A comprehensive understanding of intrinsic error growth mechanisms is essential for improving tropical cyclone (TC) track predictability. To investigate how region-specific initial condition errors influence forecast uncertainty, we employed convection-permitting ensemble forecasts using the HWRF model with perturbations applied in three spatial domains: (1) the inner core including outer rainbands (0–350 km), (2) the near environment (350–1300 km), and (3) the far environment (1300–3500 km). Two contrasting typhoon cases are analyzed — Chan-hom (2020), which exhibited large track uncertainty, and Maysak (2020), characterized by relatively high predictability.
The results demonstrate case-dependent sensitivities: for Chan‑hom, perturbations introduced in the inner-core region result in the most substantial track spread and error growth. In contrast, for Maysak, near-environment perturbations dominate uncertainty early in the forecast, with far-environment perturbations becoming more influential at longer lead times. Mechanism diagnostics reveal that in Chan-hom with relatively weak inertial stability, initial vortex perturbations efficiently propagate outward through enhanced advection effect by upper-level outflow and low-level inflow, enabling rapid modulation of the surrounding environment. This energy transmission leads to significant growth in uncertainties of mid-tropospheric geopotential height and deep-layer steering wind, which in turn amplify track divergence. For Maysak, strong inertial stability as the vortex rapidly intensifies and weak vortex–environment coupling suppress outward perturbation energy transfer, limiting the influence of inner-core errors.
 

typhoon,Ensemble Forecast; Predictability; Prediction Error