Astronomy and Coffee

Wavelets and Galapagos

Today the morning was a small walk in the town, very nice sunny (but very cold) day and a discussion about application of wavelets to the detection of overdensities in images generated from galaxy position catalogues. Still needs some "tuning" but it can be done. Eventually the a trous is not the best function, but the idea remains :-) In the afternoon application of GALAPAGOS from scratch (still happening due to WCS problems) so that I can check every step with Marco Barden before coming back to Germany.

About yesterday, couldn't give much detail, because we were leaving to have some Gluehwein in the Christimas market and dinner.

Marco's talk about morphological classification on STAGES. 3 main approaches. 1 Eye classification, 2 non-parametric and 3 parametric methods.

1- Eye inspection, 5000 galaxies were classified by 7 classifiers (not all of them received the 5000 galaxies, they were randomly distributed, but each was inspected my 3 of the 7 members). Many classes, many flags and a big dispersion, as expected.

2- As non-parametric they as using CAS (concentration, asymmetry and "clumpyness"). Another option would be the GINI coefficient. I have to search more about it, because we're getting right in this point about now!

3- Parametric, they used GALFIT to get the parameters of the galaxies (that's exactly the part of Marco, the GALAPAGOS). GALFIT already have version 3.0 (still "under request" basis) and it can do Fourier fitting. So using higher order terms, non-symmetric structures, such as lobes, can be modeled. The physical interpretation of that is a whole other story!

In the afternoon I had the introduction to GALAPAGOS. That's the name of the code that gets the 80 ACS pointings of STAGES, detect and do single-sersic fit in all the objects present there. It's an impressive code. Almost fully automatic, really minimum interaction is required, basically selection of the not many spurious objects that can appear in the catalogue (a couple of weeks in the case of STAGES, a couple of ours per cluster in our case - 4 pointings per cluster). The amount of details to be taken in consideration in an analysis like this is big. Things like GALFIT requires a mask in the regions that should not be used in the fitting, or the objects that should be simultaneously fitted and for some thousands of objects those things can't be done "one by one". GALAPAGOS manage to do this ... very cool! At the end it prepares the "post stamps" (or thumbnails, as other people like to call), the masks and the input files. The other main issue they took in consideration was how to optimize and to parallelize all that procedure. His paper with the description is in preparation, almost ready!

posted by Cris Da Rocha @ 9:45 PM