We examine the physical mechanism of the negative and positive storm at middle latitude in August 1992, based on incoherent scatter radar (ISR) observations over Millstone Hill (42.6 • N, 288.5• E) and a first-principles ionospheric model. The exospheric temperature T ex , thermospheric composition and neutral winds, inferred from the ion temperature T i profile using the ion energy balance calculation (e.g., Bauer et al., 1970;Oliver, 1979) and from the electron density N e profile using an ISR data assimilation method (Zhang et al., 2001(Zhang et al., , 2002, are employed to investigate the storm effects. The derived thermospheric information shows that the negative phase on August 5 is attributed to both the large poleward wind and the reduced [O] ] ratio at F 2 -layer. For the daytime positive storm on August 4, the thermospheric composition perturbation, in addition to the enhanced equatorward wind, plays a significant role. This study also suggests that the data assimilation technique can provide useful information to understand some physical mechanisms of the ionospheric storm when direct experimental data are not available.