-
During 1975–2020, there were 470 RI-TCs over the WNP, representing an annual mean of 10.2 and its time series shows an obvious upward trend, which is significant at the 90% confidence level (Fig. 1a). The interannual variation is large, ranging from 4 in 1978 to 18 in 2015. Out of the 470 RI-TCs, 289 make landfall along the coast of East Asia. The mean annual number of landfalling RI-TCs is 6.3 and its time series also shows a significant upward trend (confidence level of 99%) (Fig. 1b), which is partly related to the upward trend of the annual number of RI-TCs as suggested by the high correlation (r = 0.61) between the two time series.
Figure 1. Time series of (a) the annual number of RI-TCs, (b) the annual number of landfalling RI-TCs, (c) percentage of RI-TCs that make landfall, and (d) annual PDI (units: 104 kt3). The dashed lines indicate the linear trends.
To remove the effect of the overall RI-TC activity on the number of landfalling RI-TCs, the variation in the percentage of RI-TCs that make landfall is examined. On average, 62.7% of the RI-TCs make landfall along the coast of East Asia and the percentage shows a gradual increase as indicated by its significant upward trend (confidence level of 99%) (Fig. 1c). The lowest percentage (22.2%) is found in 1986, with only 2 out of the 9 RI-TCs making landfall. In 2020, all the RI-TCs made landfall along the coast of East Asia, giving a percentage of 100%, which was the highest during the study period. These results demonstrate that the increase in the annual number of landfalling RI-TCs is not only related to the increase of RI-TCs over the entire WNP but also the percentage of these TCs making landfall. Guan et al. (2018) also found an increase in the percentage of TCs with at least typhoon intensity making landfall during 1974–2013. The correlations of the landfalling RI-TCs frequency with the total RI-TCs frequency and percentage of RI-TCs that make landfall are similar (correlation coefficients being 0.61 and 0.62, respectively), and the relative weight analysis shows that their contributions to the R-squared values of the multiple regression model are also similar (49.8% and 50.2%, respectively), suggesting that the two factors are of equal importance in controlling the annual number of landfalling RI-TCs.
To better understand the change in the percentage of RI-TCs that make landfall, it is useful to investigate the characteristics of the RI-TCs with and without landfall. An examination of the tracks of non-landfalling RI-TCs shows that most of these TCs follow the recurving track, move towards the ocean southeast of Japan and dissipate over water without landfall (not shown). Only a few move towards the coast of East Asia and dissipate over water without landfall. Therefore, the percentage of RI-TCs that make landfall partly depends on the genesis location and the subsequent track. Normally, a TC that forms further to the west has a higher chance to make landfall. Indeed, the percentage of landfalling RI-TCs is significantly correlated with the annual mean longitude of genesis location of all RI-TCs (including both landfalling and non-landfalling) (r = –0.48, confidence level of 95%). In addition, the mean genesis longitude of landfalling RI-TCs (141.5°E) is further to the west than that of non-landfalling RI-TCs (150.9°E) and the difference is significant at the 99% confidence level. The latitude and longitude of genesis location of all RI-TCs show an increasing trend (confidence level of 98%) and a decreasing trend (confidence level of 97%), respectively, indicating a northwestward shift in genesis location (Figs. 2a and 2b), consistent with the result from Zhao et al. (2018). A similar trend in genesis location is found for landfalling RI-TCs. However, no trend is detected for the genesis longitude of non-landfalling RI-TCs, suggesting that the genesis locations of non-landfalling RI-TCs are generally confined to a longitudinal band and a westward shift in genesis location will, therefore, increase the chances for an RI-TC to make landfall.
Figure 2. Time series of the mean (a) latitude and (b) longitude of genesis location and the mean (c) latitude and (d) longitude of LMI location and (e) the mean LMI (units: kt) of RI-TCs (including both landfalling and non-landfalling). The dashed lines indicate the linear trends.
While the chances of an RI-TC making landfall is partly related to the genesis longitude, the actual number of landfalling RI-TCs also depends on the total number of RI-TCs over the WNP. As discussed above, the annual number of landfalling RI-TCs is highly correlated with the total number of RI-TCs (r = 0.61) but its correlation with the mean genesis longitude is less significant (r = –0.20). Using these two factors as predictors for the annual number of landfalling RI-TCs, the multiple regression model gives a correlation of 0.70, and the contributions of the total number of RI-TCs and mean genesis longitude to the R-squared values are 84.4% and 15.6%, respectively. Thus, the upward trend in landfalling RI-TCs is mainly due to the increase in RI-TCs over the WNP and the role of the northwestward shift in genesis location is to increase the percentage of RI-TCs that make landfall along the coast of East Asia.
The annual number of landfalling RI-TCs along the coast of East Asia has been shown to have a significant increasing trend. To better measure the destructive potential in coastal areas, the annual PDI, which depends on both landfall frequency and intensity, is also examined. The annual PDI shows a significant increasing trend at the 99% confidence level (Fig. 1d), indicating an increasing threat posed by RI-TCs to the coastal areas of East Asia. The climatological mean of the annual PDI is 554.9 × 104 kt3. Over the study period, the annual PDI has increased by 160%. The interannual variation of annual PDI is very large, with the lowest value (127.7 × 104 kt3) in 1983 and the highest value (1363.7 × 104 kt3) in 2006. Mei and Xie (2016) showed that the increase in the intensity of landfalling typhoons is due to enhanced intensification rates. Guan et al. (2018) also examined the PDI at land, defined as the sum of PDI when the TC center is over land for each TC, and found an increasing trend over the period 1974–2013. Our result is therefore consistent with that of Guan et al. (2018) although the definitions of PDI are different and only the RI-TCs are considered in the present study.
The increasing trend of the annual PDI is obviously related to the increase in the annual number of landfalling RI-TCs. Since the annual PDI also depends on landfall intensity, the contribution of landfall intensity to the long-term change of annual PDI is investigated. The time series of annual mean landfall intensity also shows an increasing trend (Fig. 3a), but it is not significant (confidence level of 82%). It should be noted that there was only one landfalling RI-TC in 1978, making landfall in the Philippines with an intensity of 125 kt and giving an exceptionally high annual mean landfall intensity. If this year is excluded in the trend analysis, the confidence level rises to 92%, which is statistically significant. Therefore, the increase in annual mean landfall intensity could contribute to the increasing trend of annual PDI.
Figure 3. Time series of (a) the annual mean landfall intensity (units: kt) and (b) the time interval (units: h) between LMI occurrence and landfall (original and adjusted). The dashed lines indicate the linear trends.
The annual PDI is significantly correlated with the annual number of landfalling RI-TCs and annual mean landfall intensity (correlation coefficients being 0.66 and 0.53, respectively, both are significant at the 99% confidence level). Using these two factors as predictors for the annual PDI, the multiple regression model gives a correlation of 0.92 and their contributions to the R-squared values are 59.1% and 40.9%, respectively. Thus, the increasing trend of annual PDI is mainly due to the increase in the annual frequency of landfalling RI-TCs, with the increase in annual mean landfall intensity playing a secondary role.
The intensity at landfall may be related to the LMI location of a landfalling RI-TC. If the LMI occurs close to the coast, weakening may not be significant during the time between LMI and landfall so that the intensity at landfall may be close to its LMI. In some cases, a TC actually attains its LMI near its landfall so that the landfall intensity is equal to the LMI. This situation is usually found for those RI-TCs making landfall in the Philippines. Thus, an RI-TC with its LMI location close to the coast may pose a severe threat to the coastal area. In contrast, a TC with its LMI location far away from the coast may substantially weaken before making landfall so that the intensity at landfall is lower. Therefore, the long-term change in LMI location may exert a significant influence on the landfall intensity and hence the annual PDI. Indeed, the trend analysis of the LMI occurrence of RI-TCs (both landfalling and non-landfalling) shows an increasing trend in latitude (confidence level of 79%) and a decreasing trend in longitude (confidence level of 99%), indicating a possible northwestward shift in the LMI location (Figs. 2c and 2d), consistent with the results from the previous studies (Park et al., 2014; Zhao et al., 2018; Wang and Toumi, 2021). This implies that the LMI location has moved closer to the coast of East Asia, which may partly explain the increase in landfall intensity. Park et al. (2014) also showed that the LMI location of TCs with at least tropical storm intensity has moved closer to East Asian coastlines over the period 1977–2010, which is the main reason for the increase in landfall intensity over East China, Korea, and Japan. In addition, the annual mean LMI of RI-TCs also shows a significant increasing trend (confidence level of 99%) (Fig. 2e), which may contribute to the increase in landfall intensity. Song et al. (2021) also found a similar trend, which is primarily linked to a significant increase in the mean intensification rate prior to the LMI.
To further investigate the possible relationship between LMI location and landfall intensity, the variation in the time interval between LMI occurrence and landfall (the time when LMI occurs minus time at landfall) is examined. A shorter time interval generally implies an LMI location closer to the coast. The landfall intensity is negatively correlated with the time interval (r = –0.51) suggesting that the landfall intensity is generally higher if the LMI location is closer to the coast. The trend analysis shows an insignificant trend in the annual mean time interval (Fig. 3b). The average LMI location of landfalling RI-TCs is near (19.4°N, 126.4°E), which is close to the coast of the Philippines and Taiwan Island, but far away from the coasts of Japan, the Korean Peninsula, and East China. The LMI usually occurs at a lower latitude because a TC generally experiences higher VWS and cooler water as it moves northward, leading to weakening. Therefore, the mean time intervals between LMI occurrence and landfall for the TCs making landfall in Japan, the Korean Peninsula, and East China (69.0 h, 77.3 h, and 54.9 h, respectively) are longer than those making landfall in the Philippines and Taiwan Island (16.4 h and 23.9 h respectively). The mean time intervals for South China and Vietnam (21.4 h and 17.3 h, respectively) are also shorter because the LMI is usually located over the SCS. Thus, the annual mean time interval depends on the preferred landfall regions in that year. A year with a higher portion of TCs making landfall in the northern domain generally yileds a longer mean time interval.
For example, the mean time interval in 1983, which consists of four landfalling RI-TCs (Japan: 2; East China: 1; South China: 1), is 100.5 h. In contrast, the mean time interval is generally shorter for a year in which the preferred landfall regions are to the south. A typical example occurred in 1980 when the four landfalling RI-TCs had a southern bias (the Philippines: 2; Taiwan Island: 2), thus resulting in a very short mean time interval (4.5 h). To remove this effect, the anomaly of the time interval for a region is obtained by subtracting the time interval from the climatological mean corresponding to that region. The time series of the adjusted time interval, which should reflect the actual change in the time interval between LMI occurrence and landfall, shows a significant decreasing trend (confidence level of 93%) (Fig. 3b). In other words, the time interval has actually shortened, leading to a higher landfall intensity.
To investigate the changes of annual PDI in different regions, the East Asia region is divided into three sub-regions and the landfalling TCs are accordingly grouped as south TCs (South China, Vietnam, and the Philippines), middle TCs (East China and Taiwan Island), and north TCs (Japan and the Korean Peninsula). The annual PDI of south TCs shows a significant upward trend (confidence level of 99%), which is related to the increasing trends of the annual frequency (confidence level of 86%) and the annual mean landfall intensity (confidence level of 97%), suggesting that the latter is the dominant factor (Table 1). A significant upward trend (confidence level of 99%) is also found for north TCs, which is mainly due to the increase in annual frequency (confidence level of 99%). The role of landfall intensity is minimal, as no trend exists in the annual mean landfall intensity. The trend in the annual PDI of middle TCs (confidence level of 63%) is not as significant as those of south and north TCs. It can be concluded that the increase in annual PDI in East Asia is mainly due to the increasing frequency of RI-TCs making landfall in the southern (South China, Vietnam, and the Philippines) and northern parts (Japan and the Korean Peninsula) of East Asia and the increase in annual mean landfall intensity for the former also plays an important role.
Annual PDI Annual number of
landfalling RI-TCsAnnual mean landfall
intensityTime interval between
LMI and landfallAll TCs Upward
(99%)Upward
(99%)Upward
(82%)Downward
(98%)North TCs Upward
(99%)Upward
(99%)Upward
(24%)Downward
(37%)Middle TCs Upward
(63%)Upward
(48%)Upward
(66%)Downward
(39%)South TCs Upward
(99%)Upward
(86%)Upward
(97%)Downward
(95%)Table 1. Linear trends of the annual PDI, the annual number of RI landfalling TCs, the annual mean landfall, and the time interval between LMI and landfall intensity for all TCs, north TCs, middle TCs, and south TCs (see text for the definitions). Percentages in the parenthesis are the confidence levels, with those ≥ 90% in bold.
-
The important stages of a landfalling RI-TC include genesis, RI, LMI, and landfall. It is useful to investigate the spatial distribution of the trends in each of these stages of RI-TCs (both with landfall or without landfall) and their possible impacts on the annual PDI. The major genesis area of RI-TCs is near (5°−15°N, 130°−160°E) and the spatial distribution of genesis frequency shows a decreasing trend in the southeastern part of the WNP (5°−10°N, 145°−180°E) and an increasing trend north and northwest of the major genesis area, indicating a northwestward shift in genesis location (Fig. 4a), which is consistent with the increasing trend (northward shift) in mean genesis latitude and the decreasing trend (westward shift) in mean genesis longitude of RI-TCs (see Figs. 2a and 2b). The shift of genesis location towards the coast of East Asia increases the chance of an RI-TC to make landfall and hence the percentage of landfalling RI-TCs as indicated by the significant correlation between the mean genesis longitude and the percentage of RI-TCs that make landfall (r = –0.48, confidence level of 95%). Indeed, such an increase in genesis frequency leads to an increasing frequency of the RI-TCs that form at higher latitudes and follow the recurving path towards Japan and the Korean Peninsula or a straight path towards Taiwan Island and east China (Fig. 4b). An increasing frequency of RI-TCs moving across the SCS and making landfall in south China is also observed. These results are consistent with the upward trend in the annual number of landfalling RI-TCs along the coast of East Asia (see Table 1).
Figure 4. Spatial distribution of the linear trends in (a) genesis frequency, (b) track density, (c) RI occurrence, (d) LMI occurrence, and (e) annual PDI (units: kt3 yr−1) of RI-TCs (including both landfalling and non-landfalling). Red and blue shadings indicate the areas with positive and negative trends significant at the 90% confidence level respectively. The purple dashed rectangular box in (a) and (b) and the orange dashed rectangular box in (c) indicate the major area for genesis and RI occurrence respectively. The purple dots in (d) indicate the LMI locations of all the landfalling RI-TCs with the time interval between LMI occurrence and landfall ≤ 24 h occurring between 1975 and 2020.
Because of the upward trend in the annual frequency of RI-TCs, the frequency of occurrence of RI should increase and the spatial distribution of RI occurrence shows an increasing trend in most parts of the WNP (Fig. 4c). Such trends are more significant over the SCS, northeast of Taiwan Island, and southeast of Japan, which are outside the major RI region (10°−23°N, 123°−150°E). This implies a westward and northward expansion of the RI region and a higher frequency of RI-TCs undergoing RI near the coast of East Asia. The increasing trend in the RI occurrence also implies an increasing trend in the RI-TCs attaining the maximum intensity near the coast of south China and Vietnam (Fig. 4d). Liu and Chan (2020) also found an increase in maximum landfall intensity near the south China coast during 2012–18, which is related to the increase in the annual frequency of the RI-TCs making landfall in south China. An increasing trend in LMI occurrence is also observed near East China.
To further examine the impact of the change of LMI location on annual PDI, the landfalling RI-TCs with a short time interval between LMI occurrence and landfall (≤ 24 h) is examined. This type of landfalling RI-TC is of particular importance because it may make landfall shortly after the occurrence of LMI, leaving very little time for typhoon preparation and evacuation. The areas with the increasing trend in LMI occurrence generally coincide with the major region of the occurrence of these RI-TCs (Fig. 4d). Thus, the westward shift in LMI locations leads to an increasing frequency of the landfalling RI-TCs with a short time interval between LMI occurrence and landfall (confidence level of 91%) (Fig. 5a) as well as an increase in their mean landfall intensity (confidence level of 99%) (Fig. 5b), leading to the higher annual PDI in these regions (Fig. 4e). Note also the increasing trend in annual PDI near the coast of Japan and the Korean Peninsula, which is largely related to the increasing frequency of RI-TCs affecting these regions.
Annual PDI | Annual number of landfalling RI-TCs | Annual mean landfall intensity | Time interval between LMI and landfall | |
All TCs | Upward (99%) | Upward (99%) | Upward (82%) | Downward (98%) |
North TCs | Upward (99%) | Upward (99%) | Upward (24%) | Downward (37%) |
Middle TCs | Upward (63%) | Upward (48%) | Upward (66%) | Downward (39%) |
South TCs | Upward (99%) | Upward (86%) | Upward (97%) | Downward (95%) |