Investigation of TC track uncertainty using multiple ensembles for the official TC forecast

The official tropical cyclone information in Korea includes a deterministic forecast position of TC center and its uncertainty with the 70% probability circle, which is statistically determined by previous 3 year’s operational track errors. Therefore, the current probability circle does not represent situational uncertainty. In this study, it is investigated for using an ensemble prediction system (EPS) to represent the TC position uncertainty with three different methods: circle (CIR), ellipse with an along-track and a cross-track axes (EAC), ellipse with eigenvector axes (EEV). Five single EPSs, ECMWF, NCEP, UKMO-UM, JMA and KMA-UM, and two multiple ensembles, a simple one (SME) and a calibrated one (CME) which coincides the ensemble means, were evaluated. The methods and the ensembles were verified for 5 days with the hit rate which is defined as the percentage of the observed TC central positions within circles or ellipses.
In order to verify the new probability areas as well as the operation, the hit rate which is defined as the percentage of the observed TC central positions within 70% probability circle or ellipses were used. The operational radii have over 70% hit rate, around 0.8 for all forecast times. It means that the official forecast skill is getting better year by year and the current circle is overestimated. EPS based circle or ellipse showed better performance apart from the EAC. In more detail, the CME for both circle and ellipse method outperformed the operational method until 48 forecast hours. Since the five single EPSs were under-spread at this time, the multiple ensembles could overcome this shortage. After 72 forecast hours, SME and CME are too overspread, so that a single EPS is more likely to be consistent with the 70% probability area.
Although it is definitely sure that the EPS based one is better, there are still limitations to use them. It is difficult to say which method is the best because performance of methods is different according to the forecast time and to get other organizations’ EPS data in real time. Nevertheless, utilizing ensemble for TC track is valuable information since EPS can provide the best method for estimating uncertainty.