What is the SNR that I need to achieve to obtain reliable galaxy properties?

What is the optimal set of observables to derive these properties for my galaxy sample?

What are the uncertainties and degeneracies expected between these properties for the SNR and observables chosen?


These are  some of the most common questions that arise while designing a new project or preparing an observing proposal on galaxy evolution. Unfortunately, very few studies have analyzed in detail and in a systematic way the effects of the observational errors and the set of observables on the uncertainties and degeneracies between the different parameters affecting the Spectral Energy Distributions of galaxies (a noteworthly exception is the optimization of the Photometric Redshifts Technique; see e.g. Kodama, Bell & Bower 1999; Bolzonella, Miralles & Pello 2000; Wolf, Meisenheimer & Roser 2001). OPTIMIZER is a powerful tool designed to optimize the observing time in those studies on galaxy evolution based in the comparison of photometric data with the predictions of evolutionary synthesis models. It allows to derive the Signal-to-Noise ratios and set of observables needed to determine precise age, formation timescale, dust extinction, and metallicity values for a given galaxy.


In order to address these questions we have artificially generated a large sample of galaxies (~9000) with different star formation histories, ages, metallicities, dust content, etc. We have considered exponential star formation histories parameterized by the timescale of formation. The actual UV-optical-NIR mass-to-light ratios (M/L) and colors for these galaxies have been derived using the predictions of evolutionary synthesis models (Bruzual & Charlot 2001) including the contribution of the nebular continuum (see Table 1 for the range in galaxy properties of the sample generated). Then, these M/L and the corresponding colors have been perturbed adopting different observational errors and compared back with the evolutionary synthesis models grouped in different sets (see Table 2 for the sets of observables considered). The comparison has been performed using a combination of Monte Carlo simulations, a Maximum Likelihood Estimator, and Principal Component Analysis (see Gil de Paz et al. 2000c).  The ranges in galaxy properties used for this comparison (see Table 1) are slightly wider than those used for generating the sample.  Using this procedure we have been able to compute the uncertainties and degeneracies between the galaxy physical properties as function of (1) the set of observables available, (2) the observational errors, and (3) the galaxy properties, including star formation history, age, extinction, metallicity, and redshift (z = 0.0, 0.7, & 1.4).


Once these quantities were derived we computed the mean differences between the output and input values along with the mean 1-sigma spread, at fixed interval in the input properties. In Figure 1 we show the results obtained before and after computing the mean differences and 1-sigma errors for a subsample of 500 nearby galaxies. Due to the relevance of the K-band luminosities in order to derive stellar masses in nearby (Aragón-Salamanca et al. 1993; Gil de Paz et al. 2000c) and intermediate-redshift galaxies (see Brinchmann & Ellis 2000 and references therein), the mean differences between the derived and input K-band M/L were also computed. The values for the mean uncertainties in the derived properties are given in Table 3. For a detailed description of the results shown in the graphs given below see Gil de Paz & Madore (2001).


Aragón-Salamanca et al. 1993, MNRAS 262, 764.
Bolzonella, Miralles & Pelló 2000, A&A 363, 476.
Brinchmann & Ellis 2000, MNRAS 536, 77.
Bruzual & Charlot 2001, in preparation (GISSEL99).
Gil de Paz & Madore 2001, AJ, accepted (astro-ph/0201013)
Gil de Paz et al. 2000a, A&AS 145, 377.
Gil de Paz, Zamorano & Gallego 2000b, A&A 361, 465.
Gil de Paz et al. 2000c, MNRAS 316, 357.
Kodama, Bell & Bower 1999, MNRAS 302, 152.
Wolf, Meisenheimer & Röser 2001, A&A 365, 660.

If you are interested in OPTIMIZING your observations send me an e-mail to: agpaz(at)astrax.fis.ucm.es

Last modified on: September 11th 2006