Morphology on large surveys with SExtractor + PSFEx

B.  Moraes,  M.  Elidaiana,  M.  Makler,  E.  Ber1n,  

and  the  CS82  collabora1on  

Aldée  Charbonnier    

Valongo  Observatory  Rio  de  Janeiro    

ESO,  Chile,  November  22th,  2013  

Outline  

SExtractor  +  PSFEx  CS82  survey  COSMOS  field  

General  procedure  

SExtractor:    hGp://www.astromaLc.net/soMware/sextractor  PSFEx:      hGp://www.astromaLc.net/soMware/psfex    

3  

SExtractor  v.1  

image   background   filtered  

apertures   segmentaLon  

4  

PSFEx  ID  card  

  Solving  in  direct  space  (≠  Fourier)    AutomaLcally  selects  the  point  like  sources    PSF  modeled  as  a  linear  combinaLon  of  basis  funcLons    Weighted  χ2  minimizaLon    PSF  varia1ons  may  be  decomposed  on  a  polynomial  basis    HomogenizaLon  is  available    

5  

PSF  modelling  

Automa1c  selec1on  of  point  

like  sources  

PSF  modeled  over  the    

field  of  view  

6  

PSFEx  checkplots  

Distribu1on  of  point  like  sources   FWHM  

+  others  (chi2,  ellip1city..)  

7  

SExtractor  v.i  

Automated  detecLon  of  objects  

2D  Model-‐fiUng  of  stars  and  galaxies  -‐  Point  sources  modeled  as  local  PSF  +  amplitude  -‐  Galaxies:  size  and  orienta1ons  convolved  with  the  PSF  -‐  Background  

  PSF  modeled  using  PSFEx    Pixel  grid  size  depends  on  sampling  and  on  the  object    Models:  background,  point  source,  de  Vaucouleurs,  Sérsic,  exponen1al    Levenberg-‐Marquardt  minimizaLon  algorithm    IniLal  parameter  guesses  based  on  SEx-‐v.1  measurements    Residuals  are  non  linear      chi2  modified  for  improved  robustness  

towards  outliers  

MulL  galaxy  fit/deblending  

BerLn  &  Arnouts  1996  

BerLn  (in  prep.)  

BerLn  (in  prep.)  

8  

SExtractor  v.2  sex -dp!

9  

SExtractor  v.2  

Residuals  Exponen1al  profile  CFHT/MegaCam  

Output  structural  parameters:  -‐  MAG_MODEL  -‐  REFF  -‐  ELLIPTICITY  -‐  errors  -‐  chi2  of  fijng  -‐  PSF  properLes  -‐  star/galaxy  

discriminators  

10  

Star/Galaxy  classifier  

CLASS_STAR   SPREAD_MODEL  

11  

Pros-‐cons  

  Fast    SExtractor  and  PSFEx  compaLble    Robust  PSF  modeling    developed  for  large  area    tested  on  simulaLons  (DES  …)    tested  on  true  data  (BCS,  CS82  …)    

  background  treatment                  (e.g.  GALAPAGOS  

 Barden  et  al.  2012)    brightness  profiles  available  

12  

CFHT  –  ‘Stripe  82’  data  

CS82  survey  in  a  nutshell    France-‐Canada-‐Brazil  project    CFHT,  MegaCam      173  science  quality  poinLngs  with  slight  overlaps    ‘Stripe  82’    goal:  weak  and  strong  lensing  survey    i-‐band    service  mode  insuring  good  seeing  condiLon  

CS82  observing  condi1ons  

  CS82  mean  seeing  =  0.6’’    Depth:  ~24.1  in  2-‐arcsec  circular  apertures    Source  number  density:  ~17.25  per  arcmin2  (standard  cuts)    

15  

CS82  first  results   16  

Morphology  catalog  produc1on    ~  16,700,000  objects  (~  150,000/1le)    Deeper  than  SDSS    Choice  of  configura1on    

 v  deblending  v  deepness  v  computaLon  Lme  v  size  of  catalogs   4  profiles:  de  Vaucouleurs,  exponenLal,  Sérsic,  deV+exp    177  Lles  x  4  profiles  x  ~  12h/profile/Lle    Masking    Cuts:  S/N,  flagged  objects,  star/galaxy  

Aspect  ra1o  before  …   …  and  ajer  model-‐fiUng  

0   0  1   1  

Moraes  et  al.  in  prep  

17  

SEx-‐PSFEx  configura1ons  

VIGNET  image  cut  size  

es1ma1on  of  the  internal  sta1s1cal  errors  due  to  the  choice  of  the  configura1ons  

PSF_SIZE  Image  size  of  the  PSF  

Modeled  magnitude  (deV)   Effec1ve  radius  (deV)  

MINSN  PSF  star  selecLon   …  

18  

SPREAD_MODEL  as  a  sanity  check  

  coadds:  different  seeings,  bad  es1ma1on  of  the  PSF    Point  like  sources  selec1on  crucial  

19  

Lensfit:  ellip1city  comparison  

Lensfit:    Miller  et  al.  2007  

Effec1ve  radius   S/N  

Magnitude  

Very  good  reconstruc1on  (single  exposure  vs  coadds)  

20  

Comparison  with  SDSS:  exp  

Galaxies  –  S/N  >  5  –  SDSS  DR9:  fracDeV_i  <  0.02  (matching)  Area  of  14  sq.  degrees      

Good  agreement  for  MAG  <  21  

21  

Comparison  with  SDSS:  deV  

Galaxies  –  S/N  >  5  –  SDSS  DR9:  fracDeV_i  >  0.98  (matching)  Area  of  14  sq.  degrees      

Good  agreement  for  MAG  <  21  

22  

BOSS/CS82  galaxies  

SDSS  BOSS  sample  

14  sq.  degrees  

Bias  in  the  BOSS  sample  

23  

COSMOS  field  

The  COSMOS  field  CFHT  Deep2  field  -‐  MegaCam  (y  band)   Hubble  COSMOS  -‐  ACS  (I  band)  

seeing  ~  1’’   seeing  ~  0.09’’  

25  

CFHT-‐D2:  catalog  produc1on  

#  objects  aMer  masking  and  S/N  >  5:  -‐  point-‐like:  ~  38,700  -‐  extended:  ~  174,200  

SExtractor  run:  ~  30.5  h  Same  diagnosLcs  as  CS82  

Chi  2  –  PSF  modelling  SPREAD_MODEL  

26  

Catalogs  comparison  

•  COSMOS  catalog:  produced  with  GIM2D  •  CFHT  D2  catalog:  our  pipeline  •  Objects  selec1on:  eliminaLon  of  stars,  good  quality  fit  

Good  agreement  

27  

Catalogs  comparison  

  Ellip1city  and  effec1ve  radius  similar    The  PSF  is  correctly  reconstructed  

Ellip1city   Effec1ve  radius  

28  

Conclusions:  2  points  

SExtractor  +  PSFEx:    EsLmaLon  of  the  PSF,  variaLon  over  the  field  of  view    The  model-‐fijng  is  working  well    Very  compeLLve  in  term  of  computaLon  Lme    Extensive  tests  to  assess  its  quality      CFHT  –  Stripe  82  data:    Data  of  very  good  quality  (mean  seeing  0.6”)  deep  (mag  limit  ~  24)    Originally  a  lensing  survey    Morpholgy  catalogs  produced  (deV,  exp,  Sersic,  deV+exp)    RedshiM  z=[0.2;0.7]    Galaxy  evoluLon/morphology  studies  to  come…  

kecélebackup  

The  ‘Stripe  82’  

40 30 20 10 0RA

1

0

1

DEC

S82m0m

S82m0p

S82m10m

S82m11m

S82m11p

S82m12m

S82m12p

S82m13m

S82m13p

S82m14m

S82m14p

S82m15m

S82m15p

S82m16m

S82m16p

S82m17m

S82m17p

S82m18m

S82m18p

S82m19p

S82m1m

S82m1p

S82m20p

S82m21m

S82m21p

S82m22m

S82m22p

S82m23m

S82m23p

S82m24m

S82m24p

S82m25m

S82m25p

S82m26m

S82m26p

S82m27m

S82m27p

S82m28m

S82m28p

S82m29m

S82m29p

S82m2m

S82m2p

S82m30m

S82m30p

S82m31m

S82m31p

S82m32m

S82m32p

S82m33m

S82m33p

S82m34m

S82m34p

S82m35m

S82m35p

S82m36m

S82m36p

S82m37m

S82m38m

S82m39m

S82m3m

S82m3p

S82m4m

S82m4p

S82m5m

S82m5p

S82m6m

S82m7m

S82m7p

S82m8m

S82m8p

S82m9m

W4m0m1

W4m0m2

W4m1m1

W4m1m2

W4p1m1

W4p1m2

W4p2m1

W4p2m2

0 10 20 30 40RA

1

0

1

DEC

S82p0m

S82p0p

S82p10m

S82p10p

S82p11m

S82p11p

S82p12m

S82p12p

S82p13m

S82p13p

S82p14m

S82p14p

S82p15m

S82p15p

S82p16m

S82p16p

S82p17m

S82p17p

S82p18m

S82p18p

S82p19m

S82p19p

S82p1m

S82p1p

S82p20m

S82p20p

S82p21m

S82p21p

S82p22m

S82p22p

S82p23m

S82p23p

S82p24m

S82p24p

S82p25m

S82p25p

S82p26m

S82p26p

S82p27m

S82p27p

S82p28m

S82p28p

S82p29m

S82p29p

S82p2m

S82p2p

S82p30m

S82p30p

S82p31m

S82p31p

S82p32m

S82p32p

S82p33m

S82p33p

S82p34m

S82p34p

S82p35m

S82p35p

S82p36m

S82p36p

S82p37m

S82p37p

S82p38m

S82p38p

S82p39mS82p39p

S82p3m

S82p3p

S82p40m

S82p40p

S82p41m

S82p41p

S82p42m

S82p42p

S82p43m

S82p43p

S82p44p

S82p45p

S82p46p

S82p47p

S82p48p

S82p4m

S82p4p

S82p5m

S82p5p

S82p6m

S82p6p

S82p7m

S82p7p

S82p8m

S82p8p

S82p9m

S82p9p

Strategy  of  observa1on  

  -‐40°  <  RA  <  +45°  and  -‐1°  <  Dec  <  +1°      avoiding  massive  stars  

CS82  magnitudes  

MAG_AUTO  –  MAG_PSF   MAG_MODEL  –  MAG_PSF  

CalibraLon  of  the  data  done  with  MAG_AUTO  MAG_AUTO  is  missing  flux  at  the  wings  (Annunziatella  et  al.  2012)                                      offset  MAG_AUTO  –  MAG_PSF  observed  

CS82/SDSS  

Point  like  sources  +  exp/deV  profiles:  offset  due  to  the  calibraLon  with  MAG_AUTO    deV  profile:  addiLonal  offset  due  to  the  truncaLon  of  the  profile  by  the  SDSS  pipeline  (info.  By  Claire  Lackner)  

Ellip1city:  Lensfit  vs  Sextractor  

Sersic   De  Vaucouleurs   exponen1al  

Lensfit  

SExtractor  

Comparison  on  CS82  1  1le  (p28m)