The modulation of tropical cyclone (TC) activity by the western North Pacific (WNP) monsoon break is investigated by analyzing the subseasonal evolution of TCs and corresponding circulations, based on 65 years of data from 1950 to 2014. The monsoon break has been identified as occurring over the WNP in early August. The present results show that TC occurrence decreases (increases) remarkably to the east of the Mariana Islands (southeast of Japan) during the monsoon break, which is closely related to local anomalous midtropospheric downward (upward) motion and lower-tropospheric anticyclonic (cyclonic) circulation, in comparison with the previous and subsequent convective periods in late July and mid-August. These changes of TC activity and the corresponding circulation during the monsoon break are more significant in typical monsoon break years when the monsoon break phenomenon is predominant. The reverse changes of TC activity to the east of the Mariana Islands and to the southeast of Japan during the monsoon break are closely associated with the out-of-phase subseasonal evolutions over these two regions from late July to mid-August, which are both contributed to greatly by 10–25-day oscillations. Finally, the roles of midlatitude and tropical disturbances on 10–25-day oscillations are also discussed.
The western North Pacific (WNP) is the most active region for tropical cyclone (TC) occurrence. Associated with extreme precipitation and strong wind, TCs exert substantial impacts on human life and socioeconomics, especially in the countries of the WNP region, or along the coast, where TCs pass by frequently (Pielke et al. 2008; Zhang et al. 2009).
The activity of TCs over the WNP, such as their genesis position and occurrence frequency, shows obvious seasonal evolution (Chia and Ropelewski 2002; Yang 2005; Molinari and Vollaro 2013). The genesis position of TCs in winter locates at its annual southernmost position, generally over the tropical area, then gradually extends poleward toward the subtropical area in spring and early summer before reaching its annual northernmost position in late summer. Correspondingly, the occurrence frequency of TCs remains small in winter, then increases in the summer season, and finally reaches its annual peak from July to October.
The peak season of TC activity is consistent with the period of the WNP summer monsoon (WNPSM). In late July, the monsoon is established over the subtropical WNP, accompanied by an abrupt enhancement of convection over the region to the east of the Mariana Islands, a phenomenon called the “convection jump” by Ueda and Yasunari (1996). This enhanced convective activity is coincident with a strong cyclonic circulation, which is related to the northward extension of the monsoon trough. The key region of the convection jump defined by Ueda and Yasunari (1996) is 15°–25°N, 150°–160°E, which has been modulated as 10°–20°N, 140°–160°E by Xu and Lu (2015). This monsoon onset in late July signifies the arrival of the annual rainfall peak over the WNP. The WNP summer monsoon persists and begins to withdraw southward until late September (Wang and LinHo 2002). On the other hand, the convection jump in late July exhibits an interannual variation. In some years, the phenomenon of convection jump is very significant, but is relatively vague in some other years (Ueda and Yasunari 1996).
The enhancement of TC activity during the period of WNPSM can be explained by the monsoon trough. Monsoon troughs over the WNP are considered as a crucial factor affecting TC activity. Monsoon troughs are associated with strong lower-tropospheric positive vorticity and convergence, weak vertical wind shear, and high midtropospheric humidity, all favoring TC formation and development (Gray 1968; Briegel and Frank 1997; Chen et al. 2006; Wu et al. 2012; Cao et al. 2014; Zong and Wu 2015). Therefore, more than 70% of TCs are formed within the monsoon trough (Ritchie and Holland 1999; Chen et al. 2004; Molinari and Vollaro 2013). On the other hand, monsoon troughs exhibit a quick northward extension in late July and then prevail over the subtropical WNP, corresponding well to the evolution of the WNPSM (Ueda et al. 1995; Wu and Wang 2001; Wu 2002; LinHo and Wang 2002).
Therefore, over the WNP, the seasonal change of TC activities concurs well with the monsoon evolution. Recently, Xu and Lu (2015) identified a monsoon break phenomenon over the WNP, which occurs climatologically in early August, subsequent to the monsoon onset in late July to the east of the Mariana Islands, and is followed by a second rainfall enhancement in mid-August. The monsoon break is accompanied by drastically reduced rainfall, suppressed convection, and a weakened monsoon trough. Sometimes (around one in three years), the monsoon break is particularly prominent, during which the precipitation is even less than that before the monsoon onset. Therefore, it is speculated that the TC activities over the WNP may be modulated by the remarkable changes in convection and circulation related to the monsoon break.
Furthermore, 10–25-day oscillations, which become remarkably strengthened over the WNP during late July to mid-August (Ko and Hsu 2006; Chen and Sui 2010), are known to play a crucial role in forming the monsoon break (Xu and Lu 2015). In addition, previous studies have revealed that the activities of TCs over the WNP are very different between the convective and suppressed phases of quasi-biweekly oscillations (Ko and Hsu 2006; Wang et al. 2009; Gao and Li 2011; Jin et al. 2012; Li and Zhou 2013; Zhao et al. 2015). Therefore, the role of quasi-biweekly oscillations in modulating TC activities during the monsoon break is another issue to be examined in this study.
The objectives of this study are to examine the influences of the monsoon break on the activities of TCs over the WNP and discuss the role of 10–25-day oscillations. The organization of the paper is as follows: The data and methods are described in section 2. The subseasonal variation of TC activity around the monsoon break over the 65-yr study period (1950–2014) is presented in section 3. This analysis period is about twice the length of that in Xu and Lu (2015). In section 4, the modulation of TCs by the monsoon break, the role of 10–25-day oscillations, and the possible mechanisms involved are investigated, based on the more prominent monsoon break cases. The final section summarizes and concludes the study.
2. Data and methodology
The daily data for 850-hPa wind, 200-hPa wind, and 500-hPa vertical velocity were from the National Centers for Environmental Prediction–National Center for Atmospheric Research (NCEP–NCAR) Reanalysis-1 dataset (Kalnay et al. 1996), extending from 1950 to 2014. This period is longer than the study period of 1979–2012 in Xu and Lu (2015), enabling us to increase the sample size of TC cases. The daily outgoing longwave radiation (OLR) data were from the National Oceanic and Atmospheric Administration (NOAA) satellite (http://www.esrl.noaa.gov/psd): the interpolated OLR (Liebmann and Smith 1996) for the period 1979–2013 and the uninterpolated OLR for 2014. The daily circulation and OLR data have a horizontal resolution of 2.5° × 2.5°, and values on 29 February in leap years were removed. Lanczos bandpass filtering was performed to obtain the components of the 10–25-day oscillations.
The TC data over the WNP were taken from the Joint Typhoon Warning Center (http://www.usno.navy.mil/NOOC/nmfc-ph/RSS/jtwc/best_tracks/wpindex.php) at 6-h intervals (0000, 0600, 1200, and 1800 UTC) from 1950 to 2014. The TC data on 29 February in leap years were also removed. The daily TC activity on each grid, with resolution of 2.5° × 2.5°, is obtained by counting the number of TCs occurring in the 10° × 10° latitude–longitude box that centers on this grid. A TC is counted if it appears at any of the four times in one day but is counted only once even if it appears at several times on that day. A high value of TC activity indicates more TCs existing around that grid. Thus, the TC activity reflects the density distribution of TC occurrence. In this study, all TCs from tropical depressions to typhoons are included.
Based on the feature of convection evolution around the monsoon break, three stages are defined, in the same way as in Xu and Lu (2015), to facilitate the comparison of circulations and TC activity between the monsoon break and the previous and subsequent periods. Days 215–220 (3–8 August) are selected as stage 2 to represent the WNP monsoon break when convection becomes remarkably suppressed over (10°–20°N, 140°–160°E), the so-called key region in Xu and Lu (2015). Stage 1 (days 205–214, or 24 July–2 August) and stage 3 (days 221–230, or 9–18 August) cover 10 days prior to, and subsequent to, stage 2, respectively, representing the late-July convection jump and the second enhancement of convection in mid-August. Considering every stage has different lengths, the daily averaged TC activity over each stage is derived to facilitate the comparison. The Student’s t test is utilized to examine whether the TC activity during the monsoon break is statistically different from those during stages 1 and 3.
3. Modulation of TC activity by the climatological monsoon break during 1950–2014
Figure 1 shows the differences of 500-hPa vertical velocity and 850-hPa wind between the monsoon break (stage 2) and the adjoining periods (stages 1 and 3) during 1950–2014, which is about twice the length of the study period in Xu and Lu (2015). The differences are shown to highlight the changes of convection and circulations during the monsoon break. Because of a lack of OLR data prior to 1979, we use the 500-hPa vertical velocity instead of OLR to estimate convection. A positive anomaly of 500-hPa vertical velocity, which represents anomalous downward motion, appears to the east of the Mariana Islands [i.e., the monsoon break region in Xu and Lu (2015)] and is accompanied by an anticyclonic anomaly located slightly to its west. The anomalous downward motion and the anticyclonic anomaly denote, respectively, the suppression of the convection and the weakening of the monsoon trough, which indicates that the climatological monsoon break phenomenon is also significant in the extended period.
It is also noted that there is a negative anomaly center of 500-hPa vertical velocity, although statistically insignificant, to the north of the monsoon break region. The band of anomalous upward motion overlays the region of southwesterly and southerly winds associated with a cyclonic anomaly centering to the southeast of Japan. These results indicate the reverse changes of convection and circulation to the southeast of Japan during the monsoon break, even though the changes are not as remarkable as those over the monsoon break region. We also repeated the above analysis using the data during 1950–78 and 1979–2012, separately (figures not shown). The latter period is the same as in Xu and Lu (2015). During these two periods, there is also anomalous downward (upward) motion to the east of the Mariana Islands (to the southeast of Japan), which is generally similar to that found during 1950–2014.
Figure 2 shows the spatial distributions of the TC activity averaged over every stage during 1950–2014, and the differences between stage 2 and the adjoining stages. As described in section 2, for a particular day, the value of TC activity on each grid represents the number of TCs occurring in the 10° × 10° latitude–longitude box centering on this grid. Considering every stage has different lengths, the daily averaged TC activity over each stage is derived. In stages 1 and 3, when convection is enhanced over the monsoon break region, the distribution of TC activity appears like an ellipse, with mainly zonal elongation from the South China Sea to about 150°E along 10°–35°N. In comparison, in stage 2, the values of TC activity decrease obviously around the monsoon break region, accompanied by a remarkable suppression of convection, while they increase remarkably to the southeast of Japan, leading to a southwest–northeast-tilted distribution of TC activity. These changes are also reflected by the notably westward retreat (northeastward extension) of the contour of 6 to the east of the Mariana Islands (to the southeast of Japan) in comparison with stages 1 and 3. The changes of TC activity during the monsoon break can be seen more clearly from the difference plotted in Fig. 2d. There is a dipole structure along 140°–160°E, with a negative and significant anomaly around the monsoon break region and a positive anomaly to the southeast of Japan.
We have also used the “effective degrees of freedom” when performing the Student’s t test to examine the statistical significance of changes in circulations and TCs during the monsoon break and found that the results are almost the same as the present ones (figures not shown), because of the high independence of samples in this study.
To quantitatively estimate the changes of TC activity around the monsoon break, we define the north region with positive difference as “region N” (27.5°–37.5°N, 140°–155°E) and the south region with negative difference as “region S” (12.5°–25°N, 140°–160°E), and further investigate the evolution of TCs over these two regions. Region S locates slightly northward of the monsoon break region (10°–20°N, 140°–160°E), defined by Xu and Lu (2015).
Figure 3 shows the daily number of TCs appearing in region S and region N from mid-July to late August. The number of TCs over region S is less than 8 in mid-July and increases remarkably to above 16 in late July when the monsoon onset occurs but rapidly turns to decrease at the end of July and reaches a minimum during early August when the monsoon break occurs (Fig. 3a). There are only 5 TCs on day 219, and this low value is close to that prior to the monsoon onset in mid-July. After the monsoon break, the number of TCs increases again and exceeds 16 during mid-August. The mean values of TC number in the three stages are, respectively, 14.8, 8.7, and 15.4, and the reduction during the monsoon break is statistically significant at the 99% confidence level.
In contrast, the number of TCs over region N remains low during mid and late July, then rapidly increases to above 10 during early August when the monsoon break occurs, which is about twice the number of TCs during mid and late July (Fig. 3b). Subsequent to the monsoon break, the TC number shows a decrease but with a strong fluctuation in mid-August. The mean values of TC number during the three stages are, respectively, 4.9, 9.3, and 6.4, and the increase during the monsoon break is statistically significant at the 90% confidence level.
All kinds of TCs are involved in this study. We have divided the TCs into tropical depressions and TCs as strong as tropical storms. The results indicate that both the categories of TCs show consistent changes during the monsoon break, featured by the decrease (increase) over region S (N) (figures not shown).
In addition, we have also repeated the above analyses using OLR and TC data from 1979 to 2014 (figures not shown). The changes of circulations and TCs during the monsoon break are consistent with the present results, except for a smaller sample size of TCs. For instance, the mean values of TC number over region S (N) during the three stages are, respectively, 9.1 (3.3), 5.2 (5.5), and 9.8 (2.9).
The changes of TC number over regions S and N around the monsoon break are closely related to the evolution of large-scale circulation and convection. Over region S, the monsoon break is characterized by anomalous downward motion, which coincides with weakened convection, upper-level divergence, and low-level convergence and is accompanied by anomalous 850-hPa anticyclonic circulation, which coincides with negative relative vorticity and a weakened monsoon trough (Fig. 1). These conditions suppress the activity of TCs there (e.g., Gray 1968; Yumoto and Matsuura 2001; Wu et al. 2012; Gao and Li 2012). In contrast, over region N, the upward motion and anomalous 850-hPa cyclonic circulation during the monsoon break provide favorable conditions for the activity of TCs there.
4. Modulation of TC activity in typical monsoon break years
The above results demonstrate the changes of TC activity during the monsoon break. However, these changes only reach a moderate statistical significance. The weakness of the statistical significance may be because of two possible reasons: On the one hand, the sample size of TCs is small in such short periods for stages 1–3. For example, we calculated the TC number in each year, and it ranges from 2 to 16 over the WNP from year to year. As for much smaller domains [i.e., over region S (N)], the TC number ranges from 0 to 6 (0 to 4). On the other hand, the WNP monsoon break exhibits a significant year-to-year variation, as revealed by Xu and Lu (2015). About 30% of years can be classified as typical monsoon break years (called “type A” in their study), when the monsoon break concurs with the climatological monsoon break but is much more significant. However, there is a similar proportion of years (called “type B” in their study) in which the monsoon break phenomenon is relatively vague, and convection even tends to show an out-of-phase evolution with that in the climatology. The large year-to-year variations result in a weakening of the monsoon break intensity in the climatological sense. Therefore, focusing on the typical monsoon break years may highlight the modulation of TC evolution by the monsoon break.
To test this hypothesis, we first define a monsoon break index in the same way as in Xu and Lu (2015), but replacing OLR with 500-hPa vertical velocity: that is,
where the values of ω are averaged over the monsoon break region [(10°–20°N, 140°–160°E); the same as in Xu and Lu (2015)], and i means the ith year starting from 1950, taking values from 1, 2, …, 65. A positive (negative) index indicates that the monsoon break in a particular year tends to be phase locked (out of phase) with the climatological monsoon break.
Figure 4 shows the time series of the standardized monsoon break index by using 500-hPa vertical velocity (black line). The index by using OLR, as in Xu and Lu (2015), is also shown for comparison (red line). The correlation coefficient between these two indexes is 0.87 during 1979–2014, indicating good coherence between them. Therefore, the present index, by using 500-hPa vertical velocity, can be used to quantitatively measure the interannual variation of the monsoon break over a longer time range. Based on the index, 17 typical monsoon break years are selected during 1950–2014 (1958, 1959, 1967, 1969, 1977, 1981, 1982, 1986, 1990, 1991, 1994, 1996, 1997, 1998, 2002, 2012, and 2014) when the index exceeds 0.7 standard deviations, as in Xu and Lu (2015).
The monsoon break index exhibits two interesting decadal changes. One is an obvious change around 2003, with indexes turning from positive into negative. In addition, the intensity of the variability tends to be smaller before the end of the 1970s, compared to the latter period, which is consistent with the fact that there are only 5 typical years out of the first half of the period, much less than that (12 typical years) in the second half of the period, despite the negative phase since 2003. But is the decadal change around the end of the 1970s a real one or only an artificial one due to the differences in observational data used in the reanalysis? And what mechanism is responsible for the decadal change around 2003? These questions are interesting but beyond the scope of the present study.
Figure 5 shows the differences of 500-hPa vertical velocity and 850-hPa wind between the monsoon break (stage 2) and the adjoining periods (stages 1 and 3) in typical years. There is a positive (negative) anomaly of vertical velocity and an anticyclonic (cyclonic) anomaly over the subtropical (midlatitude) WNP. These anomalies show much stronger intensity and larger scales than those in the climatology (Fig. 1). In particular, the amplitude of the anomalous downward (upward) motion is greater than 5 × 10−2 Pa s−1 (−3 × 10−2 Pa s−1) over region S (N), and the maximum speed of the westerly along 20°–30°N associated with the anticyclonic and cyclonic anomalies is about 6.0 m s−1; both the values are 3 times larger than their climatological counterparts. Therefore, it can be concluded that the circulation changes during the monsoon break are indeed remarkably intensified in typical years.
Figure 6 shows the spatial distributions of the TC activity averaged over every stage in typical years and the differences between stage 2 and the adjoining stages. In stages 1 and 3, region S is an active zone for TCs, while very few TCs exist over region N. The large values of TC activity also show zonal elongation, particularly for stage 3. In contrast, in stage 2, the values of TC activity increase significantly over region N, where a center appears with its amplitude greater than 3. However, TC activity is remarkably suppressed over region S, which is almost a “TC-free area.” In the difference field (Fig. 6d), the amplitudes of both the positive and negative anomalies are over ±2.5 in regions N and S. The amplitudes are over half of those shown in Fig. 2d, even though the number of typical years is only about of that for the entire analysis period (17 vs 65). Therefore, the changes of TC activity during the monsoon break are statistically more significant in typical years, in comparison with the entire analysis period.
Figure 7 shows the daily TC number appearing in regions S and N from mid-July to late August in typical years. The TC number over region S increases from 0–2 in mid-July to as high as 8 in late July, then rapidly decreases to 0–2 again in early August as the monsoon break appears, and subsequently increases to 8 again during mid-August (Fig. 7a). In particular, during the monsoon break, the number of TCs remains at no more than 2, and the reduction in TC number is statistically significant at the 99% confidence level in comparison with stages 1 and 3. In contrast, the evolution of TC number over region N is almost out of phase with that over region S (Fig. 7b). The TC number remains at about 2 during late July, then rapidly increases to as high as 8 during the monsoon break period in early August, and subsequently decreases again in mid-August. In comparison with stages 1 and 3, the statistical significance of the increase in TC number during the monsoon break reaches the 95% confidence level.
Figure 8 shows the TC tracks on each stage in typical monsoon break years. Considering every stage has different length, the middle 6 days for stages 1 and 3 are selected to facilitate the comparison with stage 2. In stages 1 and 3, the majority of TCs tend to propagate northwestwards, and more than 10 tracks appear over region S, but very few tracks appear over region N. In contrast, at stage 2, most of northwestward tracks concentrate to the west of 140°E, and more eastward recurving tracks occur north of 25°N. The number of TC tracks over region N increases to 8, while nearly none of tracks appear over region S. These features of TC tracks on each stage and the changes during the monsoon break resemble those of TC activity. Thus, this result further confirms the modulation of TCs by the monsoon break.
We also examined the evolutions of circulation and TCs, including the TC activity and TC tracks, in typical years in Xu and Lu (2015), and found that the changes in both circulation and TCs during the monsoon break are very similar to those shown in Figs. 5–8.
In addition, considering the decadal signal of the monsoon break index, especially the decreasing trend since the 1990s, we have also selected typical years by the decadal-removed index. There are only 3 years (1974, 1998, and 2008) that are different from the original typical years, and the results are very consistent with above results (figures not shown). Considering the consistency between the original and interannual variations, we show only the results obtained by the original data.
As shown above, the characteristics of the monsoon break in typical years are similar to those in the climatology, but the intensity of differences shown during the monsoon break is much stronger. The changes of TC activity and corresponding circulation during the monsoon break are more significant. Therefore, focusing on typical years can help to reveal the mechanisms responsible for the modulation of TCs by the monsoon break.
Figure 9a shows the time series of 500-hPa vertical velocity averaged over the monsoon break region (blue line) and region N (red line) in typical years. The vertical lines indicate the ±0.5 standard deviations on each day, which depict the extent of differences among cases. Taking the case difference into consideration, the out-of-phase features of convection evolution over these two regions are significant during stages 1–3. Specially, the upward motion is enhanced over the monsoon break region in late July when the monsoon onset occurs. Then it becomes remarkably weakened during the monsoon break period in early August and enhanced again in mid-August. This evolution of vertical velocity is highly consistent with that of convection in Xu and Lu (2015). In contrast, the upward motion over region N remains weak during late July, then suddenly enhances during the monsoon break period in early August, and becomes weakened again in mid-August. The correlation coefficient between the convection evolutions over these two regions comes to be −0.55 during stages 1–3.
From late July to mid-August, 10–25-day oscillations contribute greatly to the subseasonal evolution over the monsoon break region (Xu and Lu 2015) and may also influence that over region N, if considering the out-of-phase feature of subseasonal evolutions over these two regions (Fig. 9a). Figure 9b shows the time series of 10–25-day filtered vertical velocity, which are essentially consistent with those of the unfiltered field (Fig. 9a). Considering the differences between cases, the two time series are remarkably out of phase from late July to mid-August. In particular, the positive (negative) phase over the monsoon break region (region N) in early August corresponds to the suppressed (enhanced) upward motion during the monsoon break. In comparison with stages 1 and 3, the value of 10–25-day vertical velocity increases (decreases) by 3.8 × 10−2 Pa s−1 (1.7 × 10−2 Pa s−1) in stage 2 over the monsoon break region (region N), which is 76% (66%) of the unfiltered field. These results indicate that 10–25-day oscillations not only contribute greatly to the monsoon break, which is consistent with Xu and Lu (2015), but also contribute greatly to the out-of-phase subseasonal evolution over region N.
As the subseasonal evolutions over the monsoon break region and region N are almost out of phase, and are both closely associated with 10–25-day oscillations, it is suspected that they may be linked through the propagation of oscillations. Therefore, we calculated the lead–lag correlation coefficients between the two time series of filtered vertical velocity. The largest correlation coefficient is −0.87 at time lag 0, indicating that the out-of-phase evolutions over these two regions are almost simultaneous.
We hypothesize that this nearly simultaneous relationship between these two regions results from there being roughly the same amount of cases when region N leads the break region as when the break region leads region N. To test this hypothesis, we select the cases of strong 10–25-day oscillations over both regions during stages 1–3, or more precisely, when the 10–25-day anomalies are out of phase between the two regions and both are greater than 0.5 standard deviations of their respective time series. This criterion yields 14 cases in which the peak/nadir over region N leads the nadir/peak over the monsoon break region, simply called “north-lead-south cases” for short, and 15 opposite cases, simply called “south-lead-north cases” for short. Day 0 is designated when the maximum or minimum vertical velocity oscillation occurs over the monsoon break region for each case, and all the variables for minimum oscillation cases are multiplied by −1 before performing the composite analysis in order to match with the maximum oscillation cases. The negative and positive days stand for the days before and after day 0, respectively.
Figure 10 shows the evolution of composite 10–25-day filtered 200-hPa wind. For the north-lead-south cases (Fig. 10a), an obvious cyclonic anomaly appears over Europe at day −9. Then, as the energy disperses eastward, a well-defined wave train forms along 40°N over the Eurasian continent at day −6. At day −3, the energy propagates eastward to East Asia and the WNP to the east of Japan, and there is a strong cyclonic anomaly over the northeast of region N. This upper-tropospheric cyclonic anomaly is favorable for the enhancement of convection over region N (Sakamoto and Takahashi 2005; Sato et al. 2005). On the other hand, over the North Pacific, there tends to be another wave train characterized by a westward phase propagation. There is a well-defined wave train over the North Pacific at day −9, and the above-mentioned cyclonic anomaly over the northeast of region N at day −3 tends to be related to the cyclonic anomaly around the date line at day −9. This westward propagation of intraseasonal oscillations resembles somewhat the result of Lu et al. (2007). The present result suggests that both the wave trains over the Eurasian continent and over the North Pacific may play a role, implying the complexity of subseasonal oscillations over region N. By contrast, for the south-lead-north cases (Fig. 10b), the wave trains are not well organized over the Eurasian continent and North Pacific. Although there is a clear wave train over the North Pacific at day −3, it does not exhibit westward phase propagation. The cyclonic anomaly appears over the north of region N at day 0, a delay of roughly three days in comparison with the north-lead-south cases, consistent with the lag of convection enhancement over region N.
The eastward energy dispersion of disturbances over the Eurasian continent for the north-lead-south cases can be more clearly illustrated by Fig. 11a, which shows the filtered 200-hPa meridional wind averaged over 40°–42.5°N. The disturbances propagate from Europe eastward to East Asia, which resembles the so-called “Silk Road pattern” well (Lu et al. 2002; Enomoto et al. 2003; Ding and Wang 2005; Yasui and Watanabe 2010; Chen et al. 2013). This eastward propagation can be explained by the “waveguidability” of the Asian upper-tropospheric westerly jet (Hoskins and Ambrizzi 1993). There is also similar eastward propagation for the south-lead-north cases (Fig. 11b), but the disturbances are weaker, and the propagation tends to exhibit both energy dispersion and phase propagation.
During summer, 10–25-day oscillations generally propagate from the equatorial WNP northwestward to the East China Sea (Ko and Hsu 2006; Chen and Sui 2010). Figure 12 shows the filtered 850-hPa vorticity along the general propagation path as in Xu and Lu (2015): that is, a line from (0°N, 155°E) to (30°N, 130°E). Consistent with previous studies, 10–25-day oscillations generally propagate northwestwards for both categories of cases. However, this northwestward propagation is clearer for the south-lead-north cases. This difference in propagation between the two categories is consistent with the results shown in Figs. 10 and 11: that is, 10–25-day oscillations over region N for the north-lead-south cases are additionally affected by midlatitude perturbations over the Eurasian continent and North Pacific and then affect oscillations over the monsoon break region.
It should be mentioned that the midlatitude wave trains and tropical disturbance exert an influence on the subseasonal evolution over region N and the monsoon break region for all the cases, either the north-lead-south cases or south-lead-north cases. In some cases, the midlatitude effects may be predominant or appear in advance, while the tropical effects do so in other cases.
5. Conclusions and discussion
Based on 65 years of data from 1950 to 2014, this study examined the modulation of TC activity by the WNP monsoon break. In early August, when the climatological monsoon break occurs, the TC number to the east of the Mariana Islands (12.5°–25°N, 140°–160°E) decreases remarkably to about half of that during the adjoining periods in late July and mid-August, respectively. In contrast, the TC number to the southeast of Japan (27.5°–37.5°N, 140°–155°E) exhibits a rapid increase during the monsoon break to about twice that over the adjoining periods. The reduction (increase) in TC number to the east of the Mariana Islands (southeast of Japan) during the monsoon break is closely associated with local anomalous downward (upward) motion and a lower-tropospheric anticyclone (cyclone).
However, the above-mentioned modulation of TCs possesses only moderate statistical significance because of the small sample size of TCs from late July to mid-August and the substantial year-to-year variability of the monsoon break. Therefore, we focused on typical years when the monsoon break is particularly prominent and reexamined the changes of TC activity and circulation in these years. The result showed that the changes of TC activity during the monsoon break in typical years show similar spatial patterns to those of the entire 65 yr but are statistically more significant, which is attributed to the remarkably intensified changes of corresponding circulations.
In addition, the composite analyses for typical years indicated that the subseasonal evolutions of convection (estimated by vertical velocity) tend to be out of phase to the east of the Mariana Islands and to the southeast of Japan during the period from late July to mid-August, which can explain the reverse changes of TC activity over these two regions around the monsoon break. A further analysis denoted that 10–25-day oscillations contribute greatly to the subseasonal evolutions over these two regions and show the convective and suppressed phase during the monsoon break to the east of the Mariana Islands and to the southeast of Japan, respectively. This result is in agreement with previous studies, in which it was indicated that TCs tend to occur in the convective phase of quasi-biweekly oscillations (Ko and Hsu 2006; Gao and Li 2011; Jin et al. 2012; Li and Zhou 2013; Zhao et al. 2015).
Furthermore, the present results suggest that the subseasonal evolutions of convection to the southeast of Japan and to the east of the Mariana Islands can be affected by the wave trains in the midlatitudes and the northwestward propagation of 10–25-day oscillations in the tropics. The midlatitude wave trains include the Silk Road pattern–like disturbances over the Eurasian continent and westward phase propagated disturbances over the North Pacific. When the midlatitude disturbances are predominant, the subseasonal evolution of convection to the southeast of Japan tends to lead that to the east of the Mariana Islands, and vice versa.
The fact that the 10–25-day oscillations tend to be phase locked with the seasonal cycle is interesting. In this study, we suggest that the 10–25-day oscillations tend to be at the wet phase in late July, concurring with the convection jump and monsoon onset, and transfer to the dry phase in early August and thus contribute to the monsoon break. Then the reason for the phase lock of 10–25-day oscillations is transferred to that for the convection jump. Several mechanisms have been suggested to explain the phase lock of the convection jump, although there is no consensus on this issue. Wu and Wang (2001) emphasized the contribution of the match between the climatological intraseasonal oscillations and slow seasonal evolution. On the other hand, Ueda et al. (2009) suggested that the intense convection over other regions in June suppresses the convection over the subtropical WNP, and the gradual tropospheric moistening triggers the monsoon onset in late July.
The present results suggest that the monsoon break to the east of the Mariana Islands affects the local weather and is related to the weather in a remote location (i.e., TCs to the southeast of Japan). They also indicate that there are relationships between various time scales, including interannual variability, seasonal evolution, subseasonal oscillation, and synoptic activities around the monsoon break. Therefore, the monsoon break may be a good case for us to further understand the interaction between the different time scales over the WNP. The possible decadal change since 2003 shown in Fig. 4 may further extend the relevant and involved time scales to longer ones. In addition, the monsoon break may also be a case that can be used to further investigate the tropical–extratropical interaction over the WNP.
The authors greatly appreciate the comments from three anonymous reviewers. This work was supported by the National Natural Science Foundation of China (Grant 41320104007).