Galaxy Evolution Explorer
The individuals directly involved with the acquisition of GALEX data and/or construction of the GALEX-GAMA catalogs:
This web page details the construction of version 2 of the GALEX ultraviolet source catalogs and auxiliary products overlapping the GAMA survey footprint. The GAMA footprint definitions used for this project are defined as three primary regions (GAMA-1) and possible future extensions to the survey (GAMA-2).
The GALEX ultraviolet data is a combination of archival data and pointed observations from the GALEX guest investigator program GALEX-GAMA: UV/Optical/Near-IR/Far-IR/Radio observations of ~100k galaxies (GI5-048; PIs: R. Tuffs, C. Popescu, US-PI: M. Seibert). Note that the GI program was designed and proposed to cover only the original GAMA survey area (i.e. three regions of 36 square degrees each) as shown in figure 2 of the GALEX proposal.
Archival data has been used to extend the ultraviolet coverage to the current GAMA-1 and future GAMA-2 footprint definitions as much as possible.
All GI and archival data overlapping the GAMA-2 footprint and available as of March 30, 2010 were compiled for the project. This consists of 347 GALEX tiles (circular fields 1 degree wide). If interested one may view a list of the GALEX tile names, centers, and exposure times used as input for the G09, G12, G15, G02, and G23 fields.
All GALEX data have been reprocessed at Caltech with a new version of the pipeline (v7ops). There are many improvements and changes with v7ops that will cause differences with version 1 of the GALEX-GAMA catalogs. The most significant are:
A final general note on GALEX data. Version 2 of the GALEX-GAMA catalogs are created using the methods and techniques of the GALEX CATalog Team (GCATT). This is a large project underway to produce a definitive GALEX source catalog. This means that many of the column names and units change in version 2. This will be detailed further.
UNIQUE AREA (Primary vs Secondary):
GALEX tiles overlap one another. This, of course, is inevitable when trying to cover a contiguous area with circular fields of view. Furthermore, because we include archival survey data which is observed on a sky grid system and pointed observations (which are not on the sky grid system) the overlap is often very hap-hazard. There are several ways to handle this such that we include only unique detections and allow a precise footprint computation. We use a method that optimizes the areal coverage and weights tiles with FUV coverage highest.
For each GAMA region the GALEX tiles are rank-sorted by FUV exposure time and secondary rank-sorted by NUV exposure time. This ensures that any tile with FUV coverage takes precedence over any tile with only NUV coverage (note that the FUV detector on GALEX is non-operational at this time).
Once sorted, the footprint of each tile is derived from the relative response map (effective exposure time map) by comparing it to all neighboring tiles with a higher rank-order. In this way a "primary map" is made for each tile indicating which parts of the field of view are to be considered "primary" and which parts are to be considered "secondary".
The figure below displays an example FUV (top left) and NUV (top right) primary map for a single tile. Red represents the primary area for the tile. The lower two panels are the corresponding relative response (effective exposure time) for each band.
Five types of mosaiced maps have been produced for each of the GAMA fields using HEALPix to aid in the analysis of the GALEX-GAMA data set. Specifically, these are:
Because GALEX coverage is neither uniform in exposure time or fully contiguous over the GAMA fields, it is important to be able to derive an accurate footprint for any sample selections driven by the above quantities.
Each UV based map is built by ingesting the primary designated region of each tile's corresponding map into HEALPix format. The HEALPix maps are built at ~13 arc-second resolution (nested, nside = 214 = 16384). These maps allow the calculation of footprint area to better than 0.1%.
Additionally, we have used HEALPix format maps of the Schlegel, Finkbeiner & Davis (1998) Milky Way foreground reddening maps from the LAMBDA project archive to extract color excess (E(B-V)) maps for each GAMA field rebinned at the same pixel scale as the other UV maps (~13 arc-sec pixels, the resolution of course remains only ~6 arc-min).
For those who do not wish to use HEALPix format maps, Gnomonic projection standard fits images (produced from the HEALPix maps) have also been produced. The standard fits images are slightly lower resolution (15 arc-sec).
Images of the maps for each GAMA field are displayed below. Click on the images to view them at high resolution. The GAMA-2 footprint boundaries are indicated by dashed lines and the H-Atlas boundaries are indicated by dotted lines.
The following filters are applied to the GALEX pipeline produced MCAT (band merged catalog) file for each tile.
BLIND CATALOG COLUMNS:
Simple nearest neighbor matching within 4 arc-seconds has been performed between the GALEX blind catalogs and the GAMA Master Catalog (catmast_v2.fits). The resulting catalog (glx_pricat_v2c-gama_catmast_v2.fits.gz) corresponds as one-to-one with the the catmast_v2 catalog.
The subset of this master catalog that forms the the galaxies catalog (catgama_v4) have been extracted and the resulting catalog (glx_pricat_v2c-gama_catgama_v4.fits.gz) corresponds as one-to-one with the catgama_v4 catalog.
In addition to all the columns found in the blind UV catalogs a number of additional matching related columns have been added.
In order to match the UV blind catalog to the optical GAMA catalogues, we employ an advanced matching routine that takes into account multiple matches for each UV and optical source and seeks to reconstruct the original UV flux of a given optical source.
We illustrate the routine for an example optical target (e.g. a GAMA galaxy, see figure). Firstly, shape information of the target in the optical (i.e. ellipticity, position angle, and effective radius along the semimajor axis) is extracted from the GAMA master catalogue, and used to define a target area within which UV counterparts will be deemed to be associated with the optical source. To take into account scatter in apparent positions from real positions in the UV and possibly unavailable optical shape information, the minimal search region, i.e. optical target area, is a circle of radius 4" (the matching radius used in the simple matching procedure, see above). The question that is then to be answered (see flowchart) is which UV sources in the blind catalogue have their central coordinates within this optically defined search region. The number, the UV ID and the total flux of all such UV sources are extracted from the blind UV catalogue and written (in case needed in user applications), to the final matched catalogue.
Since it is possible that a UV source may itself be extended, or have more than one potential close-lying optical counterpart in the GAMA master catalogue, we also need to take into account the possibility that one or more of the UV sources in the search region around the optical target may themselves be related to further optical sources. This eventuality is catered for in the second step of the advanced matching procedure (see flowchart). In this step, a UV area is constructed for each UV source within the optical search region, using the position angle, ellipticity and semimajor axis as given in the UV blind catalogue (Sextractor's Kron-radius). The GAMA MasterCatalogue (which includes both galaxies and stars) in then searched for optical sources within the UV area of each of the UV sources. In the case there is only one UV source in the optical search region which has only one optical source within its UV radius, and this optical source is the target, the whole flux of the UV source is associated with the target. This is by far the most common case. If, however, there are more than one potential optical counterparts to one or more of the UV sources in the optical search region, the UV flux of each UV source is split among all potential optical counterparts for that source, weighted inversely by angular offset. In recognition that this may not be the true flux contribution, and users may for specific applications like to apply other methods (eg using prior astrophysical information) we have included a string listing the IDs of all optical and all UV sources involved in the advanced matching process for each target. This redistribution of the flux is done for all UV counterparts of the target. The match is considered as unambiguous if there is only one UV counterpart which has only the target as its optical counterpart. In all other cases the reported UV flux is to be considered as an estimate of the true flux distribution, as the true flux distribution of the optical counterparts is not known. This estimate is the most likely one considered over a population of galaxies but is necessarily imprecise for targets considered individually.
Catalogue entries (details see tag description below):
The full UV blind catalogue entry will be given for unambiguous matches. In cases where there is more than one UV source within the target area, the full UV blind catalogue entry for the UV source closest to the target is given.
COMPARING SIMPLE VS ADVANCED MATCHING
These color magnitude diagrams (NUV-r vs r, contour level = [0.02,0.04,0.08,0.16,0.32] per cent of the according subsample) for GAMA-galaxies (as defined in TilingCatv20) illustrate the advantage of the redistributed flux of the advanced match method compared to the simple match flux. The color magnitude diagram for unique matches (i.e. no redistribution of the flux necessary since it is a one-on-one match) is shown on the left. This subset is considered to represent the true color magnitude distribution, as there are no potentially contaminating sources involved in the matching process. For ambiguous matches (i.e. non one-on-one match), the central image shows the color magnitude diagram using the NUV simple match flux, while the right image uses the NUV advanced match flux. The outer contour of the unique match color magnitude diagram is overlaid in red for comparison. It is easily seen that the simple match color magnitude diagram is biased towards the blue by about ~0.2mag while the advanced match color magnitude distribution is roughly identical to the one for unique matches.
The right image illustrates how the curve-of-growths method derives the flux of a galaxy for the example of the GAMA-galaxy CATAID 601395. The map shows a contour plot of the masked data within the UV radius of the galaxy and the unmasked data outside of the galaxy (see next paragraph for detail). The actual curve of growth plots the sum of the enclosed flux as a function of radius. The leftt image (obtained with the "SDSS DR7 Finding Chart Tool" on the SDSS web pages) shows the optical SDSS data for the same region of sky (roughly same scaling) for comparison.
For a given galaxy (listed in the galaxy catalogue v4, displayed as green circle in the right image), the curve-of-growths algorithm imports prior knowledge from the r-band about the optical position, and if available about position angle and ellipticity (as listed in the galaxy catalogue v4). Furthermore, the positions of nearby objects are extracted from the MasterCatv2 (displayed as blue stars for nearby stars, and as red triangle for nearby galaxies). These nearby objects are masked with a circular aperture with radius=1.1*FWHM_of_NUV_beam=1.1*5.3arcsec (masked area shown as dashed-dotted "circles" on the right image). In order to measure the flux of this given galaxy, the COG algorithm sums up the flux within elliptical annuli of 3arcsec width centred of the r-band position of the galaxy, interpolating for masked regions. The UV size of the galaxy (right image: dashed circle in the COG map and right vertical line on the COG plot) is defined by the annulus with mean pixel flux less equal the background flux (estimated from unmasked pixel of the region around the galaxy). Since the annuli are binned and since this definition of the radius is very sensitive to minor flux in the outer regions, we also report the half-light radius (right image: left vertical line on the COG plot). (Note: The star below and to the right of the galaxy is just on the edge of the UV size of the galaxy. The part inside the galaxy is masked and the interpolation is shown on the map, while the original data is shown for the part outside of the galaxy - demonstrating nicely the effect and necessity of the masking.)
This algorithm is done separately for the NUV and FUV data, using the same mask for confusion objects. The error of the flux measurement is the propagated error resulting from the background noise per pixel. This error ignores the possibility of a signal dependent noise component in the measurement.
Very bright neighouring sources (e.g. bright stars) may still have a significant UV flux outside of the masked area, leading to a bias in the flux and size estimate. As stars are generally less bright in the FUV than in the NUV, chances are that the masking is sufficent in the FUV even though it might not be in the NUV. Therefore, comparing the galaxy size estimates of NUV and FUV, it is possible to identify (the majority of) cases where a bright star has been misinterpreted as part of the galaxy in the NUV, leading to a much larger galaxy radius in the NUV than in the FUV (criterion: 30 arcsec difference). Since the flux measurement is most likely contaminated in such a case, the default value -99 for "no measurement" is reported for all COG values in the catalogue.
The background noise is a poisson noise with verly low mean, resulting in many pixel with zero photon counts. As the current version of the background estimate does not import UV coverage information, it does not distinguish between "not observed" and "observed, zero photon counts". In order to garantee a stabe background estimate, we limit the allowed number of pixel with value zero (due to eigher of the two reasons) to a maximum of 40% in the NUV and 88% in the FUV, otherwise the default value -99 for "no measurement" is reported for all COG values in the catalogue. This is a very conservative aproach, as it may reject observed galaxies on the edge of the UV tiles. In the future, this will be avoided by incorporating exposure time maps.
WHEN TO USE COG:
The advantage of the COG method for crowded regions of the sky is obvious as it masks potentially contaminating neighbouring objects (rather than redistributing the UV flux among several optical, potentially non-UV sources as done by the advanced matching). For isolated, unresolved objects, the flux estimates of the GALEX pipeline using beam convolved shape fits are expected to be more precise. The plots below motivate the the criteria (r-band effective radius larther than 10 arcsec or non one-on-one matches) we defined for when to use COG flux rather than redistributed flux.
The left plot shows the ratio of COG flux and redistributed flux as a function of the r-band effective radius, with the error bars indicating the inner quartile for each bin. Color coding: one-on-one matches: black, one optical and two UV sources or two optical and one UV source involved in the advanced matching("low multiplicity matches": red, many sources involved in the advanced matching ("high multiplicity matches"): blue. The COG flux is slightly fainter than the redistributed flux for one-on-one matches. For low multiplicity matches, the dominating case is two optical sources and one UV sorce involved in the match. As not all optical sources are actually UV bright, the redistributed flux of galaxies (on average larger structures that get flux taken away) is slightly biased low - counterbalacing the slight underestimate of the COG flux. For large sources (r-band effective radius larger than 10 arcsec), the redistributed flux of the low multiplicity matches is increasingly biased low; clearly indicating the superiority of the COG estimate for large sources. The high multiplicity matches are affected even stronger by this bias.
The left plot shows the ratio of COG flux and redistributed flux as a function of the NUV flux (same color coding).
COMPARING SIMPLE MATCHING VS COG:
OBSERVING CAMPAIGN GAMA-II/H-ATLAS/DINGO FIELDS:
DOWNLOADING THE DATA:
All catalogs and maps are password protected at this time. Please contact Ellen Andrae if you need the username and password.
The older version, i.e. GALEX-GAMA products untill 2012 are available here.