Project

General

Profile

Comparisons of different SExtractor galaxy model fits

-- Huan Lin, 30 Jul 2013

Some tests comparing model magnitudes and detmodel colors for 3 different SExtractor galaxy model fits (Sersic, disk-only, and bulge+disk) were done using coadd tile DES2247-4414, which contains the cluster RXJ 2248 in the lower left corner. The r-band and i-band coadd images were taken from DESDM's "Configuration Test 11" reductions, here, and the specific run number is 20130710000011_DES2247-4414. I used SExtractor 2.18.10 for the following tests, and ran SExtractor in dual-image mode, using the same detection image (DES2247-4414_det.fits), weights (extension [1] of the respective images), and PSF files as used in the DESDM run indicated.

The main overall conclusions are:
  • The disk-only MAG_MODEL shows noticeably less scatter vs. MAG_AUTO, particularly at faint magnitudes. However, the disk-only model magnitudes are biased about 0.1-0.2 mag fainter vs. the Sersic model results, for Sersic indices greater than about 4.
  • The Sersic and bulge+disk models behave similarly.
  • The r-i DETMODEL colors for the 3 model types are very similar.

Detailed results and plots are given below.

Comparisons of model vs. auto magnitudes, as a function of magnitude

3 different galaxy models were used: (1) Sersic (i.e., the Sersic index is a free parameter), (2) disk-only, and (3) bulge+disk. I adopted the consistency between MAG_MODEL and MAG_AUTO as a metric for comparing the different galaxy model magnitudes, noting that MAG_AUTO currently appears to have better S/N compared to MAG_MODEL, and is thus serving here as a rough "truth" magnitude. The results below were done using galaxies, which were selected in each filter using the cuts SPREAD_MODEL > 0.002 and FLAGS = 0.

The 2 figures below show the results for the r and i filters for this coadd tile. The top 3 panels in each figure show scatter plots of the difference between MAG_MODEL and MAG_AUTO for each fitting model, along with lines indicating the median difference and +/- 34-percentile ranges. I've indicated with vertical red lines in each panel roughly where the turnover is located (r ~ 24.5 and i ~ 24) in the MAG_AUTO number count histogram (lower left panel in each figure). The main conclusions are:
  • The disk-only MAG_MODEL results usually show the least scatter vs. MAG_AUTO, in particular at faint magnitudes where the scatter is significantly better cf. the other 2 galaxy models. However, at magnitudes brighter than about 21, the disk-only scatter is worse than those of the other 2 models. (See bottom right panel.)
  • The behaviors of the Sersic and bulge+disk models are quite similar, though the bulge+disk model shows slightly less scatter vs. MAG_AUTO. (See bottom right panel.)
  • All 3 models show negative biases vs. MAG_AUTO (i.e., in their median offsets), which is understood as MAG_AUTO is known to miss a fraction of the total flux (as well as missing proportionally more flux for fainter objects). The bias behavior is again very similar for the Sersic and bulge+disk models, while the disk-only model shows noticeably less bias vs. MAG_AUTO, probably indicating that the disk-only MAG_MODEL is missing some flux itself. (See bottom middle panel.)

Comparisons of model vs. model magnitudes, as a function of Sersic index

The top 2 panels in the figure below show scatter plots of the difference between the disk-only and Sersic model magnitudes (left) and between the bulge+disk and Sersic model magnitudes (right), as a function of the Sersic index from the Sersic model fit. Results are for i-band galaxies (same selection as before) with MAG_AUTO < 24. The bottom 2 panels then compare the median offsets (left) and +/- 34-percentile ranges. We again see that the Sersic and bulge+disk models give similar results, while the disk-only model is a bit biased and is fainter by about 0.1-0.2 mag in the median relative to the Sersic model, for Sersic index values of about 4 and above. Also, the disk-only model shows bigger scatter vs. the Sersic model for larger Sersic indices.

Comparisons of r-i DETMODEL colors, as a function of magnitude

The top 2 panels in the figure below show scatter plots of the difference between the disk-only and Sersic model r-i DETMODEL colors (left) and between the bulge+disk and Sersic DETMODEL colors (right), as a function of MAG_AUTO. Results are for i-band galaxies (same selection as before). The bottom 2 panels then compare the median offsets (left) and +/- 34-percentile ranges. We see that the the different models give very similar DETMODEL colors, with the main item of note being the slightly larger scatter of the disk-only vs. Sersic results at bright magnitudes MAG_AUTO < 20.