GEMS: Galaxy Fitting Catalogs and Testing Parametric Galaxy Fitting Codes: GALFIT and GIM2D
Häussler, Boris; McIntosh, Daniel H.; Barden, Marco;Bell, Eric F.; Rix, Hans-Walter; Borch, Andrea;Beckwith, Steven V. W.; Caldwell, John A. R.;Heymans, Catherine; Jahnke, Knud; Jogee, Shardha;Koposov, Sergey E.; Meisenheimer, Klaus;Sánchez, Sebastian F.; Somerville, Rachel S.;Wisotzki, Lutz; Wolf, Christian. GEMS: Galaxy Fitting Catalogs and Testing Parametric Galaxy Fitting Codes: GALFIT and GIM2D. The Astrophysical Journal Supplement Series. 2007, Vol. Volume 172, Issue 2, pp. 615-633, p. -2007.
In the context of measuring the structures of intermediate-redshift galaxies with HST ACS surveys, we tune, test, and compare two widely used fitting codes (GALFIT and GIM2D) for fitting single-component Sérsic models to both simulated and real galaxy data. Our study focuses on the GEMS survey with the sensitivity of typical HST survey data, and we include our final catalog of fit results for all 41,495 objects detected in GEMS. Using simulations, we find that fitting accuracy depends sensitively on galaxy profile shape. Exponential disks are well fit and have small measurement errors, whereas fits to de Vaucouleurs profiles show larger uncertainties owing to the large amount of light at large radii. Both codes provide reliable fits with little systematic error for galaxies with effective surface brightnesses brighter than that of the sky; the formal uncertainties returned by these codes significantly underestimate the true uncertainties (as estimated using the simulations). We find that GIM2D suffers significant systematic errors for spheroids with close companions owing to the difficulty of effectively masking out neighboring galaxy light; there appears to be no work-around to this important systematic in GIM2D's current implementation. While this crowding error affects only a small fraction of galaxies in GEMS, it must be accounted for in the analysis of deeper cosmological images or of more crowded fields with GIM2D. In contrast, GALFIT results are robust to the presence of neighbors because it can simultaneously fit the profiles of multiple companions as well as the galaxy of interest. We find GALFIT's robustness to nearby companions and factor of >~20 faster runtime speed are important advantages over GIM2D for analyzing large HST ACS data sets.
In the context of measuring the structures of intermediate-redshift galaxies with HST ACS surveys, we tune, test, and compare two widely used fitting codes (GALFIT and GIM2D) for fitting single-component Sérsic models to both simulated and real galaxy data. Our study focuses on the GEMS survey with the sensitivity of typical HST survey data, and we include our final catalog of fit results for all 41,495 objects detected in GEMS. Using simulations, we find that fitting accuracy depends sensitively on galaxy profile shape. Exponential disks are well fit and have small measurement errors, whereas fits to de Vaucouleurs profiles show larger uncertainties owing to the large amount of light at large radii. Both codes provide reliable fits with little systematic error for galaxies with effective surface brightnesses brighter than that of the sky; the formal uncertainties returned by these codes significantly underestimate the true uncertainties (as estimated using the simulations). We find that GIM2D suffers significant systematic errors for spheroids with close companions owing to the difficulty of effectively masking out neighboring galaxy light; there appears to be no work-around to this important systematic in GIM2D's current implementation. While this crowding error affects only a small fraction of galaxies in GEMS, it must be accounted for in the analysis of deeper cosmological images or of more crowded fields with GIM2D. In contrast, GALFIT results are robust to the presence of neighbors because it can simultaneously fit the profiles of multiple companions as well as the galaxy of interest. We find GALFIT's robustness to nearby companions and factor of >~20 faster runtime speed are important advantages over GIM2D for analyzing large HST ACS data sets.