Enhancing the predictive capabilities for tropical cyclones (TCs) necessitates comprehensive investigations into their predictability. This study employs convection-allowing ensemble forecasts using the Hurricane Weather Research and Forecasting model, integrating perturbed initial conditions to examine error growth and initial condition error-related predictability (IE-predictability) of TCs. Distinctively, our research concentrates on how the predictability of TC track and intensity is influenced by initial errors in various zones: (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).Contrary to many prior studies, our findings suggest that the most critical region for initial errors affecting TC track forecasts is within the TC inner structures. For Typhoon Chan-hom (2020), characterized by significant track forecast discrepancies, the initial inaccuracies in the inner core and outer rainbands substantially affect the TC's proximate environment and track, more so than errors in the near and far environments. Regarding TC intensity, the inner core emerges as the most sensitive area. The surface wind configuration of the TC inner core at larger scales (wavenumbers 0–2) maintains predictability for over three days, whereas its structure at finer scales is only predictable for a few hours. Plain language summaryThe forecasting of tropical cyclones (TCs) has seen significant advancements over recent decades. A critical question arises: to what extent can we further enhance the accuracy of TC predictions? Addressing this requires in-depth exploration of TC predictability. In our study, we utilized convection-allowing ensemble forecasts based on the Hurricane Weather Research and Forecasting (HWRF) model, incorporating perturbed initial conditions to analyze the development of forecast errors and the predictability related to these initial errors in TCs. A novel aspect of this research is its emphasis on the impact of initial errors in different regions on the predictability of TC track and intensity. Our key findings are twofold: (1) The region most impacted by initial errors for TC track predictions may not always be the surrounding environment. Instead, errors within the TC's inner core and outer rainbands, owing to their pronounced interplay, can lead to significantly greater track forecast errors compared to those in the near and far environments. (2) As for TC intensity, the inner core emerges as the most susceptible region. The large-scale components (wavenumbers 0–2) of the TC's inherent vortex flow can be forecasted accurately for over four days, largely due to the predictability of the synoptic-scale environment. In contrast, the smaller-scale components (wavenumbers >2) demonstrate predictability for only a few hours.