DESI
Dark Energy Spectroscopic Instrument (DESI)
Description
The Dark Energy Spectroscopic Instrument (DESI) was built by the Department of Energy (DOE) to conduct a Stage-IV dark energy experiment (Instrument Overview Paper). Over the course of the survey (2021-2026), the DESI collaboration will obtain spectra of about 40 million galaxies and quasars as well as 10 million stars (DESI collaboration 2016a,b).
For more complete information about the DESI data, please visit https://data.desi.lbl.gov/doc/. General information about the instrument, experiment and science results can be found at https://www.desi.lbl.gov/.
The NOIRLab Astro Data Lab will serve data from each DESI public data release, starting with the Early Data Release (EDR). The main spectroscopic and photometric catalogs are hosted at Data Lab and available for queries. The full-depth spectra ("HEALPix coadds") are searchable and retrievable from SPARCL as described in the Data Access section. These value-added services are provided for convenience to the astronomy community, with a reminder that any use of DESI data must be accompanied by the official DESI acknowledgments. Other types of spectra (per exposure, per night, tile coadds) are only available as files via NERSC.
Both the desi_edr.zpix and desi_dr1.zpix tables have been crossmatched against our default reference datasets within a 1.5 arcsec radius, nearest neighbor only. These tables will appear with x1p5 in their name in our table browser. Example: desi_dr1.x1p5__zpix__gaia_dr3__gaia_source.
Scientific Goals
The DESI collaboration seeks to understand the nature of dark energy and its evolution. In addition, the DESI data enable a suite of other scientific discoveries and research thanks to the legacy value of the immense spectroscopic dataset. The DESI collaboration maintains an ADS library of DESI papers.
DESI Spectra
The DESI collaboration produces one-dimensional (1-D) spectra that have been sky-subtracted, flux-calibrated, and modeled with a best-fitting template to identify the spectral type (spectype) and measure the redshift (z) with the Redrock pipeline (Bailey et al., in preparation). There are a few categories of spectra at different stages of data combination. Namely:
- Per exposure (spectra from a single exposure at the telescope)
- Per night (coadded spectra obtained on multiple exposures during the same night)
- Per tile (coadded spectra obtained for the same tile but possibly from multiple nights)
- Per healpix (coadded spectra per position on the sky grouped by HEALPix (with n=64); which can be from multiple nights and/or multiple tiles)
Furthermore, the spectra are treated separately per SURVEY:
- cmx = commissioning
- sv1 = first phase of Survey Validation (SV)
- sv2 = operations testing ("dress rehearsal" for sv3)
- sv3 = one-percent survey (covering 1% of area; final phase of SV)
- special = dedicated secondary targets or test targets
- main = main DESI survey (starting with DR1)
And per PROGRAM:
- dark = dark time for the main targets: Luminous Red Galaxies (LRG), Emission-Line Galaxies (ELG), and Quasi-Stellar Objects / Quasars (QSO)
- bright = bright time for the Bright Galaxy Survey (BGS) and the Milky Way Survey (MWS)
- backup = when observing conditions were worse than the bright conditions
- other = none of the above
The SPARCL database only includes DESI spectra that have been coadded per healpix. All other spectra are available in a file format directly through the public release at NERSC. See the DESI Data Access page for details.
The following figure illustrates a DESI tile (divided into 10 petals) and a DESI rosette (divided into HEALPix for illustration purposes). The left-hand panel shows the sky positions of the fibers from a single tile (pointing) color-coded by PETAL_LOC values as labeled. The right-hand panel shows the fiber sky positions from Rosette 1, which was obtained by observing the same area of the sky with multiple dithered tiles, color-coded per HEALPIX value (figure credit: S. Juneau / DESI collaboration / NSF NOIRLab Astro Data Lab).
Data Releases
First Data Release (DESI DR1)
The First Data Release of DESI contains spectra and catalogs for more than 18 million unique targets from the first 13 months of the main survey ("Year 1") as well as a reprocessing of the same commissioning and survey validation (SV) data that were previously released as part of EDR. Information about the data release is presented in the DR1 Overview paper (DESI Collaboration et al. 2025).
The DR1 data are released under the Creative Commons Attribution 4.0 International License (CC BY 4.0). Use of DESI data requires including the citation and acknowledgment text given on the Data License and Acknowledgments page.
| DESI DR1 Summary | |
|---|---|
| Main Survey | |
| Number of useful⁽¹⁾ spectra | 18,659,804 |
Galaxies (SPECTYPE==GALAXY) | 13,049,402 |
Quasars (SPECTYPE==QSO) | 1,553,713 |
Stars (SPECTYPE==STAR) | 4,056,689 |
| Special Observations | |
| Number of useful⁽¹⁾ spectra | 141,473 |
Galaxies (SPECTYPE==GALAXY) | 83,961 |
Quasars (SPECTYPE==QSO) | 3,624 |
Stars (SPECTYPE==STAR) | 53,888 |
| Commissioning & Survey Validation | |
| Number of useful⁽¹⁾ spectra | 1,613,846 |
Galaxies (SPECTYPE==GALAXY) | 1,046,508 |
Quasars (SPECTYPE==QSO) | 88,505 |
Stars (SPECTYPE==STAR) | 478,833 |
| DESI Instrument | |
| Spectral coverage⁽²⁾ | 360--982.4 nm |
| Spectral resolution | 2000 (at 360 nm) -- 5500 (at 980 nm) |
| Wavelength system | vacuum barycentric |
| Photometric bands (Legacy Surveys DR9) | g, r, z, W1, W2, W3, W4 |
| Approximate area | |
| Main Survey / Backup | 2,726 sq. deg. |
| Main Survey / Bright | 9,739 sq. deg. |
| Main Survey / Dark | 9,528 sq. deg. |
Footnotes: (1) "Useful spectra" are defined as having ZCAT_PRIMARY==True, OBJTYPE=='TGT', and ZWARN==0, which selects all unique, non-sky targets with no known redshift-fitting failures. The Main Survey additionally excludes PROGRAM=='other'; (2) Original spectra are split on three spectrograph arms: blue (B), red (R), infrared (Z). The version available in the SPARCL database has been coadded across cameras into a single array spanning the full BRZ spectral range.
The catalog tables listed below include a redshift catalog per healpix (zpix; recommended for most analyses). We consider the zpix catalog as the primary spectroscopic catalog (analogous to specObjAll for SDSS) but some of the information such as photometry and observing conditions need to be obtained by joining with other catalog tables. This Jupyter notebook features several example queries that show how to join tables.
| DESI DR1 Tables | |
|---|---|
| Table Name | Description |
| exposure | Summary quantities for every DESI exposure |
| fiberassign | Quantities obtained when a DESI target is assigned to a fiber |
| frame | Summary quantities for each petal of the DESI instrument in a given exposure; in normal operation there are ten frames for every exposure |
| photometry | Photometric quantities from LS DR9 for every TARGETID |
| potential | For a given tile, this table lists all targets that could have received a fiber assignment |
| target | The quantities obtained when photometric objects are analyzed in the target selection process |
| tile | Summary quantities for every DESI tile (pointing on the sky) |
| zpix | Redshift and other quantities for spectra that have been grouped and coadded by HEALPixel |
| ztile | Redshift and other quantities for spectra that have been grouped and coadded by tile |
| DESI DR1 Value-Added Catalogs (VACs) | |
|---|---|
| Table Name | Description |
| agnqso | Value-Added Catalog of DESI DR1 galaxies and quasars with spectral and infrared classification diagnostics |
Known issues with Data Release 1 are documented here.
Early Data Release (DESI EDR)
The Early Data Release of DESI contains spectra and catalogs from commissioning and Survey Validation (SV) campaigns (Survey Validation paper; EDR Overview paper).
| DESI EDR Summary | |
|---|---|
| Number of healpix coadded spectra⁽¹⁾ | 2,044,588 |
| Number of useful⁽²⁾ spectra | 1,769,157 |
| Number of useful spectra of unique targets | 1,712,004 |
Galaxies (SPECTYPE==GALAXY) | 1,125,635 |
Quasars (SPECTYPE==QSO) | 90,241 |
Stars (SPECTYPE==STAR) | 496,128 |
| Spectral coverage⁽³⁾ | 360-982.4 nm |
| Spectral resolution R | 2000 (at 360 nm) - 5500 (at 980 nm) |
| Wavelength system | Vacuum |
| Photometric bands (Legacy Surveys DR9) | g, r, z, W1, W2, W3, W4 |
| DESI EDR approximate area | ∼1,390 square degrees |
Footnotes: (1) Number of spectra for fibers placed on targets (OBJTYPE="TGT"), which excludes sky fibers and known faulty fibers; (2) Useful spectra defined as having ZWARN=0, which means no errors due to instrument or data and no flags from spectral fitting issues; (3) Original spectra are split on three spectrograph arms: blue (B), red (R), infrared (Z). The version available in the SPARCL database has been coadded across cameras into a single array spanning the full BRZ spectral range.
The catalog tables listed below include a redshift catalog per healpix (zpix; recommended for most analyses) and a redshift catalog per tile (ztile; can be useful for instrument-dependent analyses). We consider the zpix catalog as the primary spectroscopic catalog (analogous to specObjAll for SDSS) but some of the information such as photometry and observing conditions need to be obtained by joining with other catalog tables.
| DESI EDR Tables | |
|---|---|
| Table Name | Description |
| exposure | Summary quantities for every DESI exposure |
| fiberassign | Quantities obtained when a DESI target is assigned to a fiber |
| frame | Summary quantities for each petal of the DESI instrument in a given exposure; in normal operation there are ten frames for every exposure |
| photometry | Photometric quantities from LS DR9 for every TARGETID |
| potential | For a given tile, this table lists all targets that could have received a fiber assignment |
| target | The quantities obtained when photometric objects are analyzed in the target selection process |
| tile | Summary quantities for every DESI tile (pointing on the sky) |
| zpix | Redshift and other quantities for spectra that have been grouped and coadded by HEALPixel |
| ztile | Redshift and other quantities for spectra that have been grouped and coadded by tile |
Data Access
The DESI data are accessible by a variety of means:
SPARCL
SPectra Analysis & Retrievable Catalog Lab (SPARCL) at NOIRLab provides flexible access to spectra from large optical and near-infrared surveys. SPARCL currently contains spectra from DESI and SDSS. The SPARCL client package is pre-installed in Data Lab and can also be installed in a user's local computing environment. See the How to Use SPARCL Jupyter Notebook.
Data Lab Table Access Protocol (TAP) service
TAP provides a convenient access layer to the DESI catalog database. TAP-aware clients (such as TOPCAT) can point to https://datalab.noirlab.edu/tap, select the desi_dr1 database, and see the database tables and descriptions. You can also view the DESI tables and descriptions in the Data Lab table browser.
Data Lab Query Client
The Query Client is available as part of the Data Lab software distribution. The Query Client provides a Python API to Data Lab database services. These services include anonymous and authenticated access through synchronous or asynchronous queries of the catalog made directly to the database. Additional Data Lab services for registered users include personal database storage and storage through the Data Lab VOSpace.
The Query Client can be called from a Jupyter Notebook on the Data Lab Notebook server. Example notebooks are provided to users upon creation of their user account (register here), and are also available to browse on GitHub at https://github.com/astro-datalab/notebooks-latest.
Jupyter Notebook Server
The Data Lab Jupyter Notebook server (authenticated service) contains multiple examples of how to access and visualize the DESI catalogs and spectra:
- How to Use SPARCL
- Introduction to DESI First Data Release (DR1) at the Astro Data Lab
- Comparing SDSS and DESI spectra using SPARCL
- Stacking SDSS Spectra of Galaxies Selected from the BPT Diagram
- Multi-wavelength Image Cutouts and SDSS Spectra of Active Galaxies with Extreme Emission-Line Ratios
- Obtain spectra with SPARCL and plot them with Jdaviz
- Obtain spectra with SPARCL and plot them with Prospect
Bulk File Access
To download catalog files and/or spectra files from the public data releases NERSC or for bulk transfer using Globus see the Data Access page from the DESI Data Documentation website.
Acknowledgments
Acknowledgments must be used following the complete instructions from the DESI Data Documentation website Acknowledgments page.
If using the Astro Data Lab and/or SPARCL, please also include their respective acknowledgments as described below:
Acknowledgment of SPARCL used jointly with the Astro Data Lab
If you use SPARCL jointly with the Astro Data Lab platform (via JupyterLab, command-line or web interface) in your published research, include the text below in your paper:
This research uses services or data provided by the SPectra Analysis and Retrievable Catalog Lab (SPARCL) and the Astro Data Lab, which are both part of the Community Science and Data Center (CSDC) program at NSF National Optical-Infrared Astronomy Research Laboratory. NOIRLab is operated by the Association of Universities for Research in Astronomy (AURA), Inc. under a cooperative agreement with the National Science Foundation.
If publishing in a AAS journal, also add the keywords: \facility{Astro Data Lab} and \software{SPARCL (Juneau et al. 2024)} and cite:
- Juneau et al., "SPARCL: SPectra Analysis and Retrievable Catalog Lab", Conference Proceedings for ADASS XXXIII, 2024 https://doi.org/10.48550/arXiv.2401.05576
Acknowledgment of SPARCL
If you use SPARCL in your published research, please include the text below in your paper:
This research uses services or data provided by the SPectra Analysis and Retrievable Catalog Lab (SPARCL), which is part of the Community Science and Data Center (CSDC) program at NSF National Optical-Infrared Astronomy Research Laboratory. NOIRLab is operated by the Association of Universities for Research in Astronomy (AURA), Inc. under a cooperative agreement with the National Science Foundation.
If publishing in a AAS journal, also add the keyword: \software{SPARCL (Juneau et al. 2024)} and cite:
- Juneau et al., "SPARCL: SPectra Analysis and Retrievable Catalog Lab", Conference Proceedings for ADASS XXXIII, 2024 https://doi.org/10.48550/arXiv.2401.05576
