Probability Ellipse for Tropical Cyclone Track Forecasts with Multiple Ensembles

Abstract

Pacific as the Regional Specialized Meteorological Center (RSMC) (Regional Specialized Meteorological Center Tokyo-Typhoon Center 1989), has adopted probability circles to represent track uncertainty. A probability circle is defined as the domain that the center of a TC is expected to exist within at a given

Development of Interpretable Probability Ellipse in Tropical Cyclone Track Forecasts Using Multiple Operational Ensemble Prediction Systems

“…The origin, abscissa, and ordinate of the two‐dimensional coordinate system are defined as the t ‐hour ensemble mean, along‐track direction, and cross‐track direction, respectively. The along‐ and cross‐track coordinates are determined in the same manner as the along‐ and cross‐track errors (Heming, 2017; Leonardo & Colle, 2017; Peng et al., 2017), except that the best track positions were replaced by the ensemble means, as in Kawabata and Yamaguchi (2020) (see Figure S3 in Supporting Information ). The along‐track coordinate of the position of an EPS member is positive (negative) if the member is ahead of (behind) the ensemble mean, whereas the cross‐track value is positive (negative) if the ensemble member is to the right (left) of the mean forecast track.…”

“…However, some extreme cases were indicated in the 96-hr and 120-hr forecasts, that is, the cross-track radius was more than four times the along-track radius (Figure S5 in Supporting Information S1). These cases typically result from track bifurcations in the track, as described by Kawabata and Yamaguchi (2020).…”

“…The uncertainty of a TC track forecast may not resemble a circle; in fact, it may be larger in one direction than in another (Hamill et al, 2011;Heming et al, 2019). Recently, researchers introduced the concept of probability ellipses, originally used to explain the track error characteristics (e.g., Neumann & Hope, 1972;Peng et al, 2017), to represent the anisotropy of track uncertainties (e.g., Hansen et al, 2011;Kawabata & Yamaguchi, 2020;Komaromi et al, 2021;Zhang & Yu, 2017). Hansen et al (2011) included the spread along and across a track as potential predictor variables for the along-and cross-track radii, respectively.…”

“…Instead of using eigenvalue decomposition, they used the along‐track and cross‐track variance as the two axes of the ellipse. Kawabata and Yamaguchi (2020) used multiple ensembles, namely, EPSs from the ECMWF, NCEP, United Kingdom Met Office, and Japan Meteorological Agency, to reduce the area of an ellipse with respect to that of a probabilistic circle. Specifically, they selected three parameters of the ellipse, namely the major axis, minor axis, and the orientation of the axes, to minimize the area of the ellipse as well as to include 70% of the members of the ensemble forecast.…”

The Influence of Large-Scale Environment on the Extremely Active Tropical Cyclone Activity in November 2019 over the Western North Pacific

“…The TCs with tropical storm intensity or above (maximum sustained surface wind speed of 35 kt or above) were taken into consideration. Note that there are several organizations that maintain their own historical TC records for the WNP basin (e.g., JTWC, JMA, the Hong Kong Observatory (HKO) of China, and CMA_STI), but previous studies have suggested that the differences in TC tracks and TC genesis are generally small, particularly in the satellite-observation areas [19][20][21][22]. To further improve our analyses, we used the three best track data sets available from the JTWC, CMA_STI, and JMA, and found consistent results for the record high TC frequency in November 2019.…”