PaperlayerView exact revision on arXiv

Source-linked AI summary

GALAPAGOS: From Pixels to Parameters

Marco Barden, Boris Häußler, Chien Y. Peng, Daniel H. McIntosh, Yicheng Guo

arXiv:1203.1831v1astro-ph.IMastro-ph.CO

TL;DR

Large imaging surveys require many coordinated steps to produce galaxy structural catalogues, including detection, fitting, sky estimation, and duplicate handling. GALAPAGOS integrates these operations using SExtractor and Galfit, and simulations find it extremely robust for parameter recovery, although outcomes depend on the input setup and some stellar cases remain difficult.

  • Problem

    Producing galaxy-parameter catalogues from complete imaging surveys requires coordinating detection, fitting, sky estimation, catalogue assembly, and duplicate-source handling.

  • Method

    GALAPAGOS integrates SExtractor detection, Sersic fitting with Galfit, flux-growth-curve sky estimation, catalogue compilation, and ranked processing from a single setup.

  • Results

    GALAPAGOS is extremely robust in recovering structural parameters in extensive simulations.

  • Takeaways & Limitations

    The pipeline provides an automated route from survey images to fitting catalogues while combining complementary cold and hot source detections.

  • Takeaways & Limitations

    Results depend critically on user-defined parameters, and fitting saturated stars with Sersic models may prevent Galfit from converging within a reasonable time.

Abstract

from arXiv · show

To automate source detection, two-dimensional light-profile Sersic modelling and catalogue compilation in large survey applications, we introduce a new code GALAPAGOS, Galaxy Analysis over Large Areas: Parameter Assessment by GALFITting Objects from SExtractor. Based on a single setup, GALAPAGOS can process a complete set of survey images. It detects sources in the data, estimates a local sky background, cuts postage stamp images for all sources, prepares object masks, performs Sersic fitting including neighbours and compiles all objects in a final output catalogue. For the initial source detection GALAPAGOS applies SExtractor, while GALFIT is incorporated for modelling Sersic profiles. It measures the background sky involved in the Sersic fitting by means of a flux growth curve. GALAPAGOS determines postage stamp sizes based on SExtractor shape parameters. In order to obtain precise model parameters GALAPAGOS incorporates a complex sorting mechanism and makes use of modern CPU's multiplexing capabilities. It combines SExtractor and GALFIT data in a single output table. When incorporating information from overlapping tiles, GALAPAGOS automatically removes multiple entries from identical sources. GALAPAGOS is programmed in the Interactive Data Language, IDL. We test the stability and the ability to properly recover structural parameters extensively with artificial image simulations. Moreover, we apply GALAPAGOS successfully to the STAGES data set. For one-orbit HST data, a single 2.2 GHz CPU processes about 1000 primary sources per 24 hours. Note that GALAPAGOS results depend critically on the user-defined parameter setup. This paper provides useful guidelines to help the user make sensible choices.

1 INTRODUCTION

GALAPAGOS automates the survey-to-catalogue workflow by combining SExtractor detection, postage-stamp preparation, Galfit Sersic modelling, and catalogue compilation. Its modular design supports independent processing stages and incorporates user-provided PSF information and sky estimation.

  • Pipeline overview: SExtractor detects sources and supplies shape information, while Galfit models two-dimensional galaxy light profiles using Sersic and other supported components.Galfit can fit neighbouring sources simultaneously and convolves model profiles with the supplied PSF.
  • Pipeline overview: GALAPAGOS automates source detection, postage-stamp creation, sky estimation, Galfit fitting, and catalogue compilation from one setup.The pipeline is divided into four independently executable blocks, allowing selected stages to be repeated without rerunning the entire analysis.
  • Pipeline overview: The code requires a user-provided, background-subtracted, normalized PSF image with sufficient signal-to-noise and untruncated features.GALAPAGOS itself does not create the PSF image required by Galfit.
  • Pipeline overview: Postage-stamp dimensions are estimated from SExtractor parameters before Galfit fits each extracted object.The fitting stage also requires identifying neighbours, selecting initial parameter guesses, and preparing the sky background.
  • Pipeline overview: GALAPAGOS measures sky background from the full survey image with a flux-growth curve rather than fitting sky as a free Galfit parameter.This avoids unreliable sky estimates when neighbouring contamination is present and reduces fit freedom.

3 COMPONENTS

The component methods address source extraction across wide dynamic ranges and organize the resulting detections for downstream analysis. GALAPAGOS uses SExtractor-derived Kron measurements and combines complementary cold and hot catalogues.

  • 3.1 SExtractor: GALAPAGOS combines cold and hot SExtractor catalogues to retain well-deblended bright sources while recovering fainter detections.Cold detections are imported first, and hot detections inside enlarged cold-source Kron ellipses are discarded.
  • 3.1 SExtractor: Figure 2 illustrates the catalogue-combination rule with cold detections on the left, hot detections on the right, and retained objects marked by arrows.Kron ellipses define the spatial test used to exclude redundant hot detections.
  • 3.1 SExtractor: SExtractor Kron radii provide source extents that GALAPAGOS uses to estimate postage-stamp sizes and assess which image pixels are affected by source light.For stars and galaxies convolved with Gaussian seeing, the Kron radius encircles about 90% of the flux.
  • 3.1 SExtractor: Wide surveys challenge SExtractor because sources span a broad range of sizes and luminosities, forcing a trade-off between deblending bright objects and detecting faint ones.Classic pencil-beam settings can split large bright galaxies, whereas wide-area settings may sacrifice depth.

HDR SExtraction

GALAPAGOS combines cold and hot SExtractor runs to detect bright and faint sources, then compiles and fits a deduplicated catalogue across survey tiles.

  • HDR SExtraction: Cold sources enter the catalogue first, while hot detections inside enlarged cold Kron ellipses are discarded as likely duplicates.Parameter B09 controls the enlargement of cold Kron ellipses; B09=1.1 enlarges them by 10%.
  • HDR SExtraction: HDR SExtraction combines a cold run that properly deblends bright objects with a hot run tuned to detect fainter sources.The resulting mode is called High Dynamic Range (HDR) SExtraction.
  • Catalogue Compilation: Catalogue combination uses image coordinates, source-to-boundary distances, and nonzero-flux image borders to remove duplicate detections across overlapping tiles.This boundary-based procedure also accommodates nonrectangular images such as HST data.
  • Catalogue Compilation: For sources larger than the overlap region, the procedure can remove legitimate detections or spurious deblends, although this ambiguity rarely occurs in practice.The problem is avoided completely when the largest survey source is smaller than the tile overlap.
  • Postage Stamps: Postage stamps reduce fitting time and memory use, with sizes based on enlarged Kron radii while neighbouring galaxies remain included in the fit.For one-orbit HST surveys, a scaling factor of 2.5 worked well.
  • Sky Estimation: Galapagos estimates a fixed sky background before fitting because insufficient empty sky in a postage stamp can bias fitted flux and structural parameters.Its flux-growth method averages elliptical annuli while excluding other sources and stops when recent slope measurements level off.

Simultaneous Fits

GALAPAGOS fits nearby sources together when their light profiles interact, while adapting parameters for sources whose influence or position makes unrestricted fitting unsuitable.

  • Simultaneous Fits: Sources may require simultaneous fitting because nearby objects can influence one another’s recovered parameters.A source that is tertiary relative to one primary can still materially affect a secondary, motivating refitting the secondary as a primary.
  • Simultaneous Fits: Previously fitted neighbours can be inserted as fixed components, with coordinates transformed through world-coordinate information when fits come from another image.This exception accelerates fitting while preserving the neighbour’s location in the current pixel system.
  • Simultaneous Fits: When a secondary lies partly outside the postage stamp and lacks a prior fit, GALAPAGOS fixes its position, axis ratio, and position angle while freeing n and Re.This reduces unnecessary degrees of freedom using SExtractor-provided shape information.
  • Simultaneous Fits: Bright, large contributors are treated as secondary sources even when they lie off the current survey image or lack overlapping Kron ellipses.They receive the same parameter treatment as ordinary secondary sources.
  • Simultaneous Fits: Tertiary-source masks can be partial when a tertiary overlaps a secondary, or complete when it overlaps neither the primary nor a secondary.The masking scheme distinguishes which portions of tertiary sources remain relevant to the fit.

Bad Pixel Masks

GALAPAGOS uses bad-pixel masks to exclude tertiary and otherwise unsuitable image regions from GALFIT, while preserving relevant overlaps with fitted sources.

  • Bad Pixel Masks: The masking procedure enlarges each tertiary’s Kron ellipse by a user-specified factor and offset before excluding it from the fit.The enlargement factor may differ from the one used for the sky map.
  • Bad Pixel Masks: GALAPAGOS masks secondary objects downgraded by magnitude as tertiaries, but preserves portions overlapping the primary or other secondaries.The criterion differs for extended and point sources and is controlled by user parameters.
  • Bad Pixel Masks: Pixels with zero weight-map values are also masked because they correspond to regions with zero exposure time.These rules produce the bad-pixel masks illustrated in Figure 10.

Parameter Constraints

GALAPAGOS stabilizes Sersic fitting by constraining GALFIT parameters to physically or procedurally acceptable ranges, including consistency with SExtractor magnitudes.

  • Parameter Constraints: GALFIT supports absolute, relative, and inter-parameter constraints, such as restricting Sersic indices to an acceptable range.These constraints can relate parameters to supplied input values or to other fitted components.
  • Parameter Constraints: GALAPAGOS imposes 0.2 < n < 8.0, 0.3 < Re < E11, and E12 < mGALFIT − mSExtractor < E13 as global fitting constraints.The bounds on n and Re, together with the magnitude-difference limit, are intended to stabilize Sersic fitting.
  • Parameter Constraints: These parameter restrictions are part of GALAPAGOS characteristics designed to increase fitting efficiency and robustness.The paper introduces them alongside other mechanisms for improving execution and stability.

Sorting and Parallel Computation

GALAPAGOS orders fits from bright to faint and combines dependency-aware sequential processing with parallel execution across sufficiently separated sources and survey tiles.

  • Sorting and Parallel Computation: The pipeline detects sources, cuts postage stamps, estimates sky and runs GALFIT, then compiles the galaxy catalogue.These stages organize the transition from source extraction to parallelized fitting and final catalogue production.
  • Sorting and Parallel Computation: Bright-to-faint ordering lets GALAPAGOS omit negligible faint neighbours from bright-source fits and reuse fitted bright-neighbour parameters for fainter sources.Holding known neighbour parameters fixed reduces degrees of freedom and substantially accelerates large catalogues.
  • Sorting and Parallel Computation: Sources sufficiently far apart in the processing queue can start as separate processes on different CPU cores without materially affecting one another.This recovers parallelism while retaining the rank-ordered fitting strategy.
  • Sorting and Parallel Computation: GALAPAGOS also parallelises processing across survey tiles by running multiple code instances on different computers, reducing total time by a factor of n.A batch file specifies which tiles each instance processes, although neighbouring-tile dependencies must be handled.
  • Sorting and Parallel Computation: The fitting workflow is split into a dependency-aware first stage and a tile-level second stage that can run images out of order and simultaneously.The first stage processes sources spanning tile overlaps; the second processes each tile in decreasing brightness.
  • Sorting and Parallel Computation: Neighbouring tiles are defined by the n closest survey images, with more distant tiles included at survey edges.For n = 8, the immediate neighbourhood forms a 3 × 3 checkerboard pattern.
  • Sorting and Parallel Computation: Sequential processing incurs overhead because complete science, weight, and segmentation images may be repeatedly loaded for sky fitting.The relevant sequential fraction can be defined by sources spanning areas larger than the survey-tile overlap.

Neighbouring Tiles

GALAPAGOS manages neighbouring sources and detection failures by limiting relevant contributors, classifying errors by their effect on nearby fits, and optionally removing them at appropriate pipeline stages.

  • Neighbouring Tiles: GALAPAGOS evaluates only nearby sources that can affect the current object's fitting or background estimation, reducing unnecessary contributor checks.Users can limit the reach of bright sources and select a specified number of neighbouring tiles for calculations.
  • Detection Failures: Detection failures are divided into critical errors removed before fitting and catalogue errors removed afterward because they can influence neighbouring fits.Critical failures include over-deblends, cosmic rays, and edge detections; catalogue failures include bright image defects such as diffraction spikes and bleeding.
  • Detection Failures: GALAPAGOS can remove user-supplied critical or catalogue failures at specified stages, while untreated failures remain modelled as ordinary sources.Most detection errors are faint enough not to influence neighbours and are therefore usually classified as critical failures.
  • Detection Failures: The critical-versus-catalogue classification is subjective, and omitting manual correction can leave a small fraction of sources affected.The pipeline still runs without manual treatment, but users must accept that residual effects may remain.

Treatment of Stars

GALAPAGOS addresses difficult stellar and galaxy-fitting cases through saturated-star handling, selective neighbour fitting, and extensive simulations of parameter, sky, and neighbour recovery.

  • Treatment of Stars: Bright saturated stars are modelled with high-index Sersic profiles instead of ordinary PSFs to represent their extended tails, although convergence may still take too long.The paper does not pursue a more elaborate stellar model because GALAPAGOS focuses on galaxy properties.
  • Treatment of Stars: Restricting fitted secondary sources to objects bright enough to influence the fit resolves convergence problems around saturated stars.For STAGES stars, objects more than two magnitudes fainter than the primary were treated as tertiaries rather than fitted secondaries.
  • Simulation Results: Around 3.7 million simulated galaxies were successfully fitted after excluding undetected objects, constrained fits, and crashed fits.The simulations contained about 7 million galaxies; approximately 3 million fell below the detection threshold, about 280,000 hit fitting constraints, and 293 fits crashed.
  • Simulation Results: Parameter recovery worsens for faint galaxies and higher Sersic indices, while the brightest galaxies also show increased deviations.The simulations examine magnitude, effective radius, and Sersic index across surface-brightness and magnitude regimes.
  • Sky Recovery: GALAPAGOS recovers local sky values with very small offset and scatter, largely independent of Sersic index and substantially better than SExtractor.For bright n=3–5 galaxies, GALAPAGOS deviations are about 0.04/0.03 with scatter about 0.02, compared with SExtractor values around 2.3/1.3 and scatter about 0.4.
  • Neighbour Effects: Recovered parameters show no dependence on neighbour brightness or distance in the tested simulations, supporting the applied deblending and fitting scheme.The analysis separates contamination by bright neighbours from contamination by close neighbours.

5 PERFORMANCE

GALAPAGOS was evaluated on the 80-tile STAGES survey, where Galfit fitting dominated processing time and parallel execution handled the workload.

  • 5 PERFORMANCE: 430 hours of wall-clock time were required to process the 80-tile STAGES mosaic containing approximately 75,000 sources.The data include large, peculiar, and challenging objects for source extraction and Galfit modelling.
  • 5 PERFORMANCE: Approximately 260 hours were spent fitting the fainter 95% of sources using eight 2.2 GHz CPU cores in parallel.The reported duration is set by the slowest of the eight parallel processes.
  • 5 PERFORMANCE: Approximately 150 hours were spent fitting the brightest 5% using four 2.4 GHz CPU cores.Bright objects take considerably longer to fit, motivating separate treatment of the brightest sources.
  • 5 PERFORMANCE: 13.5 hours were required for SExtractor, 2.5 hours for postage-stamp creation, and 0.7 hours for catalogue compilation.These stages consumed an almost negligible fraction of total processing time compared with Galfit fitting.
  • 5 PERFORMANCE: Reported processing times exclude setup and parallel-fitting preparation overheads, and stage fractions can vary substantially with survey layout and configuration.The authors caution that the timing profile is not universal across survey designs.

6 SUMMARY

GALAPAGOS automates source detection, single-Sérsic modelling, and catalogue construction for survey data, with simulations indicating robust parameter recovery. Its results remain sensitive to user-defined setup choices, and future work targets richer structural models and spatially varying PSFs.

  • 6 SUMMARY: GALAPAGOS automates source detection and single-Sérsic modelling while combining results into a duplicate-cleaned catalogue.It incorporates SExtractor and Galfit, supports HDR extraction, cuts postage stamps, estimates local sky, and uses multi-core processing.
  • 6 SUMMARY: GALAPAGOS is intended for survey applications ranging from single-tile deep observations to wide-area shallow surveys.Galfit’s support for user-supplied PSFs permits use with both space- and ground-based data.
  • 6 SUMMARY: Extensive simulations found GALAPAGOS extremely robust in recovering structural parameters.The reported fitting quality depends on the chosen input parameters, including the SExtractor setup.
  • 6 SUMMARY: Future development includes consistent two-component bulge-disc fitting, automated bar and Fourier-mode fitting, and potentially variable PSFs.The current implementation uses one PSF for convolving model profiles across the survey.
  • 6 SUMMARY: GALAPAGOS is freely available for download from the authors’ webpage.The paper identifies the software as an openly downloadable implementation.

APPENDIX A: CODE SETUP AND CONTROL

The appendix describes GALAPAGOS’s script-driven setup, including file-location definitions, SExtractor configuration, optional bad-detection removal, and parallel-processing controls.

  • APPENDIX A: CODE SETUP AND CONTROL: GALAPAGOS is controlled by scripts whose startup file manages execution and divides parameters into six sections, A through F.Static sections define file locations and Galfit options, while the remaining sections control other processing stages.
  • APPENDIX A: CODE SETUP AND CONTROL: The setup examples include SExtractor convolution configuration and file-location declarations used by the startup scripts.The appendix shows the relevant configuration snippets rather than only describing them procedurally.
  • APPENDIX A: CODE SETUP AND CONTROL: The file list records each tile’s science image, weight image, output directory, and output-file prefix.The example uses one row per survey pointing and supports a hypothetical ten-tile survey.
  • APPENDIX A: CODE SETUP AND CONTROL: Users can refine SExtractor output by identifying bad detections, creating tile-and-position lists, and rerunning the relevant processing block.Galapagos rejects detections within a specified radius around listed positions on the subsequent run.
  • APPENDIX A: CODE SETUP AND CONTROL: Parallel processing is configured by supplying tile lists so separate computers can process different survey sections after postage-stamp preparation and bright-source fitting.The appendix illustrates tile lists containing the filenames assigned to each pipeline.

A5 An Example Sequence

A complete GALAPAGOS run proceeds from startup and file-list setup through source extraction, postage-stamp preparation, staged Galfit fitting, parallel batches, and final catalogue compilation.

  • A5 An Example Sequence: The complete sequence begins by setting up the startup script and file list, then running SExtractor with optional HDR mode.Users may identify and manually list bad detections before rerunning the extraction block.
  • A5 An Example Sequence: After extraction, GALAPAGOS cuts postage stamps, fits the brightest galaxies, creates batch lists and scripts, and reruns Galfit in parallel.The example file list specifies science images, weight images, output directories, and output prefixes for ten tiles.
  • A5 An Example Sequence: The final step is to run catalogue compilation after parallel processing finishes.The sequence places catalogue compilation after all distributed Galfit jobs are complete.
  • A5 An Example Sequence: The setup examples show a startup script and a file-location list for a survey with ten tiles.The file-location list uses four columns for science image, weight image, output directory, and output prefix.
  • A5 An Example Sequence: When the survey is small enough, postage-stamp preparation and later fitting steps can be combined by setting the brightest-galaxy fraction to 100%.Alternatively, one batch file can contain all tiles, although the authors recommend that option mainly for testing.

APPENDIX B: STARTING PARAMETERS

Appendix B documents GALAPAGOS’s configurable startup parameters for source detection, postage-stamp creation, sky preparation, fitting, parallel processing, and catalogue output.

  • Parameter blocks: GALAPAGOS’s startup file controls executable paths, input/output locations, detection settings, postage-stamp generation, sky preparation, and catalogue construction.Blocks B, C, D, E, and F respectively configure SExtractor, postage stamps, sky maps, Galfit-related files, and combined catalogues.
  • Postage stamps: Postage-stamp dimensions are calculated by enlarging SExtractor isophotes with a configurable scale factor.The C03 parameter specifies the enlargement applied to Kron ellipses when calculating stamp sizes.
  • Sky preparation: Sky preparation uses configurable isophote enlargements, offsets, distances, and source-contribution thresholds to construct the sky map.Parameters D03–D10 define sky-map scaling, neighbouring-source treatment, sky-isophote geometry, and allowed contributing sources.
  • Model fitting: The fitting configuration controls Galfit output naming, PSF and mask files, constraint files, neighbouring tiles, and parallel-process limits.Output filenames combine a global prefix, configurable object or mask prefixes, and the SExtractor detection number; Galfit outputs contain image, model, and residual extensions.
  • Catalogue output: The final catalogue block reads Galfit header results into the source catalogue and writes the combined FITS output after removing additional bad detections.The combined SExtractor catalogue and final FITS catalogue are placed in the configured catalogue directory.
Loading 1203.1831v1…