Data assimilation is a mathematical technique for updating model fields using observational information. Initially, data assimilation was developed primarily for numerical weather prediction (NWP) purposes, but the technique is now applied in a wide range of scientific disciplines including atmospheric chemistry, climate research, oceanography and space weather. Even though different forecast models are used in the different disciplines, the so-called analysis equations used for updating the forecast states are essentially the same, and the overall data assimilation systems are therefore very similar across the disciplines. The purpose of this presentation is to provide the space weather community with a brief survey of the state of the art of "classical" NWP data assimilation research and development. Among the topics to be discussed are: data assimilation for sparse vs. dense observations, model and forecast bias, state-dependent error covariance modeling.