI spectroscopic datasets thus form the backbone of current and future large-scale structure cosmology and galaxy evolution studies.