In Pursuit of LSST Science Requirements: A Comparison of Photometry Algorithms

Abstract

We have developed an end-to-end photometric data processing pipeline to compare current photometric algorithms commonly used on ground-based imaging data. This testbed is exceedingly adaptable, and enables us to perform many research and development tasks, including image subtraction and co-addition, object detection and measurements, the production of photometric catalogs, and the creation and stocking of database tables with time-series information. This testing has been undertaken to evaluate existing photometry algorithms for consideration by a next-generation image processing pipeline for the Large Synoptic Survey Telescope (LSST). We outline the results of our tests for four packages: The Sloan Digital Sky Survey's (SDSS) Photo package, Daophot and Allframe, DoPhot, and two versions of Source Extractor (SExtractor). The ability of these algorithms to perform point-source photometry, astrometry, shape measurements, star-galaxy separation, and to measure objects at low signal-to-noise is quantified. We also perform a detailed crowded field comparison of Daophot and Allframe, and profile the speed and memory requirements in detail for SExtractor. We find that both Daophot and Photo are able to perform aperture photometry to high enough precision to meet LSST's science requirements, and less adequately at PSF-fitting photometry. Photo performs the best at simultaneous point and extended-source shape and brightness measurements. SExtractor is the fastest algorithm, and recent upgrades in the software yield high-quality centroid and shape measurements with little bias towards faint magnitudes. Allframe yields the best photometric results in crowded fields.