Further extension of skillful prediction of tropical cyclones (TCs) relies on in-depth studies about the intrinsic predictability of TCs. In this study, convection-resolving ensemble forecasts based on the Hurricane Weather Research and Forecasting model were adopted with perturbed initial conditions to study the error growth and intrinsic predictability of TCs. The new aspect of our study is the focus on the sensitivity of TC track and intensity predictability to initial errors in different regions, including (1) the inner core and outer rainbands (0-350 km), (2) the near environment (350-1300 km), and (3) the far environment (1300-3500 km). The results of TC track predictability show that the most sensitive region of initial errors for TC track forecasts is case-dependent. For the TC case with striking track forecast errors (e.g., Typhoon Chan-hom, 2020), the initial errors in the combined region of the TC inner core and outer rainbands produce the largest track uncertainties compared to those in the near and far environment. However, for the TC case with a highly predictable track (e.g., Typhoon Maysak, 2020), the most sensitive region of initial errors is the near environment at early lead times and the far environment later. By contrast, the most sensitive region for TC intensity is the inner core for both cases. The surface wind structure of TC inner core at larger scales (wavenumbers 0-2) can be predicted for more than 3.5 days, while the structure at smaller scales can only be predicted for a few hours.