Kawabata and Yamaguchi (2020)
Kawabata, Y., and M. Yamaguchi, 2020: Probability ellipse for tropical cyclone track forecasts with multiple ensembles. J. Meteor. Soc. Japan, 98, 821-833.
Graphical Abstract with highlights
The effectiveness of the probability ellipse for tropical cyclone (TC) track forecasts is investigated with multiple ensembles from the Japan Meteorological Agency (JMA), the European Centre for Medium-Range Weather Forecasts (ECMWF), the U.S. National Centers for Environmental Prediction (NCEP), and the Met Office in the United Kingdom (UKMO), for all TCs from 2016 to 2018. The multiple ensembles composed of these four global ensembles are capable of predicting the situation-dependent uncertainties of TC track forecasts appropriately in both the along-track (AT) and cross-track (CT) directions.
In recent decades, various TC track forecasting techniques have been developed, and the value of dynamical ensemble forecasts to estimate the uncertainties associated with TC track forecasts has been studied. However, many operational TC forecasting centers still don’t use ensemble-based forecast uncertainty information in their operational TC warnings. JMA is one of the leading meteorological centers in this area and has adopted multiple ensemble-based probability circle in June 2019 (Fukuda and Yamaguchi 2019). They found that the mean probability circle radius and, therefore, the mean area differed little, whether determined by the conventional statistical, single-ensemble, or multiple-ensemble method. Thus, we introduced an elliptic shape and examined how much the forecast area of the probability ellipse decreases, compared with that of the circle.
Figure 1 shows examples of TC track predictions along with the probability circle and ellipse, including 70 % ensemble members. Individual members of each TC forecast have variations, whose extent is represented as an ensemble spread. The ensemble spread in Fig. 1a is larger in the AT direction than in the CT direction, indicating that the uncertainty as to when the typhoon will approach (i.e., its speed of movement) is larger than the uncertainty as to where it will move (i.e., its direction of movement). In contrast, the ensemble spread in Fig. 1b is larger in the CT direction than in the AT direction, indicating that the direction of TC movement is more uncertain than the speed of that.
Table 1 shows the area ratio of the probability ellipse to the circle, calculated by the total area of all TCs during 3-year verification period. All of the evaluated area ratios are smaller than 1. The probability ellipse can potentially reduce area by 16, 15, 24 %, on average, at the forecast time (FT) of 3, 4, and 5 days. If the area ratios are evaluated only for cases that the ratio of the ensemble spread in the AT direction to that in the CT direction (AT/CT) is 1.5 or more, the area is reduced by 28, 24, and 31 % at the FT of 3, 4, and 5 days. Similarly, for cases that the ratio of the ensemble spread in the CT direction to that in the AT direction (CT/AT) is 1.5 or more, the area is reduced by 14, 12, and 28 % at the FT of 3, 4, and 5 days. In particular, the area of ellipse is smaller when ensemble members are scattered in AT direction.
The use of a probability circle involves the implicit assumption of an isotropic error distribution, whereas the introduction of the probability ellipse makes it possible to provide information as to which is more uncertain; the direction or the speed of TC movement. This additional information would be useful for decision-makers and disaster preparedness planning communities. Moreover, narrowing warning areas of TC track forecasts by the probability ellipse enables us to enhance disaster prevention/mitigation measures. The results of this study do not immediately lead to the JMA adopting the probability ellipse instead of the circle. Future challenges include training forecasters on the utilization of ensembles, as use of advanced probabilistic information requires changes in operational working practice, and developing ensemble-based and hazard-based forecasts for strong winds, heavy precipitation, and storm surge as TC-related disaster may happen outside the probability circle or ellipse.
Fukuda, J., and M. Yamaguchi, 2019: Determining 70 percent probability-circle radii of tropical cyclone track forecasts with multiple ensembles. SOLA, 15, 250−256. https://doi.org/10.2151/sola.2019-045