Recent advances in high-resolution, full physics numerical models and continued increase in computer powers have made it possible to simulate the full lifecycle of tropical cyclones with resolved fine structure of the inner core over multi-day integrations. To use these high-resolution models for hurricane prediction in practice, we face a challenge due to the lack of high-resolution observations to initialize these models over the open ocean. Previous studies have experimented with various vortex initialization techniques including bogusing of idealized vortices (Kurihara et al., 1993) and vortex relocation (Liu et al., 1997). We apply a similar method as in Liu et al., in which we integrate the high-resolution model forward and relocate a spun up vortex to the best track location at the initial time of a forecast period, for a number of model simulations of Hurricanes Bonnie (1998), Georges (1998), Floyd (1999), and Lili (2002). These simulations are done with a multi-nested, vortex-following, high-resolution version of MM5 with the finest grid resolution at 1.67 km for the inner-most model domain. Our results show various degrees of sensitivities of the storm evolution to the initial vortex structure in both inner-core and outer vortex blending radius in all cases. Variations of initial vortex structure seem to alter the eyewall replacement process in Floyd, rainfall asymmetry in Bonnie, and intensity change in Georges and Lili. The objective of this study is to evaluate and understand the physical processes responsible for these sensitivities and, ultimately, to improve vortex initialization methodology for high-resolution hurricane prediction models. To address this objective, we will conduct the following analyses and model simulations. First, we will analyze the symmetric structure of our simulations using the framework established in Willoughby (1979) and Shapiro and Willoughby (1982), and we will evaluate the extent to which the symmetric secondary circulations and tangential wind tendencies agree with the theoretically predicted circulations and tendencies based on the heating and eddy momentum fluxes diagnosed from the model solutions. Second, we will perform a set of simulations in which we vary the horizontal scale of an idealized initial vortex and evaluate the extent to which the symmetric tangential circulations change in accordance with what we expect from the scaling arguments presented in Shapiro and Willoughby (1982).