In Wang (2015), Figs. 14 and 15 are incorrect. According to the caption, Fig. 14 shows an example of rainfall distribution in Taiwan using 13 thresholds that yield fixed areal coverage in percent (from 99% to 1%, inferred from the fraction of rain gauge sites). Subsequently, Fig. 15 presents the averaged threat score (TS) of day-1, day-2, and day-3 QPFs by the 2.5-km Cloud-Resolving Storm Simulator (CReSS), as a function of areal coverage (i.e., event base rate), for groups with different observed rainfall amount (“top 5” then “A” to “D,” from the most to least rain). However, because of a miscommunication, the percentages of rain areas were computed using only the number of sites with nonzero rainfall, rather than the total number of all the sites in Taiwan, and thus these two figures are incorrect in Wang (2015). The correct Figs. 14 and 15 are presented in this corrigendum.
As in Fig. 3c, but plotted using 13 rainfall thresholds (mm, scales on right) that yield rain areas covering 99%, 95%, 90%–10% (every 10%), 5%, and 1% of all rain gauge sites over Taiwan (as labeled next to the color scale).
Citation: Monthly Weather Review 144, 8; 10.1175/MWR-D-16-0180.1
As in Fig. 13, but for averaged TS for (a) day 1 to (c) day 3 for groups A–D and the top-5 cases at rainfall thresholds covering certain percentages of all sites over Taiwan, from 99%, 95%, and 90% (low threshold) to 10%, 5%, and 1% (high threshold). The averaged rainfall thresholds (mm) for the five groups are also given at the bottom.
Citation: Monthly Weather Review 144, 8; 10.1175/MWR-D-16-0180.1
The differences before and after the correction are small for both Figs. 14 and 15. This is expected for Fig. 14, as the example was very rainy and the rain area at the lowest threshold (0.05 mm) covered nearly all of Taiwan. For Fig. 15, its purpose is to demonstrate that the dependency of categorical statistics such as the TS on the overall rainfall amount (i.e., higher scores and better skill for larger events) holds true even when the thresholds are adjusted to give fixed rain-area size (and areal coverage). In other words, the phenomenon of higher scores for larger events arises not only due to the positive correlation between the scores and rain-area size, but also because the models [at least the one studied in Wang (2015)] indeed tend to be more skillful in predicting larger events, which are often associated with a closer typhoon track (such as a direct hit) and/or more favorable conditions at large and synoptic scale. As shown in Fig. 15 here, the TS curves for larger events remain above those for smaller ones in most cases (except for curve “B” on day 3) although they are now closer over the size range from about 80% to 20% (across mainly the middle thresholds). Therefore, the relevant conclusion in Wang (2015) remains valid and unchanged despite the error.
Acknowledgments
The contribution from Mr. S.-C. Wang and Ms. Y.-W. Wang to this corrigendum is acknowledged. This study is jointly supported by the Ministry of Science and Technology of Taiwan under Grants MOST-103-2119-M-003-001-MY2, MOST-103- 2625-M-003-001-MY2, MOST-105-2111-M-003-003-MY3, and MOST-105-2625-M-003-001.
REFERENCE
Wang, C.-C., 2015: The more rain, the better the model performs—The dependency of quantitative precipitation forecast skill on rainfall amount for typhoons in Taiwan. Mon. Wea. Rev., 143, 1723–1748, doi:10.1175/MWR-D-14-00137.1.
