Based on 4-dimensional variational (4D-VAR) data assimilation, a bogus data assimilation (BDA) method has been developed recently by Zou and Xiao (2000) to improve the initial condition for tropical cyclone simulation. Given a specified sea level pressure (SLP) distribution, the BDA process can lead to a better initial typhoon structure. Xiao et al. (2000) expanded their work by assimilating the wind field data into the model. Using the observed data, such as the minimum SLP, the maximum wind speed, radius of maximum wind speed, the moisture, and/or reflectivity information, and based on the presumed idealized distribution of the SLP and/or 3-D wind field, a dynamically consistent initial structure is produced. By comparing simulations with different data used for BDA, their result indicates that the assimilation of only the pressure field is more effective than the assimilation of only the wind field. However, using a similar approach, Pu and Braun (2001) showed that the assimilation of wind field would be more useful than that of pressure field, while assimilating both the wind and pressure fields would provide the best results. To understand the above conflicting results, we employ the Observing-Systems Simulation Experiments (OSSE's) to study the relative role of the wind and pressure fields in initialization of tropical cyclones. We found that the concept on the geostrophic adjustment can provide a useful explanation to the above issues. Results from the simulated track and intensity suggest that the assimilation of the wind field is more effective than the assimilation of the pressure field. In other words, the wind field is critical for maintaining a correct initial vortex structure of a tropical cyclone, while the model's memory on the pressure field is very short. To evaluate the geostrophic adjustment process in the BDA, we estimate the ratio of the vortex scale to the radius of Rossby deformation (L/LR, Shapiro and Montgomery 1993) in our OSSE's, as well as in the experiments in Xiao et al. (2000) and Pu and Braun (2001). The calculated ratios of L/LR in these three works have approximately the same value of 0.3, that is much small than 1. Therefore the geostrophic adjustment appears to favor the pressure field to adjust to the wind field. This result is consistent with the simulated result in our OSSE's and Pu and Braun (2001). We believe that as far as the geostrophic adjustment is concerned, at least based on the framework of Schubert et al. (1980) and Shapiro and Montgomery (1993), with the small value of L/LR in all the analyses above, the wind field should play a more important role than the pressure field for the initialization of tropical cyclones.