Deep Lens Survey Catalog Access

The catalogs tabulate positions and magnitudes of 5 million galaxies (and some stars). Each object also has a photometric redshift—including full probability distribution p(z)—a morphological likelihood of being a point source, and lots of parameters measured by SExtractor in each of four filters. If your analysis requires measurements more sophisticated than SExtractor can provide, please download the images. The primary exception to this is that we have carefully measured the colors (see below) in order to optimize the photometric redshifts. The catalogs also include about 1200 spectroscopic redshifts we have acquired, and if you want more please check out the list of overlapping surveys. If you are interested in the time domain please contact us regarding access to the individual exposures.

The best way to access our catalogs is with a MySQL query browser such as MySQL Workbench, although any MySQL client will do (black belts can even scriptize their DLS queries with, for example, Python's mysqldb module). Regardless of the specific client, the connection info is:

  • server: matilda.physics.ucdavis.edu
  • port: 3306 (the default)
  • username: guest
  • password: (none)

Once you are in, you will see two databases, RC1c_public which contains the data release and DLSref which contains supporting items which will not change with the data release version (eg stored procedures to make searches easier). Select RC1c_public to get a list of tables and so that we can omit the "RC1c_public." prefix from the examples below. Before getting in to the specifics of each table, a few general notes. First, please practice your queries with "limit 10" appended until you are satisfied with the query. This will avoid overloading the server with queries which mistakenly ask for millions of results, and it will save you time as well. Second, please use objid rather than sky position to join tables. Joining on objid is better than joining on sky position because deblended objects can have very similar positions but will always have a unique objid. We detect in R (the deepest and highest-resolution band) and use Sextractor and use dual-image mode on the other bands to guarantee that objects with the same objid are the same object. (A side effect of this is that if you want to do something highly specialized like search for z dropouts, you can't use this catalog; download the images instead.) And never use x,y positions because these are not unique. Should you need RA and Dec, they are available in most tables as either alpha and delta or ra and dec, sometimes with a filter name appended. See also the searching by position note below.

A brief description of each table follows (note that table names are case-sensitive, but field names are not).

  • PhotoObj: this is the table most users will be interested in, containing the magnitudes you should use for general-purpose astronomy: Bdered, Vdered, Rdered, and zdered are the dereddened (using SFD98) magnitudes. Important: see magnitude system note below. Undereddened mags are also available in this table; in either case, the mags come from our heavily modified version of ColorPro to provide accurate colors in the face of PSF variation so avoid the Sextractor table if you care about colors.

    The PhotoObj table also includes important flags you must pay attention to. The excluded field tells you if the object fell in an exclusion region around a bright star or at the edge of the field. Needless to say, you shouldn't use these objects unless you really know what you are doing (eg, you need to look at an object you know exists based on other observations). Read more about exclusion regions below. Second, the FLAGSB, FLAGSV, FLAGSR, and FLAGSz fields are the Sextractor flags for those filters (here is a cheat sheet listing their meanings). Due to our long exposures, some low-redshift galaxies are saturated, so you should check this. Sextractor sets bit two for saturation (eg, FLAGSR=4 if no other bits are set). We do not use anything with flags greater than 3 for our own science. As a summary of the sanity checks you should do, let's say you want colors of all objects which are not in exclusion zones, not saturated, and don't have any other serious Sextractor problem. The query is:
    select * from PhotoObj where FLAGSB<4 and FLAGSV<4 and FLAGSR<4 and FLAGSz<4 and 
        excluded=0 limit 10
    Finally, the PhotoObj table contains a star/galaxy indicator. Roughly speaking, dlsqc_prob is the probability that the object is a star, but see the dlsqc table below for a more accurate description. PhotoObj.dlsqc_prob is just a copy of dlsqc.dlsqc_prob provided here so you don't have to bother joining tables.
  • Sextractor: this contains lots of Sextractor outputs for each object. Most astronomers will be familiar with these outputs, so we do not describe them here, but a few points are worth noting: (1) We detect in R and run Sextractor in dual-image mode so it is not recommended to use the magnitudes in this table to derive colors; better colors are available in the DerivedMags table. (Another consequence of dual-image mode is that the positions are identical in all filters.) (2) The depth is variable so if your science is sensitive to this you may make use of Sextractor.thresholdR, which describes the detection threshold (in ADU) at the position of each object.
  • Bpz: contains the photometric redshifts, with the z_b field containing the most probable redshift. To get RA, dec, and redshift you must join with PhotoObj using objid like this:
    select p.alpha,p.delta,b.z_b from PhotoObj p, Bpz b where p.objid=b.objid limit 10
    If you want to use photometric redshift information beyond just Bpz.z_b (and you should), you can find information about the meaning of the other Bpz parameters at the BPZ home page (see also Dan Coe's BPZ page). For information on the accuracy and precision of the DLS photometric redshifts, see Schmidt & Thorman (2013).
  • Probs: these are the p(z) probability distributions for each galaxy. To identify a galaxy uniquely, you must join with the Bpz table on subfield and numberr. If your favorite galaxy is at RA,dec of 80.46255,-48.66329, then to get its p(z) you query:
    select p.alpha,p.delta,p.objid,r.* from PhotoObj p, Bpz b, Probs r where 
          abs(p.alpha-80.46255)<0.0003 and abs(p.delta+48.66329)<0.0003 and 
          p.objid=b.objid and b.subfield=r.subfield and b.numberr=r.numberr
    (Note that 0.0003 deg ~1".) The probs come back as 500 floating-point fields representing the probability of the galaxy being at z=0.01, z=0.02,....z=5.00. This query takes quite a long time (~2.5 minutes) because it requires joining three tables. We are working on streamlining this process by including objid in the Probs table directly.
  • Dlsqc: contains a probability (dlsqc_prob) that the object's R-band second moments are consistent with those of a point source. (We use adaptive second moments as defined by Smith et al 2001, and these appear in the table as ixxe,iyye and ixye.) Note that real point sources can have second moments very inconsistent with a point source, due to diffraction spikes, saturation, bleeding, etc. On the other hand, having second moments consistent with the PSF is not proof that the object is a point source, if the second moments have a large uncertainty (ie, the object is faint). Where you make the cut depends on your science and could be a function of magnitude. This computation was done in R band only because it is both deeper and higher-resolution than the other bands.

    As an example, let's look for things which are almost certainly stars:
    select * from Dlsqc where Dlsqc_prob >0.3 limit 10
    (if you remove the limit you'll find 105873 objects, but the query will take longer). Note that this table doesn't contain RA/Dec so if you want their positions you will have to join on objid with another table. However, avoid the mistake of finding the RA/Dec and then using that to join with yet another table. Always use objid to join tables: it's faster and more reliable. Let's say we want to look at the photometric redshifts of the objects which match the morphology of stars:
    select * from Dlsqc,Bpz where Dlsqc_prob >0.3 and Dlsqc.objid=Bpz.objid
    z_b shows quite a range. This illustrates an important point for aficionados: we did not attempt to match colors to stellar templates, so stars will get assigned meaningless galaxy photometric redshifts. In a deep high-latitude survey stars are a small fraction of all objects, but they do contaminate the galaxy sample unless you specifically exclude them. Looking at the m_0 column of this query result tells us their R magnitudes (because the R magnitude was used as BPZ's magnitude prior parameter which it calls m_0). Most of these are rather bright and so might be excluded by your magnitude cuts anyway if you are looking at faint galaxies; some of the fainter ones might actually be unresolved galaxies, but this is something you have to fine-tune based on your science goals.
  • StarDB: this is a very pure (but incomplete) list of stars, useful for PSF modeling. The most useful columns will be alpha and delta (RA and Dec), which allow you to locate the star on an image. Don't use this sample blindly; some stars may be saturated or otherwise unsuitable for PSF modeling even though they are really stars. See the data release paper for details of how we selected objects into this table.
  • Spectra: actually contains spectroscopic redshifts rather than spectra. These are redshifts obtained by the DLS team as part of various followup efforts, mostly on galaxy clusters. They are not part of the survey proper but are offered here to maximize the usefulness of the data. The columns you will be most interested in are alpha/delta (ra/dec), z, and of course objid if you want to join with any other table.

Searching by position: there is a procedure DLSref.fAngularDIstance which does the spherical trig to determine the distance from a reference point, but it is quite slow if you ask it to scan the entire database. You can speed it up dramatically by defining a bounding box before you call this procedure. Let's say you want to search within 0.01 degrees of 140 degrees RA, +30 Dec:

SELECT * FROM PhotoObj p where p.alpha<140.01 and p.alpha>139.99
                           and p.delta>29.99 and p.delta<30.01
                           and DLSref.fAngularDistance(140,30,p.alpha,p.delta)<0.01
Technically, the size of the bounding box in RA should be at least 0.01 divided by the cosine of 30 degrees, or 0.0115467. Any larger bounding box in conjunction with DLSref.fAngularDistance will also give the same results, so you can be lazy and just throw in a factor of two on the RA; this will work for any DLS field. If you do that you get exactly the same results as with DLSref.fAngularDistance alone, but about 100 times faster.

Important magnitude system note: The DLS filters are Johnson-Cousins BVR and SDSS z. The BVR filters are calibrated by comparison to Landolt standards, and the z filter was calibrated by comparison to SDSS standards. As a result, the BVR magnitudes are in the Vega system and the z magnitudes are in the AB system. The following chart shows what you must add to the Vega magnitudes to obtain AB magnitudes:

B-0.105
V0.006
R0.210

and if you want to change the z magnitude to a Vega one, you should subtract 0.549.

Exclusion zones: We generated two masks for each bright star: a tall skinny mask for the bleed and a circular mask for scattered light. There are also some masks which define the edges of the survey. You can see the masks using the Navigator.