Back up: des_balrog_y3

Balrog FAQ

Q: What is Balrog?
A: Balrog is a set of software which: Adds artificial stars and galaxies to images and then reprocesses them, to come up with a input->output matrix of things such as detection efficiency vs magnitude, biases in magnitude,color, photo-z, shear vs. crowding, magnitude.

Q: What is DES Balrog Y3?
A:DES Balrog for Y3 is specifically aimed at running balrog over the whole footprint (10000) tiles of DES Y3, using the same detection and measurement codes for magnitudes and shears and photo-zs that were used to generate the released Y3A2 dataset.

Q: How is it different from Balrog Y1?
A: Balrog for Y1 operated on coadd images, laying down objects in the coadd and recovering them. Balrog for Y3 lays down the objects in the single epoch exposures and then recombines the exposures into coadds (or MEDs files) the same way they were done for Y3 and redetects and measures there.

Q: Where's the basic Balrog paper link?
A: Here's the Y1 Balrog paper link:

Q: When are the meetings?
A: Currently Thursdays at 3pm ET

Q: When will Balrog be done?
A:Hopefully shortly after the Y5A2 data are released to the collaboration the Y5 version of Balrog will be released to help scientists with their Y5 cosmology results. Before that we hope there will be a Y3 version of Balrog.

Q: How is Balrog used by the DES collaboration?
A:By weak lensing, photo-z, large scale structure and possibly other groups for refining their selection function and understanding biases.

Q: When will Balrog start?
A: It has already started!

Q: Where are Balrog plots?
A: The sanity check testing plots are here: Balrog Sanity Check Testing. Many more plots are coming.

Q: How can I get the Balrog outputs?
A: For now the test data outputs are being stored at FNAL at des41.fnalgov:/data/des41.b/data/Balrog. Some limited subset will
be put on an https site for those who don't login to the FNAL computers.

Q: Where can I get the Balrog test data?
A: The Balrog test data are based upon the Y3A1 immask images and coadds and associated catalogs. They are available here: (use the DESSCI y3a2_file_archive_info table to find the full file paths to a particular image/catalog) for the baseline set. The Balrog catalogs with injected objects detected and measured and the truth tables are found currently here: (with FNAL login).

Q:What is BFD?
A: This is Bayesian Fourier Domain, a method of measuring moments of flux in galaxies. We are going to try and additionally measure BFD moments on galaxies that come out of Balrog. This will require a bit more processing, (perhaps 1 hour/tile). Working on getting the algorithm in place and tested. There's a single tile test done at /data/des41.b/data/BFD_TEST/

Q: What are SOF, MOF, metacal/shear?
A: These are the key measured quantities on galaxies. SOF = single object fit (no neighbor subtraction but neighbor masking) CM_MAG (Composite Model magnitudes), where the composite model is a linear superposition of a DeVauc. spheroid (sersic n=4) and an exponential disk(sersic n=1). MOF (Multiobject fit) is like SOF but with neighbors subtracted. Both MOF and SOF mask neighbors, only MOF subtracts neighbors as part of the fit. Shear is also called metacal/ngmix and other things. This is how DES measures shapes of galaxies. In Y1 there was also im3shape.

Q: Are there other balrog type things going on?
A: Yes: HSC/LSST has this nice video:

Q: Is there a Y3 Balrog Paper?
A: Yes, please contact us if you wish to help!

Q: What is truth -- as far magnitudes of galaxies goes?
A: This is difficult because the light from galaxies is very extended and reaches very low S/N in the outer parts. An estimate of the total light from a galaxy needs to be cut off at some radius short of infinity. 'Fitted models' of an assumed shape (like a linear combination of the DeVauc+exponential disk models used in the CM_MAG estimate) are about the best we can do these days. When we start with a truth table which has a CM_MAG, and a CM_T (x^2+y2 size parameter for a galaxy), we need to truncate the estimate of total mag somewhere short of sqrt(CM_T) in order in order to avoid being dominated by noise in the outer parts, especially for very low surface brightness extended objects with large CM_T. There will be an effort to adjust our truth table CM_MAG values to account for this in Balrog. But that hasn't happened yet.