 Top level directory for web access:
 Top level directory for web access to dr3.1 sweeps (see known issues for a description of dr3.1):

https://portal.nersc.gov/cfs/cosmo/data/legacysurvey/dr3/sweep/3.1/
 Top level directory local to NERSC computers (for collaborators):

/global/cfs/cdirs/cosmo/data/legacysurvey/dr3/
 Top level directory local to NERSC computers for dr3.1 sweeps (see known issues for a description of dr3.1):

/global/cfs/cdirs/cosmo/data/legacysurvey/dr3/sweep/3.1/
Summary Files
surveybricks.fits.gz
FITS binary table with the RA, DEC bounds of each geometrical "brick" on the sky. This includes all bricks on the sky, not just the ones in our footprint or with coverage in DR3. For that information, see the next file description.
HDU1 (only HDU)  tags in the
surveybricks.fits.gz
file
Column 
Type 
Description 


char[8] 
Name of the brick. 

int32 
A unique integer with 1to1 mapping to 

int16 
A "priority" factor used for processing. 

int32 
Dec row number. 

int32 
Number of the brick within a Dec row. 

double 
RA of the center of the brick. 

double 
Dec of the center of the brick. 

double 
Lower RA boundary. 

double 
Upper RA boundary. 

double 
Lower Dec boundary. 

double 
Upper Dec boundary. 
surveybricksdr3.fits.gz
A FITS binary table with information that summarizes the contents of each brick, as of DR3.
Column 
Type 
Description 


char[8] 
Name of the brick. 

double 
RA of the center of the brick. 

double 
Dec of the center of the brick. 

int16 
Median number of exposures in the unique area (i.e. 

int16 
Median number of exposures in the unique area of the brick in \(r\)band. 

int16 
Median number of exposures in the unique area of the brick in \(z\)band. 

int32[6] 
Histogram of number of pixels in the unique brick area with 0, 1, 2, 3, 4, or > 5 exposures in \(g\) 

int32[6] 
Histogram of number of pixels in the unique brick area with 0, 1, 2, 3, 4, or > 5 exposures in \(r\) 

int32[6] 
Histogram of number of pixels in the unique brick area with 0, 1, 2, 3, 4, or > 5 exposures in \(z\) 

int16 
Total number of 

int16 
Total number of 

int16 
Total number of 

int16 
Total number of 

int16 
Total number of 

int16 
Total number of 

float32 
Median PSF size, in arcsec, evaluated at the 

float32 
Median PSF size, in arcsec, evaluated at the 

float32 
Median PSF size, in arcsec, evaluated at the 

float32 
Median SFD dust map E(BV) extinction, in magnitudes, evaluated at 

float32 
Median Milky Way dust transparency in \(g\)band, based on 

float32 
Median Milky Way dust transparency in \(r\)band, based on 

float32 
Median Milky Way dust transparency in \(z\)band, based on 
Note that, for the nexphist
rows, pixels that are masked by the NOIRLab Community Pipeline as, e.g., cosmic rays or saturation, do NOT count toward the number of exposures. More information about the morphological types and DECAM_MW_TRANSMISSION
can be found on the catalogs page.
surveyccdsdecals.fits.gz
A FITS binary table with almanac information (e.g. seeing, etc.) about each individual CCD image.
This file contains information regarding the photometric and astrometric zero points for each CCD of every DECam image that is part of the DECaLS DR3 data release. Photometric zero points for each CCD are computed by identifying stars and comparing their instrumental magnitudes (measured in an approximately 7 arcsec diameter aperture) to colorselected stars in the PanSTARRS "qz" catalog.
The photometric zeropoints (zpt
, ccdzpt
, etc)
are magnitudelike numbers (e.g. 25.04), and
indicate the magnitude of a source that would contribute one count per
second to the image. For example, in an image with zeropoint of 25.04
and exposure time of 30 seconds, a source of magnitude 22.5 would
contribute
\(30 * 10^{((25.04  22.5) / 2.5)} = 311.3\)
counts.
Column 
Type 
Description 


char[35] 
Name listed in the object tag from the CCD header. For propid 2014B0404, in most cases, this has the form: DECaLS_<tile_number>_<filter>, but this can differ for older and nonDECaLS data. 

int32 
Unique DECam exposure number, eg 348224. 

float 
Exposure time in seconds, eg 30. 

char[1] 
Filter used for observation, eg "g", "r", "z". 

float 
Seeing in arcseconds determined by fitting a 2dimensional gaussian to the median PSF of stars on the CCD, eg 1.1019. 

char[10] 
Date of observation start, eg "20140815". Can be combined with 

double 
Date of observation in MJD (in UTC system), eg 56884.99373389. 

char[15] 
Time of observation start, eg "23:50:58.608241". 

char[13] 
Hour angle of the observation (HH:MM:SS) 

float 
Airmass, eg 1.35. 

char[10] 
Proposal ID of the program that took this image, eg "2014B0404". 

float 
Median zero point for the entire image (median of all CCDs of the image), eg 25.0927. 

float 
Average sky level in this image, in ADU, eg 36.9324. 

float 
Average gain for this CCD, eg 4.34. 

float 
[use "seeing" instead] 

float 
Astrometric header value: X reference pixel. 

float 
Astrometric header value: Y reference pixel. 

double 
Astrometric header value: RA of reference pixel. 

double 
Astrometric header value: Dec of reference pixel. 

float 
Astrometric header value: transformation matrix. 

float 
Astrometric header value: transformation matrix. 

float 
Astrometric header value: transformation matrix. 

float 
Astrometric header value: transformation matrix. 

int16 
CCD number (see DECam layout), eg 1. 

char[3] 
CCD name (see DECam layout), eg "N10", "S7". 

float 
Zeropoint for the CCD (AB mag). 

float 
Zeropoint for amp A (AB mag). 

float 
Zeropoint for amp B (AB mag). 

float 
(ignore) 

float 
Photometric rms for the CCD (in mag). 

float 
Sky rms (in counts) 

float 
Mean sky background in AB mag/arcsec^{2} on each CCD; measured from the CPprocessed frames as 2.5*alog10( 

float 
Mean sky count level per pixel in the CPprocessed frames measured (with iterative rejection) for each CCD in the image section [500:1500,1500:2500] 

float 
Median astrometric offset for the CCD <PS1DECaLS> in arcsec. 

float 
Median astrometric offset for the CCD <PS1DECaLS> in arcsec 

float 
(ignore) 

int16 
Number of stars found on the CCD. 

int16 
Number of stars matched to PS1 (and used to compute the photometric zero points and astrometric offsets). 

int16 
Number of stars in amp A matched. 

int16 
Number of stars in amp B matched. 

float 
Median (gi) color from the PS1 catalog of the matched stars. 

float 
Outside temperature in ^{o}C listed in the 

char[5] 
The camera that took this image; "decam". 

char[12] 
Exposure ID string, eg "00348224S29" (from 

int16 
FITS HDU number in the 

char[61] 
Path to FITS image, eg "decam/CP20140810_g_v2/c4d_140815_235218_ooi_g_v2.fits.fz". 

int16 
Width in pixels of this image, eg 2046. 

int16 
Height in pixels of this image, eg 4096. 

double 
Telescope boresight RA of this exposure (deg). 

double 
Telescope boresight Dec of this exposure (deg). 

double 
Approximate RA center of this CCD (deg). 

double 
Approximate Dec center of this CCD (deg). 
surveyccdsnondecals.fits.gz
As for surveyccdsdecals.fits.gz, but for areas of the sky covered with DECam by surveys other than DECaLS.
surveyccdsextra.fits.gz
As for surveyccdsnondecals.fits.gz but for some additional regions of the sky. The "nondecals" and "extra" files are currently split up simply to make them easier to store on github.
ccdsannotateddecals.fits.gz
A version of the surveyccdsdecals.fits.gz file with additional information gathered during calibration preprocessing before running the Tractor reductions.
Includes everything listed in the surveyccdsdecals.fits.gz files plus the following:
Column 
Type 
Description 


boolean 
True if this CCD was considered photometric and used in the DR3 reductions 

boolean 
We blacklisted certain programs (Proposal IDs) from other PIs where there were a large number of images covering a single patch of sky, because our pipeline code didn't handle the extreme depth very well. True if this CCD was not blacklisted, ie, was used. 

int[4] 
If only a subset of the CCD images was used, this array of x0,x1,y0,y1 values gives the coordinates that were used, [x0,x1), [y0,y1). 1 for no cut (most CCDs). 

double 
RA coordinate of pixel (1,1) 

double 
Dec coordinate of pixel (1,1) 

double 
RA coordinate of pixel (1,H) 

double 
Dec coordinate of pixel (1,H) 

double 
RA coordinate of pixel (W,H) 

double 
Dec coordinate of pixel (W,H) 

double 
RA coordinate of pixel (W,1) 

double 
Dec coordinate of pixel (W,1) 

float 
Maximum distance from RA,Dec center to the edge midpoints, in RA 

float 
Maximum distance from RA,Dec center to the edge midpoints, in Dec 

double 
RA coordinate of CCD center 

double 
Dec coordinate of CCD center 

float 
Median perpixel error standard deviation, in nanomaggies. 

float 
Our pipeline (not the CP) estimate of the sky level, average over the image, in ADU. 

float 
Standard deviation of our sky level 

float 
Max of our sky level 

float 
Min of our sky level 

float 
Pixel scale (via sqrt of area of a 10x10 pixel patch evaluated in a 5x5 grid across the image), in arcsec/pixel. 

float 
Standard deviation of pixel scale 

float 
Max of pixel scale 

float 
Min of pixel scale 

float 
PSF norm = 1/sqrt of N_eff = sqrt(sum(psf_i^2)) for normalized PSF pixels i; mean of the PSF model evaluated on a 5x5 grid of points across the image. Pointsource detection standard deviation is 

float 
Standard deviation of PSF norm 

float 
Norm of the PSF model convolved by a 0.45" exponential galaxy. 

float 
Standard deviation of galaxy norm. 

float 
PSF model second moment in x (pixels^2) 

float 
PSF model second moment in y (pixels^2) 

float 
PSF model second moment in xy (pixels^2) 

float 
PSF model major axis (pixels) 

float 
PSF model minor axis (pixels) 

float 
PSF position angle (deg) 

float 
PSF ellipticity 1  minor/major 

float 
Percent humidity outside 

float 
Outside temperate (deg C). 

int32 
DECaLS tile number, if this was a DECaLS observation; or 0 for data from other programs. 

uint8 
DECaLS tile pass number, 1, 2 or 3, if this was a DECaLS observation, or 0 for data from other programs. Set by the observers; pass 1 is supposed to be photometric with good seeing, pass 3 unphotometric or bad seeing, and pass 2 in between. 

float 
Mean SFD E(BV) extinction in the DECaLS tile, or 0 for nonDECaLS data. 

char[6] 
Community Pipeline (CP) PLVER version string 

float 
SFD E(BV) extinction for CCD center 

float[6] 
Extinction for DECam filters ugrizY 

float[4] 
Extinction for WISE bands W1,W2,W3,W4 

float 
5sigma PSF detection depth in AB mag, using PsfEx PSF model 

float 
5sigma galaxy (0.45" round exp) detection depth in AB mag 

float 
5sigma PSF detection depth in AB mag, using Gaussian PSF approximation (using 

float 
5sigma galaxy detection depth in AB mag, using Gaussian PSF approximation 
ccdsannotatednondecals.fits.gz
As for ccdsannotateddecals.fits.gz, but for areas of the sky covered with DECam by surveys other than DECaLS.
ccdsannotatedextra.fits.gz
As for ccdsannotatednondecals.fits.gz but for some additional regions of the sky. The "nondecals" and "extra" ccds files are currently split up simply to make them easier to store on github.
dr3depth.fits.gz
A concatenation of the depth histograms for each brick, from the
coadd/*/*/*depth.fits
tables. These histograms describe the
number of pixels in each brick with a 5sigma AB depth in the given
magnitude bin.
dr3depthsummary.fits.gz
A summary of the depth histogram of the whole DR3 survey. FITS table with the following columns:
Column 
Type 
Description 


float 
Lower limit of the depth bin 

float 
Upper limit of the depth bin 

int 
Number of pixels in histogram for point source depth in g band 

int 
Number of pixels in histogram for canonical galaxy depth in g band 

int 
Number of pixels in histogram for point source depth in r band 

int 
Number of pixels in histogram for canonical galaxy depth in r band 

int 
Number of pixels in histogram for point source depth in z band 

int 
Number of pixels in histogram for canonical galaxy depth in z band 
The depth histogram goes from magnitude of 20.1 to 24.9 in steps of 0.1 mag. The first and last bins are "catchall" bins: 0 to 20.1 and 24.9 to 100, respectively. The histograms count the number of pixels in each brick's unique area with the given depth. These numbers can be turned into values in square degrees using the brick pixel area of 0.262 arcseconds square. These depth estimates take into account the smallscale masking (cosmic rays, edges, saturated pixels) and detailed PSF model.
External Files
 The DECaLS photometric catalogs have been matched to the following external spectroscopic files from the SDSS, which can be accessed through the web at:

https://portal.nersc.gov/cfs/cosmo/data/legacysurvey/dr3/external/
 Or on the NERSC computers (for collaborators) at:

/global/cfs/cdirs/cosmo/data/legacysurvey/dr3/external/
Each row of each externalmatch file contains the full record of the nearest object in our Tractored survey
imaging catalogs, matched at a radius of 1.0 arcsec. The structure of the imaging catalog files
is documented on the catalogs page. If no match is found, then OBJID
is set to 1.
surveydr3specObjdr13.fits
HDU1 (the only HDU) contains Tractored survey photometry that is rowbyrowmatched to the SDSS DR13 spectrosopic pipeline file such that the photometric parameters in row "N" of surveydr3specObjdr13.fits matches the spectroscopic parameters in row "N" of specObjdr13.fits. The spectroscopic file is documented in the SDSS DR13 data model for specObjdr13.fits.
surveydr3DR12Q.fits
HDU1 (the only HDU) contains Tractored survey photometry that is rowbyrowmatched to the SDSS DR12 visually inspected quasar catalog (Paris et al. 2017) such that the photometric parameters in row "N" of surveydr3DR12Q.fits matches the spectroscopic parameters in row "N" of DR12Q.fits. The spectroscopic file is documented in the SDSS DR12 data model for DR12Q.fits.
surveydr3Superset_DR12Q.fits
HDU1 (the only HDU) contains Tractored survey photometry that is rowbyrowmatched to the superset of all SDSS DR12 spectroscopically confirmed objects that were visually inspected as possible quasars (Paris et al. 2017) such that the photometric parameters in row "N" of surveydr3Superset_DR12Q.fits matches the spectroscopic parameters in row "N" of Superset_DR12Q.fits. The spectroscopic file is documented in the SDSS DR12 data model for Superset_DR12Q.fits.
surveydr3DR7Q.fits
HDU1 (the only HDU) contains Tractored survey photometry that is rowbyrowmatched to the SDSS DR7 visually inspected quasar catalog (Schneider et al. 2010) such that the photometric parameters in row "N" of surveydr3DR7Q.fits matches the spectroscopic parameters in row "N" of DR7qso.fit. The spectroscopic file is documented on the DR7 quasar catalog description page.
Tractor Catalogs
In the file listings outlined below:
brick names (<brick>) have the format <AAAa>c<BBB> where A, a and B are digits and c is either the letter m or p (e.g. 1126p222). The names are derived from the (RA, Dec) center of the brick. The first four digits are \(int(RA \times 10)\), followed by p to denote positive Dec or m to denote negative Dec ("plus"/"minus"), followed by three digits of \(int(Dec \times 10)\). For example the case 1126p222 corresponds to (RA, Dec) = (112.6°, +22.2°).
<brickmin> and <brickmax> denote the corners of a rectangle in (RA, Dec). Explicitly, <brickmin> has the format <AAA>c<BBB> where <AAA> denotes three digits of the minimum \(int(RA)\) in degrees, <BBB> denotes three digits of the minimum \(int(Dec)\) in degrees, and c uses the p/m ("plus"/"minus") format outlined in the previous bullet point. The convention is similar for <brickmax> and the maximum RA and Dec. For example 000m010010m005 would correspond to a survey region limited by \(0^\circ \leq RA < 10^\circ\) and \(10^\circ \leq Dec < 5^\circ\).
subdirectories are listed by the RA of the brick center, and subdirectory names (<AAA>) correspond to RA. For example 002 corresponds to brick centers between an RA of 2° and an RA of 3°.
<filter> denotes the g, r or z band, using the corresponding letter.
Note that it is not possible to go from a brick name back to an exact (RA, Dec) center (the bricks are not on 0.1° grid lines). The exact brick center for a given brick name can be derived from columns in the surveybricks.fits.gz file (i.e. brickname
, ra
, dec
).
tractor/<AAA>/tractor<brick>.fits
FITS binary table containing Tractor photometry, documented on the catalogs page.
Users interested in database access to the Tractor Catalogs can contact the Astro Data Lab [1] at datalab@noirlab.edu.
Sweep Catalogs
sweep/3.1/sweep<brickmin><brickmax>.fits
Note that previous versions of the sweeps are available in sweep/3.0/sweep<brickmin><brickmax>.fits for reasons detailed on the known issues page.
The sweeps are lightweight FITS binary tables (containing a subset of the most commonly used
Tractor measurements) of all the Tractor catalogs for which BRICK_PRIMARY==T
in rectangles of RA, Dec. Includes:
Name 
Type 
Units 
Description 


int32 
Brick ID [1,662174] 


char 
Name of brick, encoding the brick sky position, eg "1126p222" near RA=112.6, Dec=+22.2 


int32 
Catalog object number within this brick; a unique identifier hash is 


char[4] 
Morphological model: "PSF"=stellar, "SIMP"="simple galaxy" = 0.45" round EXP galaxy, "EXP"=exponential, "DEV"=deVauc, "COMP"=composite. Note that in some FITS readers, a trailing space may be appended for "PSF ", "EXP " and "DEV " since the column data type is a 4character string 


float64 
deg 
Right ascension at equinox J2000 

float32 
1/deg² 
Inverse variance of RA (no cosine term!), excluding astrometric calibration errors 

float64 
deg 
Declination at equinox J2000 

float32 
1/deg² 
Inverse variance of DEC, excluding astrometric calibration errors 

float32[6] 
nanomaggies 
DECam model flux in ugrizY 

float32[6] 
1/nanomaggies² 
Inverse variance oF DECAM_FLUX 

float32[6] 
Galactic transmission in ugrizY filters in linear units [0,1] 


uint8[6] 
Number of images that contribute to the central pixel in each filter for this object (not profileweighted) 


float32[6] 
Profileweighted χ² of model fit normalized by the number of pixels 


float32[6] 
arcsec 
Weighted average PSF FWHM per band 

float32[6] 
Profileweight fraction of the flux from other sources divided by the total flux (typically [0,1]) 


float32[6] 
Profileweighted fraction of pixels masked from all observations of this object, strictly between [0,1] 


float32[6] 
Fraction of a source's flux within the blob, near unity for real sources 


float32[6] 
1/nanomaggies² 
For a \(5\sigma\) point source detection limit, \(5/\sqrt(\mathrm{DECAM\_DEPTH})\) gives flux in nanomaggies and \(2.5[\log_{10}(5 / \sqrt(\mathrm{DECAM\_DEPTH}))  9]\) gives corresponding magnitude 

float32[6] 
1/nanomaggies² 
As for DECAM_DEPTH but for a galaxy (0.45" exp, round) detection sensitivity 

boolean 
True for objects whose center is on the brick; less strong of a cut than BRICK_PRIMARY 


int16[6] 
Bitwise mask set if the central pixel from any image satisfy each condition 


int16[6] 
Bitwise mask set if the central pixel from all images satisfy each condition 


float32[4] 
nanomaggies 
WISE model flux in W1,W2,W3,W4 

float32[4] 
1/nanomaggies² 
Inverse variance of WISE_FLUX 

float32[4] 
Galactic transmission in W1,W2,W3,W4 filters in linear units [0,1] 


int16[4] 
Number of images that contribute to the central pixel in each filter for this object (not profileweighted) 


float32[4] 
Profileweight fraction of the flux from other sources divided by the total flux (typically [0,1]) 


float32[4] 
Profileweighted χ² of model fit normalized by the number of pixels 


float32[5] 
Difference in χ² between successively morecomplex model fits: PSF, SIMPle, DEV, EXP, COMP. The difference is versus no source. 


float32 
Fraction of model in deVauc [0,1] 


boolean 
Is there a Tycho2 (very bright) star in this blob? 


float32 
arcsec 
Halflight radius of deVaucouleurs model (>0) 

float32 
1/arcsec² 
Inverse variance of SHAPEDEV_R 

float32 
Ellipticity component 1 (documented on the catalogs page) 


float32 
Inverse variance of SHAPEDEV_E1 


float32 
Ellipticity component 2 (documented on the catalogs page) 


float32 
Inverse variance of SHAPEDEV_E2 


float32 
arcsec 
Halflight radius of exponential model (>0) 

float32 
1/arcsec² 
Inverse variance of R_EXP 

float32 
Ellipticity component 1 (documented on the catalogs page) 


float32 
Inverse variance of SHAPEEXP_E1 


float32 
Ellipticity component 2 (documented on the catalogs page) 


float32 
Inverse variance of SHAPEEXP_E2 


float32 
mag 
Galactic extinction E(BV) reddening from SFD98, used to compute DECAM_MW_TRANSMISSION and WISE_MW_TRANSMISSION 
Image Stacks
Image stacks are on tangentplane (WCS TAN) projections, 3600 × 3600 pixels, at 0.262 arcseconds per pixel.
coadd/<AAA>/<brick>/legacysurvey<brick>ccds.fits
FITS binary table with the list of CCD images that were used in this brick.
Same columns as surveyccds*.fits.gz
, plus:
Column 
Type 
Description 


int16 
Minimum x image coordinate overlapping this brick 

int16 
Maximum x image coordinate overlapping this brick 

int16 
Minimum y image coordinate overlapping this brick 

int16 
Maximum y image coordinate overlapping this brick 

int16 
Minimum x brick image coordinate overlapped by this image 

int16 
Maximum x brick image coordinate overlapped by this image 

int16 
Minimum y brick image coordinate overlapped by this image 

int16 
Maximum y brick image coordinate overlapped by this image 

float 
Same as 

float 
Same as 

char[6] 
Community Pipeline (CP) version 

char[16] 
Git version of the sky calibration code 

char[16] 
Git version of the WCS calibration code 

char[16] 
Git version of the PSF calibration code 

char[16] 
CP version of the input to sky calibration 

char[16] 
CP version of the input to WCS calibration 

char[16] 
CP version of the input to PSF calibration 
coadd/<AAA>/<brick>/legacysurvey<brick>image<filter>.fits
Stacked image centered on a brick location covering 0.25° × 0.25°. The primary HDU contains the coadded image (inversevariance weighted coadd), in units of nanomaggies per pixel.
NOTE: These are not the images used by Tractor, which operates on the singleepoch images.
NOTE: that these images are resampled using Lanczos3 resampling.
coadd/<AAA>/<brick>/legacysurvey<brick>invvar<filter>.fits
Corresponding stacked inverse variance image based on the sum of the inversevariances of the individual input images in units of 1/(nanomaggies)² per pixel.
NOTE: These are not the inverse variance maps used by Tractor, which operates on the singleepoch images.
coadd/<AAA>/<brick>/legacysurvey<brick>model<filter>.fits.gz
Stacked model image centered on a brick location covering 0.25° × 0.25°.
The Tractor's idea of what the coadded images should look like; the Tractor's model prediction.
coadd/<AAA>/<brick>/legacysurvey<brick>chi2<filter>.fits
Stacked χ² image, which is approximately the summed χ² values from the singleepoch images.
coadd/<AAA>/<brick>/legacysurvey<brick>depth<filter>.fits.gz
Stacked depth map in units of the pointsource flux inversevariance at each pixel.
The 5σ pointsource depth can be computed as 5 / sqrt(depth_ivar) .
coadd/<AAA>/<brick>/legacysurvey<brick>galdepth<filter>.fits.gz
Stacked depth map in units of the canonical galaxy flux inversevariance at each pixel. The canonical galaxy is an exponential profile with effective radius 0.45" and round shape.
The 5σ galaxy depth can be computed as 5 / sqrt(galdepth_ivar) .
coadd/<AAA>/<brick>/legacysurvey<brick>nexp<filter>.fits.gz
Number of exposures contributing to each pixel of the stacked images.
coadd/<AAA>/<brick>/legacysurvey<brick>image.jpg
JPEG image of calibrated image using the g,r,z filters as the colors.
coadd/<AAA>/<brick>/legacysurvey<brick>model.jpg
JPEG image of the Tractor's model image using the g,r,z filters as the colors.
coadd/<AAA>/<brick>/legacysurvey<brick>resid.jpg
JPEG image of the residual image (data minus model) using the g,r,z filters as the colors.
Raw Data
See the raw data page.
Footnotes