|
|
Due to the difficulties and slow process of spectral acquisition, most cosmological surveys will map the large scale structure through photometric techniques only. To correctly estimate the distances and distribution of galaxies using photometric data only, reliable photo-z methods have to be implemented. It is known, that the precision of the resulting photometric redshifts is mainly affected by the photometric errors and the colour coverage of the training set itself. However, it was not investigated before, how the large variation in emission lines hinders or increases this precision. Since the emission line strengths depend on a large set of parameters, thus vary significantly from spectrum to spectrum, their prediction involves difficult radiative transfer treatments, which makes theoretical studies cumbersome. As such, the currently available route for such investigation is through the statistical modelling of existing large datasets.
In this talk, I will introduce the semi empirical method that we developed for generating mock catalogues that mimic the real distribution of galaxies, and the general treatment with which we could generate emission lines with realistic ratios on top of galaxy continua. Moreover, I will show that through generating mock catalogues with this technique, and by applying basic machine learning methods for the photo-z estimation, we could determine how the emission lines affect the precision across various redshift ranges. Apart from presenting the basic framework and describing the results we obtained for the effects of emission lines, I will also introduce an extension for this method that could be used for the general classification of emission line galaxies.
Here you can play back the recording of the presentation
