Spectroscopic analysis of hot, massive stars in large spectroscopic surveys with de-idealised models
Upcoming large-scale spectroscopic surveys with e.g. WEAVE and 4MOST will provide thousands of spectra of massive stars, which need to be analysed in an efficient and homogeneous way. Usually, studies of massive stars are limited to samples of a few hundred objects which pushes current spectroscopic analysis tools to their limits because visual inspection is necessary to verify the spectroscopic fit. Often uncertainties are only estimated rather than derived and prior information cannot be incorporated without a Bayesian approach. In addition, uncertainties of stellar atmospheres and radiative transfer codes are not considered as a result of simplified, inaccurate or incomplete/missing physics or, in short, idealised physical models. Here, we address the question of "How to compare an idealised model of complex objects to real data?" with an empirical Bayesian approach and maximum a {\it posterior} approximations. We focus on application to large scale optical spectroscopic studies of complex astrophysical objects like stars. More specifically, we test and verify our methodology on samples of OB stars in 30 Doradus region of the Large Magellanic Clouds using a grid of FASTWIND model atmospheres. Our spectroscopic model de-idealisation analysis pipeline takes advantage of the statistics that large samples provide by determining the model error to account for the idealised stellar atmosphere models, which are included into the error budget. The pipeline performs well over a wide parameter space and derives robust stellar parameters with representative uncertainties.