Determination of stellar parameters with GAIA

The GAIA Galactic survey satellite will obtain photometry in 15 filters of over 10(9) stars in our Galaxy across a very wide range of stellar types. No other planned survey will provide so much photometric information on so many stars. I examine the problem of how to determine fundamental physical parameters (T-eff, log g, [Fe/H] etc.) from these data. Given the size, multidimensionality and diversity of this dataset, this is a challenging task beyond any encountered so far in large-scale stellar parametrization. I describe the problems faced (initial object identification, interstellar extinction, multiplicity, missing data etc.) and present a framework in which they can be addressed. A probabilistic approach is advocated on the grounds that it can take advantage of additional information (e.g. priors and data uncertainties) in a consistent and useful manner, as well as give meaningful results in the presence of poor or degenerate data. Furthermore, I suggest an approach to parametrization which can use the other information GAIA will acquire, in particular the parallax, which has not previously been available for large-scale multidimensional parametrization. Several of the problems identified and ideas suggested will be relevant to other large surveys, such as SDSS, DIVA, FAME, VISTA and LSST, as well as stellar parametrization in a virtual observatory.