This file describes the catalogues for the H-ATLAS Data Release 3. Data Release 3 contains 2 catalogues of the near-infrared candidate counterparts to sources from the SGP field. The sources are those listed in the catalogue released in DR2, and everything about the sources and the Herschel images is described in the documentation for DR2 (see above). The method used to find the near-infrared counterparts is briefly described below (Section 3). The method is fully described in Ward et al. (2022). ********************************************************************** 1. The Catalogues ********************************************************************** The main catalogue HATLAS_SGP_DR3_CAT contains the best candidate near-infrared ID to each SPIRE source and is recommended for most science purposes. We also provide a catalogue HATLAS_SGP_DR3_CAT_ALLIDS containing all possible counterparts within a 15-arcsec radius of each SPIRE source. Column descriptions for the 2 catalogues are in the file HATLAS_SGP_DR3_CAT.COLUMNS. ********************************************************************* 2. Estimated Redshifts for the Herschel Sources ********************************************************************* One addition to the basic catalogues of Herschel sources that we released in DR2 is that we have now estimated redshifts for all the sources by fitting a template to their flux densities at 250, 350 and 500 microns. These estimates are listed in the catalogues, and the method used to make the estimates is described in Ward et al. (2022). ************************************************************************ 3. Near-IR IDs from VIKING DR5 ************************************************************************ IDs from the VIKING Data Release 5 survey are found using a Likelihood Ratio analysis of all the potential counterparts within 15 arcsec of all SPIRE sources. Although all potential candidates are provided in the catalogues, we recommend applying a cut of RELIABILITY>=0.8 as described in Ward et al. (2022). 3.1 Star/Galaxy Separation ************************************************************************ The VIKING positions are matched to the closest KiDS object to within 0.5 arcsec to obtain g and i band magnitudes. These are used to separate stars from extragalactic sources using a cut on the g-i/J-Ks colour-colour diagram. Further details on the star-galaxy classification method can be found in Ward et al. (2022). The GS_KEY column defines how each near-infrared object is classified and by which condition: 0 = Stellar - pStar > 0.95 1 = Galaxy - Location on g-i/J-Ks colour-colour diagram 2 = Stellar - Location on g-i/J-Ks colour-colour diagram 3 = Galaxy - J-Ks > 0.98 4 = Stellar - J-Ks < 0.42 5 = Stellar - pStar > 0.70 6 = Galaxy - All remaining objects We include in the catalogues a simplified flag, GS_FLAG, which reduces GS_KEY to 2 options: 0 = Star 1 = Galaxy During the Likelihood Ratio analysis, extragalactic sources and stars are treated separately. 3.2 Likelihood Ratio Analysis ************************************************************************ The Likelihood Ratio (the ratio of the likelihood that a counterpart with a given magnitude and distance from the source position is the true ID and the likelihood that a chance object lies in the same location with the same properties) is calculated for all near-infrared objects within 15 arcsec of each SPIRE source. The LR is calculated in the following way: LR = f(r) q(m)/n(m) where q(m) is the normalised magnitude distribution of real counterparts, n(m) represents the magnitude distribution of background objects and f(r) represents the probability distribution of the source positional errors. Based on the assumption that HATLAS sources are point-like and symmetric in RA and Dec, we use a Gaussian profile for f(r) with a standard devitation sigma_pos. To estimate q(m) we use a set of random positions to create a background sample of near-infrared objects that we subtract from the magnitude distribution of objects surrounding SPIRE positions. This is normalized and scaled by the value Q0, which is our estimate of the fraction of Herschel sources with counterparts in the VIKING catalogue (whether or not we can distinguish which these are). We make this estimate by comparing the fraction of sources wth no potential counterparts with the fraction of sources in a catalogue of random positions that also appear blank on the VIKING images. A full description of the method is given in Ward et al. (2022) and Fleuren et al. (2012). For each source we use a Gaussian form for f(r) with sigma_pos given by sigma_pos = 0.66*FWHM/SNR The value of the constant of proportionality in this relationship was given by Ivison et al. (2007) as 0.6, but we have replaced this by a value of 0.66, which we derived using the 'blanks method'. Using our estimate of f(r) for each source, and the estimates of q(m) and n(m), we estimated the value of LR for each near-IR object within 15 arcsec of the Herschel source. The reliability (the probability of being the true counterpart) of each potential match depends on the Likelihood Ratio of all possible counterparts to a given source, such that the jth candidate to a source has a reliability given by: R_j = LR_j/[Sum_i{LR_i} + (1-Q0)]. Here we account for all i possible counterparts to a given source and account for the probability that the true counterpart is absent from VIKING (1-Q0). The reliabilities range from 0 - 1, though we recommend a value >=0.8 to define a sample of reliably matched sources. 3.3 Redshifts ************************************************************************ Spectroscopic and photometric redshifts are obtained from the Herschel Extragalactic Legacy Project (HELP). We consider the closest match within 0.5 arcsec to the near-infrared positions as being the correct photometric/spectroscopic counterpart. The catalogue of spectroscopic redshifts is formed from the merged catalogues of the 2dF, 6dF, 2MRS Huchra et al. (2012) and SRSS2 Da Costa et al. (1998) surveys. The source of each redshift in the catalogue is given by Z_SOURCE. Codes are as follows: 1 = 2dF 2 = 6dF 4 = 2MRS 8 = SRSS2 Other numbers indicate that the source has a redshift in more than one redshift catalogue. For example, Z_SOURCE = 5 indicates that there are redshifts in the both the 2dF and 2MRS survey (5=4+1). The Z_QUAL flag gives an indication of the quality of the estimated redshift using the method of Colless et al. (2001). HELP spectroscopic redshifts have the following quality flags: Q = 1 - no redshift could be estimated Q = 2 - a possible but doubtful redshift estimate Q = 3 - means a probable redshift (notionally 90% confident) Q = 4 - means a reliable redshift (notionally 99% confident) Q = 5 - means a reliable redshift and high quality spectrum We recommend using spectroscopic redshifts with Z_QUAL >= 3 for science purposes. For the surveys where the exact reliability measurements are available, their reliability value is given under REL, which can take a value in the range 0 - 1. A final flag, AGN, is present to identify those spectroscopic redshifts that have information identifying their spectrum as a QSO or AGN object. The flag is 1 when this is true, 0 otherwise. The photometric redshifts obtained from HELP are estimated using a template-fitting method as presented in Duncan et al. (2018a), with additional estimates being incorporated using a machine-learning approach as outlined in Duncan et al. (2018b). The SED templates used for the redshifts come from the default library of SEDs of EAZY (Brammer, van Dokkum & Coppi (2008), the XMM-COSMOS template library of Salvato et al. (2009) and the Atlas of Galaxy SEDs Brown et al. (2014). The Machine Learning estimates use the SGP spectroscopic sample (plus additional spectroscopic redshifts from three GAMA fields) as a training set. The near-infrared photometric redshifts used in the Data Release 3 catalogues are the median value of the primary redshift peak above the 80% Highest Posterior Density (HPD) credible interval. Details of how the posterior redshift distributions are obtained can be read in Shirley et al. (2021). The minimum and maximum values, Z_PHOTMIN and Z_PHOTMAX, represent the edges of the credible interval. The errors on these estimates are made assuming that the credible interval is Gaussian in shape, centered on some mean redshift value. On this assumption, the error, Z_PHOTERR, represents 1 sigma uncertainties calculated using 80% HPD credible interval = 1.282 sigma. 3.4 Lensing Probabilities ************************************************************************ For the HATLAS_SGP_DR3_CAT catalogue, we present a probability, LENSEDPROB, that an HATLAS source is being lensed by its respective VIKING ID (Reliablity>0.8) for those sources where the probability is greater than 94%. This value represents the lensing probability threshold that creates a lensed source catalogue (>94%) with the lowest expected percentage of incorrectly classified, unlensed sources. Details of the method used to determine these lensing probabilities can be found in Ward et al. (2022). ************************************************************************* 4. Nearby Galaxies ************************************************************************* The analysis described above is unreliable for extended Herschel sources. For the extended sources, we replaced the results of the analysis above with photometry for sources in the 2MASS Extended Source Catalogue within 5 arcsec of the Herschel position (Maddox et al. 2018). As above, we then looked for spectroscopic redshifts for the 2MASS source in the HELP spectroscopic redshift database.