Extracting stellar parameters from observed spectra: the role of cross-correlation and minimum distance methods

One may consider cross-correlation routines applied to stellar spectra as a way to determine not only the radial velocities but also to assess the properties of stellar atmospheres. We show that this is not the case. High value of the cross-correlation coefficient between the observed and a template spectrum does not mean that the template corresponds to correct values of stellar temperature, gravity or metallicity. Cross-correlation operates in normalized space, so it is not sensitive to scaling or shifting of stellar flux. It is argued that cross-correlation is useful to efficiently determine the velocity shift between the observed and the template spectrum. But minimal distance methods need to be used to determine which template gives the best match to the observed spectrum, and so derive values of physical parameters. To illustrate the point we use data from the RAVE project which are being obtained at very similar spectral resolution and wavelength domain as will be the case for GAIA’s RVS.