The full suite of DECaPS2 data products are document in a datamodel that parallels the release file structure. Some of the main products are highlighted below.

DECaPS provides two types of catalogs: individual-image catalogs and band-merged catalogs. The individual catalogs are the outputs of crowdsource and CloudCovErr.jl on the individual DECam CCDs. The band-merged catalogs are derived from the individual-image catalogs, merged so that different detections of the same source are collected and averaged together in the same catalog entry. Most users will want the band-merged catalogs.

All catalogs are available for download. The raw and DECam Community Pipeline (CP) processed images are formally available through the NOIRLab Astro Data Archive, but since some images were processed by the CP multiple times, we also provide the exact set of images used to create the catalog here.

crowdsource analyzes each DECaPS CCD individually to determine the position and fluxes of all sources in the image. CloudCovErr.jl then processes the residuals to correct the flux and flux uncertainties in the presence of nebulosity by interpolating the unmodeled (sharp) background behind the star. The resulting output catalog has the following fields:

NameTypeUnitDescription
Xfloat64pixelsx coordinate
Yfloat64pixelsy coordinate
FLUXfloat32ADUraw flux
PASSNOfloat64iteration at which source was found
DXfloat32pixelsx uncertainty
DYfloat32pixelsy uncertainty
DFLUXfloat32ADUflux uncertainty (statistical only)
QFfloat32"quality factor" based on PSF-weighted fraction of pixels contributing to the detection. 0 < qf <1 ranges bad < qf < good
RCHI2float32average \(\chi^2\) per pixel, weighted by PSF
FRACFLUXfloat32fraction of flux in this object's PSF that comes from this object
FLUXLBSfloat32[DEFUNCT] flux with attempt at local background subtraction (Backwards compatability only)
DFLUXLBSfloat32[DEFUNCT] fluxlbs uncertainty (Backwards compatability only)
FWHMfloat32pixelsfull-width at half-maximum of PSF
SPREAD_MODELfloat32sextractor-like spread_model; positive means the source is broader than a PSF
DSPREAD_MODELfloat32uncertainty in sextractor spread_model
FLUXISOfloat32ADUflux derived from linear least squares fit to neighbor-subtracted image; significant difference from ordinary flux indicates a convergence issue
XISOfloat64pixelsx coordinate derived from linear least squares fit to neighbor-subtracted image; significant difference from ordinary x indicates a convergence issue
YISOfloat64pixelsy coordinate derived from linear least squares fit to neighbor-subtracted image; significant difference from ordinary y indicates a convergence issue
FLAGSint32DECam Community Pipeline flags at central pixel from upstream image processing
SKYfloat32ADUsky level at central pixel
RAfloat64degreeright ascension
DECfloat64degreedeclination
DECAPSIDint64unique object id (internal to each data release)
GAINfloat32e-/ADUemperical gain (counts times inverse variance)
PRNfloat32probability central pixel in "nebulosity" class
PRLfloat32probability central pixel in "light nebulosity" class
PRRfloat32probability central pixel in "regular" class
PREfloat32probability central pixel in "error" class
DCFLUXfloat32ADU[CloudCovErr] flux uncertainty (using full background covariance)
DCFLUX_DIAGfloat32ADU[CloudCovErr] flux uncertainty (using diagonal covariance)
CFLUXfloat32ADU[CloudCovErr] background-corrected flux
FDB_RESfloat32ADU[CloudCovErr] bias offset from residuals
FDB_PREDfloat32ADU[CloudCovErr] bias offset from predicted background
CCHI2float32[CloudCovErr] \(\chi^2\) of predicted background evaluated for "good" pixels
KCOND0float32[CloudCovErr] initial number of "good" pixels (representative of true background)
KCONDfloat32[CloudCovErr] final number of "good" pixels (representative of true background)
KPREDfloat32[CloudCovErr] number of "hidden" pixels (background interpolation region)
DNTfloat32[CloudCovErr] bitmask on corrected flux solution

The DECaPS band-merged catalogs have undergone photometric calibration and tabulate a number of quantities for each source.

NameTypeUnitDescription
RAfloat64degreeright ascension
DECfloat64degreedeclination
POSSTDEVfloat32arcsecstandard deviation in position of object between different detections
RA_OKfloat64degreeright ascension
DEC_OKfloat64degreedeclination
POSSTDEV_OKfloat32arcsecstandard deviation in position of object between different detections
NDETint16total number of detections associated with object
NDET_OKint16total number of detections associated with object
NMAGint16[6]total number of detections associated with object per band
NMAG_OKint16[6]total number of detections associated with object per band
MEANfloat32[6]Mgymean flux per band
STDEVfloat32[6]Mgystandard deviation of flux measurements per band (can be NaN if 1 detection)
ERRfloat32[6]Mgyaverage dflux (uncertainty) per band
MEDIANfloat32[6]Mgymedian flux per band
Q25float32[6]Mgy25th percentile flux
Q75float32[6]Mgy75th percentile flux
NDET_CFLUXint16total number of detections associated with object
NDET_CFLUX_OKint16total number of detections associated with object
NMAG_CFLUXint16[6]total number of detections associated with object per band
NMAG_CFLUX_OKint16[6]total number of detections associated with object per band
MEAN_CFLUXfloat32[6]Mgymean cflux per band
STDEV_CFLUXfloat32[6]Mgystandard deviation of cflux measurements per band (can be NaN if 1 detection)
ERR_CFLUXfloat32[6]Mgyaverage dcflux (uncertainty) per band
MEDIAN_CFLUXfloat32[6]Mgymedian cflux per band
Q25_CFLUXfloat32[6]Mgy25th percentile cflux
Q75_CFLUXfloat32[6]Mgy75th percentile cflux
MAGLIMITfloat32[6]AB Magvery rough magnitude limit at which this detection would be only 5 sigma, -2.5 * log10(5 * dflux) + zeropoint
EPOCHRANGEfloat64dayMJD difference between earliest and latest detection
EPOCHRANGE_OKfloat64dayMJD difference between earliest and latest detection
EPOCHMEANfloat64J2000Mean MJD
EPOCHMEAN_OKfloat64J2000Mean MJD
DNT_ALLint32[6]bits from CloudCovErr solutoin thrown in all (ok) detections in a given band
DQFLAGS_ALLint64[6]bits from DECam Community pipeline/crowdsource solution thrown in all (ok) detections in a given band
FRACFLUX_AVGfloat32[6]average fraction of flux in this object's PSF that comes from this object
RCHI2_AVGfloat32[6]average \(\chi^2\) per pixel, weighted by PSF
QF_AVGfloat32[6]average "quality factor" based on PSF-weighted fraction of pixels contributing to the detection (0 < qf <1 ranges bad < qf < good)
PRN_AVGfloat32[6]average probability of nebulosity classification at the detection location per band
PRL_AVGfloat32[6]average probability of light nebulosity classification at the detection location per band
PRR_AVGfloat32[6]average probability of regular classification at the detection location per band
PRE_AVGfloat32[6]average probability of error mode classification at the detection location per band
CCHI2_AVGfloat32[6]average \(\chi^2\) of predicted background evaluated for "good" pixels
SNR_AVGfloat32[6]average flux/dflux
DNT_ANYint32[6]bits from CloudCovErr solutoin thrown in any (ok) detections in a given band
DQFLAGS_ANYint64[6]bits from DECam Community pipeline/crowdsource solution thrown in any (ok) detections in a given band
OBJ_IDint64unique DECaPS2 object ID

For quantities with an additional "_OK" field, the first is an average over all detections and the "_OK" field is limited to only "OK" detections. Detections are "OK" if no bad flag bits were set for the center pixel, the quality factor and RCHI2 pass a cut, and the flux uncertainty is nonzero. See the DECaPS2 release paper for these cuts and bad flag lists. The fields with a "_AVG", "_ANY", and "_ALL" were computed only over "OK" detections. The weights used for "_AVG" differ slightly from those used to compute the flux mean, using 0.1 instead of 0.0921 to regularize multiplicative errors.

Fields that are arrays of 6 elements in one dimension are "per band" with the band ordering ugrizY. However, no u-band observations were taken during DECaPS and thus the first element in that dimension should always be zero.

All fluxes are reported in "maggies" (Mgy) which are equivalent to 3631 Jy and are a convenient unit so that the AB magnitudes of the stars are given by \(-2.5\log_{10}\mathrm{flux}\). The absolute calibration is based on the Tonry et al. (2012) absolute calibration of PS1. Subsequent analyses (Scolnic et al., 2014, 2015) of the PS1 survey found offsets relative to AB. To shift the DECaPS magnitudes to the Scolnic et al. (2015) absolute calibration, offsets of 0.020, 0.033, 0.024, 0.028, and 0.011 mag must be added to the DECaPS grizY magnitudes, respectively.

The crowdsource flags are combined with those from the DECam Community Pipeline into a single bitmask. Bits 0 - 8 are inherited from the CP, 20 - 24 indicate a special region on the CCD or sky, and 30 - 31 indicate a change in the crowdsource source identification parameters.

Flags
Bit Description Exclude Catalog? Exclude Injections?
0 No problms No No
1 Bad pixel Yes Yes
3 Saturated Yes Yes
4 Bleed trail Yes Yes
5 Cosmic ray Yes Yes
6 Low weight Yes No
7 Difference detection No Yes
8 Long streak Yes No
20 Additional bad pixel Yes Yes
21 Nebulosity No No
22 S7 amplifier B Yes No
23 Near bright star No Yes
24 Near galaxy No Yes
30 No deblend No No
31 Sharp No No

Excluded flags mean that detections with that flag set are not considered "ok" in the computation of the band-merged catalogs or for determining the distribution of sources to use in synthetic injection tests.

The flags from CloudCovErr.jl are saved as a separate bitmask field known as dnt.

dnt
Bit Meaning
- No problems
0 Few "good" pixels, used pixels beyond radial mask
1 Few "good" pixels, force outermost row/column of pixels "good"
2 [Not Used]
3 Any pixel in PSF model for a source is (even infinitesimally) negative
4 Min/Max PSF < -1e-3
5 Min/Max PSF < -1e-1
6 [Not Used]

Bits beyond 1 should not appear in the band-merged catalog because we explicitly excluded cases with PSF models that had even infinitesimally negative pixels during the photometric calibration. Such cases are extrordinarily rare and occur when the PSF model cannot be well estimated because of clouds obscuring the image or unmasked saturated sources distorting the PSF. Only order 10 CCDs (out of 1.3 million) that were otherwise considered photoemtric were excluded as a result of sources with dnt bits larger than 1.