(Zhou et al., 2018) performed empirical orthogonal function (EOF) analysis on the 850-hPa EKE averaged over June-November within the domain (0°-35°N, 100°-160°E) during 1958-2014. Accordingly, they identified two dominant modes of interannual variation in synoptic disturbance activities over the WNP. The first mode is called the northeast pattern, which depicts the interannual variation in synoptic disturbance over the subtropical WNP. The second mode is called the southwest pattern, which shows the interannual variation in synoptic disturbance around the Philippines. The regions of red and green contours in Fig. 1 show the regions where the first and second modes explain more than half of the total interannual variance of synoptic disturbances, respectively, i.e., the main regions of the dominant modes. Based on the principal components (PC1 and PC2) of these two modes, which are shown in Zhou et al. (2018, Figs. 2c and d), we select eight years with the highest PC values as positive PC years, and the eight years with the lowest PC values as negative PC years. This process yields positive PC1 years (1962, 1990, 1992, 1996, 1997, 2002, 2004, 2014), negative PC1 years (1963, 1968, 1973, 1977, 1995, 1998, 2008, 2010), positive PC2 years (1964, 1967, 1971, 1974, 1985, 1990, 2009, 2011), and negative PC2 years (1963, 1966, 1968, 1972, 1982, 1997, 2002, 2004).
Figure 1 shows the composite differences in TC occurrence frequency between the positive and negative PC years. The TC occurrence differences associated with PC1 and PC2 are both characterized by a dipole pattern between the subtropical WNP and the northern SCS-Philippines corresponding to the regions with high interannual variabilities of synoptic disturbances for PC1 and PC2, respectively (Zhou et al., 2018). For PC1, there is a positive TC occurrence anomaly over the subtropical WNP and a relatively narrow negative anomaly over the northern SCS (Fig. 1a). Conversely, the PC2-related TC occurrence anomalies are characterized by a negative anomaly over the WNP and a positive anomaly over the northern SCS-Philippines (Fig. 1b). These results suggest that TCs will appear more frequently in the subtropical WNP when both PC1 is positive and PC2 negative, and in the northern SCS-Philippines when both PC1 is negative and PC2 is positive. Therefore, in the following section, we focus on the opposite combination of the northeast and southwest patterns to highlight the anomalies in TC activity.
Figure 2 is a scatterplot of PC1 and PC2 from 1958 to 2014. Based on the values of PC12+PC22, or the Euclidean distance from the coordinate origin, eight years with the greatest distances are selected in the second and fourth quadrants, respectively, to represent the opposite combination of
the northeast and southwest patterns. Accordingly, the years of positive northeast pattern and negative southwest pattern (hereafter referred to as N+S- years) are 1962, 1972, 1976, 1979, 1982, 1997, 2002 and 2004, and the years of negative northeast pattern and positive southwest pattern (hereafter referred to as N-S+ years) are 1964, 1970, 1978, 1983, 1988, 1995, 2008 and 2009, as displayed by the solid circles in Fig. 2. We also conducted the same analysis with different sample sizes (e.g., 5 or 10 years) and the results were qualitatively consistent.
Figures 3a and b show the composite 850-hPa EKE during the N+S- and N-S+ years. As expected, in the N+S-
years, the synoptic disturbances are active over the WNP corresponding to the region with large interannual variability of synoptic disturbances for PC1 [referring to Zhou et al. (2018, Fig. 2a)]; whereas, in the N-S+ years the active region is shifted to the Philippines, coincident with the counterpart for PC2 [referring to Zhou et al. (2018, Fig. 2b)]. Figure 3c displays the difference characterized by a dipole pattern, with a positive anomaly over the WNP and a negative anomaly around the Philippines. The distribution confirms that the synoptic disturbance activities during the N+S- and N-S+ years exhibit the opposite variation between the northeast and southwest patterns.
Figures 4a and b show the composite TC occurrence frequency during the N+S- and N-S+ years. During the N+S- years, TCs mainly occur over the WNP, with the active region expanding northward to 30°N. Meanwhile, in the N-S+ years, TCs mostly appear over the SCS and Philippines, with the active region confined to the south of 30°N but expanded westward to the west of 110°E. Their difference is characterized by a positive anomaly over the WNP and a negative anomaly around the Philippines (Fig. 4c). This distribution is similar to that of the synoptic disturbances shown in Fig. 3, implying there is a consistent change in TC occurrence and synoptic disturbance activity during the N+S- and N-S+ years. Considering that the distribution of TC occurrence frequency is primarily a reflection of TC motion (e.g., Wu and Wang, 2004), the difference in TC occurrence frequency depicted in Fig. 4 may suggest TCs take distinct tracks over the WNP between the N+S- and N-S+ years.
Figures 5a and b display the TC tracks during the N+S- and N-S+ years. During the N+S- years, more TCs tend to recurve northward, thus leading to an increase in TC activity over the WNP. In contrast, in the N-S+ years, more TCs take straight-moving tracks, resulting in an increase in TC activity around the Philippines. Furthermore, the variation in TC track may impact the TC landfall location. The increase in the number of recurving TCs during the N+S- years may cause more TCs to make landfall over the northern coastal regions of the WNP, including eastern China, the Korean peninsula and Japan. Meanwhile, the increase in straight-moving TCs during the N-S+ years may induce more TCs to strike the southern coastal regions of the WNP, including southern China, Vietnam and the Philippines. It is thus anticipated that TCs making landfall over the northern and southern regions may be different between the N+S- and N-S+ years.
If the latitude of 25°N is used to divide the southern and northern regions according to previous studies (e.g., Kim et al., 2008; Li et al., 2017), Figs. 5c and d show the difference in landfalling TC numbers in these two regions during the N+S- and N-S+ years. In the southern region, the number of landfalling TCs is 54 in the N+S- years, apparently smaller than the number (98) in the N-S+ years, which is a decrease of 45%. By comparison, the number of landfalling TCs in the northern region is 45 in the N+S- years, which is an increase of 47% when compared to that in the N-S+ years (24). These results indicate that the number of landfalling TCs in the northern and southern regions differs largely between the N+S- and N-S+ years.
To examine whether these variations are statistically significant, the prevailing track and landfalling number for each case are compared in Fig. 6. For the prevailing track, we use the method of (Wu and Wang, 2004) and (He et al., 2015), who defined the track based on the proportion of TCs over the region of concern to TCs over the source region. We define the source region as the domain (7.5°-25°N, 120°E-180°), and the regions of concern as (7.5°-22.5°N, 107.5°-120°E) and (25°-45°N, 120°-145°E) for straight and recurving tracks, respectively.
Figure 6 confirms the significant differences in TC tracks and landfalls between the N+S- and N-S+ years, despite the case-to-case variations. For the straight-moving track, the mean proportion is 16.5% in the N+S- years, but this increases to 33.8% in the N-S+ years, and their difference (-17.3%) is significant at the 99% confidence level. In contrast, the mean proportion for the recurving track is 59.2% in the N+S- years, but this reduces to 41% in the N-S+ years, and their difference (18.2%) is also significant at the 99% confidence level. In good agreement with these differences in TC tracks, landfalls also show distinction between the N+S- and N-S+ years (Figs. 6c and d). The differences in both the tracks and landfalls between the N+S- and N-S+ years are significant at the 95% confidence level.
We also analyzed TC track and landfall variations associated with PC1 and PC2, respectively, and the result (not shown) indicated that these variations are similar to, but weaker than, those associated with the combination of PC1 and PC2. This result confirms that the combination of PC1 and PC2 can highlight the differences in TC tracks and landfalls.