Interdecadal climate changes occurring during the latter half of August (LA) in central Japan are described, and the associated changes in rainfall, typhoon tracks, and circulation patterns over East Asia and the western North Pacific (WNP) regions are investigated. Since 1984, rates of sunshine and temperature have increased, while rainfall has decreased significantly during LA in central Japan. In contrast, rainfall over the Southwest Islands, northern part of Taiwan, and in the south of the middle and lower reaches of the Yangtze River in China has increased. Changes in the positions and tracks of typhoons are responsible for these changes. Prior to 1983, many typhoons approached the central-western part of Japan during LA, while after 1984, most typhoons were deflected away from Japan and moved northwestward to Taiwan and China. The North Pacific subtropical high during LA extended more westward after 1984, which affected the interdecadal changes in sunshine, temperature, rainfall, and typhoon tracks, not only in central Japan, but also over East Asia and the WNP regions.
Recently, many kinds of decadal-to-interdecadal-scale climate changes have been detected in various regions of the world. One example of these changes is a sudden interdecadal climate shift that occurred over the North Pacific in the late 1970s (e.g., Nitta and Yamada 1989; Trenberth 1990). This kind of sudden shift is called a climate regime shift (Minobe 1997), which might be related to the Pacific decadal oscillation (PDO; e.g., Mantua et al. 1997). Although some hypotheses have been proposed, its mechanism remains controversial. Associated with this shift, interdecadal changes occurring in the late 1970s in the East Asian summer monsoon season have also been pointed out by many researchers (e.g., Nitta and Hu 1996; Gong and Ho 2002).
A majority of the previously mentioned studies on interdecadal changes in the East Asian summer monsoon season are based on June–August (JJA) mean data. However, studies of climate changes over East Asia using these 3-month mean data seem to be too simple to clarify the mechanism of these changes in detail. This is due to the fact that there is a climatological intraseasonal oscillation in this region (e.g., Wang and Xu 1997; LinHo and Wang 2002; Chen et al. 2004). Actually, there is an abrupt wet/dry cycle over East Asia, due to a northward stepwise shift of the baiu/mei-yu frontal zone from May to July (e.g., Yoshino 1965) and a southward stepwise shift of the Akisame (autumnal rain) frontal zone from August to October (e.g., Matsumoto 1988). In addition, typhoons often attack this area in a specific period, which also contributes to the climatological seasonal variation of rainfall. Therefore, further analyses using subseasonal to submonthly time-scale datasets are needed in order to understand the precise mechanisms underlying interdecadal changes. Moreover, Matsumoto (1992) revealed that some abrupt seasonal transitions of circulation systems and convective activities occur almost simultaneously over broad regions of the whole monsoon area of Asia and northern Australia. Therefore, investigating the interdecadal changes of seasonal marches is also important for improving understanding of the Asian–Australian monsoon system.
To cite the studies from this viewpoint, Yamakawa (1988) studied seasonal and secular variations in the prevailing pressure patterns over East Asia for the period 1941–85 and found a recent delay in the end of the baiu season (early summer rainy season in Japan) and a shortening of the midsummer dry season occurring since the 1970s. Sato and Takahashi (2001) took note of sunshine duration in midsummer (early August) in Japan, and they showed that the sunshine duration in early August had a decreasing trend for the period 1959–95, especially over the Sea of Japan side of central Japan. They also suggested that a delay of the northward movement of the baiu frontal zone was likely to be associated with strengthening of the polar air mass (i.e., the Okhotsk high) around northern Japan. Inoue and Matsumoto (2003) investigated the seasonal and secular changes of sunshine duration in Japan for the period 1951–2000. They found that the end of the baiu season has been delayed since 1980, which is consistent with the previous studies described above. During the latter half of August (i.e., 16–31 August) (LA), on the other hand, a sunshine increase was observed in central-western Japan after the early to mid-1980s. In other words, the long-term trends of sunshine duration in this region are different, even within the same month (August). Since this region has a relatively rapid seasonal cycle, and is sandwiched between midlatitude and subtropical circulation regimes in summer, this result implies an existence of interdecadal modulation of seasonal cycle there.
In the present study, we take note of the changes occurring during LA and further investigate the interdecadal climate changes observed in the 1980s during LA in central Japan. Then we examine the interdecadal changes in rainfall patterns over East Asia during LA.
Typhoons and tropical cyclones bring much rain and extraordinarily strong wind and sometimes cause severe disasters over East Asia. Ho et al. (2004) and L. Wu et al. (2005) documented interdecadal variations in typhoon activity in the western North Pacific (WNP) region and their influence over East Asia during the typhoon season (June–September or June–October). Typhoons themselves are also important phenomena for seasonal cycles over East Asia, and August is the month in which tropical cyclone and typhoon activity over the WNP is the strongest of the year (e.g., Chan 2005). Yoshino and Kai (1977) and Yamakawa (1988) noticed that the “typhoon-type” pressure pattern in the vicinity of Japan appeared most frequently during mid- to late August. Therefore, it is possible that statistics of typhoon activity during LA have also changed since the early 1980s. We, thus, also analyze the interdecadal changes of typhoon tracks and circulation patterns over East Asia and the WNP regions.
Section 2 describes data utilized in the present study. In section 3, we analyze observational station data from Japan and clarify the changes observed during LA. Section 4 examines the changes in rainfall patterns over East Asia, and section 5 describes the changes in typhoon tracks over the WNP region. Comparisons of atmospheric circulation in the midtroposphere are attempted in section 6. Major conclusions and discussions are described in the last section.
Daily station data for sunshine duration and rainfall in Japan were obtained from the surface daily product (SDP) data CD-ROMs and the annual report CD-ROMs, published or released by the Japan Meteorological Agency (JMA). We used 115 observational stations in Japan, where continuous observation data are available for the period 1961–2000 without any location movement (Fig. 1).
During the late 1980s, the JMA replaced the Jordan sunshine recorders (which use a closed can with a pinhole on one side to let sunlight onto light-sensitive paper) with rotating mirror sunshine recorders (which use a specially designed rotating mirror that reflects the sun’s direct component onto a spectrally flat pyroelectric sensor). Referring to Katsuyama (1987), the observed values before the replacement were calibrated as
where SJ is the observed value of daily sunshine duration obtained by the Jordan sunshine recorder before the replacement, and SR is a calibrated value corresponding to the rotating mirror sunshine recorder. This relationship was determined based on the results of simultaneous observational comparisons between the two sunshine recorders.
We calculated rates of sunshine (values of sunshine duration divided by possible duration of sunshine) and used that instead of sunshine duration because of the latitudinal difference of possible duration of sunshine (PDS). PDS (h day−1) is calculated as
where θ(°) is the latitude of the station, and δ(°) is the solar declination parameter determined by the date of the day.
We also used daily rainfall data from 123 stations in China, compiled by the China Meteorological Administration, 18 stations in Taiwan, compiled by the Central Weather Bureau, and 15 stations in South Korea, compiled by the Korea Meteorological Administration (Fig. 1).
For typhoon statistics, we utilized “Regional Specialized Meteorological Center (RSMC) best-track data” over the WNP region that was obtained from the Web site of the RSMC Tokyo-Typhoon Center (http://www.jma.go.jp/jma/jma-eng/jma-center/rsmc-hp-pub-eg/RSMC_HP.htm). For the present study, “typhoons” were defined as tropical cyclones of which maximum sustained wind speeds are greater than 17 m s−1, and the 4-times daily data (0300, 0900, 1500, and 2100 Japan Standard Time) were used.
As for upper-air wind data, the 40-yr European Centre for Medium-Range Weather Forecasts (ECMWF) Re-Analysis (ERA-40) daily data (Uppala et al. 2005) were used. These data are available from the data server of the ECMWF Web site (http://data.ecmwf.int/data/d/era40_daily/), with grid intervals of 2.5° (latitude) × 2.5° (longitude). The ERA-40 seems to be better than another long-term reanalysis dataset, the National Centers for Environmental Prediction–National Center for Atmospheric Research (NCEP–NCAR) reanalysis, in investigations of interdecadal climate variations over East Asia (Inoue and Matsumoto 2004; R. Wu et al. 2005).
3. Interdecadal changes detected from observational station data in Japan
As described in section 1, sunshine duration during LA has increased over central-western Japan since the early to mid-1980s, which is the opposite tendency (sunshine decrease since the 1980s) to that in late July and the former half of August over there (Inoue and Matsumoto 2003). To highlight the change during LA, we sought a “core region” of this change, where the changes were most clearly recognized. After examining interdecadal variations of sunshine during LA at each station, 24 stations located on the Pacific Ocean side of central Japan were selected. These stations are plotted as open circles in Fig. 1. Figure 2a shows the interannual variation in rates of sunshine during LA, averaged over the 24 stations. As already indicated by Inoue and Matsumoto (2003), the rates of sunshine abruptly increased in the early to mid-1980s. This increase appears to be a discontinuous change in the early 1980s, rather than a linear trend. To determine the discontinuous year of this shift, the Lepage test (Lepage 1971) was conducted. The Lepage test is a nonparametric test that investigates significant differences between two samples. Yonetani (1992, 1993) showed that this statistical test is very useful for detecting discontinuous climate changes. The result of the Lepage test (sample numbers n1 = n2 = 12) is shown in Fig. 2b. The Lepage statistic, HK, which indicates a degree of discontinuity, has a peak in 1983/84 that is significant at a 95% confidence level. We also checked the results of the Lepage test when the sample numbers (n1 and n2) changed from 10 to 18, and we confirmed that the peak in 1983/84 was unchanged (figures not shown). Thus, we divided the analyzed years into two periods: 1961–83 and 1984–2000. Hereafter, we mainly focus on the difference between the two periods during LA.
Figure 3 compares seasonal variations of rates of sunshine during the warm season, averaged in 1961–83 and 1984–2000 over the 24 stations. Climatologically, after a little-sunshine season during mid-June to mid-July (the baiu rainy season in central-western Japan), a dry and very sunny season (midsummer) begins in late July (Maejima 1967; Inoue and Matsumoto 2003). In 1961–83, the very sunny season had its highest peak in early August, but it did not last long. Sunshine rather abruptly decreased after LA and decreased again toward another little-sunshine season during mid-September to early October (the Akisame rainy season or autumnal rains in central-western Japan). The most obvious difference between 1961–83 and 1984–2000 was a sunshine increase during LA. The peak of sunshine shifts to LA, and midsummer (very sunny season) has recently been prolonged to as late as early September.
Figure 4 shows the difference in rates of sunshine (1984–2000 minus 1961–83) in Japan during LA. Rates of sunshine have increased at a large number of stations in central Japan, and significantly increased stations at a 95% confidence level according to the Student’s t test are distributed over the Pacific Ocean side of central Japan. In contrast sunshine has decreased at some stations in Hokkaido, southern and western Kyushu, and the Southwest Islands. This geographical pattern is similar to the opposite situation of the recent sunshine changes in early August in Japan (Sato and Takahashi 2001). This distribution implies an intensification of the influence of the North Pacific Subtropical High (NPSH) over this region during LA, which will be discussed in section 6.
The interdecadal changes during LA described above were observed not only in sunshine but also in temperature and rainfall. Figure 5a shows seasonal variations of pentad mean temperature as in Fig. 3. Temperature has increased through the warm season, in particular after LA. We checked the Student’s t test for the eight 15-day (16-day) periods from June to September (1–15 June, 16–30 June, . . . , and 16–30 September). Although temperature has increased in all periods, those before 1–15 August are not significant at a 95% level. The periods between 16–31 August and 16–30 September pass a 95% confidence level, and only LA passes a 99% confidence level. Figure 5b compares the seasonal variations in rainfall averaged for the same stations. During LA, the rainfall decreased obviously after 1984. The rainfall in 1984–2000 during LA (4.2 mm day−1) was approximately half of that in 1961–83 (8.0 mm day−1), and this decrease is significant at a 99% confidence level according to the Student’s t test. Both the temperature increase and the rainfall decrease occurred abruptly in 1983/84 (Figs. 5c,d), and the discontinuity is confirmed by the Lepage test at a 95% confidence level (figures not shown for both time series).
4. Interdecadal changes of rainfall over East Asia during LA
Figure 6 shows a spatial pattern of the 1983/84 change in rainfall during LA over East Asia. Stations where rainfall decreased are concentrated on the Pacific Ocean side of central Japan. In contrast, rainfall over the southern portion of the Southwest Islands, northern part of Taiwan, and southern part of China (the middle and lower reaches of the Yangtze River Valley and to the south) increased. In particular, stations with significant rainfall increases were found to be located in 1) the southwestern-most part of Japan (the Sakishima Islands) and northern Taiwan (24°–26°N, 120°–126°E; denoted as SI–NT), and 2) the middle and lower reaches of the Yangtze River Valley (27°–32°N, 110°–122.5°E; denoted as YRV). Rainfall has also increased in Korea, though the increase is not significant.
Figure 7 compares the seasonal variations in rainfall, averaged over the above two regions between 1961–83 and 1984–2000. In 1961–83, over the SI–NT region (Fig. 7a), rainfall during LA showed a temporal minimum, in particular at the end of this time period. In 1984–2000, on the other hand, LA became one of the highest rainfall periods in the warm season. The increase during LA is significant at a 99% confidence level, according to the Student’s t test. Over the YRV region (Fig. 7b), there is a primary peak during June to early July (the mei-yu season). After the end of the mei-yu season, a relatively dry season starts in mid-July. Another rainfall peak, although not very striking, exists in August over this region. This peak seems to shift from mid- to late August. Therefore, rainfall at several stations within the YRV region has increased significantly during LA. The increase during LA is significant at a 90% confidence level, according to the Student’s t test.
5. Changes of typhoon tracks over the WNP region during LA
In the previous section, we showed the interdecadal change of rainfall over East Asia. Since typhoons bring much rain in the warm season over East Asia, the 1983/84 changes in typhoon activity over the WNP region during LA are examined in this section. Figures 8a and 8b show genesis locations and tracks of typhoons over the WNP region during LA in 1961–83 and 1984–2000, respectively. In 1961–83 (Fig. 8a), typhoon tracks were concentrated in the vicinity of the Southwest Islands, and the Pacific Ocean side of central-western Japan. A number of typhoons occurred over the tropical WNP, moved northward along approximately 130°E, and then curved east-northeastward along the southern coast of central-western Japan. In 1984–2000 (Fig. 8b), on the other hand, there were many typhoons that moved northwestward in the vicinity of Taiwan and over the northern part of the South China Sea (SCS). Typhoons in the vicinity of central-western Japan have obviously decreased.
To quantitatively examine the spatial variations in typhoon positions, the typhoon passage frequency within each 5° × 5° grid box from the original 6-hourly datasets was calculated. Since probability density functions of the typhoon passage frequency within each grid are quite different from the normal distribution, significance levels were estimated from the Monte Carlo simulations, rather than the Student’s t test, by using 10 000 randomized time series. The details are as follows: First 9999 randomly sorted 40-yr time series of the typhoon passage frequency values are prepared, and then averages of the former 23-yr values minus those of the latter 17-yr values are calculated. From the rank of the original difference value among 10 000 samples, the significance level is estimated. Figure 8c shows the difference in the typhoon passage frequency (1984–2000 minus 1961–83) during LA. The typhoon passage frequency has significantly decreased over central-western Japan and its neighboring ocean to the south (25°–35°N, 125°–140°E; denoted as CWJ). On the other hand, the frequency has increased over the northern SCS and in the vicinity of Taiwan and the Philippines (15°–25°N, 105°–130°E; denoted as SCS–TP).
Figure 9 shows the seasonal variations in typhoon passage frequency in 1961–83 and 1984–2000 over the two box areas. In the CWJ area (Fig. 9a), a peak during mid- to late August in 1961–83 is clear. In 1984–2000, on the other hand, the peak disappears, and the typhoon passage frequency during LA decreases by half. This decrease during LA is significant at the 95% level estimated from the Monte Carlo simulation. In recent years, a peak before LA (late July to early August) has become obvious. Over the SCS–TP area (Fig. 9b), the most active typhoon season during the period of 1961–83 existed during September. However, the beginning of this season has become earlier in recent years. In 1984–2000, the typhoon passage frequency during LA increased twice as much as that of 1961–83 (significant at a 99% confidence level estimated from the Monte Carlo simulation), in sharp contrast to the situations in the CWJ area.
These changes in the typhoon passage frequency are closely related to the interdecadal changes in rainfall in central Japan. Figure 10 shows seasonal variations in the accumulated rainfall, averaged at the 24 stations (same as those of Figs. 2, 3 and 5) in central Japan, when typhoons existed within the CWJ area. A drastic decrease in the typhoon-influenced rainfall is recognized during LA. This decrease is significant at a 99% confidence level. This result indicates that the interdecadal decrease in rainfall over central Japan during this season is strongly influenced by the interdecadal change in typhoon activity over the WNP.
6. Changes of upper-air circulation
Typhoon tracks are essentially controlled by large-scale atmospheric circulation patterns in the midtroposphere. Therefore, the changes in typhoon tracks shown in the previous section may be related to large-scale circulation changes over the WNP. Steering flow that influences tropical cyclone motion has investigated by many researchers (e.g., George and Gray 1976; Chan and Gray 1982). Holland (1984) suggested that the pressure-weighted layer mean flow from 850 to 300 hPa appeared to correlate reasonably well with tropical cyclone movement. Here, we calculated and compared this pressure-weighted flow in 1961–83 and 1984–2000 during LA (Fig. 11). In 1961–83, a westward extension of the NPSH was not strong, and southerly winds prevailed along the western periphery of the NPSH to the south of Japan. In 1984–2000, in contrast, the westward extension of the NPSH was stronger, and the ridge line near Japan (130°–140°E) was located slightly more northward than it had been prior to the early 1980s (i.e., south of 30°N in 1961–83; north of 30°N in 1984–2000). The difference in wind vector (1984–2000 minus 1961–83; figure not shown) indicates that the wind change in the south of Japan, which indicates stronger anticyclonic circulation in the vicinity of central Japan, is significant at a 95% confidence level.
To compare the seasonal retreat of the western peripheries of the NPSH, time–longitude cross sections of the meridional wind component of the pressure-weighted (300–850 hPa) steering flow along 30°N from July to September in 1961–83 and 1984–2000 were compared in Fig. 12. In 1961–83, the southerly component was the strongest over eastern China (the middle and lower reaches of the Yangtze River Valley) and/or the western part of the East China Sea (115°–125°E) until early August. Then, a sudden eastward shift of the southerly wind maximum occurred in mid-August and stagnated in the south of Japan (135°–140°E) until September. In 1984–2000, on the other hand, the first retreat from eastern China and/or the western part of the East China Sea occurred as early as late July to early August, but the maximum of the southerly wind component (western periphery of the NPSH) remained in the vicinity of Kyushu (around 130°E) until the end of August. The southerly wind component to the south of Japan (135°–140°E) was not strong, even in late August, and the second eastward movement occurred in early to mid-September.
These results imply that the changes in typhoon tracks (and possibly the changes in rainfall, sunshine, and temperature over East Asia) during LA are attributable to the recent stagnant tendency of the seasonal eastward retreat of the NPSH. During LA prior to 1983, the western periphery of the NPSH retreated eastward earlier. Thus, the seasonal changes in sunshine decrease and rainfall increase were brought about directly. In addition, it gave favorable conditions for many typhoons to approach central-western Japan. Therefore, rainfall supplied from the typhoons was also abundant over this region. After 1984, in contrast, the western periphery of the NPSH was stagnant over the East China Sea and in the vicinity of Kyushu until the end of August. The conditions of much sunshine and little rainfall thus continued even into late August. As most typhoons are deflected northwestward from the WNP toward Taiwan and the southern part of China, the midsummer dry season in central-western Japan now lasts until early September, and the rainfall in the SI–NT and YRV regions has increased during LA.
7. Conclusions and discussion
We clarified the interdecadal climate changes occurring during the latter half of August (LA) in central Japan, and investigated their associated changes in rainfall, typhoon tracks, and circulation patterns over East Asia and the western North Pacific (WNP) regions. Since 1984, rates of sunshine and temperature have increased, while rainfall has decreased significantly in central Japan during LA. At the same time, rainfall over the Southwest Islands, northern part of Taiwan, and the middle and lower reaches of the Yangtze River Valley has increased. These changes are attributable to the clear changes in the positions and tracks of typhoon in the WNP. Prior to 1983, many typhoons approached the central-western part of Japan during LA. After 1984, on the other hand, most typhoons were deflected away from Japan and moved northwestward to Taiwan and China. Changes in the circulation patterns in the midtroposphere over the WNP region indicate that, over the oceanic area in the south of Japan, the North Pacific subtropical high during LA has extended more westward since 1984. This is related to the delay of the eastward retreat of the subtropical high during LA in recent years. These changes in the circulation patterns affect the interdecadal changes in sunshine, temperature, rainfall, and typhoon tracks, not only in central Japan, but also over East Asia and the WNP regions.
As mentioned in section 3, the changes in LA discovered in the present study are opposite to those that occur in late July and early August, a fact pointed out by Sato and Takahashi (2001) and Inoue and Matsumoto (2003). Sato and Takahashi (2001) suggested that long-term changes, such as sunshine decrease in early August, were mainly due to the recent slower northward movement of the baiu frontal zone in the vicinity of Japan as well as the strengthening of the polar air mass over northern Japan (i.e., the Okhotsk high). However, the influences of the baiu (mei-yu) frontal zone and the Okhotsk high on Japan are no longer strong during LA, because the thermal contrast in the vicinity of the Sea of Okhotsk (i.e., warmness of the Siberian landmass and coolness of the North Pacific Ocean), which is connected to the strength of the Okhotsk high (Nakamura and Fukamachi 2004; Tachibana et al. 2004), disappears until LA (figure not shown). Therefore, some other mechanisms, which make the opposite impact to those in late July and early August, may bring the interdecadal changes during LA.
The changes of typhoon activity and the NPSH imply that the interdecadal changes observed in central Japan seem to be a reflection of interdecadal changes of seasonal marches over the East Asian and WNP monsoon region, rather than local phenomena. The next step is to clarify the reason for the changes in the NPSH. From Fig. 9b, typhoon passage frequency has changed over the tropical WNP region on an intraseasonal time scale. Since the intraseasonal variation of monsoon activity over the WNP is seasonally phase locked (Nakazawa 1992), it is implied that the phase-locked tropical intraseasonal variation has changed interdecadally, and may be related to the recent stronger NPSH in the vicinity of Japan (Fig. 11) through the Rossby wave response, known as the Pacific–Japan (PJ) pattern (Nitta 1987). Meanwhile, LA is a transitional season from a subtropical regime to a midlatitude regime over central and western Japan. It is possible that the relationship between this steady seasonal transition in the midlatitude and the phase-locked intraseasonal variation over the tropical WNP region has changed on the interdecadal time scale. However, interdecadal climate jumps in the early 1980s are not well known over other regions of the world. Therefore, whether a relationship with other mechanisms of climate changes, such as PDO or global warming, exists or not is an important question to be solved. To clarify the whole mechanism, further investigations are needed in the future.
We appreciate valuable comments and suggestions from the two anonymous reviewers. We are also thankful to Mr. Hideaki Shoji of the University of Tokyo for our use of the typhoon data file, which he obtained and processed, Dr. Rena Nagata of Tokyo Metropolitan University for access to the daily rainfall data in Taiwan and Korea, and Dr. Kiyotoshi Takahashi of the Meteorological Research Institute, Dr. Shinji Nakagawa of the JMA, and Dr. Yasushi Agata of the University of Tokyo for helping with the provision of the daily rainfall data in China.
The Grid Analysis and Display System (GrADS), distributed from the Center for Ocean–Land–Atmosphere Studies (COLA), and the Generic Mapping Tools (GMT), distributed from the School of Ocean and Earth Science and Technology (SOEST) at the University of Hawaii, were utilized for the drawing part of the figures.
Part of this study is financially supported by the Grant-in-Aid for JSPS Fellows from the Japanese Ministry of Education, Culture, Sports, Science and Technology (16-11316) and by the Core Research for Evolutional Science and Technology (CREST) of the Japan Science and Technology Agency (JST).
* Current affiliation: Current affiliation: Doctoral Program in Sustainable Environmental Studies, Graduate School of Life and Environmental Sciences, University of Tsukuba, Tsukuba, Japan
+ Current affiliation: Department of Geography, Graduate Schools of Urban Environmental Sciences, Tokyo Metropolitan University, Hachioji, Japan
Corresponding author address: Tomoshige Inoue, Doctoral Program in Sustainable Environmental Studies, Graduate School of Life and Environmental Sciences, University of Tsukuba, 1-1-1, Tennodai, Tsukuba, Ibaraki 305-8572, Japan. Email: firstname.lastname@example.org