Accurate data assimilation is critical to the successful prediction of hazardous weather.  In this talk, I will briefly discuss three convective-scale simulations that have been initialized via 3DVAR data assimilation of real observational data.  Accurate storm-scale predictions from each of these simulations prompted the nesting of very high-resolution simulations within the convective-scale simulations.  These very high-resolution simulations were qualitatively accurate in their portrayal of tornado-scale features when compared with observations and damage surveys.  The majority of this presentation will focus on the analysis of one of these simulations – the 8 May 2003 Oklahoma City tornadic supercell.  I will discuss tornadogenesis in this simulation and in particular discuss a novel role of surface drag therein.  I will also discuss the origin of internal momentum surges in the simulation.