Tropical cyclone (TC) activity over the western North Pacific (WNP) exhibits a significant interdecadal variation during 1960–2011, with two distinct active and inactive periods each. This study examines changes in TC activity and atmospheric conditions in the recent inactive period (1998–2011). The overall TC activity shows a significant decrease, which is partly related to the decadal variation of TC genesis frequency in the southeastern part of the WNP and the downward trend of TC genesis frequency in the main development region.
The investigation on the factors responsible for the low TC activity mainly focuses on the effect of vertical wind shear and subtropical high on multidecadal time scales. A vertical wind shear index, defined as the mean magnitude of the difference of the 200- and 850-hPa horizontal zonal winds (10°–17.5°N, 150°E–180°) averaged between June and October, is highly correlated with the annual TC number and shows a significant interdecadal variation. Positive anomalies of vertical wind shear are generally found in the eastern part of the tropical WNP during this inactive period. A subtropical high area index, calculated as the area enclosed by the 5880-gpm line of the June–October 500-hPa geopotential height (0°–40°N, 100°E–180°), shows a significant upward trend. A high correlation is also found between this index and the annual TC number, and a stronger-than-normal subtropical high is generally observed during this inactive period. The strong vertical wind shear and strong subtropical high observed during 1998–2011 together apparently lead to unfavorable atmospheric conditions for TC genesis and hence the low TC activity during the period.
The 2010 tropical cyclone (TC) season for the western North Pacific (WNP) is characterized by its exceptionally low TC activity. The number of TCs with at least tropical storm intensity is 14 and the number of typhoons is 8, with both numbers being the lowest since 1960. In fact, WNP TC activity shows a significant decline starting from 1998. It is therefore very important to find out the possible causes for this inactive period (1998–2011), which is not only useful for the seasonal forecast of TC activity but also helpful in understanding the long-term variations of TC activity in the WNP.
The variability of the TC activity over the WNP has received much attention during the last decade. Many studies have been performed on the interannual variations in TC activity (e.g., Chan 1985, 2000; Dong 1988; Chen et al. 1998; Chia and Ropelewski 2002; Wang and Chan 2002). Some more recent studies have also examined the interdecadal variations of TC activity. In particular, Matsuura et al. (2003) studied the interdecadal variability of TC frequency in the WNP between 1951 and 1999 and identified two low-frequency periods (1951–60 and 1973–85) and two high-frequency periods (1961–72 and 1986–94), which are related to the zonal wind stress in the tropical western Pacific and the sea surface temperatures (SSTs) in the tropical western and central Pacific. Chan (2008) found a strong multidecadal variation of intense typhoon occurrence during 1960–2005 in the WNP, which is related to similar variations of oceanographic and atmospheric conditions. Yeh et al. (2010) showed a decadal change in the relationship between TC frequency and tropical Pacific SST, with a positive correlation during the period 1990–2000 but a negative correlation during the period 1979–89. They showed that the oceanic conditions (SST and heat content) are more important in the period 1979–89 while the atmospheric conditions (low-level vorticity and vertical wind shear) are the dominant factors in the period 1990–2000.
Most of these previous studies focused on the interannual and interdecadal variations of TC activity but very few studies have tried to identify the causes for the recent low TC activity in the WNP. Maue (2011) showed that the recent low global TC activity is related to the large-scale climatic oscillations including El Niño–Southern Oscillation (ENSO) and the Pacific decadal oscillation (PDO). The present study, therefore, represents an attempt to examine the interdecadal variation of the TC activity over the WNP with the focus on the period 1998–2011 and investigates the atmospheric conditions responsible for this recent low TC activity.
The data and methodology employed in this study are described in section 2. Section 3 discusses the interdecadal variations of the TC activity and the spatial distribution of TC genesis over the WNP. The effects of vertical wind shear and subtropical high on TC activity are presented in section 4. The summary and discussion are then given in section 5.
2. Data and methodology
The 6-hourly best-track positions of TCs over the WNP are obtained from the Joint Typhoon Warning Center (JTWC; https://metocph.nmci.navy.mil/jtwc/best_tracks/). Since the data before 1960 are likely to have larger uncertainties according to JTWC, only the data between 1960 and 2011 are employed. In this study, TC activity is defined as the annual number of TCs with at least tropical storm intensity to minimize subjectivity in the identification of tropical depressions. The climatological mean and standard deviation of the TC number during this period are 26.8 and 5.0, respectively. The TC activity in a given year is therefore classified as normal if the TC number is within half a standard deviation of the mean (i.e., between 25 and 29). A TC season is considered as below normal (above normal) if the annual TC number is ≤24 (≥30). The location of TC formation is defined as the position first reaching tropical depression intensity.
Monthly 500-hPa geopotential heights and zonal winds at 850 and 200 hPa are extracted from the 40-yr European Centre for Medium-Range Weather Forecasts (ECMWF) Re-Analysis (ERA-40) dataset, which has a horizontal resolution of 2.5° latitude × 2.5° longitude. The heights are used to estimate the strength and extension of the subtropical high and the winds for calculating the vertical wind shear. Since this dataset is only available up to August 2002, the ECMWF Interim Re-Analysis (ERA-Interim) dataset, with a horizontal resolution of 1.5° latitude × 1.5° longitude, from September 2002 to December 2011 are used (Simmons et al. 2007). The combined dataset therefore gives the data covering the period 1960–2011.
The Niño-3.4 SST anomaly is obtained from the Climate Prediction Center website. The PDO index is obtained from the Joint Institute for the Study of the Atmosphere and Ocean of the University of Washington.
b. Regime shift detection
The regime detection algorithm developed by Rodionov (2004) is adopted to detect the regime shifts in a time series. The method identifies a significant change in the sequential running means with a certain cutoff 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 yr is used and only those regime shifts significant at the 0.1 significance level or lower are considered.
3. Variations of tropical cyclone activity
a. Interdecadal variation of TC activity
The multidecadal variation in TC activity is obvious from its time series (Fig. 1). Based on the regime detection algorithm (with a cutoff length of 10 yr and a significance level of 0.05), two active (above normal) periods: 1960–74 (period A1) and 1989–97 (period A2) and two inactive (below normal) periods: 1975–88 (period B1) and 1998–2011 (period B2) can be identified. The mean number in period A1 is 29.3 and the fluctuation is relatively large compared with the other three periods as indicated by the largest standard deviation among the four periods (Table 1). The highest number is 39 in 1964 while the lowest number is 19 in 1969, with three years (1969, 1970, and 1973) having below-normal TC activity. The first inactive period begins in 1975 with a mean TC number of 24.6, and only below-normal (6 yr) or normal TC activity (8 yr) (Table 1). Period A2 is very active, with a mean TC number of 31.1, which is higher than that in period A1. There are 8 yr with an above-normal number and 1 yr with a normal number (Table 1). The second inactive period (period B2) begins in 1998, with a mean TC number of 23.2, which is slightly lower than that of period B1. Out of the 14 TC seasons, 9 are below normal, 4 are normal, and only 1 is above normal (Table 1). It should be noted that the percentage of the TC seasons with below-normal TC activity is 64%, which is higher than that in period B1 (43%). Moreover, the two years with the lowest TC numbers (1998 and 2010) are both in period B2. Thus, the period B2 is unprecedentedly inactive since the record began in 1960 and it is therefore of importance to examine the atmospheric conditions responsible for such changes. The focus of this paper is therefore on period B2.
b. Changes in the spatial distribution of TC formation location
Because the annual TC number over the entire WNP shows a significant decrease in period B2, it is of interest to examine the changes in spatial distribution of formation location in this period. The mean anomalous number of TC formation in each 10° × 10° box (Fig. 2) shows that for the region between 0° and 10°N, a decrease in TC formations is found east of 140°E, with a total decrease of 1.4. The decrease is more significant for the region between 10° and 20°N, with the maximum decrease between 130° and 160°E. On the other hand, an increase in the number of TC formation is generally observed north of 20°N. Thus, the inactive TC activity in period B2 is mainly due to the decrease in the number of TC formation in the southeast quadrant (0°–20°N, 140°E–180°) of the WNP, with a total decrease of 3.4. On the other hand, a northward shift in formation location is also observed between 120° and 160°E, with fewer (more) TCs forming south (north) of 20°N. These changes in the spatial distribution of TC formation are related to the decadal variations and trend of TC formation in different regions, which are discussed further below.
1) Long-term trend
The number of TC formations in the region 10°–20°N, 130°–150°E (which is defined as the main development region) has a significant downward trend (Fig. 3a). The mean numbers in the periods A1, B1, A2, and B2 are 8.5, 6.2, 7.8, and 5.3, respectively, with those in the two active periods (A1 and A2) higher than those in the two inactive periods (B1 and B2).
The mean TC number in period B2 is 1.7 lower than the climatological mean (7.0). Of the 14 yr in this period, 8 are associated with below-normal number (≤5) and 6 are associated with normal number (6–8). Thus, the downward trend of TC formation in this region contributes to the decrease of total TC number over the entire WNP in period B2.
Ho et al. (2004) examined the interdecadal variability of summertime (June–September) typhoon tracks over the WNP and found that the typhoon passage frequency decreased significantly over the Philippine Sea (5°–15°N, 135°–155°E) in the period 1980–2001 as compared with the period 1951–79, which is related to the westward expansion of the subtropical high. This region is close to the main development region defined in the present study. Thus, the decrease in typhoon passage frequency identified by Ho et al. (2004) may be related to the decreasing trend of TC genesis frequency found in the present study.
2) Interdecadal change
In addition to the decreasing trend of TC genesis frequency in the main development region, there are some regions with interdecadal variations of TC genesis frequency, which also contribute to a change of TC activity over the entire WNP. The regime detection algorithm (section 2b) identifies a significant interdecadal variation in the number of TC formations in the region 0°–10°N, 150°E–180° (Fig. 3b). The first inactive period is between 1960 and 1975, with a mean number of 1.6. The mean number rises to 4.3 in the active period 1976–97. The second inactive period is from 1998 to 2011, with the mean number of 1.4. Note that this period covers the same years as period B2. The correlation between the TC formation number and the June–October Niño-3.4 index is 0.67, suggesting a possible relation with ENSO (Chan 1985, 2000; Wang and Chan 2002; Camargo et al. 2007). The numbers in El Niño years are generally above the climatological mean except for 1963, 1969, 2006, and 2009 while in La Niña years, the numbers are below the climatological mean except for 1964. Thus, the number of TC formations in this region also exhibits an interannual variation, which is partly controlled by the ENSO event. It is interesting to point out that in the period 1998–2011, there are 7 yr with no TC forming in this region, including El Niño year 2006, which has not been observed in the previous El Niño years during the period of study. Moreover, only one TC formed in El Niño year 2009. Thus, the enhancement of TC formation due to El Niño appears to be not significant in this period. On the other hand, there are two ENSO-neutral years (2005 and 2008) with no TC formation and this only happened twice in the previous ENSO-neutral years (1960 and 1983). Thus, the decrease in TC formation appears to be more significant for the latter half (2005–11) of this period, which may be related to other climatic oscillations such as the PDO.
The number of TC formations in the region 10°–20°N, 160°E–180° also shows a remarkable interdecadal variation (Fig. 3c). The mean number in the first active period (1960–72) of 1.4 reduces to 0.6 in the first inactive period (1973–90). No TC formed in this region in 10 out of the 18 yr in this period. The second active period is between 1991 and 2004 and the mean number rises to 1.9. The TC formation becomes inactive again from 2005. It is interesting to note that no TC formed in this region for seven consecutive years, which has not occurred since the record began in 1960. The absence of TC formation partly explains the low TC activity in the latter part (2005–11) of period B2.
The above results suggest that the frequency of TC genesis shows a remarkable decrease in these two regions (0°–10°N, 150°E–180° and 10°–20°N, 160°E–180°) during the period 2005–11, which partly reflects the suppressed TC activity over the entire WNP. A westward shift in TC formation locations can clearly be seen by comparing the TC tracks in the two subperiods of period B2. For the first subperiod (1998–2004), many TCs formed in these two regions, with a mean of about 4.3 TCs per year (Fig. 4a). In contrast, only two TCs are found in the second subperiod (2005–11) (Fig. 4b).
The PDO is an important climatic oscillation that has a significant influence on the Pacific climate (Mantua et al. 1997) and TC activity (Wang et al. 2010; Aiyyer and Thorncroft 2011) and therefore it is worth investigating its possible effect on TC genesis. The correlation between the June–October PDO index and the TC genesis number in these two regions is 0.49. The correlation is higher (r = 0.61) if the August–October PDO index is used, indicating a greater influence of late summer PDO on TC activity.
To examine the PDO effect further, a year is classified as the PDO+ (PDO−) year if the mean June–October PDO index is ≥0.5 (≤−0.5). The mean TC genesis numbers in the PDO+ years (12 cases) and the PDO− years (17 cases) are 5.8 and 1.9, respectively, and the difference is statistically significant at the 99% confidence level. Thus, the above results suggest that the TC genesis number in these two regions is generally higher (lower) in a PDO+ (PDO−) year. Since some PDO+ (PDO−) years may be associated with El Niño (La Niña) events, the changes in TC genesis number may be due to the ENSO event rather than the PDO. To separate out the ENSO effect, only the ENSO-neutral years or the ENSO years with weak strength in the months of June–October are considered. The mean TC genesis number for the PDO+ years (1976, 1979, 1981, 1983, 1985, 1992, 1993, 1995, and 2003) is 4.9, which is still significantly higher than the mean number of 2.1 for the PDO− years (1961, 1962, 1967, 2000, 2001, 2008, and 2011).
The time series of the PDO index also shows an interdecadal variation (Fig. 5). The PDO was generally in its cold phase in the period 1960–75. A regime shift occurred in 1976 and the PDO changed to its warm phase. A possible shift to the cold phase was observed in 1998. The mean numbers of TC formations in the two regions (0°–10°N, 150°E–180° and 10°–20°N, 160°E–180°) in the periods 1960–75, 1976–97, and 1998–2011 are 2.8, 5.5, and 2.3, respectively. The mean number in the warm period is significantly higher than those in the two cold periods. Thus, the PDO may have a significant effect on the TC genesis frequency in the southeast quadrant of the WNP on an interdecadal time scale. The low TC genesis frequency in the period B2 may be related to the frequent occurrence of La Niña events (1998, 1999, 2007, and 2010) and the cold PDO events (2000, 2001, 2008, and 2011).
Wang and Chan (2002) also found the ENSO-related interannual variation of TC formation in the southeast quadrant (5°–17°N, 140°E–180°) and the present study is therefore consistent with their findings. We further show that an interdecadal variation of TC formation also exists in this region, which may be related to other climatic oscillations such as the PDO. To summarize, the suppressed TC activity in period B2 is mainly due to the downward trend of TC formation in the main development region and the decadal changes of TC formation in the southeast quadrant of the WNP.
4. Atmospheric conditions
It has been shown that the WNP TC activity and the spatial distribution of TC formation locations have significant decadal variations and the period 1998–2011 is identified as an inactive period. The next part of this study is to identify the possible factors responsible for such variations. The variations of vertical wind shear (VWS) and subtropical high are examined first and their possible relationships with TC activity are then discussed.
a. Vertical wind shear
In general, westerly shear dominates in the eastern part of the tropical WNP and easterly shear over the western part as well as the South China Sea. Because the meridional shear is relatively weak, the VWS in this study is defined as the magnitude of the difference between the 200- and 850-hPa zonal winds. The correlation map between the June–October (during which most of the TCs occur) VWS and the annual TC number shows negative correlations over the eastern part of the WNP, with maximum amplitude near 15°N, 160°E (Fig. 6a). Thus, TC activity tends to be lower (higher) if the VWS in this region is stronger (weaker).
To study the relationship between VWS and TC activity, a vertical wind shear index (hereafter referred to as VWSI) is defined, which is calculated as the standardized mean June–October VWS in the region 10°–17.5°N, 150°E–180°. The correlation between VWSI and the annual TC number over the WNP is −0.64, which is statistically significant at the 99% confidence level. In other words, a stronger (weaker) VWS over the eastern part of the WNP, which corresponds to a positive (negative) value of VWSI, is associated with below-normal (above normal) annual WNP TC activity.
The time series of VWSI shows a significant interannual variation (Fig. 7a) and is negatively correlated with the June–October Niño-3.4 index (r = −0.46), suggesting a possible relation with ENSO. The VWS is generally weaker (stronger) in an El Niño (a La Niña) year. Indeed, the mean VWSI values in El Niño and La Niña years are −0.46 and 0.81, respectively, and the difference is statistically significant at the 99% confidence level.
The time series of VWSI also exhibits a significant interdecadal variation. The mean VWSI in periods A1, B1, A2, and B2 are −0.33, 0.21, −0.35, and 0.52, respectively. Although there is a large interannual variation associated with ENSO events, the interdecadal variation may be reflected by the changes of VWSI in ENSO-neutral years. The composite of the mean VWS in period A1 shows negative anomalies over the tropical WNP east of 150°E (Fig. 8a). However, such anomalies are not significant, which may be related to the large fluctuation in El Niño and La Niña years as indicated by the time series of VWSI (see Fig. 7a). It is worth noting that the VWSI in the six ENSO-neutral years (1960, 1961, 1962, 1966, 1967, and 1968) are all negative and the composite pattern for these ENSO-neutral years (not shown) actually shows the significant negative anomalies in this region. The atmosphere tends to have a weaker VWS in the absence of an ENSO effect during this period, which partly explains the above-normal TC activity in ENSO-neutral years. The changes of TC activity in the ENSO years are largely related to the changes of VWS associated with ENSO events. In period B1, the VWS anomalies are positive in the eastern part of the tropical WNP, with maximum anomalies occurring near 15°N, 155°E (Fig. 8b). In contrast to period A1, the VWSI in the seven ENSO-neutral years (1977, 1978, 1979, 1980, 1981, 1983, and 1985) are all positive, suggesting a possible change in the atmospheric conditions not favorable for TC genesis. In period A2, negative anomalies of VWS are generally in the eastern half of the WNP, with maximum anomalies occurring in the central Pacific (Fig. 8c). Compared with the last active period, such anomalies are more significant and located more eastward. The VWSI is generally negative except for 1995 and correspondingly the TC activity is generally above normal in this period.
Since the focus of this study is on period B2 (1998–2011), the atmospheric conditions in this period will be discussed in more detail. In period B2, the state of the atmosphere becomes unfavorable for TC genesis again, with obvious positive anomalies of VWS over the eastern part of the tropical WNP, with the maxima locating near 12°N, 180° (Fig. 8d). The mean VWSI is 0.52, which is higher than the last inactive period (period B1). Since the variation of VWS is related to ENSO, it is more appropriate to examine the spatial patterns during La Niña, neutral, and El Niño years separately. In period B2, there are five La Niña years (1998, 1999, 2000, 2007, 2010, and 2011). Compared with the previous La Niña years in the period 1960–97, the enhancement of VWS over the eastern part of the WNP is more prominent, including the key region used to define the VWSI (Fig. 9a). Indeed, the VWSI in 2010 is the highest since 1960, which may be one of the reasons for the record-breaking low TC number in this year. The composite for the ENSO-neutral years (2001, 2003, 2005, and 2008) also shows significant positive anomalies between 160°E and 150°W as compared with the previous ENSO-neutral years (Fig. 9b). The VWSI in 2001 and 2003 is still negative but becomes positive in 2005 and 2008. It should be noted that the VWSI in 2008 is the highest among the ENSO-neutral years. Negative anomalies are also found for El Niño years but are located more eastward (Fig. 9c). These results suggest that the VWS in the eastern half of the WNP tends to be enhanced in both ENSO and non-ENSO years. As suggested in the previous section, the decrease in TC activity is more significant in the second half of period B2 (2005–11). It is, therefore, useful to compare the VWS patterns in these two subperiods (Fig. 10). It can be seen that the patterns are similar but the positive anomalies on the right are much more significant for the second subperiod, which is consistent with the variations of TC activity in these two subperiods.
It has been shown that the interdecadal changes of the VWS in the eastern part of the WNP may contribute toward the interdecadal variations of the annual TC activity over the entire WNP. To demonstrate the effect of VWS on TC genesis, the difference in spatial distribution of TC formation locations between weak and strong VWS years is examined. A year is defined as a weak (strong) VWS year if the VWSI is <−0.5 (>0.5). During the period of study, there are 19 weak VWS years and 14 strong VWS years. The TC genesis in the region east of 145°E is generally higher for the weak VWS years, with the largest difference near 15°N, 150°E (Fig. 11a). The mean numbers of TC formations in the region 0°–20°N, 145°–175°E in weak and strong VWS years are 9.6 and 3.8, respectively. Recall from section 3b that the numbers of TC formations in the regions 0°–10°N, 150°E–180° and 10°–20°N, 160°E–180° have significant interdecadal variations. The correlations between the VWSI and these two TC numbers are −0.38 and −0.37, respectively, which suggest that the VWS is partly responsible for such variations. All these results therefore suggest that the influence of VWS on the WNP TC activity is likely through the alteration of the number of TC formations in the southeastern part of the WNP.
In the studies of the relationship between VWS and TC activity, some define the VWS as the vertical shear of zonal winds (Ho et al. 2005; Yeh et al. 2010), while the others utilize both the zonal and meridional winds (Chu 2002; Aiyyer and Thorncroft 2011). In the present study, the VWS is defined as the magnitude of the difference of the 200- and 850-hPa horizontal zonal winds. Our results are not sensitive to the definition that we chose. The time series of the VWSI based on these two definitions show similar variations and the differences in their yearly values are small. Thus, those results based on the VWSI would be similar if both the zonal and meridional components are used. For example, a significant correlation still exists between the VWSI and TC number (r = −0.61). In addition, no significant change is found for the VWS composite patterns (Figs. 8–10).
In the WNP, the VWS has been considered as one of the factors modulating the TC activity. However, only a few studies have been done to examine its interdecadal variation and the possible relationship with decadal changes of TC activity. Chan (2008) found that the VWS is one of the dynamic factors responsible for the decadal variations of intense typhoon occurrence in the WNP, with smaller (larger) VWS in periods with above-normal (below normal) intense typhoon occurrence. Aiyyer and Thorncroft (2011) examined the variability of July–October VWS and its possible relationship with TC activity. They identified the decadal mode of low-frequency VWS which is significant within the central and western Pacific. This mode may be related to the PDO and no significant relation is found between the July–October accumulated cyclone energy (ACE) and the VWS over the western Pacific (10°–25°N, 120°–160°E). The decadal variations of VWS identified by the present study therefore confirm the interdecadal variability of the VWS in the WNP. We also identified a significant relationship between wind shear and number of tropical storms other than the relationship with intense typhoon occurrence suggested by Chan (2008). However, our results are different from those of Aiyyer and Thorncroft (2011). This discrepancy may be related to the difference in the TC parameters examined by the two studies. Their study examined the ACE, which is affected by TC number, intensity, and life span, while our study focused on the TC number. In addition, the areas examined are also different, with one focusing on the area west of 160°E and the other focusing on the southeastern part of the WNP. Nevertheless, it may be concluded that the vertical wind shear in the WNP exhibits interannual and interdecadal variations, which not only control the variations of intense typhoon occurrence but also the overall TC activity. The proposed VWS index is believed to be able to reflect the changes of VWS and the possible impact on TC activity.
b. Subtropical high
The correlation map between the June–October 500-hPa geopotential height and the annual TC number shows negative correlations over the entire WNP, with maximum magnitude between 15° and 25°N (Fig. 6b). Thus, the variations of TC activity are related to the western Pacific subtropical high. The characteristics of the subtropical high can be described by its strength, area, and position. In this study, we employ a subtropical high area index (SHAI). The June–October 500-hPa geopotential height data are first interpolated into the 1° × 1° grid. The number of grid points enclosed by the 5880-gpm line on the grid within the region 0°–40°N, 100°E–180° is counted. The anomalous number is then standardized to give the index. Thus, a positive (negative) value of SHAI indicates a larger (smaller) areal coverage of the subtropical high, implying a stronger (weaker) subtropical high. The SHAI is negatively correlated with the annual TC number, with a coefficient of −0.51 (statistically significant at the 99% confidence level). When the subtropical high is stronger, the sinking motion associated with the subtropical high tends to suppress the convective activity. The easterly anomalies associated with the more southward extent of the subtropical high may also enhance the tropical easterly trade winds, leading to the weakening of the monsoon trough. The atmospheric conditions are not favorable for TC genesis and the TC activity is therefore generally reduced. No significant relationship is found between ENSO and SHAI.
The time series of SHAI shows a significant increasing trend, with an obvious upward jump in 1979 (Fig. 7b). Some studies also found the westward expansion of the western Pacific subtropical high since the late 1970s (e.g., Zhou et al. 2009). The mean SHAI in periods A1, B1, A2, and B2 are −0.68, 0.04, 0.01, and 0.74, respectively, showing an obvious upward trend. The SHAI in period A1 is generally negative, indicating a weaker subtropical high. The mean positions of the western and southern edges are 142°E and 23°N, respectively (see Fig. 12a). An upward jump in SHAI appears in 1979 and the mean SHAI in periods B1 therefore rises to 0.04, indicating a strengthening of the subtropical high. The mean positions of the western and southern edges shift to 136°E and 19°N, respectively. The mean SHAI in period A2 is similar to period B1, with a similar position of the western edge but a slightly southward expansion of the subtropical high.
A significant strengthening of the subtropical high occurs in period B2, as indicated by the upward jump of the mean SHAI to 0.74. The SHAI is generally positive except for 2000 and 2004 (Fig. 7b). Note that 1998 and 2010 are the two years with the highest SHAI and the exceptionally strong subtropical high may partly explain the very low TC activity in these years. The mean position of the western edge shifts to 131°E, which is the most westward among the four periods. This also suggests that the subtropical high is generally stronger than the normal, which is probably related to the inactive TC seasons in this period. It should also be noted that the subtropical high in the second subperiod (2005–11) of period B2 is stronger than the first subperiod (1998–2004) (Fig. 12b).
The influence of subtropical high on TC genesis can be clearly seen from the difference in spatial distribution of TC genesis between weak (SHAI < −0.5) and strong (SHAI > 0.5) subtropical high years. In general, the TC genesis is higher in most parts of the WNP in weak subtropical high years (Fig. 11b). The enhancement of TC genesis is more significant in the region 10°–30°N, 125°–145°E, which is near the western edge of the mean position of subtropical high. Thus, the change in westward extent of subtropical high may have an effect on TC genesis in this region and hence the TC activity in the entire region. When the subtropical high extends more westward and southward, the TC genesis in this region tends to be reduced. In contrast, TC genesis tends to be enhanced when the subtropical high is located more eastward and northward. In fact, the numbers of TC formations in the weak and strong subtropical high years are 10.4 and 7.4, respectively, and the difference is statistically significant at the 95% confidence level.
The changes in atmospheric circulation associated with the variations of subtropical high are generally consistent with the pattern of its regional impact on TC genesis. The strength and extent of the subtropical high may influence the low-level winds at its southern flank as suggested by the negative correlations between the SHAI and the 850-hPa zonal winds along 15°N (Fig. 13a). A higher value of SHAI indicates an increase in the extent of subtropical high. The easterly anomalies associated with the more southward extent of the subtropical high may enhance the tropical easterly trade winds, leading to the weakening of the monsoon trough, which is not favorable for TC genesis. The typical examples can found be in 1998 and 2010, the two years with the highest SHAI and the lowest TC number. The southern edge of the 5880-gpm line is near 10°N (not shown), which is much more southward than its climatological position (see Fig. 12b). The easterly trade winds dominate the entire tropical WNP and the westerly winds can only be found in the areas west of 120°E. As a result, the monsoon trough is much weaker than normal and the TC activity is therefore very low. In contrast, when the value of SHAI is lower, the subtropical high tends to be weaker and more northward. Its influence on the tropical winds is less significant and the equatorial westerlies may penetrate more northward and eastward, resulting in a stronger monsoon trough. Indeed, the influence of the subtropical high on the monsoon trough can be seen from the correlation map between the SHAI and the 850-hPa relative vorticity (Fig. 13b). Note that the two regions with the significant correlations—one being east of the Philippines and the other centered at near 20°N, 150°E—generally match the regions with the significant impact of the subtropical high on TC genesis (see Fig. 11b).
It is worth noting that the regional impact on TC genesis of the subtropical high is different from that of VWS (cf. Figs. 11a and 11b). The major impact of VWS is on the southeastern part of the WNP. The number of TC formations in the region 0°–20°N, 145°–175°E is highly correlated with the VWSI (r = −0.60) but is only weakly correlated with the SHAI (r = −0.18). In contrast, the impact of the subtropical high is mainly on the western part of the WNP. The correlation between the number of TC formations in the region 10°–30°N, 125°–145°E and the SHAI is −0.39 but no correlation is found for the VWSI (r = −0.09). Thus, the combined effect of VWS and the subtropical high may explain the variations of TC genesis in most parts of the WNP and hence the TC activity over the whole WNP. Indeed, the multiple correlation between the annual number of TC number in the whole WNP and the two indices rises to −0.69, which is higher than the correlations using the individual index (the individual correlations being −0.64 and −0.53, respectively).
It has been shown that the decadal changes in VWS and subtropical high may contribute toward the variations of the WNP TC activity on multidecadal time scales. In period A1, the weaker VWS and subtropical high are the primary factors for the high TC activity in this period. The subtropical high is near normal in both periods B1 and A2 and, therefore, may not have a significant effect on the TC activity in these periods. Instead, the changes in VWS are the major factor for the low TC activity in period B1 and high TC activity in period A2. In period B2, the stronger VWS and subtropical high both contribute to the exceptionally low TC activity in this period.
5. Summary and discussion
Tropical cyclone (TC) activity, measured as the annual number of tropical storms, over the western North Pacific (WNP) has shown to exhibit a significant interdecadal variation during the last five decades, with two active periods (1960–74 and 1989–97) and two inactive periods (1975–88 and 1998–2011). The focus of this study is on the recent inactive period 1998–2011, namely, period B2. The TC activity in period B2 is unprecedentedly low, with most of the TC seasons having below-normal or near-normal TC activity. Variations in the spatial distribution of TC genesis location are also identified. The number of TC formation in the southeastern part of the WNP has a significant interdecadal variation, with the recent inactive period covering most parts of period B2. The decrease in TC genesis frequency is more significant for the latter half (2005–11), with no TC forming in the region 0°–20°N, 160°E–180°. On the other hand, a significant downward trend of TC genesis frequency is also observed in the main development region (10°–20°N, 130°–150°E) and the number of TC formation in period B2 is found to be significantly below the long-term mean. Therefore, the decrease in TC genesis frequency in these two regions partly contributes to the low TC activity in period B2.
The second part of this study is to identify the factors responsible for the interdecadal variation of the TC activity and hence the reasons for the low TC activity in period B2. A vertical wind shear index (VWSI) is defined as the mean magnitude of the difference of the 200- and 850-hPa horizontal zonal winds in the region 10°–17.5°N, 150°E–180° averaged between June and October. The high correlation between the VWSI and the TC number suggests that the vertical wind shear has a significant effect on the interannual variation of TC activity. The VWSI also shows a significant interdecadal variation, which partly explains the interdecadal variation of the TC activity. Positive anomalies of vertical wind shear are generally found in the eastern part of the tropical WNP in period B2. The variation of the strength and extent of the subtropical high is represented by the subtropical high area index, which is calculated as the area enclosed by the 5880-gpm line of the June–October 500-hPa geopotential height within the region 0°–40°N, 100°E–180°. This index shows a significant increasing trend, with an obvious upward jump in 1979 and is negatively correlated with the annual TC number. A stronger-than-normal subtropical high is generally observed in the inactive period 1998–2011. The sinking motion associated with the subtropical high tends to suppress the convective activity. The easterly anomalies associated with the more southward extent of the subtropical high may enhance the tropical easterly trade winds, leading to the weakening of the monsoon trough. The atmospheric conditions are therefore not favorable for TC genesis and this partly explains the below-normal TC activity in this period. Thus, it may be concluded that the strong vertical wind shear and the strong subtropical high are the major factors responsible for the low TC activity in the period 1998–2011.
The recent low TC activity in the period 1998–2011 is primarily related to the decadal change of the TC activity. It is important to know the time of occurrence of the possible regime shift from this inactive period to a new active period and analyses of the past TC activity, therefore, provide useful information for such a regime shift. A wavelet analysis (not shown) of the annual TC number gives a distinct oscillation period of 16–32 yr, suggesting a possible length of 8–16 years for each period (active or inactive). Indeed, the lengths of the two active periods (1960–74 and 1989–97) and the first inactive period (1975–88) are 15, 9, and 14 yr, respectively, all are within this range. The recent inactive period (1998–2011) has lasted for at least 14 yr and therefore a possible regime shift may occur in the coming few years but it should be noted that the length of the period could be altered due to global warming or decadal changes of climatic oscillations governing the TC activity.
The VWS in the eastern part of the tropical WNP has shown to have a significant interdecadal variation and is generally stronger in the recent inactive period (1998–2011). However, the physical causes for such variations are not well documented. Possible reasons include decadal changes of ENSO. Recent studies showed an increasing frequency of occurrence of central Pacific warming (CPW; also known as El Niño Modoki) during the last few decades (Ashok and Yamagata 2009). It is, therefore, interesting to see if this is related to the changes in VWS and the recent decline in TC activity. Chen and Tam (2010) examined the different impacts of El Niño Modoki and canonical ENSO events on TC frequency over the WNP and found that the summertime (June–October) TC frequency is significantly positively correlated with the ENSO Modoki index, suggesting a higher TC frequency in a year with an El Niño Modoki event. Thus, the increasing frequency of occurrence of CPW might mitigate the other effects that cause the recent decline in TC activity. Indeed, the annual TC numbers in the two recent strong El Niño Modoki years (2002 and 2004) are 26 and 30, respectively, which are higher than most of the years in the period 1998–2011 (see Fig. 1). Kim et al. (2011) also examined the difference in TC activity between east Pacific warming (EPW) and CPW and showed that the VWS in the northwestern part of the WNP is stronger in EPW years but near normal in CPW years. However, no significant difference in the spatial pattern of VWS in the tropical WNP is found between EPW and CPW years. Thus, the decadal change in the VWSI may not be related to the increasing frequency of occurrence of the CPW.
We have shown that the PDO may have a significant effect on the TC genesis frequency in the southeast quadrant of the WNP on an interdecadal time scale. However, its relation with the TC number in the entire WNP is insignificant, which suggests that the PDO may only play a secondary role in modulation of the TC activity in the WNP. Some recent studies also proposed the possible effect of PDO on TC activity but no relevant physical mechanism is given (Wang et al. 2010; Aiyyer and Thorncroft 2011). Thus, further studies are required to investigate the PDO effect more thoroughly and find out the physical mechanism involved.
It has been proposed that the interdecadal variations of the VWS and subtropical high are the main factors responsible for the decadal changes of TC activity. However, there may be other atmospheric and oceanic factors contributing to these changes, which require further investigation. In addition, the causes for the interdecadal variations of the VWS and subtropical high are still not clear and, therefore, further studies are required to investigate the physical mechanism responsible for the changes in the VWS and subtropical high. Moreover, it is also important to identify the preseason (the spring or the preceding winter) atmospheric and oceanic conditions that are related to the strength of VWS and the subtropical high in the TC season. These precursors may then be used as potential predictors for the seasonal forecast of TC activity.
Comments from the two anonymous reviewers were helpful in improving the manuscript and are gratefully acknowledged.