Updates on the Fake Object Pipeline for HSC Survey
- 1. Updates on Fake Object Pipeline Song Huang (Kavli-IPMU) Alexie Leauthaud, Surhud More, Ryoma Murata, Claire Lackner, & Steve Bickerton Special Thanks to the Pipeline Team
- 3. Fake Galaxy Tests by DES Team http://arxiv.org/pdf/1507.08336v1.pdf http://arxiv.org/pdf/1411.0032v2.pdf ufig+SExtractor Balrog https://github.com/emhuff/Balrog
- 4. We are on GitHub https://github.com/clackner2007/fake-sources
- 5. https://github.com/clackner2007/fake-sources/blob/master/ doc/fakePipe_manual_v0.3.pdf We have….something that looks like a manual
- 6. Log of Recent Tests
- 7. Some Technical Details • Image simulation using GalSim • Inject fake object onto single CCD image; Then run hscPipe on these images. • Can simulate: Star (using pipeline PSF); Galaxies using Exponential/de Vaucouleurs/Sersic/Double Sersic models; Can add g1/g2 shear; • Image of fake galaxy is truncated at 10xRe; It will be cropped at the edge of each CCD; Random Gaussian noise is added. • Can help you generate random locations within a Tract; Visit; or a (RA, DEC) range; Or help you put fakes close to real objects. • Will add a FAKE mask plane to the image
- 8. What You Need Before Running It • Have access to computer with hscPipe installed; to HSC data reduced by the pipeline. • Besides GalSim, also depends on Astropy. • Prepare an input catalog with RA, DEC, and model information for each band (Be creative) • Select the Tract/Patch you wan to work on; And find out which Visits contribute to it. You Need to Remember: • The test is done using **Perfect** PSF and sky background. • Since we have a lot of **fake** objects on the images, we skip the stack.py (uber-calibration) phase.
- 9. Model Selection Mag Re (arcsec) Sersic n b/a
- 10. Filter the Fake Objects Through Mask Planes With BRIGHT_OBJECT Without BRIGHT_OBJECT It will be much easier to do this with the Random Data
- 11. Typical Number of Visits in one WIDE Tract Create a lot of data…I just deleted all the single Visit data after the test is done
- 12. The FAKE Mask Plane on Coadd Images
- 13. Tests using hscPipe 4.0.1 • Based on Yasuda-san’s recent DR_S16A rerun on Master@IPMU • Only using data in HSC-I, HSC-G, HSC-R • Using Tract=8766/8767 in WIDE/XMM-LSS • Input catalog: Claire’s COSMOS catalog down to I_F814=25.2 mag • Assuming (G-I) = 0.5 mag & (R-I) - 0.0 mag • Adding 300-400 fake galaxies per Patch • 2 additional tests: BRIGHT (mag < 21.5); BLENDED (Fakes close to real objects)
- 14. Star-Galaxy Separation Magnitude difference between PSF and cModel magnitudes ~3% of fakes are classified as star: (classification.extendedness == 0)
- 15. “Blendedness” see Ryoma Murata’s talk for more details Behaviour consistent with real objects
- 16. cModel Magnitude in HSC-I cModel fails for ~0.5% of fakes
- 17. cModel Magnitude v.s. log(cModel (S/N)) Input Magnitude v.s. log(cModel (S/N)) Highly blended objects have lower S/N, as expected
- 19. Kron Magnitude in HSC-I Kron fails for ~4% of fakes
- 20. “Blendedness” and Magnitude Error
- 21. (G-I) color cModel magnitude Forced photometry
- 22. Most of these faint ones have b/a = 1.0 The open histograms are for galaxies brighter than 21.5 mag ExpModelDevModel
- 25. Next Steps • Make it more efficient…with the help from the pipeline team. • Realistic SED/colors for multiband tests. • More complicated models / Drop outs? • Anything you want!
- 26. Backup Slides
- 27. Relative difference of Re cModel.ellipse.A The open histograms are for galaxies brighter than 21.5 mag
- 28. Relative difference of Re Flux.Kron.Radius The open histograms are for galaxies brighter than 21.5 mag
- 29. Basic Photometry Star-Galaxy Separation …..Everything
- 30. Remove the “Diffuse” Objects But……