Clim Dyn (2017) 48:1963–1985 DOI 10.1007/s00382-016-3185-5

Variations in the power dissipation index in the East Asia region Kin Sik Liu1 · Johnny C. L. Chan1 

Received: 11 September 2015 / Accepted: 19 May 2016 / Published online: 25 May 2016 © Springer-Verlag Berlin Heidelberg 2016

Abstract  This study examines the variability of the power dissipation index (PDI) for different regions of the East Asia region during the period 1960–2013. The annual PDI (APDI) in the region is calculated as the sum of the PDI, defined as the cube of the maximum sustained wind speed at landfall of each tropical cyclone (TC) making landfall at that region. Upward and downward trends in APDI are found in the northern and southern parts of East Asia respectively, suggesting a possible northward shift in TC landfall locations. Interdecadal variations of the APDI can also be found in some regions. The APDI in various regions show a close relation with the PDI distribution over the western North Pacific (WNP) with three characteristic patterns. The ENSO and basin-wide mode represents the PDI patterns associated with ENSO events and the overall PDI over the WNP. The east–west dipole mode and the north–south dipole mode denote the east–west and north– south shifts of PDI respectively. Based on the steering flow (average winds within the 850–300 hPa layer) near the East Asian coast, a three-cell model for TC landfall in East Asia is proposed, which corresponds to three major modes of the atmospheric circulation in the WNP. Each of these modes shows an anomalous circulation located east of Taiwan, east of Japan and the South China Sea, respectively and each of which has a significant impact on the APDI in some regions along the coast of East Asia. A northward shift in the APDI along the East Asian coast is identified in the period 1997–2013 as a result of the change in steering flow

* Kin Sik Liu [email protected] 1



Guy Carpenter Asia‑Pacific Climate Impact Centre, School of Energy and Environment, City University of Hong Kong, Tat Chee Ave., Kowloon, Hong Kong, China

pattern, northward shift in TC genesis location and weaker vertical wind shear over the ocean near the coastal areas. Keywords  Tropical cyclone landfall · Climate variability · Power dissipation index · East Asia

1 Introduction Tropical cyclone (TC) landfalls cause severe damage to the coastal areas of East Asia each year. Investigating the trends and variations of TC landfalling activity and understanding the underlying mechanisms are therefore very important. The effect of global warming on TC activities such as frequency, intensity and landfall has also received much attention during the last decade (Knutson et al. 2010). A better understanding of the variations of TC landfalling activity may provide better projections of the future frequency of TC landfall and its intensity. While many studies have examined its variability in individual regions in East Asia [Japan (Park et al. 2011; Goh and Chan 2012; Grossman et al. 2014), Korea (Park et al. 2011; Choi et al. 2010a, b; Goh and Chan 2012), China (Liu and Chan 2003; Li and Duan 2010; Zhang et al. 2011; Lu and Zhao 2013), Taiwan (Tu et al. 2009) and the Philippines (Kubota and Chan 2009)], very few have examined the variability of TC landfall for the entire East Asia region (Chan and Xu 2009). Chan and Xu (2009) found that the annual number of landfalling TCs in each sub-region of East Asia exhibits large inter-annual and inter-decadal variations but the relationship between the TC landfalling activities in different regions has not been discussed. A further investigation is therefore required to get a more complete picture of the variations of TC landfall for all of East Asia.

13

1964

Landfall frequency is usually used to measure the TC landfalling activity in a region. However, this cannot reflect the likely damage brought by the TCs, which is more related to the landfall intensity. The damage caused by a single strong typhoon is much larger than that by several weak storms. Most studies of TC landfall focus on either landfall frequency or landfall intensity and only a few studies have considered both (Park et al. 2011; Zhang et al. 2011). Since the actual monetary loss due to strong wind roughly varies with the cube of the wind speed (Southern 1979), the power dissipation index (PDI), defined as the cube of the maximum sustained wind speed (Emanuel 2005), is commonly used as a measure of TC destructive potential for the ocean basins (Emanuel 2005; Wu et al. 2008) as well as for land areas (Park et al. 2011). The objective of this study is therefore to examine the variability of both TC landfall frequency and intensity in different regions of East Asia using the PDI as the measure of the damages associated with TC landfall. The data and methodology employed in this study are described in Sect. 2. Section 3 discusses the variations of the PDI in various regions of East Asia. Changes in spatial distribution of PDI are examined in Sect. 4. Large-scale atmospheric circulation patterns responsible for variations in the PDI are proposed in Sect. 5 and possible reasons for a northward shift in the PDI are presented in Sect. 6. The summary and discussion are then given in Sect. 7.

2 Data and methodology 2.1 Data The best-track datasets for TCs over the western North Pacific (WNP) are obtained from the Joint Typhoon Warning Center1 (JTWC), China Meteorological Administration2 (CMA) (Ying et al. 2013) and Taiwan Central Weather Bureau. The maximum sustained wind speeds obtained from these datasets are 1-min, 2-min and 10-min averaged respectively. Since the data before 1960 are likely to have larger uncertainties, only the data between 1960 and 2013 are employed. The reason for using these datasets is given in Sect. 2.2. The Niño-3.4 sea-surface temperature (SST) anomaly, defined as the average SST anomalies over the region 5°N– 5°S and 170°–120°W, is obtained from the Climate Prediction Center website.3 The Pacific decadal oscillation (PDO) index is obtained from the Joint Institute for the Study of the Atmosphere and Ocean of the University of Washington

1

  https://metocph.nmci.navy.mil/jtwc/best_tracks/.   http://tcdata.typhoon.gov.cn/en/index.html. 3   http://www.cpc.noaa.gov/data/indices/.

K. S. Liu, J. C. L. Chan

(Mantua et al. 1997), and is defined as the leading principal component of monthly sea surface temperature anomalies in the North Pacific Ocean poleward of 20°N. Monthly zonal and meridional winds at 850, 700, 600, 500, 400, 300 and 200 hPa are extracted from the ECMWF ERA-40 reanalysis dataset,4 which has a horizontal resolution of 2.5° latitude × 2.5° longitude. The average winds within the 850–300 hPa layer are used to examine the steering flow pattern while the winds at 850 and 200 hPa are used to calculate the vertical wind shear. Because this dataset is only available up to Aug 2002, the ECMWF Interim Re-Analysis (ERA-Interim) dataset, with a horizontal resolution of 1.5° latitude × 1.5° longitude, from Sep 2002 to Dec 2013 are used (Simmons et al. 2007). The combined dataset therefore covers the full study period 1960–2013. 2.2 Methodology 2.2.1 Tropical cyclone landfall The East Asia region is divided into eight individual regions: Japan, the Korean Peninsula, Zhejiang, Fujian, Taiwan, Guangdong, Vietnam and the Philippines (Fig. 1). Jiangsu and Shandong are not included because of the small annual number of landfalling TCs. In this study, a TC is considered to affect a region if it makes landfall in this region or comes within 100 km of the coastline as a TC may cause significant damage when it passes close to the coast. The value of 100 km is chosen as it is greater than or equal to the radius of maximum winds of most TCs (Weatherford and Gray 1988). If a TC makes actual landfall (with its center crossing the coastline), the intensity at landfall is defined as the 6-hourly (or 3-hourly if available) best-track intensity at or just prior to landfall. For a TC passing close to the coast without actual landfall, the intensity “at landfall” is estimated as the 6-hourly (or 3-hourly if available) best-track intensity closest to the coastline. Only those TCs with at least tropical storm intensity (intensity ≥34 kt) at landfall are counted to minimize subjectivity in the identification of tropical depressions. If a TC makes multiple landfalls in the same region, it is only counted once. Three sets of best-track data are employed in this study. Because a meteorological centre should have a better estimate of landfall intensity of a TC making landfall in its own region as it should have more observations, the landfall intensity of a TC making landfall at Zhejiang, Fujian or Guangdong is based on the best-track data from the CMA. Best-track data from the Taiwan Central Weather Bureau are used for the landfall intensity of the TCs making landfall in Taiwan. For other regions (Japan, the Korean Peninsula, Vietnam and the Philippines), best-track data from

2

13

4

  http://apps.ecmwf.int/datasets/.

1965

Variations in the power dissipation index in the East Asia region Fig. 1  Map showing the locations of the study regions. Japan, the Korean Peninsula, Zhejiang, Fujian, Taiwan, Guangdong, Vietnam and the Philippines

the JTWC are used rather than the best-track data from the Regional Specialized Meteorological Center (RSMC)— Tokyo Typhoon Center because the maximum sustained wind speed data are not available for the period prior to 1977. Data from other countries are also not easily accessible. It should be noted that although the definitions of maximum sustained wind speed are different for these datasets, no conversion is made because no direct quantitative comparison is made between the PDI in different regions. 2.2.2 Definition of power dissipation index As mentioned in the Introduction, the damage from TC landfall is measured as the PDI, which is defined as the cube of the maximum sustained wind speed at landfall (Emanuel 2005). Unless otherwise specified, the unit of PDI is 105 knots3 in this paper. The annual PDI (APDI) for a region is then calculated as the sum of the PDI of each TC making landfall at that region, which therefore depends on both the number of landfalling TCs and their intensities at landfall. Note, however, that because the APDI is equal to the cube of the intensity, the APDI will be eight times higher if the intensity is doubled, which highlights the importance of landfall intensity in consideration of the APDI.

2.2.3 Trends and interdecadal variations For each time series of the APDI of the eight regions, the long-term trend is examined by calculating the linear correlation with time. The regime detection algorithm developed by Rodionov (2004) is adopted to detect the regime shifts in a time series. This method identifies a significant change in the sequential running means with a certain cut-off length based on the Student’s t test. If the difference in the two means is significant at a certain confidence level, then a regime shift is identified. Details of the algorithm can be found in Rodionov (2004). In this study, a cutoff length of 10 years is used and only those regime shifts significant at the 90 % confidence level or above are considered. 2.2.4 Power dissipation index pattern In order to investigate the changes in the distribution of PDI over the entire WNP, the PDI patterns in each year are constructed based on the JTWC dataset. The region 0°–45°N, 100°–180°E is divided into 5° latitude × 5° longitude grid boxes. For each box, the 6-hourly best-track positions located in this box are extracted for all the TCs in a year. The PDI of each position is calculated and then summed

13

1966

up to give the yearly total PDI in this box. The anomalies of PDI at each box are then obtained to get the anomalous PDI pattern.

3 Variations of APDI in different regions Significant interdecadal variations and/or trends are identified for Japan, the Korean Peninsula, Zhejiang, Taiwan, Guangdong and Vietnam, the details of which are given below. No significant trends or interdecadal variations are found for Fujian and the Philippines (see Fig. 2d, h) and therefore, they will not be discussed. 3.1 Japan The APDI time series for Japan (main season of TC landfall is from June to October, with 97 % of the annual number of landfalling TCs) shows no statistically significant trend but a significant interdecadal variation (Fig. 2a), with two distinct high-PDI [1960–1972 (H60-72) and 1989–2007 (H89-07)] and two low-PDI periods [1973–1988 (L73-88) and 2008–2013 (L08-13)]. The mean APDI in H60-72 and H89-07 are 14.41 and 15.60, respectively (climatological mean = 11.46), with the mean numbers of landfalling TCs being 3.7 and 3.6, respectively. Both numbers are only slightly higher than the normal of 3.1, suggesting that the increase in APDI is not due to an increase in TC landfall number, but from the higher percentages of occurrence of intense typhoons (Category 2 or above, intensity ≥83 kt), the percentages being 22.9 and 30.6 %, respectively, which are much higher than those in the two low-PDI periods (with percentages of 11.1 and 0 %, Fig. 3a). The mean APDI for the two low-PDI periods (L73-88 and L08-13) are 6.58 and 4.97, respectively. Note that during L73-88, no landfalls (APDI = 0) occurred in four years (1973, 1977, 1984 and 1988). A significant decrease (confidence level of 95 %) in TC landfall number (the mean being 2.3) together with a lower percentage (11.1 %) of intense landfalling typhoons contribute to the drop in APDI during L73-88. The low APDI in L08-13 also results from a lower mean TC landfall number (2.7) and no Category 2 or higher TCs making landfall during the entire period. 3.2 The Korean Peninsula Compared with Japan, the APDI for the Korean Peninsula is much lower, with a climatological mean of 2.04 [main season for TC landfall is between July and September, with 91 % of the landfalling TCs (Choi et al. 2010a)] The trend analysis suggests a possible upward trend of APDI, with a confidence level of 97 % (Fig. 2b). An obvious jump is also

13

K. S. Liu, J. C. L. Chan

found as indicated by the significant regime shift in 1984, giving a low-PDI period during 1960–1983 (L60-83, mean APDI  = 0.89) and a high-PDI period during 1984–2013 (H84-13, mean APDI = 2.97). Thus, this upward trend may not be purely linear but is partly the result of the jump in 1984. Because very few TCs make landfall over the Korean Peninsula (a mean of 1.0 and a range from 0 to 4), the APDI largely depends on whether there are landfalling TCs or not. Indeed, the TC landfall number in H84-13 (a mean of 1.3, with seven years (23 %) having no landfalls) is significantly higher than that in L60-83 (a mean of 0.7, with 11 years (46 %) having no landfalls) at the 95 % confidence level. The changes in landfall intensity may also contribute to the upward trend of APDI. The difference in the mean landfall intensities between L60-83 and H84-13 (47 and 56 kt, respectively) is statistically significant at the 99 % confidence level. The frequency distribution of landfall intensities (Fig. 3b) also shows a higher percentage of landfalling typhoons (intensity ≥64 kt) in H84-13 (38.5 %) than that in L60-83 (17.7 %). Furthermore, three (or 7.7 %) Category 2 or above (intensity ≥83 kt) typhoons made landfall in H84-13 but none in L60-83. Indeed, all the seven TCs with the highest landfall intensity are found in H84-13. An obvious upward trend of yearly maximum intensity (the highest landfall intensity in a year) is also observed, which is statistically significant at the 90 % conference level (Fig. 4a). Thus, the upward trend of APDI for the Korean Peninsula is related to the regime shift to a high-PDI period in 1984 resulting from the increases in the number of landfalling TCs as well as in TC intensity at landfall (both mean and extreme). Choi et al. (2010a) also examined the interdecadal variation of TCs making landfall in the Korean Peninsula for the period from 1951 to 2004 and identified two highfrequency periods (1951–1965 and 1986–2004) and one low-frequency period (1966–1985). An increase in landfall intensity is also found in the period 1986–2004. The lowfrequency period (1966–1985) generally matches the lowPDI period (1960–1983) while the high-frequency period (1986–2004) generally coincides with the high-PDI period (1984–2013). Our results are therefore largely consistent with theirs. 3.3 Zhejiang The main season of TC landfall in Zhejiang is between July and September, which includes 91 % of the annual number of landfalling TCs. The time series of APDI (Fig. 2c) shows a possible upward trend (confidence level of 92 %) as well as a multi-decadal variation, with two distinct low-PDI periods (L60-84 and L08-13) and one high-PDI period (H8507). Similar to the Korean Peninsula, the APDI for Zhejiang largely depends on whether there are landfalling TCs or not.

1967

Variations in the power dissipation index in the East Asia region Fig. 2  Time series of the APDI for the regions including a Japan, b the Korean Peninsula, c Zhejiang, d Fujian, e Taiwan, f Guangdong, g Vietnam and h the Philippines. The solid lines indicate the possible trends and the dashed lines indicate the regime shifts

(a)

50 40 30 20 10 0

1960

1965

1970

1975

1980

1985

1990

1995

2000

2005

2010

1960

1965

1970

1975

1980

1985

1990

1995

2000

2005

2010

1960

1965

1970

1975

1980

1985

1990

1995

2000

2005

2010

1960

1965

1970

1975

1980

1985

1990

1995

2000

2005

2010

1965

1970

1975

1980

1985

1990

1995

2000

2005

2010

1965

1970

1975

1980

1985

1990

1995

2000

2005

2010

1965

1970

1975

1980

1985

1990

1995

2000

2005

2010

1985

1990

1995

2000

2005

2010

(b)

12 10 8 6 4 2 0

(c)

15 10 5 0

(d)

14 12 10 8 6 4 2 0

(e)

30 25 20 15 10 5 0

1960

(f)

35 30 25 20 15 10 5 0

1960

(g)

25 20 15 10 5 0

1960 120 100 80 60 40 20 0

(h)

1960

1965

1970

1975

1980

13

1968

K. S. Liu, J. C. L. Chan

80

(a)

90 TS Cat1 Cat2 Cat345

70 60 50

70 60 50 40

30

30

20

20

10

10

0

0 H60-72

L73-88

H89-07

L08-13

(c)

40

TS

40

Cat1 Cat2

35

20 15 10

10

5

0

0 L60-84

50

TS Cat1 Cat2 Cat345

25

20

60

H84-13

(d)

30

Cat345

30

70

L60-83 45

50

80

TS Cat1 Cat2 Cat345

80

40

60

(b)

H85-07

L60-83

(e)

90 TS Cat1 Cat2 Cat345

80 70 60

H84-13

(f) TS Cat1 Cat2 Cat345

50

40

40

30

30

20

20

10

10

0

0 H60-75

L76-88

H89-96

L97-11

H60-73

L74-88

H89-97

L98-11

Fig. 3  Percentage distribution of landfall intensities in high-PDI and low-PDI periods for the regions a Japan, b the Korean Peninsula, c Zhejiang, d Taiwan, e Guangdong and f Vietnam. TS, Cat1, Cat2, and

Cat345 represent the landfall intensities of tropical storm, category 1, category 2 and category 3 or above respectively

The first low-PDI period (L60-84), with a mean APDI value of 1.63 (climatological mean = 2.93), is related to a low TC landfall number (mean = 0.5 versus the normal of 0.8). Out of the 25 years, 15 (~60 %) had no landfalling TCs. During H85-07, the mean APDI rises to 4.86. Note that the changepoint year of 1985 is similar to that of the Korean Peninsula (1984), suggesting similar interdecadal variations of the APDI for Zhejiang and the Korean Peninsula. The mean TC landfall number rises to 1.3 and only four out of 23 years (17 %) had no landfalling TCs. Moreover, the percentage of TCs with landfall intensity of Category 2 or above (intensity ≥83 kt) increases from 7.7 % in L60-84 to 20.7 % in

H85-07 (Fig. 3c). Thus, the increasing numbers of landfalling TCs and intense typhoons are both responsible for the higher APDI in this period. The main reason for the lower APDI in L08-13 is that only one year had a landfalling TC. To summarize, the APDI for Zhejiang shows an upward trend, which may not be purely linear and is related to its prominent rise after 1985 although a drop is observed after 2007. The increases in both the annual number of landfalling TCs and TC intensity at landfall are responsible for the upward trend of APDI. Moreover, the landfall intensity tends to be higher in the recent decade, suggesting an increasing risk of intense TC landfall in Zhejiang.

13

1969

Variations in the power dissipation index in the East Asia region Fig. 4  Time series of the yearly maximum intensity for the regions including a the Korean Peninsula, b Taiwan and c Guangdong. The solid lines indicate the possible trends

(a) 90 80 70 60 50 40 30 1960 120

1965

1970

1975

1980

1985

1990

1995

2000

2005

2010

(b)

110 100 90 80 70 60 50 40 30 1960

120

1965

1970

1975

1980

1985

1990

1995

2000

2005

2010

1965

1970

1975

1980

1985

1990

1995

2000

2005

2010

(c)

110 100 90 80 70 60 50 40 1960

3.4 Taiwan The time series of its APDI (main season of TC landfall is between June and October, with 95 % of the landfalling TCs) shows an obvious downward trend, which is significant at the 99 % confidence level (Fig. 2e). However, no regime shift can be identified. In order to investigate this apparent continuous decrease, the entire study period is divided into two periods, 1960–1986 (H60-86) and 1987– 2013 (L87-13). The mean APDI decreases from 12.52 in H60-86 to 8.31 in L87-13 (climatological mean = 10.41). However, this decrease is not related to the mean annual number of landfalling TCs, which is the same in these two sub-periods (2.2). Instead, the percentage of TCs making landfall at intensity of Category 3 or above (≥96 kt) in H60-86 is much higher than that in L87-13 (Fig. 3d). Moreover, an obvious downward trend (confidence level

of 95 %) of yearly maximum intensity is found (Fig. 4b). Thus, the decreasing occurrence of intense landfalling TCs is the major factor responsible for the downward trend of APDI of Taiwan. 3.5 Guangdong Most of the TCs striking Guangdong occur in the months between June and October. The APDI time series shows an obvious downward trend, which is statistically significant at the 99 % confidence level (Fig. 2f). There are two distinct high-PDI periods (H60-75 and H89-96) and two low-PDI periods (L76-88 and L97-11). The APDI in the years 2012 and 2013 appear to rise again, both having the above-normal APDI, but more data are required to confirm this new high-PDI period. The mean APDI in H60-75 and H89-96 are 16.51 and 16.48, respectively (climatological

13

1970 Table 1  Trends and interdecadal variations of the APDI at different regions

K. S. Liu, J. C. L. Chan Region

Trend

Interdecadal variation

Japan

None

The Korean Peninsula

Upward (97 %)

Zhejiang

Upward (92 %)

Fujian Taiwan Guangdong

None Downward (99 %) Downward (99 %)

Vietnam

None

High (1960–1972) Low (1960–1983) Low (1960–1984) None None High (1960–1975) High (1960–1973)

The Philippines

None

Low (1973–1988) High (1984–2013) High (1985–2007)

High (1989–2007)

Low (1976–1988) Low (1974–1988)

High (1989–1996) High (1989–1997)

Low (2008–2013)

Low (2008–2013)

Low (1997–2011) Low (1998–2011)

None

The percentages indicate the confidence levels of the trends. “High” and “Low” denote the high-PDI and low-PDI periods respectively

mean = 12.50). The mean numbers of landfalling TCs (5.7 and 5.8 respectively) are also higher than the normal number of 4.7. While the mean landfall intensities in these two periods are close to normal (64 and 62 kt respectively versus the normal of 61 kt), the percentages of the TCs making landfall with typhoon intensity (≥64 kt) are higher (48.4 and 48.9 % respectively, Fig. 3e). The mean APDI in the two low-PDI periods (L76-88 and L97-11) are 10.61 and 7.17, respectively, with the 2 years 1998 and 2004 having the lowest APDI. The lower percentage (32.2 %) of the TCs making landfall with typhoon intensity is the main reason for the lower APDI during L7688 rather than the TC landfall number (4.5). In L97-11, significant decreases in both TC landfall number (the mean number being 3.2) and intensity lead to a very low mean APDI in this period. In summary, the APDI for Guangdong shows an obvious downward trend and interdecadal variation, with two highPDI periods and two low-PDI periods. The TC landfall number, which shows a downward trend, is the major factor responsible for such changes. The changes in landfall intensity, with a decrease in the percentage of the TCs making landfall with typhoon intensity and downward trend in yearly maximum intensity, with a confidence level of 90 % (Fig. 4c), play the secondary role. 3.6 Vietnam The time series of the APDI for Vietnam (main season of TC landfall is from June to November, with 95 % of the annual number) has no significant trend but a significant interdecadal variation (Fig. 2g), with two distinct highPDI (H60-73 and H89-97) and two low-PDI periods (L7488 and L98-11). Similar to Guangdong, the APDI values for Vietnam in 2012 and 2013 appear to rise again. The

13

mean APDI in H60-73 of 7.82 is only slightly higher than the normal of 6.73 and is contributed mostly by very high APDI in some years (1964, 1971, 1972 and 1973). The mean APDI during H89-97 of 11.32, on the other hand, is related to the generally higher levels in the APDI in most of the years (Fig. 2g). The mean annual TC number in H60-73 is 3.9, only slightly higher than normal (the normal value being 3.2) but the higher percentage (29.1 %) of landfalling typhoons (intensity ≥64 kt) is the reason for the higher APDI in this period (Fig. 3f). The higher APDI in H89-97 is related to the prominent increases in the number of landfalling TCs (4.6) and the percentage of landfalling typhoons (36.6 %). The mean APDI in L74-88 and L98-11 are 3.98 and 4.55, respectively, which are mainly due to significant drops (confidence level of 95 %) in TC landfall numbers (mean numbers being 2.4 and 2.3, respectively). The lower percentages of landfalling typhoons (16.7 % and 21.9 %, respectively) also lead to lower mean APDI in these two periods. It is interesting to note that the variation of the APDI of Vietnam is quite similar to that of Guangdong, with similar high and low PDI periods. Indeed, their APDI values are highly correlated, the correlation coefficient being 0.57 (confidence level of 99 %). Such a relation is partly due to the fact that some of the TCs making landfall in Guangdong may also make landfall in the northern part of Vietnam. 3.7 Summary The APDI for different regions in the East Asia region are summarized in Table 1. Trends in APDI are identified in some regions (Korean Peninsula, Zhejiang, Taiwan and Guangdong) and interdecadal variations in others (Japan, Korean Peninsula, Zhejiang, Guangdong and Vietnam),

1971

Variations in the power dissipation index in the East Asia region

Fig. 5  “Time-region” cross section of the standardized APDI for the regions (from south to north) Vietnam, Guangdong, Taiwan, Zhejiang, Korean Peninsula and Japan. Dark (light) shadings indicate the

APDI > 0.5σ (<−0.5σ). Vertical dashed lines represent the times of occurrence of the significant changes in APDI distribution

with distinct high and low PDI periods. The variations of APDI for some neighboring regions are quite similar such as the Korean Peninsula and Zhejiang with similar upward trend and regime shifts, and Guangdong and Taiwan, both with downward trends. The opposite trends between north and south suggest a possible northward shift in TC landfall locations along the coast of East Asia. In addition, the APDI for Vietnam is significantly correlated with those for Guangdong and the Philippines (correlation coefficients being 0.57 and 0.43, respectively, both are significant at the 99 % confidence level). As might be expected, a significant correlation (r  = 0.46) also exists between the APDI for Fujian and Taiwan. The shift in APDI along the coast of East Asia can be clearly seen from the “time-region” cross section of the 9-year-Gaussian-filtered standardized APDI (Fig. 5). The time series of the APDI in each region is firstly standardized and then low-pass filtered with a 9-year Gaussian filter. The time-region cross section for the regions (from south to north) Vietnam, Guangdong, Taiwan, Zhejiang, the Korean Peninsula and Japan is then obtained. Before the mid-1970s, the APDI is generally higher than normal in the southern regions (Vietnam, Guangdong and Taiwan) but lower in the northern regions (Zhejiang and Korean

Peninsula). Throughout most of the regions, the APDI is below normal from the mid-1970s to mid-1980s, probably due to the lower overall WNP TC activity but becomes above normal after the mid-1980s. After the late 1990s, significant decreases in APDI are found in Vietnam and Guangdong while the APDI in the northern regions continue to be higher, resulting in a distribution with higher APDI in the northern regions and lower APDI in the southern regions, which is opposite to that before the mid-1970s. Thus, a northward shift in APDI appears to have occurred along the coast of East Asia during the period 1960–2013.

4 PDI patterns over the WNP The empirical orthogonal function (EOF) technique is widely used to detect major modes of climate variability (Hannachi et al. 2007) and is therefore employed to obtain the characteristic PDI patterns over the WNP. An EOF analysis is performed on the anomalous fields of PDI and the time series of the PC coefficient of each EOF mode is then compared with the APDI for various regions to find out the possible relationships. The first mode represents the changes in PDI pattern associated with ENSO events. The

13

1972

K. S. Liu, J. C. L. Chan

(a)

PC1

80

(d)

60 40 20 0 -20 -40

(b)

1960 1965 1970 1975 1980 1985 1990 1995 2000 2005 2010

50

(e)

40

PC2

30 20 10 0 -10 -20 -30 -40 1960 1965 1970 1975 1980 1985 1990 1995 2000 2005 2010

(c)

40 30

PC4

(f)

20 10 0 -10 -20 -30 -40 1960 1965 1970 1975 1980 1985 1990 1995 2000 2005 2010

Fig. 6  Loading patterns of the PDI pattern for the a leading, b second and c fourth modes. Time series of the d PC1, e PC2 and f PC4 coefficients. The horizontal lines indicate the possible regime shifts

second and fourth modes reflect the east–west and north– south PDI distributions respectively. Temporal changes in these modes are likely responsible for some of the changes in APDI for some regions, which will be further examined in this section. The third mode is not discussed because it bears no clear relation with the changes in APDI. 4.1 ENSO (basin‑wide) mode The leading mode explains ~24.8 % of the total variance. Positive loadings are found over the entire WNP while negative loadings, with smaller magnitude, are found over the South China Sea (SCS) (Fig. 6a). This mode may be related to ENSO, as suggested by the high correlation

13

coefficient of 0.61 between the PC1 coefficient and the Jul–Nov Niño 3.4 index. A year is defined as the El Niño (La Niña) year if the Jul–Nov Niño 3.4 index is ≥0.5 (≤−0.5). In an El Niño (a La Niña) year, the PDI is generally higher (lower) over most parts of the WNP but lower (higher) over the SCS. The PC1 coefficient is also significantly correlated (correlation coefficient = 0.43, confidence level = 95 %) with the annual number of TCs with at least tropical storm intensity over the entire WNP, which suggests that this mode may reflect the overall TC activity over the WNP. The PC1 coefficients also show a significant interdecadal variation, with two positive periods (1960–1968 and 1991–1997) and two negative periods (1969–1990 and

1973

Variations in the power dissipation index in the East Asia region Table 2  APDI in different regions during the positive and negative years of different EOF modes

Japan EOF1

EOF2

EOF4

Fujian

Taiwan

Guangdong

Vietnam

The Philippines

Positive (14 years)

12.69

21.51

Negative (19 years)

7.91

40.84

Positive (14 years)

15.70

13.53

16.84

40.35

Negative (17 years)

8.69

8.83

9.86

23.35

Positive (15 years)

8.50

2.48

7.51

16.91

9.50

41.99

Negative (12 years)

17.21

5.27

13.02

10.12

5.33

24.76

11.46

3.72

10.41

12.50

6.73

33.78

Climatological mean

Only those with significant difference (confidence level of 90 % or above) in APDI between positive and negative years are shown. The climatological means of the APDI in each region are shown in the bottom row

1998–2013, Fig. 6d), which are similar to the overall TC activity over the WNP (see Liu and Chan 2013). However, the relationship between this mode and the APDI in different areas is not very high. Most of the positive loadings are found over the ocean, with only the edge touching Taiwan (Fig. 6a) and the APDI for Taiwan shows a slight increase in the positive years (defined as the years with PC1 coefficient >0.5σ) compared with the negative years (defined as the years with PC1 coefficient <−0.5σ) and its correlation with the PC1 coefficient is only 0.26 (Table 2). The negative loadings covering the southern part of the Philippines also suggest a possible influence on the APDI of the Philippines, with lower value in the positive years compared with the negative years and a correlation of −0.26 with the PC1 coefficient (Table 2). 4.2 East–west dipole mode The second mode explains ~10.8 % of the total variance and its main feature is an east–west dipole over the WNP (hereafter referred to as the east–west dipole mode), with positive and negative loadings west and east of 135°E respectively (Fig. 6b). The positive center is near 17°N, 127°E and the positive loadings extend to the South China coast, Taiwan and the southern part of Japan, suggesting a possible influence on the APDI of these regions. The correlation coefficients between the PC2 coefficient and the APDI for Japan, Taiwan, Guangdong and the Philippines are 0.36, 0.23, 0.30 and 0.31, respectively. While these correlations are not high, the APDI for these regions show significant differences in the positive and negative years, with confidence levels of 95 % for Japan and Guangdong and 90 % for Taiwan and the Philippines. When this mode is in its positive phase, the PDI in the areas west of 135°E

is generally higher than normal, indicating more TCs with higher intensity passing through these areas. The chance of a TC making landfall with a higher intensity at the nearby coastal areas is therefore higher. The reverse occurs when this mode is in its negative phase. Correspondingly, the mean APDI for Japan, Taiwan, Guangdong and the Philippines in the positive years (PC2 coefficient >0.5σ) are significantly higher than those in the negative years (PC2 coefficient <−0.5σ) (Table 2). Thus, the east–west dipole mode represents the east–west shift of the location of the higher PDI region and is related to the APDI of some coastal areas such as Japan, Taiwan, Guangdong and the Philippines. The time series of PC2 coefficients shows a remarkable interdecadal variation between 1960 and 2002 (Fig. 6e), with two positive (1960–1971 and 1989–1996) and two negative periods (1972–1988 and 1997–2002). The positive (negative) periods generally match the high (low) PDI periods of Japan and Guangdong (see Table 1). The decadal change of this mode may therefore partly explain the interdecadal variations of the APDI for Guangdong and Japan. 4.3 North–south dipole mode The fourth mode explains ~7.2 % of the total variance and is characterized by a north–south dipole over the western part of the WNP (Fig. 6c, hereafter referred to as the north–south dipole mode). Positive loadings cover the areas north of 20°N, including Japan, Taiwan and the east coast of China while negative loadings cover those south of 20°N, encompassing the Philippines, Vietnam and the south coast of China. This pattern suggests that this mode may represent a north–south shift of the high PDI region. Positive correlations exist between the PC4 coefficient and the APDI for Guangdong, Vietnam and the Philippines

13

1974

(0.35, 0.25 and 0.35, respectively) but negative correlations are found for Japan, Fujian and Taiwan (−0.29, −0.32 and −0.26). In the positive years (PC4 coefficient >0.5σ), the mean APDI values for Guangdong, Vietnam and the Philippines are generally above normal but those for Japan, Fujian and Taiwan are generally below normal (Table 2). The reverse is true in the negative years (PC4 coefficient <−0.5σ). Indeed, the APDI for these regions show significant differences in the positive and negative years, with confidence levels of 95 % for Japan, Fujian and Guangdong and 90 % for Taiwan, Vietnam and the Philippines. The decadal change of this mode is less obvious. This mode is generally in its negative phase between 1960 and 1966 (Fig. 6f). Then the interdecadal variation becomes less significant during the period 1967–1999 and the variation is mainly on the interannual time scale, with slightly more years having positive values. This mode is in its negative phase again during the period 2000–2005 but shifts to its positive phase during the period 2006–2013. 4.4 Summary Changes in PDI pattern over the WNP seem to be represented by three major modes. The ENSO (basin-wide) mode represents the PDI patterns associated with ENSO events and also reflects the overall PDI over the WNP. This mode may have an influence on the APDI for Taiwan and the Philippines. The east–west dipole mode gives the east–west shift of PDI and has an influence on the APDI for Japan, Taiwan, Guangdong and the Philippines. The north– south dipole mode denotes the north–south shift of PDI and has an influence on the APDI for Japan, Fujian, Taiwan, Guangdong, Vietnam and the Philippines. Indeed, the variations of APDI for these regions can be largely explained by these modes. Using the PC1, PC2 and PC4 coefficients as predictors for the APDI, the multiple regression model gives correlations of 0.49, 0.44, 0.46 and 0.47 for Japan, Taiwan, Guangdong and the Philippines respectively, all significant at the 95 % confidence level.

5 Large‑scale atmospheric circulation As discussed above, the APDI for a region depends on the number of TCs making landfall in that region and their intensities at landfall. The former is largely related to the preferred TC tracks, which are influenced by the steering flow while the latter is partly affected by vertical wind shear. To understand the variations of APDI therefore requires an investigation of the variations of the large-scale flow and vertical wind shear patterns. As the mid-tropospheric flow patterns are important to TC movement (e.g. Chan and Gray 1982), the tropospheric layer-mean winds

13

K. S. Liu, J. C. L. Chan

between 850 and 300 hPa in the months between June and October are examined. Singular value decomposition (SVD) is a common technique to identify coupled modes of variability between time series of two fields and their temporal variations (Wallace et al. 1992). In this study, a SVD analysis is performed on the anomalous fields of zonal and meridional winds and the time series of the expansion coefficients of each SVD mode is then compared with the APDI for various regions to find any possible relationships. The spatial patterns for the zonal and meridional winds of each mode then give the characteristic steering flow patterns. If a characteristic pattern dominates in a particular year, the associated steering flow may affect the TC movements and the TCs tend to follow some preferred prevailing tracks, resulting in the changes of APDI for various regions. The first SVD mode is the flow pattern associated with ENSO events but is not related to the changes of APDI and will not be discussed. The second, third and fourth SVD modes are found to be related to the changes in APDI in various regions and will be discussed below. To examine the interdecadal variations in these modes, the expansion coefficients of zonal and meridional winds are averaged to give a single time series and a regime detection algorithm is then applied to find possible shifts. 5.1 Second SVD mode The second SVD mode, which explains 25.4 % of the total squared covariance, shows a cyclonic circulation east of Taiwan centered at about 20°N, 130°E, resulting in easterly anomalies between 20°N and 35°N and westerly anomalies between 5°N and 20°N (Fig. 7a). A year is considered to be dominated by the positive (negative) phase of this mode if the expansion coefficients of both zonal and meridional winds are >0.5σ (<−0.5σ). A correlation analysis for the expansion coefficients of zonal and meridional winds and the APDI suggests a positive correlation for Zhejiang and Taiwan but a negative correlation for Guangdong. The higher correlation coefficients for zonal wind (0.36, 0.31 and −0.27 for Zhejiang, Taiwan and Guangdong respectively) suggest that the changes in zonal wind are more important for the variations of APDI for these regions. While the correlation coefficients are not high, the APDI for these regions show significant differences in the positive and negative years (Table 3). During the positive phase of this mode (12 years), the anomalous cyclonic circulation east of Taiwan tends to steer TCs towards the east coast of China and prohibits TCs from moving towards the south coast of China. As a result, the APDI for Zhejiang and Taiwan (the means being 5.01 and 13.45, respectively) are slightly higher than normal while the APDI for Guangdong (10.99) is slightly lower than normal. It should be noted that this flow pattern only suggests

1975

Variations in the power dissipation index in the East Asia region

(a) 30

(d)

Expansion coefficients of SVD2

20 10 0 -10 -20 -30

(b) 20

1960 1965 1970 1975 1980 1985 1990 1995 2000 2005 2010

(e)

Expansion coefficients of SVD3

15 10 5 0 -5 -10 -15 -20

(c)

1960 1965 1970 1975 1980 1985 1990 1995 2000 2005 2010

(f) 20

Expansion coefficients of SVD4

15 10 5 0 -5 -10 -15 -20

1960 1965 1970 1975 1980 1985 1990 1995 2000 2005 2010

Fig. 7  Homogenous correlation maps of Jun–Oct 850–300-hPa winds for the a second, b third and c fourth SVD modes. The time series of the mean values of the expansion coefficients of zonal and

meridional winds are shown in d–f. The horizontal lines indicate the possible regime shifts

more TCs approaching the east coast of China but which region (Zhejiang or Taiwan) they will likely make landfall in depends on the actual steering flow at that time. Normally, a TC making landfall in Taiwan seldom makes

landfall again in Zhejiang. Thus, there may be more TCs making landfall in Taiwan in one positive year but more in Zhejiang in another, and this explains why the increases in APDI for these two individual regions are not significant.

13

1976 Table 3  APDI in different regions during the positive and negative years of different SVD modes

K. S. Liu, J. C. L. Chan Japan SVD2

SVD3

SVD4

Korean Peninsula

Zhejiang Taiwan Guangdong Vietnam

Positive (12 years)

5.01

13.45

10.99

Negative (15 years)

1.51

8.24

15.73

Positive (12 years)

13.20

4.10

Negative (13 years)

6.50

1.56

Positive (12 years)

14.40

7.06

Negative (10 years)

8.96

3.61

12.50

6.73

Climatological mean

11.46

2.04

2.93

10.41

Only those with significant difference (confidence level of 90 % or above) in APDI between positive and negative years are shown. The climatological means of the APDI in each region are shown in the bottom row

The sum of the APDI for Taiwan and Zhejiang, however, does show a significant increase (confidence level of 95 %) in the positive years, which supports the above argument. It is also interesting to note that the increase in APDI for Taiwan (Zhejiang) is more significant before 1978 (after 1978). Before 1978 (Since 1978), the mean APDI is significantly higher for Taiwan (Zhejiang) but near normal for Zhejiang (Taiwan). Indeed, the flow patterns show significant differences before and after 1978. Before 1978, the center of the anomalous cyclonic flow is near 23°N, 145°E with northeasterly anomalies observed near 130°E, suggesting a flow pattern more favorable for TCs making landfall in Taiwan (Fig. 8a). Since 1978, the anomalous cyclonic flow shifts westward and the center is located at 23°N, 130°E. The strong easterly winds then tend to steer TCs towards the east coast of China (Fig. 8b). The position of this anomalous cyclonic circulation may therefore exhibit a decadal change that causes a significant impact on the preferred landfall region. However, the current SVD analysis cannot distinguish these patterns and considers them as one single characteristic flow pattern (SVD2). When this mode is in its negative phase (15 years), an anomalous anticyclonic circulation is found east of Taiwan so that the APDI is significantly lower for Zhejiang (1.51) and slightly lower for Taiwan (8.24), but that for Guangdong is slightly higher (15.73). The second SVD mode therefore appears to have a significant impact on the APDI for Guangdong, Zhejiang and Taiwan and the differences in APDI between positive and negative years are significant, with confidence levels of 95 % for Zhejiang and 90 % for Guangdong and Taiwan. The time series of the expansion coefficients of zonal and meridional winds for the second SVD mode does not show any significant trend or interdecadal variation and its

13

variation is mainly on the interannual time scale (Fig. 7d). Thus, the variations of this mode cannot explain the downward trend and interdecadal variation of APDI for Taiwan and Guangdong. 5.2 Third SVD mode The main feature of the third SVD mode, which explains 8.7 % of the total squared covariance, is an anomalous circulation just east of Japan, centered at about 35°N, 145°E (Fig.  7b). The expansion coefficient of meridional wind is weakly correlated with the APDI of Japan and Korean Peninsula, with correlation coefficients of 0.28 and 0.23, respectively. When this mode is in its positive phase (12 years), an anomalous anticyclonic circulation is found just east of Japan. The enhanced southeasterly flow south of Japan tends to steer the TCs towards the Korean Peninsula and Japan. The mean APDI is significantly higher for the Korean Peninsula (4.10) but is only slightly higher than normal for Japan (13.20). Since the APDI for the Korean Peninsula is less dependent on landfall intensity as compared with Japan, an increase in the number of landfalling TCs associated with this flow pattern generally leads to a higher APDI. The reverse occurs when this mode is in its negative phase (13 years) and the mean APDI for the Korean Peninsula and Japan decrease to 1.56 and 6.50, respectively, with the drop being more significant for the latter. Thus, the changes in steering flow represented by this mode are shown to have a significant effect on the APDI for the Korean Peninsula and Japan. Indeed, the differences in APDI between positive and negative years are significant (confidence levels of 95 %) for both regions (Table 3). These results are also consistent with the previous studies (Nakazawa and Rajendran 2007; Grossman et al. 2014).

1977

Variations in the power dissipation index in the East Asia region Fig. 8  Composites of the anomalous 850–300-hPa wind patterns for the positive years of SVD2 in the periods a before 1978 and b since 1978. Contour is the wind speed in ms−1. Shadings indicate anomalies significant at the 95 % confidence level

(a)

(b)

The time series of the expansion coefficients of zonal and meridional winds for the third SVD mode show significant interdecadal variations, with positive values in the periods 1960–1980 and 1998–2005 and negative values in the periods 1981–1997 and 2006–2013 (Fig. 7e). Recall that the APDI for Japan is generally lower during the negative phase. Indeed, the first half of the first negative phase period (1981–1997) generally coincides with the second half of its low-PDI period (1973–1988) while the second negative phase period (2006–2013) matches its low-PDI period (2008–2013). Thus, the flow pattern associated with the negative phase of SVD3 is partly responsible for the low-PDI periods of Japan. During the positive phase, the APDI for the Korean Peninsula is generally

higher. Although the positive phase periods (1960–1980 and 1998–2005) do not totally match its high-PDI period (1984–2013), the number of positive years since 1984 (eight cases) is higher than that before 1984 (four cases), suggesting that the higher frequency of occurrence of the positive phase of SVD3 since 1984 may be related to the increase in APDI in the Korean Peninsula after 1984. 5.3 Fourth SVD mode The fourth SVD mode explains 7.3 % of the total squared covariance and is characterized by an anomalous circulation over the SCS (Fig. 7c). The expansion coefficient of zonal wind is weakly correlated with the APDI of

13

1978

Guangdong and Vietnam, with correlation coefficients of 0.24 and 0.29, respectively. When this mode is in its negative phase (10 years), an anomalous anticyclonic circulation is found over the SCS. The westerly anomalies extending from south China to the northern part of the Philippines tend to prohibit TCs from approaching the south coast of China. Correspondingly, the APDI for Guangdong and Vietnam, with the means of 8.96 and 3.61, respectively, are significantly lower than their normal values. An anomalous cyclonic circulation is found over the SCS during the positive phase (12 years) and the easterly anomalies extending from the northern part of the Philippines to south China tend to steer TCs towards the SCS and south coast of China. However, the influence on APDI appears to be less significant and only a small increase in the APDI is found for Guangdong (the mean increasing to 14.40). Nevertheless, significant differences in APDI are found between positive and negative years for Guangdong and Vietnam (Table 3). The time series of the expansion coefficients of zonal and meridional winds for the fourth mode shows significant interdecadal variations, with two negative periods (1960–1968 and 1997–2010) and one positive period (1978–1996). This mode is not significant during the period 1969–1977 and after 2010 (Fig. 7f). Note that the second negative period (1997–2010) generally matches the lowPDI periods of Guangdong (L97-11) and Vietnam (L9811), suggesting that the decadal changes of this mode may be partly responsible for the interdecadal variations of the APDI for Guangdong and Vietnam. 5.4 Three‑cell model for TC landfall in East Asia Each of the three major SVD modes of 850–300-hPa winds, representing the steering flow near the East Asian coast, identified in this section apparently has an influence on APDI in the various regions. These three anomalous circulations, which are located east of Japan, east of Taiwan and over the SCS, actually form a three-cell circulation along the coast of East Asia and a three-cell model for TC landfall in East Asia is therefore proposed. The changes in the steering flow may be represented by a combination of these three modes, which subsequently reflect changes in the TC tracks and hence the APDI in various regions. For example, the anomalous steering flow pattern in 2004 is dominated by the positive phases of SVD2 and SVD3, similar to the positive phase of Pacific-Japan teleconnection pattern (Wakabayashi and Kawamura 2004; Choi et al. 2010b). The anomalous anticyclonic circulation east of Japan and the anomalous cyclonic circulation east of Taiwan give rise to strong southeasterly flow south of Japan (Fig. 9a), which tends to steer TCs towards Japan and the east coast of China (Fig. 9c) and significant increases in

13

K. S. Liu, J. C. L. Chan

APDI are therefore found in Japan and Zhejiang. Note that this flow pattern also partly explains the record-breaking number of landfalling TCs in Japan in 2004 (Kim et al. 2005). On the other hand, the stronger northwesterly flow near south China leads to a lower APDI for Guangdong. In contrast, the year 1988 is dominated by the negative phase of SVD3 and the strong northwesterly flow south of Japan tends to prohibit TCs from moving towards Japan and the Korean Peninsula (Fig. 9b), resulting in the absence of landfalling TC in these regions (Fig. 9d). While these two examples demonstrate that this model may be used to represent the changes in the steering flow and the subsequent changes in the APDI along the East Asian coast, a further study will be needed to test this model for all the years in the study period. A three-cell model for TC landfall in East Asia is proposed to explain the changes in steering flow near the East Asian coast and hence the changes in the APDI in various regions. However, the factors controlling the changes of these three anomalous circulations are still not well known. One of the possible factors may be the PDO. Liu and Chan (2008) suggested that the decadal variability of TC tracks is partly related to the PDO. Indeed, the Jun-Oct PDO index is correlated with the expansion coefficients of SVD2 and SVD4 (the correlations being −0.42 and 0.32, respectively), suggesting a possible effect of PDO on the anomalous circulations east of Taiwan (SVD2) and over the SCS (SVD4). Actually, a significant correlation exists between the PDO index and the 850–300-hPa zonal winds between 20°N and 40°N (Fig. 10). During the warm PDO phase (Jun–Oct PDO index >0.5), westerly anomalies appear in the areas along 30°N, which is favorable for the existence of an anomalous anticyclonic circulation, corresponding to the negative phase of SVD2 in which westerly anomalies also existing in the areas along 30°N (see Fig. 7a). The opposite is true for the cold PDO phase (Jun-Oct PDO index <−0.5). Thus, the circulation associated with warm (cold) PDO phase is similar to that of the negative (positive) phase of SVD2 and therefore a significant negative correlation exits between the PDO index and its expansion coefficients.

6 Possible reasons for northward shift in APDI 6.1 Shift in landfall location The shift in APDI along the East Asian coast may be related to the shift in TC landfall location and the changes in landfall intensity. To examine the former, the latitudes of TCs making landfall along the East Asian coast from Vietnam to Japan in the months between June and October are estimated to obtain the seasonal average of TC landfall

1979

Variations in the power dissipation index in the East Asia region

(a)

(c)

2004

(b)

(d)

1988

Fig. 9  The anomalous steering flow patterns for the years a 2004 and b 1988. Contour is the wind speed in ms−1. The TC tracks in these years are shown in c and d respectively. Typhoon symbols indicate the genesis locations

Fig. 10  Correlation map between the Jun–Oct PDO index and the Jun–Oct 850– 300-hPa zonal wind. Shading indicates correlation significant at the 95 % confidence level

13

1980

K. S. Liu, J. C. L. Chan Mean latitude of TC landfall

29

El Nino La Nina

28 27 26 25 24 23 22

1960 1965 1970 1975 1980 1985 1990 1995 2000 2005 2010

Fig. 11  Time series of the mean latitude of TCs making landfall along the East Asian coast in the months from June to October during the period 1960–2013. Solid circles and squares indicate El Niño and La Niña years respectively. The horizontal dashed line indicates the climatological mean and the solid lines indicate the possible regime shifts

latitude. Note that the landfalling TCs in Jiangsu and Shandong are also included for completeness. The average landfall latitude is 25.5°N, which is near the central part of Fujian (see Fig. 1). A significant northward shift is found in 1997, with the mean increasing to 26.5°N (Fig. 11), suggesting a higher percentage of TCs making landfall in the regions north of Fujian. This is consistent with the upward trends of APDI in Zhejiang and the Korean Peninsula and the downward trend of APDI in Guangdong. It is noteworthy that the mean landfall latitudes in La Nina years show a significant difference between the periods 1960–1996 and 1997–2013, being 23.9°N and 27.2°N, respectively. The change in landfall location should be highly related to the change of track pattern, which may be represented by the anomalous frequency of TC occurrence. The region 0°–45°N, 100°–180°E is divided into 5° latitude × 5° longitude grid boxes and the number of TCs with at least tropical storm intensity passing through each box during the months between June and October is then calculated. If a TC passes the same box for more than one time period, it is counted only once. An EOF analysis is then applied to the anomalous frequency of TC occurrence pattern. The leading mode, which explains 14.8 % of the total variance, represents the characteristic track pattern associated with ENSO events, which is not related to the shift in landfall location and will not be discussed. The second mode, which explains 10.1 % of the total variance, shows positive loadings extending from the Philippines to the south China coast, representing a west-northwestward straight track across the Philippines and the SCS (Fig. 12a). Negative loadings are found over the areas extending from the ocean east of the Philippines to the east coast of China, which represent a northwestward straight track toward the east coast of China. Thus, the positive (negative) phase of this mode indicates a higher (lower) frequency of TCs

13

making landfall in the south coast of China and Vietnam but a lower (higher) frequency for the east coast of China. Indeed, the PC2 coefficient is highly correlated with the APDI of Guangdong and Vietnam (correlation coefficients of 0.56 and 0.58, respectively, with confidence levels of 99 %) but weakly correlated with the APDI of Zhejiang and Taiwan (correlation coefficients of −0.25 and −0.26, respectively, with confidence levels of 90 %). The interdecadal variation of this mode is not very obvious except for a negative period found in 1997–2007 (Fig. 12b), which roughly coincides with the low-PDI periods of Guangdong (L97-11) and Vietnam (L98-11), leading to the lower APDI in the southern part of East Asia and hence the northward shift of landfall location. The shift in TC genesis locations also has a significant impact on the TC landfalling activity in East Asia region (Yonekura and Hall 2014). To examine the changes in TC genesis location, an EOF analysis is applied to the anomalous percentage of genesis frequency pattern in the months between June and October. The leading mode, which is related to the east–west shift of genesis location and the second mode, which is related to the pattern associated with ENSO events, are not related to the northward shift of APDI and will not be discussed. The third mode, which explains ~6.5 % of the total variance, shows positive and negative loadings south and north of 15°N, respectively (Fig. 13a) and therefore represents the north–south shift in TC genesis location. The time series of PC3 shows a significant shift in 1999, representing a northward shift in genesis location (Fig. 13b). Choi et al. (2015) also identified a northwestward shift in TC genesis location since the mid-1990s during the period 1977– 2011. The northward shift in genesis location may be associated with northward shift in landfall location along the East Asian coast. For example, a westward or westnorthwestward straight-moving TC tends to make landfall in South China (East China) if the TC forms in a more southward (northward) area. As the mean latitude of TC landfall in La Niña years shows a significant difference after 1997 (see Fig. 11), it is useful to examine the associated changes of TC genesis location. Compared with the La Niña years before 1997, the percentage of genesis frequency is higher in the region 15°–25°N, 135°–150°E but lower in the region 5°–15°N, 135°–150°E (Fig. 14). The former may lead to more TC landfalls in Zhejiang and the Korean Peninsula as the genesis locations come closer to the coast of East China and slight increases in APDI are actually found in these regions. The latter may result in a decrease of TCs making landfall in Guangdong and Vietnam as most of these landfalling TCs are formed in this region. Indeed, significant drops in mean APDI are found in these regions (from 19.03 to 5.69 for Guangdong and from 9.57 to 2.68 for Vietnam).

1981

Variations in the power dissipation index in the East Asia region Fig.  12  a Loading patterns of the TC occurrence pattern for the second mode. b Time series of the PC2 coefficient. The horizontal lines indicate the possible regime shifts

(a)

10

(b)

PC2

8 6 4 2 0 -2 -4 -6 -8 -10 -12 1960

1965

1970

As suggested in the previous section, the changes in steering flow pattern may be represented by the three-cell model for East Asia TC landfall. Thus, the steering flow pattern in the period 1997–2013 may be reflected by the temporal changes of the expansion coefficients of the three SVD modes. The positive phase of SVD3 found in the period 1998–2011 (see Fig. 7e) and the negative phase of SVD4 found in the period 1997–2010 (see Fig. 7f) suggest a steering flow pattern favorable for TC landfalls in Japan and the Korean Peninsula but not in Guangdong and Vietnam. Thus, there is a higher portion of TCs making landfall in the northern part of the East Asian coast and hence the northward shift in mean latitude of TC landfall. 6.2 Change in landfall intensity Since landfall intensity is another vital component of APDI, it is also important to investigate its changes and the associated changes of atmospheric conditions during the period 1997–2013 in which a significant northward shift in landfall location has been found (see Sect. 6.1). As vertical wind shear is one of the dynamic factors related to TC

1975

1980

1985

1990

1995

2000

2005

2010

development and is highly related to the decadal variations of TC intensity (Chan 2008), it is useful to examine the pattern of the vertical wind shear in this period. Because meridional shear is relatively weak, the vertical wind shear is defined as the magnitude of the difference between the 200- and 850-hPa zonal winds. A large area of negative anomalies extending from the northern part of the SCS to the ocean east of Japan is found (Fig.  15a), indicating weaker vertical wind shear in these areas. This allows the TCs approaching the coastal areas to intensify and gain higher intensity before landfall. Positive anomalies are also found over the southern part of the WNP, suggesting stronger vertical wind shear in this region, which is one of the reasons for the lower total genesis frequency over the WNP during this period (Liu and Chan 2013). The time series of the mean vertical wind shear in the region 15°–30°N, 120°–150°E clearly shows a downward shift in 2000 (Fig. 15b), which suggests a favorable atmospheric environment for TC intensification and more TCs with higher intensity near the coastal areas since then. Indeed, a dipole of anomalous frequency occurrence of typhoons is found along the East Asian coast, indicating a

13

1982

K. S. Liu, J. C. L. Chan

Fig.  13  a Loading pattern of anomalous percentage of genesis frequency for the third mode. b Time series of the PC3 coefficients. Solid lines indicate the possible regime shifts

(a)

15

PC3

(b)

10 5 0 -5 -10 -15 1960

1965

1970

1975

1980

1985

1990

1995

2000

2005

2010

Fig. 14  Difference in the pattern of anomalous percentage of genesis frequency between the La Niña years before and after 1997. Shading indicates the difference is significant at the 90 % confidence level

higher frequency of typhoons in East China and the Korean Peninsula and a lower frequency in South China and Vietnam (Fig. 16). Note that the latter is likely a result of the preferred steering flow and genesis locations not favorable for TC landfalls in South China and Vietnam although

13

weaker vertical wind shear is found in these regions. Kossin et al. (2014) show the poleward migration of the location of TC maximum intensity over the WNP during the period 1982–2012. Park et al. (2014) also showed that the location of maximum intensity of TCs has come closer

1983

Variations in the power dissipation index in the East Asia region Fig.  15  a Pattern of Jun–Oct vertical zonal wind shear anomalies in the period 1997–2013. Shadings indicate the anomalies are significant at the 95 % confidence level. b Time series of the average vertical wind shear anomalies in the region 15°–25°N, 120°–150°E as shown in the rectangular box in a

(a)

3

(b)

Wind shear anomalies in the region 15o-30oN, 120o-150oE

2 1 0 -1 -2 -3 -4 1960

1965

1970

1975

1980

1985

1990

1995

2000

2005

2010

Fig. 16  Mean anomalous frequency of typhoon occurrence in the months from June to October during the period 1997–2013. Shadings indicate the anomalies are significant at the 95 % confidence level

13

1984

to the East Asian coastline in the period 1977–2010. The present study also shows a higher frequency of typhoons in East China and the Korean Peninsula, which is therefore consistent with their findings.

7 Summary and discussion 7.1 Discussion The present study has identified a significant northward shift in the APDIs along the East Asian coast and suggests a possible growing threat of strong TC landfalls in Zhejiang, the Korean Peninsula and Japan, which is consistent with the results of previous studies (Park et al. 2011, 2014). However, it is not clear whether this trend is a direct response to global warming or just a part of normal interdecadal or even multi-decadal variation. If the former is true, a greater threat of strong TC landfalls to the northern regions of East Asia can be expected in the future. Otherwise, the reverse may occur, with higher APDI in the southern regions of East Asia, which is similar to the situation before the mid-1980s. Thus, further investigation is required to make projections of future TC landfall frequency and intensity through modeling studies. The present study provides a basis for such studies. A significant northward shift in genesis location in the period 1997–2013 has been identified, which has a significant impact on the landfalling activity in this period. It should be noted that this shift is more significant in La Niña years. As a result, the mean APDI in La Niña years for Guangdong and Vietnam show significant drops after 1997 (from 19.03 to 5.69 and from 9.57 to 2.68, respectively). In other words, their APDI are generally above normal before 1997 but become below normal after 1997, suggesting a changing relationship between ENSO and landfalling TC activity. Thus, we suggest that any seasonal forecast of TC landfall based on its relationship with ENSO (e.g. Liu and Chan 2003) should take the decadal change of genesis location into consideration especially in forecasts for South China and Vietnam. 7.2 Summary This study examines the variability of power dissipation index (PDI) for different regions of the East Asia region during the period 1960–2013. The annual PDI (APDI) for a region is calculated as the sum of the PDI, defined as the cube of the maximum sustained wind speed at landfall, of each tropical cyclone (TC) making landfall at that region. Trends in APDI are identified in the Korean Peninsula, Zhejiang, Taiwan and Guangdong while interdecadal variations of APDI were found in Japan, the Korean Peninsula, Zhejiang, Guangdong and Vietnam. The downward trends

13

K. S. Liu, J. C. L. Chan

in APDI observed in Guangdong and Taiwan, together with the upward trends found in Zhejiang and the Korean Peninsula, suggest a possible northward shift in TC landfall locations along the coasts of China and the Korean Peninsula. An Empirical Orthogonal Function analysis of the anomalous PDI distribution over the western North Pacific suggests three characteristic PDI patterns that are highly related to the variations of APDI for some regions: an ENSO (basin-wide) mode for Taiwan and the Philippines, an east–west dipole mode for Japan, Taiwan, Guangdong and the Philippines, and a north–south dipole mode for Japan, Fujian, Taiwan, Guangdong, Vietnam and the Philippines. A singular value decomposition analysis of Jun–Oct 850–300-hPa wind anomalies suggests a three-cell model for TC landfall in East Asia: a circulation east of Taiwan related to changes in APDI in Taiwan, Zhejiang, Fujian and Guangdong, a circulation east of Japan related to changes in APDI in Japan and the Korean Peninsula, and a circulation over the South China Sea related to changes in APDI in Guangdong and Vietnam. Changes in the steering flow as represented by a combination of these three modes reflect changes in TC tracks and in the APDI in the various parts of the region. A northward shift in the APDI along the East Asian coast is identified in the period 1997–2013. Within this period, the steering flow pattern is generally favorable for TC landfalls in the northern regions but not in the southern regions. The TC genesis locations tend to be more northward, facilitating TC landfalls in the northern regions. Weaker vertical wind shear over the ocean near the coastal areas allows TCs to reach higher intensity before landfall. All of these factors lead to more intense TCs making landfall in the northern part of East Asia and therefore a greater threat of strong TC landfalls is found in the regions including Zhejiang, the Korean Peninsula and Japan. Park et al. (2014) also suggested a growing threat of intense TCs to East Asia during the period 1977–2010 because the location of maximum intensity of TCs has come closer to the East Asian coastline and our results are therefore consistent with their study. Acknowledgments The authors would like to thank the Taiwan Central Weather Bureau for providing the best-track data for tropical cyclones affecting Taiwan. This project is supported by the Research Grants Council General Research Fund CityU 100113.

References Chan JCL (2008) Decadal variations of intense typhoon occurrence in the western North Pacific. Proc R Soc A 464:249–272 Chan JCL, Gray WM (1982) Tropical cyclone movement and surrounding flow relationships. Mon Weather Rev 110:1354–1374 Chan JCL, Xu M (2009) Inter-annual and inter-decadal variations of landfalling tropical cyclones in East Asia. Part I: time series analysis. Int J Climatol 29:1285–1293. doi:10.1002/joc.1782

Variations in the power dissipation index in the East Asia region Choi KS, Kim BJ, Kim DW, Byun HR (2010a) Interdecadal variation of tropical cyclone making landfall over the Korean Peninsula. Int J Climatol 30(10):1472–1483. doi:10.1002/joc.1986 Choi KS, Wu CC, Cha EJ (2010b) Change of tropical cyclone activity by Pacific-Japan teleconnection pattern in the western North Pacific. J Geophys Res 115:D19114 Choi Y, Ha KJ, Ho CH, Chung CE (2015) Interdecadal change in typhoon genesis condition over the western North Pacific. Clim Dyn. doi:10.1007/s00382-015-2536-y Emanuel K (2005) Increasing destructiveness of tropical cyclones over the past 30 years. Nature 436:686–688 Goh AZC, Chan JCL (2012) Variations and prediction of the annual number of tropical cyclone affecting Korea and Japan. Int J Climatol 32:178–189 Grossman MJ, Zaiki M, Nagata R (2014) Interannual and interdecadal variations in typhoon tracks around Japan. Int J Climatol. doi:10.1002/joc.4156 Hannachi A, Joliffe I, Stephenson D (2007) Empirical orthogonal functions and related techniques in atmospheric science: a review. Int J Climatol 27:1119–1152. doi:10.1002/joc.1499 Kim JH, Ho CH, Sui CH (2005) Circulation features associated with the record-breaking typhoon landfall on Japan in 2004. Geophys Res Lett 32:L14713. doi:10.1029/2005GL022494 Knutson TR, McBride JL, Chan J, Emanuel K, Holland G, Landsea C, Held I, Kossin JP, Srivastava AK, Sugi M (2010) Tropical cyclones and climate change. Nature Geosci 3:157–163. doi:10.1038/ngeo779 Kossin JP, Emanuel KA, Vecchi GA (2014) The poleward migration of the location of tropical cyclone maximum intensity. Nature 509:349–352 Kubota H, Chan JCL (2009) Interdecadal variability of tropical cyclone landfall in the Philippines from 1902 to 2005. Geophys Res Lett 36:L12802. doi:10.1029/2009GL038108 Li Q, Duan Y (2010) Tropical cyclone strikes at the coastal cities of China from 1949 to 2008. Meteorl Atmos Phys 107:1–7 Liu KS, Chan JCL (2003) Climatological characteristics and seasonal forecasting of tropical cyclones making landfall along the South China coast. Mon Weather Rev 131:1650–1662 Liu KS, Chan JCL (2008) Interdecadal variability of western North Pacific tropical cyclone tracks. J Clim 21:4464–4476 Liu KS, Chan JCL (2013) Inactive period of western North Pacific tropical cyclone activity in 1998–2011. J Clim 26:2614–2630 Lu X, Zhao B (2013) Analysis of the climatic characteristics of landing tropical cyclones in East China. J Trop Meteor 19(2):145–153 Mantua NJ, Hare SR, Zhang Y, Wallace JM, Francis RC (1997) A Pacific interdecadal climate oscillation with impacts on salmon production. Bull Am Meteor Soc 78:1069–1079

1985 Nakazawa T, Rajendran K (2007) Relationship between tropospheric circulation over the western North Pacific and tropical cyclone approach/landfall on Japan. J Meteorol Soc Jpn 85:101–114 Park DSR, Ho CH, Kim JH, Kim HS (2011) Strong landfall typhoons in Korea and Japan in a recent decade. J Geophys Res 116:D07105. doi:10.1029/2010JD014801 Park DSR, Ho CH, Kim JH (2014) Growing threat of intense tropical cyclones to East Asia over the period 1977–2010. Environ Res Lett 9:014008. doi:10.1088/1748-9326/9/1/014008 Rodionov SN (2004) A sequential algorithm for testing climate regime shifts. Geophys Res Lett 31:L09204. doi:10.1029/200 4GL019448 Simmons A, Uppala SM, Dee D, Kobayashi S (2007) ERA-Interim: new ECMWF reanalysis products from 1989 onwards. ECMWF Newsl 110:25–35 Southern RL (1979) The global socio-economic impact of tropical cyclones. Aust Meteorol Mag 26:175–195 Tu JY, Chou C, Chu PS (2009) The abrupt shift of typhoon activity in the vicinity of Taiwan and its association with western North Pacific-East Asian climate change. J Clim 22:3617–3628 Wakabayashi S, Kawamura R (2004) Extraction of major teleconnection patterns possibly associated with the anomalous summer climate in Japan. J Meteorol Soc Jpn 82:1577–1588. doi:10.2151/ jmsj.82.1577 Wallace JM, Smith C, Bretherton CS (1992) Singular value decomposition of wintertime sea surface temperature and 500-mb height anomalies. J Clim 5:561–576 Weatherford C, Gray WM (1988) Typhoon structure as revealed by aircraft reconnaissance. Part II: structural variability. Mon Weather Rev 116:1044–1056 Wu L, Wang B, Braun SA (2008) Implications of tropical cyclone power dissipation index. Int J Climatol 28(6):727–731. doi:10.1002/joc.1573 Ying M, Zhang W, Yu H, Lu X, Feng J, Fan Y, Zhu Y, Chen D (2013) An overview of the China Meteorological Administration tropical cyclone database. J Atmos Ocean Technol 31:287–301. doi:10.1175/JTECH-D-12-00119.1 Yonekura E, Hall TM (2014) ENSO effect on East Asian tropical cyclone landfall via changes in tracks and genesis in a statistical model. J Appl Meteor Climatol 53:406–420. doi:10.1175/ JAMC-D-12-0240.1 Zhang Q, Zhang W, Lu X, Chen YD (2011) Landfalling tropical cyclones activities in the south China: intensifying or weakening? Int J Climatol 32:1815–1924

13