Abstract

This study introduces a modified Pacific–Japan (PJ) index that exhibits a substantial periodicity of 5–16 days in the East Asian summer monsoon region. The quasi-periodic fluctuations of the PJ index can indicate changes in the large-scale circulation systems. In the PJ high phase, the wave pattern propagates northwestward from the western North Pacific tropics to an area near northern Luzon and is then forced to move westward because of a stationary, anomalous high pressure system over southern Japan. The tropical cyclones (TCs) associated with the anomalous low pressure systems tend to follow a straight-moving propagation route through the northern South China Sea. The anomalous cyclonic flow causes heavy rainfall in eastern Taiwan. However, in the PJ low phase, the wave pattern and TCs follow a recurving propagation route toward higher latitudes. The circulation pattern typically brings heavy rainfall to northern Taiwan in the PJ low phase. Therefore, wave patterns under the influence of the quasi-periodic fluctuations of the PJ pattern affect rainfall because of the changing propagation routes of the wave patterns, as well as the TC tracks.

1. Introduction

The Pacific–Japan (PJ) pattern in the western North Pacific (WNP), first reported by Nitta (1987), has a critical effect on East Asian large-scale circulation and weather (Choi et al. 2010; Ko and Tzeng 2013; Hsu and Lin 2007; Kosaka and Nakamura 2010). The PJ pattern is characterized by a strong convective band near the Philippines along 20°N and a high pressure anomaly extending from east China through the Japanese Islands to the North Pacific during the summer (Nitta 1987). This meridional wavy pattern could result from Rossby waves generated by the tropical heat source associated with intraseasonal variability and high pressure anomalies over East Asia. Hsu and Lin (2007) identified a meridionally banded tripole rainfall pattern resembling the PJ pattern during the East Asian summer. They suggested that when this meridionally banded tripole rainfall pattern is active, the zonally oriented overturning circulation driven by the positive sea surface temperature anomaly in the equatorial eastern Pacific induces heating anomalies in the tropical western Pacific. This triggers a wavelike pattern emanating northward toward extratropical East Asia. Kosaka and Nakamura (2010) used a two-layer model and successfully simulated the PJ pattern against diabatic heating. In addition, they showed that the PJ pattern could intensify anomalous convective activity over the Philippines.

In addition to the aforementioned studies on the characteristics and dynamics of the PJ pattern, Wakabayashi and Kawamura (2004) proposed a PJ index based on the empirical orthogonal function and regression analyses to streamfunction anomalies, and showed that the PJ pattern was closely related to the summer temperature anomalies of northern Japan. Using the same PJ index combined with the tropical cyclone (TC) passive frequency in the WNP, Choi et al. (2010) showed that TCs tended to move northward during the positive phase and westward during the negative phase. The positive phase was characterized by an anticyclonic circulation centered to the east of Japan as well as a cyclonic circulation centered to the east of Taiwan. In the negative phase, a stronger anticyclonic circulation was centered to the east of Taiwan. Ko and Tzeng (2013) separated the summer monsoon circulation into two types, based on wind directions. They claimed that one of the monsoon-type circulations was similar to the PJ pattern. Therefore, changes in circulation patterns can affect the weather in the East Asian monsoon region.

The migrating waves prevailing in the WNP have also attracted increasing attention in recent years. Among these waves, the synoptic (3–8 days) and higher-frequency (7–30 days or 10–25 days) branches of intraseasonal oscillations (ISOs) substantially affect East Asian weather and climate (Lau and Lau 1990; Chang et al. 1996; Hsu 2005; Ko and Hsu 2006, 2009; Ko et al. 2012; Ko and Chiu 2014; Fukutomi and Yasunari 1999, 2002). Lau and Lau (1990) identified 3–8-day synoptic waves propagating northwestward from the tropical WNP to the South China Sea. Chang et al. (1996) further examined these waves and showed that TCs tended to occur in the cyclonic circulations of these synoptic waves. The TC tracks and moving directions associated with synoptic waves were essentially straight moving. On the basis of fluctuations in wind speed between Taiwan and Japan, Ko and Hsu (2006) identified a 7–30-day wave pattern propagating north-northwestward from northeast Papua New Guinea to the East China Sea from July to August. More than 70% of the submonthly cases were associated with TCs embedded in their cyclonic circulations. The TC moving directions associated with the submonthly wave pattern were generally recurving toward the midlatitude areas. In an extension study, Ko and Hsu (2009) investigated the submonthly wave pattern under ISO westerly and easterly phases and found that the submonthly wave pattern was stronger and more organized in the westerly phase but weaker and less organized in the easterly phase. Tropical cyclones associated with submonthly wave patterns tended to be stronger in the westerly phase than the easterly phase because of the increased moisture supply generated by the ISO westerly flow. The aforementioned studies provide a link between the wave patterns and TCs; however, the connection between synoptic waves and submonthly wave patterns remains unknown.

In a composite study, Fukutomi and Yasunari (1999) used 10–25-day filtered data to explore the convection and lower-tropospheric circulation of this frequency band. They identified a wave train with a convective center near the South China Sea and an anomalous anticyclonic circulation over the North Pacific from June to July. After the peak phase of the convection, the anomalous anticyclone moved southwestward to the South China Sea and appeared to initiate a subsequent suppressed convection. The southwestward movement in this area was unique. However, in an extension study, Fukutomi and Yasunari (2002) examined a wave train in August that exhibited a similar horizontal structure to that of the circulation from June to July, but observed no southwestward propagation.

The East Asian summer monsoon region has an abundance of TCs. Figure 1 shows TC tracks for July–August–September–October (JASO) between 1979 and 2009. Because we focus only on the East Asian summer monsoon region, the TCs that exist only east of 155°E or south of 10°N are removed. The TC tracks in the WNP can be approximately divided into two types: straight moving and recurving. Although Choi et al. (2010) found the links between the PJ pattern and TC passive frequency, the connection between large-scale circulation anomalies associated with summer wave patterns and the aforementioned two TC-track types under the influence of the PJ pattern remained unknown. In addition, most studies investigating the PJ pattern have focused on mean state spatial patterns, with less attention given to the temporal variation of the PJ pattern. The present study analyzes the quasi-periodic behavior of the PJ pattern and the associated wave motions and TC tracks. The remainder of this paper is organized as follows. Section 2 describes the data and analysis procedures. Section 3 describes the mean state and temporal variability of the PJ pattern, and section 4 discusses the composite results. The TC tracks associated with the wave patterns are described in section 5, along with the rainfall patterns of Taiwan under the influence of the wave patterns. Finally, section 6 presents the conclusions.

Fig. 1.

All tropical cyclone tracks (thin black lines) except TCs that exist only east of 155°E or south of 10°N are removed in the western North Pacific for July–October during 1979–2009. The red arrows stand for the general track patterns for straight-moving and recurving tropical cyclones.

Fig. 1.

All tropical cyclone tracks (thin black lines) except TCs that exist only east of 155°E or south of 10°N are removed in the western North Pacific for July–October during 1979–2009. The red arrows stand for the general track patterns for straight-moving and recurving tropical cyclones.

2. Data and analysis procedures

Four datasets are used in this study. The circulation data were extracted from the National Centers for Environmental Prediction (NCEP) Reanalysis 1 dataset (Kalnay et al. 1996). This dataset contained 6-hourly readings (0000, 0600, 1200, and 1800 UTC) of temperature, humidity, horizontal winds, vertical velocity, and geopotential height on a 2.5° × 2.5° latitude–longitude grid. The second dataset contained the interpolated outgoing longwave radiation (OLR) data from the Climate Diagnostics Center and National Oceanic and Atmospheric Administration. The OLR data consisted of only daily readings but used the same grid spacing as the NCEP reanalysis dataset. In addition, the best track TC data, compiled by the Joint Typhoon Warning Center in Guam, were used for comparison with the circulation features. The fourth dataset comprised the precipitation data from the Taiwan Climate Change Projection and Information Platform (TCCIP). These are daily readings on a 1 km × 1 km grid. The spatial domain is 21.67°–25.41°N, 120°–122.14°E, which encompasses all of Taiwan. This dataset combines precipitation data from all of the observation stations in Taiwan and fits them to grid points over land areas according to the latent Gaussian variable method (Glasbey and Nevison 1997). This study analyzed 31 years of JASO data from 1979 to 2009. The Butterworth bandpass filter (Kaylor 1977) used by Hsu and Weng (2001) was applied to isolate the periodic signals and extract the periodic fluctuations.

3. Mean state and temporal variability

The climatological mean maps spanning 31 years of 850-hPa geopotential heights and variances during the JASO months are shown in Fig. 2. The East Asian summer monsoon region is characterized by a strong monsoon trough extending from the northern portion of the Indo-China peninsula through the area north of Luzon to 10°N, 140°E. Northeast of the eastern portion of the monsoon trough is a subtropical high with a ridge extending westward to eastern China. The variance map exhibits a small maximal area just south of Japan. A modified PJ index (PJI) is defined as follows:

 
formula

where Z850 represents the geopotential height at 850 hPa. The points shown in Fig. 2b are selected because one is within the maximal variance area along the western edge of a subtropical anticyclone and the other is near the center of the monsoon trough. The northeast–southwest orientation of these two points is similar to points observed by Wakabayashi and Kawamura (2004) and Choi et al. (2010), except that their points are farther east. Additionally, these two points are located near the axes of two major TC tracks (Fig. 1). The temporal behavior of the PJI is further examined by conducting a time–frequency analysis, which is based on the fast Fourier transform (FFT) under a assumption with the degrees of freedom as 2 (sine and cosine) × 31 (years) = 62. In addition, the red noise background including the 99.9% confidence levels has been subtracted; therefore, only the values exceeding the 99.9% confidence levels are shown. The time–frequency plot is similar to that reported by Ko and Hsu (2006). Figure 3 shows the time–frequency plot from May to October averaged between 1979 and 2009. Significant spectral maxima with a periodicity of approximately 6–14 days, emerge after early July and continue through late October. This feature indicates that the PJI tends to fluctuate every 1–2 weeks. The periodicity is close to that reported by Ko and Hsu (2006, 2009) and likely connected to the synoptic wave reported by Lau and Lau (1990). Choi et al. (2010) claimed that the PJI could be used to determine not only the intensities of the monsoon trough and subtropical high but also the TC activity in the WNP. To isolate the periodic signals, a 5–16-day bandpass filter (with half-power at 5 and 16 days) is employed to focus on the stronger signals in the frequency band. Furthermore, cases are selected using a filtered time series to explore the PJ fluctuations and wave patterns that prevail during various phases of the PJ pattern. In contrast to the unfiltered geopotential height variance, Fig. 4 shows the variance in the filtered 850-hPa geopotential height. The two points selected for PJIs are located along the maximal variance axis and represent the points of high 5–16-day fluctuations.

Fig. 2.

(a) July–October averaged 850-hPa geopotential height for 1979–2009 and (b) variance of July–October geopotential height at 850 hPa for 1979–2009. The contour interval is 5 m for (a) and 400 m2 for (b). Two points selected for computing PJ index are marked in black dots.

Fig. 2.

(a) July–October averaged 850-hPa geopotential height for 1979–2009 and (b) variance of July–October geopotential height at 850 hPa for 1979–2009. The contour interval is 5 m for (a) and 400 m2 for (b). Two points selected for computing PJ index are marked in black dots.

Fig. 3.

The time–frequency plot of the PJ indices from May to October over 1979–2009. The shading represents the spectral maxima exceeding the 99.9% confidence level.

Fig. 3.

The time–frequency plot of the PJ indices from May to October over 1979–2009. The shading represents the spectral maxima exceeding the 99.9% confidence level.

Fig. 4.

Variance of the July–October filtered geopotential height at 850 hPa for 1979–2009. The contour interval is 100 m2. Two points selected for computing PJ index are marked in black dots.

Fig. 4.

Variance of the July–October filtered geopotential height at 850 hPa for 1979–2009. The contour interval is 100 m2. Two points selected for computing PJ index are marked in black dots.

Figure 5 shows an example of selected cases using the 5–16-day filtered PJIs for 1991. The maxima and minima (negative maxima) exceeding the 31-yr JASO mean ±1.0 standard deviation are selected for the PJ high (maxima) and PJ low (minima) cases. Each case is checked for the occurrence of at least one TC in the WNP. Cases are excluded (dashed arrows) if no TC occurs in the WNP from 1 day before to 1 day after the peak time of the case. To avoid spatial bias between the two PJI points, a threshold value of 14.1 m (a climatological PJI computed according to the geopotential height difference, as shown in Fig. 2a) for the unfiltered PJIs is adopted for the PJ high cases. In other words, the unfiltered PJIs of the PJ high cases have to exceed the threshold value (14.1 m). Moreover, the unfiltered PJIs of the PJ low cases must be smaller than 0 m (another threshold value for PJ low cases). During the JASO months between 1979 and 2009, the resultant number of cases is 230 (with 322 TCs) for the PJ high phase and 218 (with 319 TCs) for the PJ low phase. The unfiltered PJIs are also compared with the filtered series, as shown in Fig. 5. The fluctuations in both series are in agreement and the correlation coefficient is 0.64.

Fig. 5.

The time series for PJ indices during July–October of 1991. The thick solid line is the filtered time series and the thin solid line stands for the unfiltered time series. The dashed lines represent the 31-yr (1979–2009) JASO-averaged PJ index ±1.0 × standard deviation of the 31-yr JASO-filtered series. The solid arrows are the selected cases and the dashed arrow represents a case without any TC in the WNP. The secondary x axis represents the dates of 1991.

Fig. 5.

The time series for PJ indices during July–October of 1991. The thick solid line is the filtered time series and the thin solid line stands for the unfiltered time series. The dashed lines represent the 31-yr (1979–2009) JASO-averaged PJ index ±1.0 × standard deviation of the 31-yr JASO-filtered series. The solid arrows are the selected cases and the dashed arrow represents a case without any TC in the WNP. The secondary x axis represents the dates of 1991.

The background mean states for both the PJ high and low cases are shown in Fig. 6. The 850-hPa geopotential height mean map of the PJ high cases (Fig. 6a) shows a low pressure center located in the northern portion of the South China Sea and a monsoon trough extending eastward and east-southeastward to 10°N, 145°E, along with a strong convection just south of the trough. Additionally, a subtropical ridge extends from the central Pacific westward toward central China. This trough and ridge system resembles the PJ pattern, as shown in Ko and Tzeng (2013). However, in the mean state map of the PJ low cases (Fig. 6b), a similar monsoon trough and subtropical ridge circulation system as that in the PJ high phase exists except that the monsoon trough is weaker and there is a break in the subtropical ridge near the oceanic area east of central China. Apparently, the trough and ridge are weaker near the PJI points as shown in Fig. 2b. The north–south zonal-wind cross sections, as shown in Figs. 6c and 6d, suggest a barotropic profile for both of the PJ phases, except that the trough is located farther south in the PJ high phase than in the PJ low phase.

Fig. 6.

Mean state of the OLR (shaded) and the 850-hPa geopotential height, winds averaged for 10-day periods for (a) PJ high-index cases and (b) PJ low-index cases. The contour interval for the geopotential height is 3 m and the 1500-m contour is highlighted in purple. (c),(d) The cross sections of the zonal wind at 130°E are shown with the contour interval of 2 m s−1.

Fig. 6.

Mean state of the OLR (shaded) and the 850-hPa geopotential height, winds averaged for 10-day periods for (a) PJ high-index cases and (b) PJ low-index cases. The contour interval for the geopotential height is 3 m and the 1500-m contour is highlighted in purple. (c),(d) The cross sections of the zonal wind at 130°E are shown with the contour interval of 2 m s−1.

4. Composite results

The composite maps of the 5–16-day filtered 850-hPa geopotential height for the PJ high phase, as shown in Fig. 7, illustrate the evolution of the wave pattern. Significant winds and an anomalous convection (filtered OLR ≤ −5 W m−2) are also displayed. The testing procedures for significant winds and adjusting the degrees of freedom are shown in the  appendix. Four days before the peak time (day −4, Fig. 7a), an anomalous high pressure center is located in the northern portion of the South China Sea, and an anomalous low pressure center is located near the oceanic area south of Japan with the cyclonic flow and convection extending southward to the east of the Philippine Sea. Two days later (day −2, Fig. 7b), the anomalous high pressure system (an anticyclonic flow) moves farther westward and weakens. The anomalous low pressure system near the south of Japan weakens, but the cyclonic flow and convection intensify to form another low pressure system and moves westward to the northern portion of Luzon. A weak anomalous high pressure system emerges east of the anomalous low pressure system. This anomalous high pressure system moves farther westward, becomes much stronger, and is situated south of Japan at day 0 (Fig. 7c) while the anomalous low pressure system and wave pattern moves farther westward. After two days (day +2, Fig. 7d), the anomalous high pressure system south of Japan weakens rapidly as the wave pattern progresses westward with an elongated anticyclonic area extending from Japan through Taiwan to Luzon. At day +4, the area south of Japan is occupied by a large anomalous low pressure system as the wave pattern weakens. During the PJ high phase, the stationary anomalous high over southern Japan, as discussed by Fukutomi and Yasunari (2002), apparently blocks and prevents the wave pattern from propagating northward.

Fig. 7.

Composite 850-hPa-filtered geopotential height at days (a) −4, (b) −2, (c) 0, (d) +2, and (e) +4 for PJ high-index cases. The contour interval is 5 m and only the wind vectors exceeding the 95% confidence levels are shown. Also shown are the composite filtered OLR less than −5 W m−2 areas (shaded).

Fig. 7.

Composite 850-hPa-filtered geopotential height at days (a) −4, (b) −2, (c) 0, (d) +2, and (e) +4 for PJ high-index cases. The contour interval is 5 m and only the wind vectors exceeding the 95% confidence levels are shown. Also shown are the composite filtered OLR less than −5 W m−2 areas (shaded).

In the PJ low phase at day −4, as shown in Fig. 8a, a wave pattern (an anomalous high pressure system) extends from the area south of Japan southeastward toward 10°N, 155°E, with an anomalous low pressure and convection centered near 17°N, 145°E. This anomalous low pressure system then moves (Fig. 8b) northwestward and intensifies as the anomalous high pressure system near southern Japan moves northeastward and weakens. In the maximal phase (day 0, Fig. 8c), the anomalous low pressure system and convection reach their maximal intensities, and the propagation route exhibits a recurving path from the lower latitudes through the area south of Japan before turning northeastward to higher latitudes. The anomalous low pressure system moves farther northward at day +2 (Fig. 8d), and the convection dissipates rapidly. The area south of Japan is occupied by a large anomalous high pressure system at day +4 (Fig. 8e), while the wave pattern continues moving northward. The wave pattern in the PJ low phase resembles the pattern reported by Ko and Hsu (2006, 2009) and exhibits a recurving route.

Fig. 8.

Composite 850-hPa-filtered geopotential height at days (a) −4, (b) −2, (c) 0, (d) +2, and (e) +4 for PJ low-index cases. The contour interval is 5 m and only the wind vectors exceeding the 95% confidence levels are shown. Also shown are the composite filtered OLR less than −5 W m−2 areas (shaded).

Fig. 8.

Composite 850-hPa-filtered geopotential height at days (a) −4, (b) −2, (c) 0, (d) +2, and (e) +4 for PJ low-index cases. The contour interval is 5 m and only the wind vectors exceeding the 95% confidence levels are shown. Also shown are the composite filtered OLR less than −5 W m−2 areas (shaded).

The west–east cross sections along 20°N of the composite 5–16-day geopotential height at days −2, 0, and +2 of the PJ high phase are shown in Figs. 9a, 9c, and 9e, respectively. At day −2, an anomalous low pressure system with a barotropic structure is centered near 123°E, and another anomalous high pressure system is near 137°E. Two days later (day 0), this anomalous low pressure system develops and moves westward followed by the anomalous high pressure system. At day +2, this anomalous low and high pressure system moves farther westward and gradually decays. Descriptions of the propagation and vertical structure of the PJ low wave pattern in south–north cross sections along 130°E of the composite 5–16-day geopotential height at days −2, 0, and +2 are displayed in Figs. 9b, 9d, and 9f, respectively. At day −2 an anomalous low pressure system with a barotropic structure is centered at 22°N and a weaker anomalous high pressure system is located at approximately 40°N. Two days later (day 0), the anomalous low pressure system moves northward and reaches its maximal phase. Another anomalous high pressure system emerges at 20°N, while the anomalous low pressure system moves farther north at day +2. As can be seen from these two cross sections cutting approximately through the axes of the two wave patterns, the anomalous low pressure system in the PJ low phase is stronger than that in the PJ high phase. Based on the composite maps and cross sections, the horizontal and barotropic vertical structure near the northern South China Sea (Fig. 9) resembles that of the 10–25-day oscillation reported by Fukutomi and Yasunari (2002).

Fig. 9.

Composite cross sections for filtered geopotential height at days (a),(b) −2; (c),(d) 0; and (e),(f) 2 for (left) PJ high-index cases along 20°N and (right) PJ low-index cases along 130°E. The contour interval is 6 m.

Fig. 9.

Composite cross sections for filtered geopotential height at days (a),(b) −2; (c),(d) 0; and (e),(f) 2 for (left) PJ high-index cases along 20°N and (right) PJ low-index cases along 130°E. The contour interval is 6 m.

5. Relationship with TCs and rainfall patterns

As discussed by Chang et al. (1996) and Ko and Hsu (2006), the summertime wave patterns in the WNP are closely related to TCs. The links of the propagation routes between the wave patterns and TCs are further demonstrated by the composite 5–16-day geopotential height and concurrent TC locations at days −4, −2, and 0 (Fig. 10). Figure 10 also shows the 5–16-day geopotential height tendency (shaded). At day −4 (Fig. 10a) in the PJ high phase, the TCs are located mostly in the southern portion of the anomalous low pressure system. A negative height tendency maximum is observed just west of the anomalous low pressure system, indicating that the anomalous low tends to move westward. The westward movement of the anomalous low and TCs is more clearly observed at day −2 (Fig. 10c). The TCs are not only concentrated near the anomalous low pressure system, but also exhibit another secondary concentration center near the negative height tendency center of the weaker cyclonic circulation near 15°N, 145°E. Crucially, the emergence of a positive height tendency maximum near southern Japan becomes a dominant feature in enhancing the anomalous high pressure area and thus prevents the anomalous low and TCs from propagating northward. The anomalous high pressure system near southern Japan becomes stronger at day 0, while those two anomalous lows and TC concentration centers continue their westward propagation near the northern portion of the South China Sea. Most of the TCs (nearly 90% or more) follow the straight-moving routes (TCs located south of 30°N) although a small portion of the TCs undergo the recurving routes toward higher latitudes along the outer rim of the anomalous high. However, at day −4 in the PJ low phase (Fig. 10b), the TCs are located in a similar anomalous low pressure area that is farther east than in the PJ high phase. A negative height tendency maximum is located northwest of the anomalous low pressure system, helping the anomalous low move northwestward. Two days later (day −2, Fig. 10d), the TCs move farther northwestward with the anomalous low, and southern Japan is occupied by a strong negative height tendency maximum, which also helps the anomalous low move farther northwestward. The anomalous low pressure system of the wave pattern reaches its maximal phase at day 0, while the TCs move farther northward with the anomalous low. Overall, the TCs in the PJ low phase are more concentrated (more than 87%) near the enhanced anomalous low pressure system while it propagates northwestward to higher latitudes, but only a small portion of TCs move westward toward Vietnam where the climatological monsoon trough is located (Fig. 6b).

Fig. 10.

Composite 850-hPa filtered geopotential height and the corresponding TC positions along with the past 1-day tracks (thin lines) at days (a),(b) −4; (c),(d) −2; and (e),(f) 0 for the (left) PJ high-index cases and (right) PJ low-index cases. The contour interval is 5 m. The shaded areas represent the geopotential height tendency (m day−1). The numbers of TCs south of 30°N/total for the PJ high cases and the numbers of TCs east of 120°E/total for the PJ low cases are also shown.

Fig. 10.

Composite 850-hPa filtered geopotential height and the corresponding TC positions along with the past 1-day tracks (thin lines) at days (a),(b) −4; (c),(d) −2; and (e),(f) 0 for the (left) PJ high-index cases and (right) PJ low-index cases. The contour interval is 5 m. The shaded areas represent the geopotential height tendency (m day−1). The numbers of TCs south of 30°N/total for the PJ high cases and the numbers of TCs east of 120°E/total for the PJ low cases are also shown.

The numbers of TCs and stronger TCs (maximal wind speed ≧66 kt, 1 kt = 0.5144 m s−1) at days −4, −2, and 0 are listed in Table 1. The percentages of stronger TCs at those three times are all greater in the PJ low phase than in the PJ high phase. Therefore, the PJ low wave pattern may have created a more favorable condition for the TCs to grow. Ko and Hsu (2009) emphasized the importance of the submonthly wave pattern associated with recurving TCs, which tend to be stronger than straight-moving TCs because their developing stages are longer. Additionally, most TCs reached peak strength before recurving toward higher latitudes (Evans and McKinley 1998; Harr 2010).

Table 1.

Numbers of TCs and stronger TCs associated with the wave patterns at days −4, −2, and 0 in the (top) PJ high phase and (bottom) PJ low phase.

Numbers of TCs and stronger TCs associated with the wave patterns at days −4, −2, and 0 in the (top) PJ high phase and (bottom) PJ low phase.
Numbers of TCs and stronger TCs associated with the wave patterns at days −4, −2, and 0 in the (top) PJ high phase and (bottom) PJ low phase.

Because Taiwan is located in the middle of the propagation routes of the two wave patterns, the rainfall patterns induced by the wave patterns and TCs are valuable for Taiwan’s weather. Figure 11 shows the composite rainfall patterns derived from the TCCIP rainfall data. At day 0 of the PJ high phase, most of the heavy rainfall occurs over the eastern portion of the island because of the southeasterly flow generated by the low pressure system and the TCs near the northern portion of the South China Sea and the mountain barrier of the Central Mountain Range, which is oriented north-northeast to south-southwest in central Taiwan. Moreover, at day 0 of the PJ low phase, the low pressure system and the TCs near southern Japan tend to generate a northerly or northwesterly flow and form heavy rainfall on the wind side of northern Taiwan. Therefore, the circulation patterns generated by the PJ high and low patterns and the TCs can influence the rainfall patterns in Taiwan.

Fig. 11.

Composite TCCIP rainfall (mm, shaded) and 850-hPa geopotential height at day 0 for the (a) PJ high-index cases and (b) PJ low-index cases. The contour interval is 5 m and only the wind vectors exceeding the 95% confidence levels are shown.

Fig. 11.

Composite TCCIP rainfall (mm, shaded) and 850-hPa geopotential height at day 0 for the (a) PJ high-index cases and (b) PJ low-index cases. The contour interval is 5 m and only the wind vectors exceeding the 95% confidence levels are shown.

The PJI concept originated in studies by Wakabayashi and Kawamura (2004) and Choi et al. (2010). Choi et al. (2010) used a PJI {hereafter referred to as the Choi Pacific–Japan (CPJ) index (CPJI), where CPJI = [Z850(35°N, 155°E) − Z850(22.5°N, 125°E)]/2} and claimed that there were more recurving TCs in the CPJ high phase, whereas there were more straight-moving TCs in the CPJ low phase. To compare the present study with that of Choi et al. (2010), the index points are replaced by the CPJI points and the case selection procedures follow the same procedures as in the present study. From the filtered CPJI time series, 183 cases are selected with 287 TCs in the CPJ high phase and 208 cases are associated with 265 TCs in the CPJ low phase. Because of the climatological bias threshold values (the unfiltered CPJI must be higher than 30.6 m for the CPJ high cases and lower than 0 m for the CPJ low cases), fewer cases are selected in the CPJ high phase. Figure 12 reveals the composite 850-hPa filtered geopotential height and height tendency at days −4, −2, and 0 determined using the CPJIs. Also shown are the corresponding TC positions at the same time with the past 1-day tracks. At day −4 in the PJ high phase, a wave pattern (Fig. 12a) extends from the tropical WNP through the oceanic area east of Taiwan to the North Pacific area east of Japan. The area between Taiwan and Japan is occupied by an anomalous high and the TCs are clustered near the anomalous low and the negative height tendency maximal area in the southeastern rim of the anomalous high. Two days later (day −2, Fig. 12c), the anomalous high moves northeastward and decays rapidly as the enhanced anomalous low and TCs progress northwestward to the area southeast of Taiwan. The negative height tendency maximum is located northwest of the anomalous low, indicating a farther northwestward propagation of the anomalous low. The anomalous low reaches its maximal phase at day 0 (Fig. 12e); the anomalous low is centered near eastern Taiwan and the associated TCs tend to follow the recurving propagation route of the wave pattern. The composite geopotential height in the CPJ low phase (Figs. 12b,d,f), however, exhibits a similar recurving wave pattern except that the values are reversed (anomalous highs become anomalous lows, and vice versa). The area east of Taiwan is occupied by an anomalous high at day 0 in the CPJ low phase. Hence, some of the TCs follow a straight-moving route along the southern rim of the anomalous high at day 0, and the other TCs propagate toward the anomalous low center near the midlatitude area east of Japan.

Fig. 12.

Composite 850-hPa filtered geopotential height based on the CPJIs in Choi et al. (2010) and the corresponding TC positions along with the past 1-day tracks (thin lines) at days (a),(b) −4; (c),(d) −2; and (e),(f) 0 for the (left) CPJ high-index cases and (right) PJ low-index cases. The contour interval is 5 m. The shaded areas represent the geopotential height tendency (m day−1).

Fig. 12.

Composite 850-hPa filtered geopotential height based on the CPJIs in Choi et al. (2010) and the corresponding TC positions along with the past 1-day tracks (thin lines) at days (a),(b) −4; (c),(d) −2; and (e),(f) 0 for the (left) CPJ high-index cases and (right) PJ low-index cases. The contour interval is 5 m. The shaded areas represent the geopotential height tendency (m day−1).

Because the points selected for the CPJI follow the recurving propagation route of the submonthly wave pattern reported by Ko and Hsu (2006), the resultant CPJ wave pattern resembles theirs. When the CPJI is high, the anomalous low near eastern Taiwan and TCs follow the recurving path and move toward higher latitudes. However, when the CPJI is low, the anomalous high near eastern Taiwan blocks TCs from propagating northward. Thus, the TCs must follow a straight-moving route to move northwestward, or propagate directly northeastward to avoid the anomalous high.

The selection of the PJ and CPJ index points yields different results. The PJ index points in the present study are selected because one is within the maximal variance center of the geopotential height and the other is near the center of the monsoon trough. These two points are also located near the axis of the maximal variance for the 5–16-day filtered geopotential height. Thus selecting these two points for the PJ index can capture the quasi-periodic-fluctuating signals and associated wave pattern. Alternatively, Choi et al. (2010) used the index (CPJ) developed by Wakabayashi and Kawamura (2004), which was based on the regression analysis of the empirical orthogonal function and streamfunction anomalies. The CPJI points are farther east and capture signals that are unlike the fluctuating PJ pattern observed in the present study. Thus, index points based on different background selection criteria can yield opposite results.

Regarding the sensitivity of the threshold value from the spatial bias while selecting the cases, a sensitivity test is performed by switching the threshold values between the PJ and CPJ cases. The resultant number of cases decreases to 155 (with 244 TCs) for the PJ high cases and increases to 213 (with 319 TCs) for the CPJ high cases (the numbers of cases remain unchanged for the PJ and CPJ low cases). The resultant 850-hPa geopotential height patterns (not shown) change only slightly, indicating that the case-selection criteria are less sensitive to the spatial biases, but more sensitive to the locations of the index points, particularly near the area between Taiwan and Japan.

6. Conclusions

Large-scale circulation systems are analyzed using a modified PJI that is equivalent to the 850-hPa geopotential height difference between a grid point near the south of Japan and another grid point near the northern South China Sea. The PJI exhibits quasi-periodic fluctuations occurring every 5–16 days during the JASO months. Cases are selected for a composite analysis of the PJ high and low phases by using a 5–16-day bandpass of filtered data. The results are summarized in schematic diagrams, as shown in Fig. 13. In the PJ high phase, a wave pattern develops and propagates northwestward from the tropical WNP to the northern portion of the South China Sea. The cyclonic anomaly (anomalous low) near the South China Sea and another cyclonic anomaly upstream near 10°N, 150°E, of the wave pattern are generally associated with TCs that tend to propagate along a straight-moving route. While one of the cyclonic anomalies of the wave pattern approaches Luzon before the peak time, a positive geopotential height tendency maximum emerges near southern Japan and forms a stationary anomalous high pressure system to prevent this wave pattern from propagating northward. This stationary anomalous high resembles that reported by Fukutomi and Yasunari (2002) and is responsible for blocking the wave pattern and TCs, forcing them westward. Consequently, the cyclonic and anticyclonic anomalies, together with the mean fields, form a typical circulation pattern, resembling the PJ pattern in the PJ high phase. Moreover, the circulation systems of the PJ high phase, along with the TCs, bring heavy rainfall to eastern Taiwan by generating a southeasterly flow on the wind side of the Central Mountain Range.

Fig. 13.

(right) Schematic diagrams for (a) PJ high phase and (b) PJ low phase. The thick lines represent the mean state geopotential height contours (H and L stand for high and low, respectively) and the thin lines represent the filtered geopotential height contours (C is for anomalous cyclone and A is for anomalous anticyclone). Also shown is the OLR minimum areas (shaded) and TC tracks (red lines). (left) The corresponding enlarged maps of Taiwan and its surrounding areas are shown in (a) and (b) and the shading is the maximal rainfall areas.

Fig. 13.

(right) Schematic diagrams for (a) PJ high phase and (b) PJ low phase. The thick lines represent the mean state geopotential height contours (H and L stand for high and low, respectively) and the thin lines represent the filtered geopotential height contours (C is for anomalous cyclone and A is for anomalous anticyclone). Also shown is the OLR minimum areas (shaded) and TC tracks (red lines). (left) The corresponding enlarged maps of Taiwan and its surrounding areas are shown in (a) and (b) and the shading is the maximal rainfall areas.

By contrast, in the PJ low phase, the wave pattern extends from the tropical WNP northwestward through Japan and turns northeastward toward the midlatitudes. The TCs associated with the cyclonic anomaly also follow this recurving propagation route. When the cyclonic anomaly, with a negative geopotential height tendency maximum just ahead of it, propagates northward near the area northeast of Luzon, the subtropical ridge near southern Japan weakens and allows the cyclonic anomaly and TCs to move farther north and penetrate the ridge toward higher latitudes. The circulation system of the PJ low phase resembles the submonthly wave pattern stated by Ko and Hsu (2006, 2009). Regarding its influence on Taiwan’s rainfall, the PJ low pattern associated with TCs typically brings heavy rainfall to northern Taiwan.

When using the CPJI described by Choi et al. (2010), the cyclonic anomaly near eastern Taiwan opens a pathway for the recurving TCs to propagate toward the midlatitudes while the CPJI is high; in the CPJ low phase, the anticyclonic anomaly near eastern Taiwan blocks the TCs from propagating northward; therefore, the TCs must follow the straight-moving path or move directly northeastward to avoid the anticyclonic anomaly. The circulation systems associated with the PJI and CPJI phases yield different results. They each are selected on the basis of their individual background statistics and criteria. However, the results derived from the selected index points are sensitive to the locations of the index points. Therefore, the selection of points for computing large-scale circulation indices is crucial for determining the wave patterns that can be detected.

Acknowledgments

The authors thank the TCCIP Project Office (NSC 100-2621-M-492-001) for providing the precipitation data used in this study. The authors also express their gratitude to NCEP for providing the global analyses and to NOAA for the OLR data. Comments from two anonymous reviewers considerably improved the quality of the manuscript and the authors thank them for their suggestions that led to better results for the relationship between the wave patterns and TCs. This study was supported by the Ministry of Science and Technology, Taiwan (Grant MOST 103-2111-M-017-001 issued to Dr. Ken-Chung Ko).

APPENDIX

Statistical Test Procedures and Adjusting Degrees of Freedom

Circulation studies typically require statistical tests on the winds to ensure that the areas with significantly stronger winds are clearer and that those with weaker winds are eliminated (Ko and Hsu 2006, 2009). The idea originated from a (1 − α/2) × 100% confidence interval of a randomly selected sample (Walpole and Myers 1993):

 
formula

where and are the population and sample means, respectively; N − 1 is the degrees of freedom (N is the sample size); σ is the population standard deviation; and is the α/2 critical value of the standard normal distribution (α = 0.05 in the current study). When the composite winds exceed the confidence interval, they are significantly greater or less than the population mean. In other words, it would be unlikely that they are the climatological mean. However, this testing procedure is based on the assumption that all variables are randomly selected; hence, the number of independent variables is equal to the degrees of freedom (N − 1). As shown in Fig. 5, this assumption might not be valid in the current study because the PJ high and low cases are selected at the maximal or minimal points, respectively. For a time series, these cases may be autocorrelated. Figure A1 shows an example of an autocorrelation plot of the filtered zonal wind at 25°N, 130°E, during the JASO months from 1979 to 2009. A correlation maximum rmax of 0.57 emerges near 4–5 days. (i.e., there is a 57% probability that the zonal wind cases are correlated). According to Madden and Julian (1971), the degrees of freedom could be adjusted according to how uncorrelated they were. Therefore, the current study adjusts the degrees of freedom (N − 1) by 1 − rmax at each grid point to avoid overestimating the degrees of freedom.

Fig. A1.

The autocorrelation plot for the filtered 850-hPa zonal wind at 25°N, 130°E from July to October, 1979–2009.

Fig. A1.

The autocorrelation plot for the filtered 850-hPa zonal wind at 25°N, 130°E from July to October, 1979–2009.

REFERENCES

REFERENCES
Chang
,
C.-P.
,
J. M.
Chen
,
P. A.
Harr
, and
L. E.
Carr
,
1996
:
Northwestward-propagating wave patterns over the tropical western North Pacific during summer
.
Mon. Wea. Rev.
,
124
,
2245
2266
, doi:.
Choi
,
K. S.
,
C.-C.
Wu
, and
E. J.
Cha
,
2010
:
Change of tropical cyclone activity by Pacific-Japan teleconnection pattern in the western North Pacific
.
J. Geophys. Res.
,
115
,
D19114
, doi:.
Evans
,
J. L.
, and
K.
McKinley
,
1998
:
Relative timing of tropical storm lifetime maximum intensity and track recurvature
.
Meteor. Atmos. Phys.
,
65
,
241
245
, doi:.
Fukutomi
,
Y.
, and
T.
Yasunari
,
1999
:
10–25 day intraseasonal variations of convection and circulation over East Asia and western North Pacific during early summer
.
J. Meteor. Soc. Japan
,
77
,
753
769
.
Fukutomi
,
Y.
, and
T.
Yasunari
,
2002
:
Tropical–extratropical interaction associated with the 10–25-day oscillation over the western Pacific during the northern summer
.
J. Meteor. Soc. Japan
,
80
,
311
331
, doi:.
Glasbey
,
C. A.
, and
I. M.
Nevison
,
1997
: Rainfall modeling using a latent Gaussian variable. Modelling Longitudinal and Spatially Correlated Data: Methods, Applications, and Future Directions, T. G. Gregoire et al., Eds., Lecture Notes in Statistics, Vol. 122, Springer, 233–242.
Harr
,
P. A.
,
2010
: The extratropical transition of tropical cyclones: Structural characteristics, downstream impacts, and forecast challenges. Global Perspectives on Tropical Cyclones: From Science to Mitigation, J. C. L. Chan and J. D. Kepert, Eds., World Scientific, 149–174.
Hsu
,
H.-H.
,
2005
: East Asian monsoon. Intraseasonal Variability in the Atmosphere–Ocean Climate System, K.-M. Lau and D. Waliser, Eds., Springer Praxis, 63–94.
Hsu
,
H.-H.
, and
C.-H.
Weng
,
2001
:
Northwestward propagation of the intraseasonal oscillation in the western North Pacific during the boreal summer: Structure and mechanism
.
J. Climate
,
14
,
3834
3850
, doi:.
Hsu
,
H.-H.
, and
S.-M.
Lin
,
2007
:
Asymmetry of the tripole rainfall pattern during the East Asian summer
.
J. Climate
,
20
,
4443
4458
, doi:.
Kalnay
,
E.
, and Coauthors
,
1996
:
The NCEP/NCAR 40-Year Reanalysis Project
.
Bull. Amer. Meteor. Soc.
,
77
,
437
470
, doi:.
Kaylor
,
R. E.
,
1977
: Filtering and decimation of digital time series. Tech. Note BN 850, Institute of Physical Science Technology, University of Maryland, College Park, 42 pp.
Ko
,
K.-C.
, and
H.-H.
Hsu
,
2006
:
Sub-monthly circulation features associated with tropical cyclone tracks over the East Asian monsoon area during July-August season
.
J. Meteor. Soc. Japan
,
84
,
871
889
, doi:.
Ko
,
K.-C.
, and
H.-H.
Hsu
,
2009
:
ISO modulation on the submonthly wave pattern and the recurving tropical cyclones in the tropical western North Pacific
.
J. Climate
,
22
,
582
599
, doi:.
Ko
,
K.-C.
, and
Y. S.
Tzeng
,
2013
:
Characteristics of summertime circulation patterns for southern Taiwan’s monsoon rainfall from July to September
.
Terr. Atmos. Oceanic Sci.
,
24
,
107
119
, doi:.
Ko
,
K.-C.
, and
P.-S.
Chiu
,
2014
:
ISO-modulating effects on the East Asian summer monsoon circulation patterns associated with southern Taiwan’s monsoon rainfall
.
Mon. Wea. Rev.
,
142
,
3163
3177
, doi:.
Ko
,
K.-C.
,
H.-H.
Hsu
, and
C.
Chou
,
2012
:
Propagation and maintenance mechanism of the TC/submonthly wave pattern and TC feedback in the western North Pacific
.
J. Climate
,
25
,
8591
8610
, doi:.
Kosaka
,
Y.
, and
H.
Nakamura
,
2010
:
Mechanisms of meridional teleconnection observed between a summer monsoon system and a subtropical anticyclone. Part I: The Pacific–Japan pattern
.
J. Climate
,
23
,
5085
5108
, doi:.
Lau
,
K.-H.
, and
N.-C.
Lau
,
1990
:
Observed structure and propagation characteristics of tropical summertime synoptic scale disturbances
.
Mon. Wea. Rev.
,
118
,
1888
1913
, doi:.
Madden
,
R. A.
, and
P. R.
Julian
,
1971
:
Detection of a 40-50 day oscillation in the zonal wind in the tropical Pacific
.
J. Atmos. Sci.
,
28
,
702
708
, doi:.
Nitta
,
T.
,
1987
:
Convective activities in the tropical western Pacific and their impact on the northern hemisphere summer circulation
.
J. Meteor. Soc. Japan
,
65
,
373
390
.
Wakabayashi
,
S.
, and
R.
Kawamura
,
2004
:
Extraction of major teleconnection patterns possibly associated with the anomalous summer climate in Japan
.
J. Meteor. Soc. Japan
,
82
,
1577
1588
, doi:.
Walpole
,
R. E.
, and
R. H.
Myers
,
1993
: Probability and Statistics for Engineers and Scientists. 5th ed. Macmillan Publishing Company, 766 pp.