Compact Source Investigation

1
Compact Source Investigation

• Graça Rocha
• IPAC/Caltech

2
CSI

• CSI - Compact Source Investigation - a
  collaboration initiated in 2006 by G. Rocha
  Pasadena (IPAC) Cambridge Santander Paris
• Coordination/collaboration on work for the ERCSC
  (Early Release Compact Source Catalog) and the
  Planck Legacy Science Catalog
• Help build the pipeline for the ERCSC
• Comparison of performance of source detection
  algorithms - a 2 step process
• For P0 build - on ercsc-single detector
  simulations - CMB Noise PointSources (no
  galaxy)
• For P1 build - on full-focal-plane simulations -
  including galactic foregrounds
• Review and discussion on QA procedures
• Improvements in the conceptual implementations of
  each code
• Planck Point Source workshop - hosted at IPAC Feb
  2008 - broad attendance (37 registered) across
  project including PIs

3
CSI-participants

• https//planck.ipac.caltech.edu/tiki-index.php?pag
  eCSI
• https//planck.ipac.caltech.edu/WebDav/CSI

4
Simulations

• ERCSC - single detector
• Test algorithms for detection of sources
• Study the effects of presence of Point Sources on
  the mapmaking
• To perform optical studies directly on the Maps -
  effective beams, etc.
• All frequencies
• scanning as one detector per frequency, noise as
  all detectors per frequency
• CMB noise (white,1/f) dipole diffuse
  foregrounds sources SZ
• Tods generated with LevelS, Plank Sky Model (PSM)
• Maps generated with M3 Springtide, Nside2048
• Extra simulations generated with grid of point
  sources to evaluate the effective point spread
  function (PSF) variations due to coverage
• Issues -
• One super-detector per frequency - Side
  effect holes in the maps near the ecliptic
  plane
• PSM - validation
• Mapmaking - validation

5
ERCSC - Maps for 30 GHz (LFI30-28a) data
6
ERCSC - Maps for 44 GHz (LFI44-24a) data
7
ERCSC - Maps for 70 GHz (LFI70-18a) data
8
ERCSC - Maps for 100 GHz (100-1a) data
9
ERCSC - Maps for 143 GHz (143-5) data
9
10
ERCSC - Maps for 217 GHz (217-1) data
10
11
ERCSC - Maps for 353 GHz (353-1) data
11
12
ERCSC - Maps for 545 GHz (545-1) data
13
ERCSC - Maps for 857 GHz (857-1) data
13
14
Validation

• Validation steps for the ERCSC sims
• Check in a systematic fashion
• the processing of the input templates
• the maps gt by computing their Power Spectra
• 1) Obtain binned map for noiseless case
• 2) Obtain the input maps (reference sky)
  smoothed by the beam and the window function (or
  just the spectra)
• 3) Obtain maps of the noise (white1/f)
• Compare the output maps with 1) and 2) - get
  maps of residuals
• Compute Power Spectra of gt output maps,
  binned maps, reference sky, and maps of residuals
• Compare Power Spectrum of residuals with that of
  3) and that due to the white noise (bluebook)
• Get Power Spectrum of 1/f residuals Power
  Spectrum of residuals (output map-binned
  noiseless map) - Power Spectrum of white noise
• We found issues with
• the input templates from PSM - eg repetition of
  Point Sources in the catalogs
• the approach used by a destriper when a large
  population of bright PS is present in the inputs
  - better to run the code with noise estimation
  off - (might require other approaches such as
  excision of very bright Point Sources in TODs
  before map-making)
• This activity involved members of the ADG -
  direct contact with the PSM team with the
  authors of the relevant codes - and weeks of work!

15
CSI - codes

• Asymmetric and Symmetric Matched Filters, Mexican
  Hat Wavelet (ifca-amf5, ifca-smf5, ifca-mhw2)
• developed by Marcos Lopez-Caniego,
  Gonzalez-Nuevo, Diego Herranz, Patricio Vielva et
  al (IFCA-Santander) - MNRAS, 369, 1603, (2006)
  MNRAS, 370, 2047, (2006) ApJS, 170, 108, (2007)
• Matched filter (pmf)
• developed by Jim Bartlett and Jean-Baptiste
  Melin (Paris) - AA, 447, 405B, (2006) AA, 459,
  341M, (2006)
• PowellSnakes (psnakes)
• - Bayesian method developed by Pedro Carvalho,
  Graca Rocha, and Mike Hobson (Pasadena,
  Cambridge) - astro-ph/08023916 (accepted for
  publication in MNRAS), 2 papers in preparation
• Codes developed at IPAC and JPL are
• Sextractor (sex) (Bertin Arnouts, AAS, 117,
  393, 1996), (Median filter (filt)), Mexican Hat
  Wavelet (mex), (Spatial wavelet (wave))
• The first ERCSC build (P0) was tagged on
  September 2008
• this build is based on the ERCSC single-detector
  CNPS simulations
• Source detection algorithm assessment are based
  on results at 30, 143, 857 GHz
• The QA criteria in this selection is based on
• Reliability - fraction of detected sources that
  are real (req 95 at ERCSC 10 sigma level)
• Completeness - fraction of real sources that are
  detected (req 90 at ERCSC 10 sigma level)

16
Results - P0

• 30GHz (10 sigma goal 790 mJy)
• 857GHz (10 sigma goal 1048 mJy)

17
Ringing effects produced by a Matched Filter vs
the MHW2 It is possible to reduce these effects
by fine tuning the estimation of the Power
spectrum of each patch in various ways masking
out very bright sources,etc
18

• Dramatic border effects introduced by the a very
  bright source in the edge of the patch after
  being filtered with a Matched Filter.
• Further work need to be done to improve the
  estimation of the P(k).
• - PwSnakes tackles this issue by considering a
  safety border and overlapping patches - while
  very bright sources are masked out

19
CSI FFP Feb 2008 Detection of PS with MHW2
Issue with very bright sources They were
introducing a ring of high SNR spurious
detections around them. Although this only
happend a dozen times in the whole sky. More
important, the optimal scale at which the MHW2
operates was not being properly estimated and
this produced a lack of detections in that patch,
as seen in the figure. This issue has been
solved, and now all the objects SNR gt 30 are
treated in an improved iterative way.
20
Spurious detections due to VBS and very dusty
regions (857GHz)

• Crosses show the truth catalog
• sources found by sextra - Cyan ellipses
• sources found by pmf - yellow circles
• sources found by mex - green circles
• 1 - Ringing of the matched filter due to real
  bright point sources induces false detections
  (this is now solved for all codes in particular
  for pmf shown)
• 2 - False detections due to very dusty regions
  (galactic plane or galactic cloud) - and is
  partly due to there being real "sources" that are
  not in the truth catalog. (shared by all codes)

21
PowellSnakes - Likelihood manifold
Position subspace (X,Y) High res antenna
(A,R) subspace (X0,Y0) of a maximum
22
PowellSnakes C 50 lt b (143GHz)

23
PowellSnakes (143GHz)

• C - 50 lt b B - 5 lt
  b lt 50 A - b lt 5

24
PowellSnakes (143GHz)

• C - 50 lt b B - 5 lt
  b lt 50 A - b lt 5

25
PMF - 143GHz
26
IFCA - MHW-143GHz
27
Input Catalog - threshold 200mJy (30GHz)
28
PMF output catalog (30GHz)
29
PowellSnakes output catalog-br95 (30GHz)
30
IFCA-MHW2 output catalog-ycr50 (30GHz)
31
P1 - 143GHz (A b lt 5 B 5 lt b lt 50 C 50 lt
b)
32
P1 and P2

• P1 build was finalized in May 2008
• Any improvements on algorithms are to be
  concluded by the end of July - in time to be
  evaluated for the P2 build
• ERCSC P2 Build (Operations) is scheduled for
  December 2008, and will be the final based on
  simulated data
• Paper in preparation to be completed by the end
  of this year - stay tuned!

33
Appendix
34
QA (P0)
35
P1 - 143GHz (A b lt 5 B 5 lt b lt 50 C 50 lt
b)