We have used an off-line chemical transport model (CTM) to diagnose the expected chemical ozone destruction in the Arctic winters of 1993/1994 to 1996/1997, and to investigate the sensitivity of the model-calculated loss to meteorological variability, chlorine and bromine loadings, denitrification, dehydration, and increased stratospheric H2O. The model was integrated on a single isentropic surface at 475 K (about 18 km or 50 hPa altitude) using analyses from the European Centre for Medium-Range Weather Forecasts. The CTM produces local depletions of up to 45% in the polar vortex in the cold winter of 1995/1996. In winter 1996/1997 the large-scale temperatures cold enough for polar stratospheric clouds did not start until early January which delayed chlorine activation. However, because the cold temperatures persisted well into March, the local O3 depletion in the model vortex was around 40% in late March. The sensitivity experiments using 1994/1995 meteorology show that the chlorine loading is much more important than bromine for controlling the polar ozone loss given the expected abundances, but the ClO+BrO ozone loss cycles are calculated to be more important than ClO+ClO. Accordingly, the local relative efficiency factor we calculate from the model (α) is large, especially in the polar region where it is around 60 for current day halogen loadings. Denitrification can increase Arctic ozone depletion through a delay in chlorine deactivation. However, very strong denitrification early in the winter causes less O3 depletion in the model, with the current heterogeneous chemistry scheme, due to enhanced recovery into HCl and inefficient chlorine reactivation. The additional inclusion of dehydration reduces the modeled Arctic O3 depletion, due to decreased heterogeneous processing rates, and the reduced occurrence of equilibrium nitric acid trihydrate particles. The modeled Arctic O3 depletion increases slightly (by 2--4% of the initial O3) using 1994/1995 meteorology when the stratospheric H2O loading is increased. This is due to slightly more polar stratospheric cloud (PSC) activity in this cold winter, although this effect would be potentially more important in a warmer winter where PSC processing was marginal. ¿ 1998 American Geophysical Union