In this section, we report the occurrence numbers of winter monsoon weather pattern during the cold season (1 October to 30 April) for each year since 1901. Figure 3(a) shows the long-term variations in the occurrence number of the winter monsoon weather pattern.
To check the reliability of our analysis, we compared variations in the occurrence number of the winter monsoon weather pattern with those of the winter monsoon-type pressure pattern identified from daily synoptic weather charts (Yamakawa and Yoshino, 2002). Figure 4 represents this relationship for the period 1981/1982 to 1999/2000. In this figure, numbers of the ‘weather pattern’ are generally lower than those of ‘pressure pattern’. Yamakawa and Yoshino (2002) did not consider the strength of the winter monsoon when classifying the daily pressure pattern types. Consequently, the ‘winter monsoon-type pressure pattern’ by Yamakawa and Yoshino (2002) is thought to include a weak winter monsoon-type pressure pattern, which usually prevails over northern Japan. When such a pattern prevails over northern Japan, precipitation is usually observed only in the northern part of the Japan Sea side region. In this study, we did not detect such precipitation patterns because there are few JMA stations in the northern part of the Japan Sea side region (Figure 1). Consequently, undetected weak winter monsoon-type pressure pattern in our study is thought to be the main reason why the number of days of ‘weather’ pattern is generally smaller than those of the ‘pressure’ pattern. The purpose of this study is to detect the typical synoptic-scale winter monsoon weather pattern, which prevails over the entire area of the Japanese islands. Therefore, we do not consider the difference in the number of days showing the ‘weather pattern’ and ‘pressure pattern’ to be a serious problem in our analysis. Although these differences exist, variations in the occurrence number of the winter monsoon weather pattern and in winter monsoon-type pressure pattern agree and show a high positive correlation (r = 0.86). In the late 1960s, JMA changed the precipitation measurement instruments from cylinder-type rain gauge instruments to tipping-bucket rain gauge instruments. In order to check whether this change affected our results, we run the standard normal homogeneity test (Alexandersson, 1986), the Buishand range test (Buishand, 1982) and the Pettitt test (Pettit, 1979) over the time series of the occurrence numbers of the winter monsoon weather pattern, shown in Figure 3(a). In the test results, we could not detect any homogeneity breaks around the late 1960s. Consequently, we confirmed that the change in the type of rain gauge instruments in the late 1960s did not affect the results of our study. These facts indicate that investigating the occurrence number of the winter monsoon weather pattern is effective in understanding the long-term variations of winter synoptic weather conditions. We then examined the features of the long-term trends and secular variations in the occurrence number of the winter monsoon weather pattern based on Figure 3(a). To identify the characteristics of the long-term trends, we applied the Mann–Kendall rank statistic (Kendall, 1938) to this time series. We found a significant (p < 0.05) decreasing trend in the occurrence number of the winter monsoon weather pattern since 1901. This implies that the winter monsoon has become weaker during the past 109 years. In order to detect abrupt increases or decreases in the occurrence numbers, we applied the Lepage test (Lepage, 1971) to the time series of the occurrence number. The Lepage test is a nonparametric test used to investigate significant differences between two samples. The Lepage test has often been used to detect discontinuous climate changes (Yonetani, 1992, 1993; Inoue and Matsumoto, 2007). The results of the Lepage test (sample number n1 = n2 = 12) are shown in Figure 3(b). The Lepage statistic (HK), which indicates a degree of discontinuity, has a peak in 1986/1987, which 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 11 to 15 (figures not shown), and we confirmed that the peak in the mid-1980s unchanged. Therefore, discontinuity in the mid-1980s is thought to be statistically robust. The occurrence number of the winter monsoon weather pattern abruptly decreased after this discontinuity in the mid-1980s (Figure 3(a)). This suggests sudden weakening of the winter monsoon after the mid-1980s. Except for this abrupt change, there were no significant changes in the occurrence number of the winter monsoon weather pattern. In the early 20th century, high occurrence numbers were observed in the early 1910s, the late 1920s and the early 1940s. Low occurrence numbers were observed in the late 1940s and the early 1950s. After the early 1950s, occurrence numbers increased until the early 1980s. This implies that the winter monsoon became stronger from the early 1950s to the early 1980s. Yamakawa (1988) indicated that occurrence frequencies of winter monsoon-type pressure pattern were increasing in the period from the 1940s to the early 1980s, which roughly corresponds to the increase in the occurrence number of the winter monsoon weather pattern observed in this study.
After the abrupt decrease in the mid-1980s, the occurrence number of the winter monsoon weather pattern was about 10 days less than those before the mid-1980s. It should be noted that this abrupt decrease almost coincides with the abrupt winter climatic changes over the Northern Hemisphere in the mid-1980s (Watanabe and Nitta, 1999; Yasunaka and Hanawa, 2002). Yasunaka and Hanawa (2008) indicated that winter temperatures in Japan abruptly increased after the mid-1980s. Jhun and Lee (2004) and Wang et al. (2009) indicated that East Asian winter monsoon weakened after the mid-1980s, corresponding to the abrupt decrease in the occurrence number of the winter monsoon weather pattern observed in this study. In the late 1990s, the occurrence numbers increased again slightly and were about 8 days higher than those in the early 1990s.