Abstract Intensity forecast errors of tropical cyclones (TCs) over the western North Pacific are investigated using forecast records from the Hurricane Weather Research and Forecasting (HWRF) model. Intensity errors increase with the lead time, growing most rapidly during the first 36 hr. Predicted intensity of intense (weak) TCs is generally underestimated (overestimated). The biases mainly stem from the initial intensity bias and the predicted TC fullness (TCF) bias, as well as from the actual intensity change. Initial intensity bias dominates during the first day, explaining 19% of intensity forecast bias. Actual intensity change exhibits a persistent negative correlation with intensity forecast bias, accounting for 16% of the bias within 5 days. In contrast, the influence of TCF forecast bias increases over time and contributes 13% to intensity forecast bias during 4–5 days. Additionally, TC initial intensity, position, and TCF, together with track errors, also affect intensity forecast errors.