Transcript photo meeting

Budapest Group
Eötvös University
MAGPOP kick-off meeting
Cassis 2005 January 27-30
HuMuDeS: Hungarian Multiwavelength Deep Survey
• 1000 sq degree of the northern
hemisphere
• High resolution spectra from radio
through optical to X-ray
• Volume limited survey up to z=3
• Unfortunately does not exist
Cassis meeting 2005 January 27-30
István Csabai
2
Astro in Hungary
•
•
•
•
•
No large telescopes
No large grants
No postdocs
Traditions: solar system, variable stars
Background: physics and computer
science
• Links: SDSS/JHU, VO
Cassis meeting 2005 January 27-30
István Csabai
3
People
• István Csabai (photo-z, data mining)
• Zsolt Frei (morphology, galaxy mergers)
• László Dobos (student, VO spectrum and filter
service)
• Bence Kocsis (PhD student, galaxy mergers)
• other PhD students:
–
–
–
–
Zoltán Lippai (merger tree code)
Merse Gáspár (morphology, disk/bulge separation)
Zsuzsanna Győry (emission line PCA)
Norbert Purger (data mining)
Cassis meeting 2005 January 27-30
István Csabai
4
Areas
•
•
•
•
•
Photometric redshift estimation
Galaxy morphology
Galaxy evolution: merging
Data mining, databases
Virtual Observatory
Cassis meeting 2005 January 27-30
István Csabai
5
Official MAGPOP tasks
• 4. Physical properties of galaxies from SDSS potometry
– Replace empirical templates with spectral synthesis models (IAP,
partly done, need improvements)
– Use morphological info in photo-z
– Compare spectral type, photo type, morphological type
classifications (Nakamura Osamu)
– Apply photo-z to SDSS (JHU, done for EDR, still improving the
method for new releases)
– Compare results from photometric and spectroscopic parameter
estimations (MPA,IAP,JHU)
– Cross correlate SDSS and GALEX, carry out statistical analysis
(JHU)
– Study galaxy enviroment, compare properties of high and low mass
galaxies (JHU,MPA)
Cassis meeting 2005 January 27-30
István Csabai
6
Other plans
• Apply photo-z for other surveys
• Galaxy merging, signatures with gravitational
waves
• Develop tools for fast data mining
(photometry)
• VO tools for SED manipulation
• Release data products for the network (BC03
service, cross matching catalogs)
Cassis meeting 2005 January 27-30
István Csabai
7
Photometric redshifts
MAGPOP kick-off meeting
Cassis 2005 January 27-30
Introduction
• Getting photometry – fast
• Getting spectra – slow
• SDSS
• 80% of time: spectra for the brightest 1% of the
objects
• 20% of time: 5 band photometry for all objects
Cassis meeting 2005 January 27-30
István Csabai
9
The idea
• The same galaxy at different redshifts
has different colors
• Invert the relation: use colors to get the
redshift
Cassis meeting 2005 January 27-30
István Csabai
10
Methods
• Empirical
– neural network
– function fit (polynom)
– nearest neighbor
– neighbors in a cell
+ no calibration
problem
+ no input parameters
Cassis meeting 2005 January 27-30
• Physical
– tempalte fitting
• maximum likelihood
estimation
• empirical seds
• model seds
+ get physical
parameters
István Csabai
14
LIGHT;
SED
TEMPLATE
OBSERVATIONS
MAGNITUDES,
COLORS
Cassis meeting 2005 January 27-30
BROADBAND
FILTERS
István Csabai
REDSHIFT
15
Major limitations
• Photometric errors
• Flat spectrum (blue galaxies)
• Lack of good templates
•
•
•
•
•
wavelength range
bulge/disk
evolution
repair (deconvolution)
model templates: parameter space is too large
• Spectral type – redshift degeneracy (morphology,
priors)
• Interpolation OK, extrapolation not: need for training
sets (SDSS photo-z plates)
Cassis meeting 2005 January 27-30
István Csabai
16
The inversion
• Principal component analysis shows that the
‘spectral space’ is low dimensional
(continuum and emission lines also) few
colors are enough
• Cause: small number of major physical
parameters
• If we would use model SED templates, we
could get not just redshift, but most of the
essential parameters of the galaxies
Cassis meeting 2005 January 27-30
István Csabai
17
PARAMETRS
GALAXY
LIGHT;
age, dust, ...
early type,
late type
SED
MODEL
OBSERVATIONS
MAGNITUDES,
COLORS
Cassis meeting 2005 January 27-30
BROADBAND
FILTERS
István Csabai
REDSHIFT
18
Plans
• Error estimates / confidence
• Hybrid method: neighbors in a cell interpolation
where available – template fitting otherwise
• Tune photoz with spectral synthesis models (Monte
Carlo library of stochastic SFH) -> help to improve
SED models
• Evolution
• Morphology prior
• Photo-z web service
• New surveys (GALEX, CFHTLS)
• Application: large scale structure
Cassis meeting 2005 January 27-30
István Csabai
20
Data mining of the color space
MAGPOP kick-off meeting
Cassis 2005 January 27-30
Introduction
• Terabytes of data
• Even current database tecnniques are
not efficient enough for handling
continous multidimensional spaces
• Example:
• given: SED, redshift range (trajectory in color
space)
• need: all nearby objects from a survey
Cassis meeting 2005 January 27-30
István Csabai
22
Indexing the color space
•
•
•
•
K-d tree
Outlier separation
PCA tree
Voronoi
Cassis meeting 2005 January 27-30
István Csabai
23