WISE Photometry (WPHOT)

Transcription

1 WISE Photometry () Tom Jarrett & Ken Marsh ( IPAC/Caltech) WISE Science Data Center Review, April 4, 2008 TJ+KM - 1

2 Overview is designed to perform the source characterization (source position & flux measurements) associated with each of the three stages of source extraction during pipeline processing. It has a strong 2MASS heritage. Profile-fitting (WPRO) Aperture Photometry & Characterization (WAPP) WISE Science Data Center Review, April 4, 2008 TJ+KM - 2

3 WSDS Scan/Frame Pipeline WISE Science Data Center Review, April 4, 2008 TJ+KM - 3

4 Multi-frame Pipeline WISE Science Data Center Review, April 4, 2008 TJ+KM - 4

5 Driving Requirements [ L4WSDC-002] The WSDC shall produce a Source Catalog derived from the images used to generate the WISE digital Image Atlas. [ L4WSDC-080] The final WISE Source Catalog shall have greater than 99.9% reliability for sources detected in at least one band with SNR > 20, where the noise includes flux errors due to zodiacal foreground emission, instrumental effects, source photon statistics, and neighboring sources. This requirement shall not apply to sources that are superimposed on an identified artifact. [eliminate?] [ L4WSDC-009] The final WISE Source Catalog shall be at least 95% complete for sources detected with SNR>20 in at least one band, where the noise includes flux errors due to zodiacal foreground emission, instrumental effects, source photon statistics, and neighboring sources. This requirement shall not apply to sources that are superimposed on an identified artifact. [ L4WSDC-010] The final WISE Source Catalog shall include sources down to SNR=5 in any band, and the completeness and reliability of sources in the Catalog shall be characterized at all flux levels. [ L4WSDC-012] Flux measurements in the WISE Source Catalog shall have a SNR of five or more for point sources with fluxes of 0.12, 0.16, 0.65 and 2.6 mjy at 3.3, 4.7, 12 and 23 µm, respectively, assuming 8 independent exposures and where the noise flux errors due to zodiacal foreground emission, instrumental effects, source photon statistics, and neighboring sources. [ L4WSDC-013] The root mean square error in relative photometric accuracy in the WISE Source Catalog shall be better than 7% in each band for unsaturated point sources with SNR>100, where the noise flux errors due to zodiacal foreground emission, instrumental effects source photon statistics, and neighboring sources. This requirement shall not apply to sources that superimposed on an identified artifact. [ L4WSDC-015] The WISE Source Catalog shall contain the measured in-band fluxes or flux upper-limits in the four WISE bands for objects detected in at least one band in the WISE Atlas Images. [L4WSDC-016] The WISE Source Catalog shall contain uncertainties in the flux measurements (one sigma) in all bands for which a source is detected. [L4WSDC-018] The WISE Source Catalog shall contain uncertainties in the coordinates measurements for each object. [ L4WSDC-043] The WSDS Pipeline processing shall detect sources down to a threshold of at least five times the image noise from the calibrated image frames, and the combined Atlas Images. [ L4WSDC-044] The WSDS Pipeline processing shall merge source detections in the four WISE bands into a single source catalog entry. [ L4WSDC-049] The WSDS Pipeline shall be robust to data missing from one or more bands. WISE Science Data Center Review, April 4, 2008 TJ+KM - 5

6 Processing Flow Prior knowledge may also include IRAS and MSX WISE Science Data Center Review, April 4, 2008 TJ+KM - 6

7 Profile-Fitting Photometry (WPRO) Based on maximum likelihood fit of PSFs to pixel values at all bands simultaneously Evolution of PROPHOT (2MASS) --- PROPHOT was multi-frame --- WPRO represents logical extension to multi-band Advantages of multi-band fit: 1. High resolution data at short wavelengths can guide the fitting procedure at the longer wavelengths where the resolution is poorer 2. No post-extraction band-merge step is required, thus avoiding cross-band matching ambiguities in crowded fields 3. Optimal band-filling (valid measurement in band(s) that may not otherwise be detected) WISE Science Data Center Review, April 4, 2008 TJ+KM - 7

8 Models for maximum likelihood estimation Measurement model: pixel value Noise model: flux PSF background noise [Poisson noise] [Flat-fielding error] [Read noise] [background noise] [PSF error] WISE Science Data Center Review, April 4, 2008 TJ+KM - 8

9 Grouping sources into blends Passive deblending: Blend group consists of primary candidate plus all neighboring candidates close enough to influence solution, i.e., all neighbors within a critical distance, r crit, which depends on nominal flux ratio of neighbor with respect to primary All components of a passive deblend correspond to candidate sources found by source detector, MDET Active deblending: Neighbor(s) close enough to primary that MDET sees only a single peak --- additional components identified using chi squared-based procedure TJ+KM - 9

10 Solution procedure Construct parameter vector: [Position of nth blend component] Maximize: [fluxes at the multiple bands] Evaluate quality of fit: Estimated fluxes and positions WISE Science Data Center Review, April 4, 2008 TJ+KM - 10

11 Evaluation of Uncertainties Uncertainty of the jth parameter estimate is given by: TJ+KM - 11

12 Active deblending Developed and validated (but not used) for 2MASS Steps involved in iterative procedure: 1. Chi squared test of current solution, with N B components: If c n 2 > (c n 2 ) crit or (c n 2 ) l > (c n 2 ) crit for any l, then add new component. 2. Fix positions for N B components & minimize c n 2 over a grid of positions for new component, based on flux-only solution. 3. Use this as a starting model for full N B +1 component solution. 4. Delta-chi test of relative likelihoods of the two models: If [ (c n 2 ) NB - (c n 2 ) NB+1 ] > (Dc n 2 ) min then new model is accepted and iteration continues until either this condition is violated or the maximum allowable number of components is exceeded, i.e. N B > (N B ) max Key parameters: (c n 2 ) crit => Completeness (Dc n 2 ) min => Reliability WISE Science Data Center Review, April 4, 2008 TJ+KM - 12

13 WPRO flow Design allows for active deblending to be turned off (e.g., during IOC activities) WISE Science Data Center Review, April 4, 2008 TJ+KM - 13

14 PSF Generation PSFs will be generated as an off-line analysis task, and not as part of the normal operation of WPRO in the pipeline processing. PSFs are expected to vary with focal plane position and with band, but they should be stable with time (but this will be tested and verified) TJ+KM - 14

15 PSF Generation Three phases: 1. Pre-launch: Based on laboratory measurements, optical models 2. Post-IOC: Based on early IOC & ops data 3. Post-data acquisition: Based on all survey data TJ+KM - 15

16 Observationally-generated PSFs PSF shape is estimated from observed images of stars Focal plane assumed to be approximately (or better than) Nyquist sampled Nonisoplanicity is allowed for using a library of PSFs covering focal plane PSF estimate for a given subregion (segment) of the focal plane: [Observed image of nth star, sinc-interpolated to fine grid] [Flux estimate] [Noise] TJ+KM - 16

17 PSF uncertainty estimation Essential for calculating flux & position uncertainties and chi squared Obtained from residuals of profile-fits to the stars used in PSF estimation Residual image of nth star is given by: PSF uncertainty image, δh j, at position r j, is then given by: [measurement noise] [weights] TJ+KM - 17

18 Steps involved in PSF generation For a given focal-plane segment: 1. Locate all stars above a given flux threshold for the particular band. -- facilitated by call to WAPP without WPRO 2. Estimate PSF and its uncertainty using this set of stars. -- filter out confusing sources within the neighborhood of each star 3. For each individual star, examine the quality of fit using χ List any stars for which χ 2 exceeds predefined threshold; discard the star with the highest χ Iterate from step 2 until all remaining stars have acceptable χ 2. TJ+KM - 18

19 PSF selection for photometry PSF segment boundaries on focal plane Estimate set of PSFs at finely-spaced intervals (subdivisions of W P ) Correlation between segments prevents discontinuities in PSF shape (and uncertainty) over focal plane W P Select appropriate PSF via table lookup Note: simple interpolation between segments spaced by W P is undesirable since uncertainty would increase between segments TJ+KM - 19

20 How many PSFs do we need? Driving requirement: Photometric accuracy on bright stars must meet the relevant Functional Requirement (=> better than 7%). dictates the accuracy to which we need to know the PSF at any given place on the focal plane. lab measurements and theoretical models will provide information on shape variation and hence the maximum acceptable segment size, W P. -- lab measurements expected later this summer TJ+KM - 20

21 Results using 2MASS+Spitzer data Field: 1 deg x 9 arcmin subfield of M67 Bands: J, H, K S from 2MASS deep field 3.6, 4.5, 5.8, 8.0 µm from Spitzer IRAC Source extraction: Simultaneous 7-band profile-fitting photometry --- same photometry algorithm as proposed for WPRO --- same PSF generation technique also Input data: coadded frames (MOPEX) TJ+KM - 21

22 M67 analysis: Input data 4 of the 7 bands 4 x 4 subfield TJ+KM - 22

23 M67 analysis: Reduced chi squared J H K S 3.6 µm 4.5 µm 5.8 µm 8.0 µm TJ+KM - 23

24 M67 analysis: Color-magnitude TJ+KM - 24

25 Aperture Photometry The Aperture Photometry System (WAPP) performs multiaperture photometry & source characterization. 2MASS heritage throughout (PROPHOT, GALWORKS). Fixed aperture photometry serves several purposes, including: (1) source flux estimation in support of profile-fitting photometry, (2) construction of curve-of-growth corrections for small aperture measurements and for normalization of profile-fit photometry, (3) more accurate flux determination for very bright sources, (4) serve as a truth measurement to test the robustness of profilefitting photometry, and (5) accurate source flux determination for small extended sources in which the PSF does not accurately model the light distribution. WISE Science Data Center Review, April 4, 2008 TJ+KM - 25

26 WAPP Input The WPRO extraction list is used as the input source list for WAPP. In this way every source extracted by WPRO using both passive and active deblending will have an aperture flux. Designed to run stand-alone (e.g., prototype does not have WPRO capability). WISE Science Data Center Review, April 4, 2008 TJ+KM - 26

27 WAPP Flow Best effort basis only TJ+KM - 27

28 Multi-Aperture Photometry bad pixels replaced with maximum likelihood values Local background annulus Nested apertures to capture curve-ofgrowth Stellar confusion? Nested apertures to capture low SB flux in the wings of point sources, and to capture extended source emission Integrated flux and source characterization limited by local background annulus size WISE Science Data Center Review, April 4, 2008 TJ+KM - 28

29 Multi-frame Photometry WISE Science Data Center Review, April 4, 2008 TJ+KM - 29

30 Multi-frame Photometry + N(m) statistics TJ+KM - 30

31 N out of M Statistics From 2MASS: The statistics of the aperture measurements are compiled in the Ndet parameter included in the PSC record. Ndet is a six-digit flag, with two digits per band that tabulates for each band the number of frames on which a source was detected with >3σ in aperture photometry, N b, and the number of frames which were available for measurement, M b. Thus, the frame detection thresholds are also referred to as the "N-out-of-M" criteria. For brighter sources, the Ndet parameter can be used as a reliability indicator. TJ+KM - 31

32 Upper Limits Negative fluxes reported (in dn units) Measured flux + 2σ (negative?) Include negative flux or upper limit with merge process? (lower the weight accordingly) TJ+KM - 32

33 Failed Measurements bad or flag pixels near core N bad > threshold (aperture) Null. Report upper limit? Set flag. TJ+KM - 33

34 Aperture Corrections Standard aperture measurements come from the smallest possible aperture to minimize the effect of photon shot noise from the background. To correct for the loss of light in the standard aperture, a curve-of-growth correction is applied to the measurements. The curve-of-growth correction is a constant factor that when added to the magnitudes measured in the standard aperture makes them equivalent to an "infinite" size aperture. TJ+KM - 34

35 Aperture Corrections As with PSF generation, aperture corrections are expected to be different for each band, and to vary across the focal plane. They are expected to be stable with time, though this will be tested and verified. TJ+KM - 35

36 Aperture Corrections The 2MASS method: integrated flux curve of growth derived from thousands of point source measurements: Avoid confusion ( b > 30 deg ; flagged sources) Avoid sources with bad (or flagged) pixels Limit to SNR sweet-spot Photometry using nested apertures TJ+KM - 36

37 Aperture Corrections The correction factor, in magnitudes, is the median difference between the 4 aperture magnitude and the magnitude in the aperture at which the magnitude differentials become indistinguishable from zero. TJ+KM - 37

38 Aperture Corrections Implementation: Derive correction and apply corrections as a post-processing (off-line) step? Apply corrections in? (must be able to back out) Note: only the standard aperture is corrected with C-of-G. TJ+KM - 38

39 Test Plan Description Test Data Sets: WISE image simulations; Spitzer NEP/SEP mini-surveys, GLIMPSE, SWIRE; M67 (2MASS & IRAC) Integrity & robustness of the algorithms Reliability (χ 2 metric; active deblending; N out of M) Completeness in confused instances (WPRO) Memory management for deep coverages Speed management (active deblending thresholding) WISE Science Data Center Review, April 4, 2008 TJ+KM - 39

40 IRAC 1, 2, 4

41 Test Description -- WAPP WAPP Annulus/Aperture testing Local background statistics; pixel-value histogram statistics. Small apertures, to test the fractional-pixel algorithm. Large apertures, to test the integrity of the system. Full range in fluxes, from noise to bright-saturated stars. Compare frame with coadd measurements WISE Science Data Center Review, April 4, 2008 TJ+KM - 41

42 Test Description -- WPRO WPRO testing: Validity of noise model: plots of χ ν 2 vs. magnitude Validity of quoted errors: repeatability tests based on multiple observations (e.g., Spitzer observations of NEP) Active deblending tests: -- Verify performance (completeness and reliability as a function of SNR, recovery of sources missed during detection step) -- Determine optimal values of deblending parameters Response to artifacts and extended sources: -- Saturated stars, diffraction spikes, latent images, etc. -- Effectiveness of χ ν 2 as discriminator WISE Science Data Center Review, April 4, 2008 TJ+KM - 42

43 After SDL payload testing, the corresponding PSFs will be incorporated into the simulated images for testing. TJ+KM - 43

44 Development Schedule Peer Review (April, 2008) v0 2/27/08 prototype (single frame, multi-band), data flow testing Input frames, masks & detection lists Local backgrounds (stats) Aperture photometry Preliminary output table v1 6/19/08 payload ground testing; prototype multi-frame Input median-filtered background images Profile-fitting (active and passive deblending, isoplanatic PSF) Other source characterization (e.g., multi-apertures) Full output table V2 2/28/09 Nonisoplanicity capability in WPRO, PSF generation software V3 8/4/09 Pre-launch version: Complete functionality, pre-launch PSF set, optimized parameters V3.5 12/30/09 Post-launch tuneup of parameters/code; update PSF model & apertures based on IOC and early operations data V4 9/20/09 Version for final processing; PSFs derived from all available data. Final algorithm modification based on lessons learned from full survey. WISE Science Data Center Review, April 4, 2008 TJ+KM - 44

45 Things to do PSF generation Parameter tuning Determine driving thresholds for active deblending Curve of Growth proto-development Masked/bad-pixel recovery Build prior catalogs Implementation of smooth background removal WISE Science Data Center Review, April 4, 2008 TJ+KM - 45

46 Issues/Concerns Focal Plane-dependent PSFs -- segment size? #PSFs? Focal Plane-dependent curve-of-growth measurements? Coadd measurements -- are they needed? Very bright stars & saturated stars -- filter artifacts? Extended sources (no plan or funds to properly deal with) Confusion from emission nebulae in the Galactic Plane WISE Science Data Center Review, April 4, 2008 TJ+KM - 46

47 Extra Slides TJ+KM - 47

48 Prototype Output TJ+KM - 48

49 WPRO Output Parameters Multi-band, single frame TJ+KM - 49

50 WPRO Output Parameters TJ+KM - 50

51 WAPP Output Parameters Multi-band, single frame + additional aperture fluxes, + characterization metrics TJ+KM - 51

52 Local Background WISE-W1 Galactic Center Simulations TJ+KM - 52

53 Local Background Local annulus assumed to possess the same background and fluctuations as within the aperture Pixel Value Distribution should be equivalent to aperture measurements (profile fitting, aperture integration), including confusion noise fluctuations: Nearby sources masked from image Upper & lower extreme values trimmed Median, mean, RMS (standard deviation) Effective gaussian median & RMS (combine the 84% and 16% quartiles to derive the histogram RMS ) TJ+KM - 53

54 Local Background Median Absolute Deviation after Iterative Trimming of Upper Tail (MADTUT) TJ+KM - 54

55 Recovery from Masked Pixels Masked pixels arise from bad pixels, flagged pixels, masked stars or galaxies. Replace with local value Replace with isophotal value (Galworks) Replace with maximum likelihood value (PSF or gaussian model) TJ+KM - 55

56 Recovery from Masked Pixels TJ+KM - 56

57 Isophotal Substitution LSB galaxy model-subtracted Stars-subtracted Shape Characterization (moments; ellipsoid model) Star subtraction / deblend SB profile Total flux & extent Half-light metrics Concentration index Plethora of isophotal & fixed aperture photometry TJ+KM - 57

58 Saturated Sources WPRO should recover the flux using the wings of the stellar profile WAPP not designed to recover from saturation -- null the measurement? Issues -- Artifacts? Multiple detections? Beta Pic How to use saturated A/D flagging? TJ+KM - 58