Ensemble mean forecasts for typhoon remain an unresolved challenge throughout the world. The critical problem is the traditional arithmetic mean (AM) as a simple point-wise statistic disregards the geographical displacement of typhoon structure in individual ensemble members. This results in the over-smoothing and distortion of typhoon structure, particularly an underestimation of typhoon intensity forecast. To resolve this problem, our study aims to develop a new ensemble mean scheme for typhoon based on the feature-oriented mean (FM). FM overcomes the limitations of AM, obtaining united forecasts for typhoon track, intensity and structure. This study will focus on the following three aspects: Develop the FM algorithm to the application for typhoon forecasts; Comprehensively evaluate the performance of FM in improving the typhoon ensemble forecast; Quantitatively estimate the relative contribution of position- and amplitude-related errors in typhoon forecasts, and examine the main influencing factors. Our study is expected to resolve the difficulty in the ensemble mean forecast of typhoon by developing a new feasible ensemble mean scheme. This study has the potential to provide new ideas for the post-processing of typhoon ensemble forecasts and improve the understanding of the sources of uncertainty in typhoon forecasting.
 

Ensemble Forecast; Predictability; Prediction Error