Characteristics of Large-Scale Atmospheric Fields during Heavy Rainfall Events in Western Japan: Comparison with an Extreme Event in Early July 2018

In order to explore large-scale atmospheric factors causing heavy rainfall events that occurred widely in western Japan, a composite analysis of atmospheric fields during the past heavy rainfall events in the region is conducted using the Japanese 55-year Reanalysis. During heavy rainfall events, atmospheric fields are characterized by an upper-tropospheric trough over the Korean Peninsula (KP), an upper-tropospheric ridge to the east of Japan, a surface high-pressure system to the southeast of Japan, and southwesterly moisture flux. The composite analysis indicates that a clear wave train due to quasi-stationary Rossby wave packet propagation (RWPP) along the polar front jet over Siberia tends to occur just before extreme events. Further analysis considering various time scale variabilities in the atmosphere reveals that surface high-pressure anomalies to the southeast of Japan are dominated by variability with a 25 – 90 day period, whereas variability with an 8 – 25 day period dominates lower-pressure anomalies over the East China Sea (ECS) in relation to the development of the upper-tropospheric trough around the KP. We also investigate atmospheric fields during an extreme heavy rainfall event that occurred in early July 2018 (HR18). Atmospheric features during HR18 are generally similar to those of the other heavy rainfall events. However, a remarkable RWPP occurred along the subtropical jet (STJ) in late June 2018 and intensified a surface high-pressure system to the southeast of Japan. Additionally, a low-pressure system with an 8 – 25 day period to the south of Japan developed in association with wave breaking induced by the remarkable RWPP along the STJ and propagated northwestward toward the ECS and then to Japan. The simultaneous development of highand low-pressure systems contributed to the extreme southerly moisture flux into western Japan. HR18 is also characterized by a sharp upper-tropospheric trough over the KP that is dominated by high frequency variability with a period < 8 days.


Introduction
Heavy rainfall events frequently occur during the rainy summer monsoon season in Japan known as the Baiu. Such events intermittently cause flooding and have serious socio-economic impacts. For example, in early July 2018 an extreme heavy rainfall event, which we refer to as HR18 (Heavy Rainfall event in 2018), occurred and seriously affected western Japan and the adjacent Tokai region, located to the east of western Japan Shimpo et al. 2019). Takemura et al. (2019) revealed that both a shallow southerly airstream caused by the surface North Pacific Subtropical High and a deeper southwesterly airstream due to enhanced convection over the East China Sea (ECS) contributed to the extensive rainfall. Yokoyama et al. (2020) thoroughly analyzed the atmospheric fields around Japan and identified the importance of an upper-tropospheric trough that stayed to the rear of the extensive rainfall area. They also discussed the cause of this extreme event in terms of both dynamical and diabatic effects. Akiyama (1975) identified the important contribution of moisture flux from the subtropical Pacific to extreme heavy rainfalls. Several case studies have focused on heavy rainfall events in Japan (e.g., Ninomiya 1978;Akiyama 1984Akiyama , 1989Kato and Goda 2001;Shibagaki and Ninomiya 2005). Ninomiya and Akiyama (1992) discussed the importance of the interaction between the multi-scale (e.g., planetary, synoptic, and mesoscale) phenomena in the occurrence of heavy rainfall. Ninomiya (2001) and Ninomiya and Shibagaki (2007) conducted Q-vector analyses of observational data and found a role for the uppertropospheric trough in the intense rainfall of July 1991.
Using a case study, Yoshida and Ito (2012) investigated the indirect effect of tropical cyclones on heavy rainfall during the Baiu season in Kyushu and discussed a contribution by a large moisture flux oriented toward the south of Kyushu. Hirota et al. (2016) examined the factors that cause an extreme rainfall event in Hiroshima, Japan, on 9 August 2014, and found significant filamentary transport of water vapor from the Indochina Peninsula to the Japanese islands. They also mentioned the importance of a cutoff low detached from the subtropical jet (STJ) over the central Pacific. Kamae et al. (2017) evaluated the contribution of atmospheric rivers (low-level moisture flows) to the hydrological cycle over East Asia and identified a relationship between heavy rainfall in the warm season and the El Niño of the preceding winter.
Furthermore, Kosaka et al. (2011) showed the statistical relationship between a 30 day Meiyu-Baiu precipitation in early summer and the El Niño-Southern Oscillation in the preceding boreal winter, Pacific-Japan teleconnection (Nitta 1987), Silk Road pattern along the Asian jet (Enomoto et al. 2003), and a wave train pattern along the polar front jet (PFJ). They also discussed the role of the Pacific-Japan teleconnection pattern as a medium between the El Niño-Southern Oscillation and Meiyu-Baiu precipitation. Hirota and Takahashi (2012) argued the importance of both southward upper-tropospheric and northward lower-tropospheric Rossby wave propagations in the formation of a tri-polar anomaly pattern with centers situated around the Philippines, China/Japan, and East Siberia, which dominantly appears in climate variations of the East Asian summer monsoon and is closely related to the interannual variations of the Baiu. However, the relationships between heavy rainfall events over western Japan in the warm season and large-scale variability, such as quasi-stationary Rossby wave packet propagation (RWPP), over Eurasia have not been elucidated.
Using the Japanese 55-year Reanalysis (JRA-55) including near-real-time data (Kobayashi et al. 2015), the present study investigates statistical large-scale atmospheric characteristics during the past heavy rainfall events that have occurred widely in western Japan since 1979. We also compare these characteristics during previous heavy rainfall events with those during HR18. This type of investigation can help elucidate the large-scale atmospheric factors that cause heavy rainfall events in western Japan. A comprehensive understanding of these factors will be useful for early warning systems and for the mitigation of adverse socio-economic effects, because these events continue to frequently occur in western Japan.
The remainder of this paper is organized as follows. Data and analysis methods are described in Section 2. Results of the composite analysis of historical events and a comparison with HR18 are provided in Section 3.
A possible mechanism that explains the atmospheric characteristics during HR18 is discussed in Section 4. Finally, a summary and conclusions are given in Section 5.

Data
In the present study, we use in-situ observational precipitation derived from the Japan Meteorological Agency (JMA) Automated Meteorological Data Acquisition System (AMeDAS) to extract past heavy rainfall events that occurred widely in western Japan. In order to analyze the atmospheric fields, we utilize surface, isobaric, total-column, and isentropic analysis fields from JRA-55 products with a horizontal resolution of 1.25° in both latitude and longitude. The National Oceanic and Atmospheric Administration Interpolated Outgoing Longwave Radiation (OLR) data (Liebmann and Smith 1996) are also used. We utilize the daily climatology defined as averages for the period 1981 -2010, which is according to a definition by the JMA, and filtered by 60 day low-pass filter (LPF) based on Duchon (1979). The details of this filter are described in Section 2.2. We define anomalies as deviations from the climatology.

Methods
In order to extract past heavy rainfall events from the historical data, we average AMeDAS daily precipitation over western Japan, using data from the 296 AMeDAS stations that are continuously available for the 40 years from 1979 to 2018, and area-averaged daily total precipitation are accumulated over a 3 day period centered around each day during the warm season (from May to September) for the study period. We then identify heavy rainfall events as 3 day precipitation that exceeds the 95th percentile (Fig. 1). Note that if two extracted dates are < 8 days apart, we consider them to be the same event, with the date of the event corresponding to the peak 3 day precipitation during the event. We exclude events when a typhoon center exists within 500 km from the measurement stations used in the analysis (gray bars in Fig. 1, 35 events) to avoid confusion between the direct effects of typhoons related to landfall and other atmospheric processes (Kamahori 2012). There are a total of 42 heavy rainfall events identified in the present study. We exclude HR18 from the composite analysis (red bar in Fig. 1). For the composite analysis, we extract 30 events from the heavy rainfall events described above and categorize them into three groups by total precipitation, namely, the highest 10 (TP10), middle 10 (MD10), and lowest 10 (LW10) events, which are represented by orange bars, yellow bars, and green bars in Fig. 1, respectively. Statistical significance is calculated using Student's t-test. The 90 % and 95 % confidence levels are used to indicate statistical significance.
In the present study, we diagnose the quasi-stationary RWPP using the wave activity flux (WAF) given by Takaya and Nakamura (2001). They derived an approximate conservation relation of the wave activity pseudo-momentum for quasi-geostrophic eddies on a zonally varying basic flow by averaging neither in time nor in space. The horizontal component of WAF is defined as follows: where U = (U, V ) is a steady zonally varying basic flow defined as the climatological horizontal winds; p is pressure normalized by 1000 hPa; and φ and λ are latitude and longitude, respectively. A prime denotes the anomalies. The stream function and radius of the Earth are noted by ψ and a, respectively. Because the rightmost term "C u M " represents the effect of the phase propagation and the present study are focusing on the quasi-stationary Rossby wave, we consider that this term can be ignored. The WAFs are derived from 3 day mean fields.
In order to evaluate the contributions of atmospheric variability over various time scales to the occurrence of rainfall events, we apply a Lanczos filter (Duchon 1979) to the JRA-55 products. This digital filtering involves transforming an input data sequence x t , where t is time, into an output data sequence y t using the linear relationship in which w k are suitably chosen weights. For example, weights for a high-pass filter (HPF) are calculated as follows: where f c and 2n + 1 are the cutoff frequency and sample size for filtering, respectively. Weights for a LPF can be obtained by subtracting those for a HPF from one. We can obtain weights for a band-pass filter (BPF) using weights for two LPFs with different cutoff frequencies. In the present study, an 8 day HPF, an 8 -25 day BPF, a 25 -90 day BPF, a 90 day LPF, and a 25 day LPF are used. Synoptic-scale eddies have 8 day or shorter periods, and the 25 -90 day period corresponds to intraseasonal variability such as the Madden-Julian Oscillation in the previous studies (e.g., Kikuchi et al. 2012

Composite analysis of 3 day mean fields and
comparison with HR18 The dates and 3 day precipitation averaged over western Japan for the top 21 heavy rainfall events are listed in Table 1. The dates in Table 1 represent the central dates for the 3 day accumulated precipitation. Notably, the precipitation during HR18 exceeds 250 mm and represents the largest value among the events. Most events affected by the landfall or passage of a typhoon occurred in September (Table 1). Most other events, by contrast, occurred in June or July, during the Baiu. The highest-precipitation non-typhoon events that rank from 3rd (205.9 mm) to 21st (139.2 mm) of the total events by precipitation, are utilized for the composite analysis as "TP10". As shown in Fig. 1, 3 day precipitation for events MD10 and LW10 are 100 -125 and 75 -100 mm, respectively. The results of the composite analysis for these events are also evaluated in the present study.
Composite atmospheric fields around Japan for TP10 are shown in Fig. 2. In the upper troposphere, positive vorticity anomalies over the Korean Peninsula (KP) and a wide area of negative vorticity anomalies centered over Japan are statistically significant (Fig.  2a), and significant westerly wind anomalies are distributed along the large gradient of the vorticity anomalies ( Fig. 2b). At 500 hPa geopotential height ( Fig. 2d), significant negative height anomalies over the KP and positive height anomalies to the east of Japan are also evident. These features represent the development of the upper-tropospheric trough over the KP and the upper-tropospheric ridge to the east of Japan, contributing to a dynamically induced midlevel ascent ahead of the trough from China to Japan, where upwelling anomalies are statistically significant at 500 hPa (Fig. 2c). Notably, these upwelling anomalies also include diabatic forcing by active convection (Fig. 2h). In the lower troposphere, horizontal distribution of significant positive vorticity anomalies is generally consistent with that of significant mid-level upwelling anomalies from China to Japan (Fig. 2e). Sea level pressure (SLP; Fig. 2f) exhibits statistically significant high-pressure anomalies to the southeast of Japan that indicate the development and persistence of a high-pressure system in the area. Low-pressure anomalies to the west of Kyushu are also statistically significant. Vertically integrated moisture flux ( Fig.  2g) indicates an anomalous southwesterly moisture inflow toward Japan along the western-to-northern fringe of the high-pressure anomalies (Fig. 2f). Statistically significant moisture flux convergence (contours in Fig. 2g) is analyzed immediately over the region of the upwelling anomalies at 500 hPa (Fig. 2c). The features in the composite maps indicate that the surface high-pressure system to the southeast of Japan plays an important role in moisture transport during heavy rainfall events, consistent with the results of Akiyama (1975).
As described above, TP10 is characterized by an upper-tropospheric deepened trough over the KP, an upper-tropospheric ridge to the east of Japan, a surface high-pressure system to the southeast of Japan, and southwesterly moisture flux in the lower troposphere. These features are also present but weaker for MD10 and LW10 (not shown).
The upper-tropospheric RWPPs preceding heavy rainfall events for TP10 and for MD10 and LW10, respectively, are shown in Figs. 3 and 4. Comparing the three groups, TP10 experiences persistent wave train along the PFJ (Figs. 3a -d). The existence of the wave packet propagation along the PFJ in summertime is consistent with related previous studies (e.g., Iwao and Takahashi 2008;Nakamura and Fukamachi 2004;Ogasawara and Kawamura 2008). The wave packets in TP10 propagate from northern Europe to eastern Siberia along the PFJ (Figs. 3a -c) and contribute to the enhancement of the ridge to the east of Japan (Fig.  3d). By contrast, the wave packets propagating along the STJ over central Asia (Fig. 3b) contribute to the enhancement of the trough over the KP and, in turn, the ridge to the east of Japan (Fig. 3d), although this wave train is not as clear as that along the PFJ. During MD10 (Figs. 4a -d), the wave packets emanating from western Europe propagate along the STJ and strengthen the trough over the KP and the ridge to the east of Japan (Fig. 4d). Although RWPPs are clearly seen over Eurasia before LW10 (Figs. 4e -g), their contribution to the enhancement of the anomalous circulation around Japan is unclear during LW10 (Fig.  4h). These results indicate that RWPPs over Eurasia, particularly along the PFJ, play an important role in extreme heavy rainfall events such as TP10.
Then, anomalous atmospheric fields during HR18 ( Fig. 5) are compared with those for TP10 (Fig. 2). In the upper troposphere ( Fig. 5a), the trough over the KP is sharper than that in TP10 (Fig. 2a). Dynamical forcing by the trough over the KP plays an important role in inducing strong upwelling over western Japan, as indicated by Yokoyama et al. (2020) and Takemura et al. (2019), although the trough is not obvious in the 500 hPa height anomaly field (Fig. 5d). This is partly associated with significantly warm conditions at mid-latitudes, as suggested by Kobayashi and Ishikawa (2019) and Takemura et al. (2019). In the mid-troposphere, strong upwelling anomalies are concentrated over western Japan (Fig. 5c), and lowertropospheric positive vorticity anomalies are distributed from the ECS to western Japan (Fig. 5e). This differs from the conditions of TP10 events, particularly over the ECS (Fig. 2e). The SLP shown in Fig. 5f indicates that high-pressure anomalies to the southeast of Japan are comparable with those of TP10 (Fig.  2f), and low-pressure anomalies over the ECS are stronger than those of TP10. Consequently, southerly moisture flux is concentrated in the region 130 -135°E  to the south of western Japan (Fig. 5g). These features indicate that both the high-pressure system to the southeast of Japan and the low-pressure system over the ECS contribute to the enhanced southerly moisture flow toward western Japan, as indicated by Takemura et al. (2019). Although the high-pressure system to the southeast of Japan is not extremely strong (Fig.   6a) and the low-pressure system over the ECS is not significant compared with other historical events (not shown), their simultaneous occurrence contributes to the intensification of the southerly flow toward western Japan in the lower troposphere. The regionally averaged meridional wind at 925 hPa during HR18 (Fig. 6b)   toward western Japan among the analyzed events and is associated with an enhanced pressure gradient in the region. Additionally, specific humidity in the region shows positive anomaly, although it is not extreme value at all (Fig. 6c). The lower-tropospheric southerly moisture flux in the region is the strongest among the analyzed events (Fig. 6d), and the consequent convergence of vertically integrated moisture flux over western Japan is greatest among the events (Fig. 6e). These enhanced moisture flux and its convergence, which are strongest compared with TP10, are consistent with the anomalous meridional wind and specific humidity in the lower troposphere. For all the analyzed events, the moisture flux convergence due to wind anomalies at 925 hPa are more correlated with precipitation over western Japan than are wind anomalies at higher levels ( Fig. S1). Therefore, the simultaneous occurrence of the high-pressure system to the southeast of Japan and the low-pressure system over the ECS is one of the most important characteristics of HR18. Furthermore, notably, the moisture flux convergence due to specific humidity anomaly at 700 hPa for HR18 is extremely large compared with those for the other rainfall events (Fig. S1c). This is consistent with the result of Yokoyama et al. (2020), who conducted a thorough analysis of the atmospheric field around Japan. They showed that the moistening in the mid-troposphere was caused by the dynamical forced ascent in association with the upper-tropospheric trough, which lingered in the region from the KP to the Sea of Japan. They also mentioned the importance of the moistening in the mid-troposphere in further development of deep cumulus convection and its organization.
The upper-tropospheric RWPP from Eurasia to Japan during HR18 is shown in Fig. 7. In late June (Figs. 7a, b), a remarkable RWPP-the strongest among the analyzed events (Fig. S2)-occurs along     the STJ and strengthens the upper-tropospheric ridge to the east of Japan. This enhanced ridge causes the subsequent formation of a surface high-pressure system to the southeast of Japan (Fig. 5f), corresponding to a formation mechanism of the Bonin high with the equivalent barotropic structure (Enomoto et al. 2003). The ridge to the east of Japan once weakens at the beginning of July (Fig. 7c) but strengthens again because of the subsequent RWPP along the PFJ accompanied by an amplified wave train on 5 -7 July (Fig. 7d). The features of the RWPP along the PFJ during HR18 are generally consistent with those in TP10, although the phases of their wave trains differ. We will discuss another role of the strong RWPP along the STJ just before HR18 in Section 4.

Analysis of atmospheric variability at various time scales
In this section, we analyze detailed circulation features, using several time filters.
Composite maps of time-filtered SLP anomalies for TP10 are shown in Figs. 8a -d (left panels). Highpressure anomalies to the southeast of Japan are evident and statistically significant in this region with all-time filters except the 8 day HPF. With the 25 -90 day BPF, which corresponds to the intraseasonal time scale, these anomalies are particularly evident (Fig.  8c). Notably, the 25 -90 day BPF anomalies of MD10 and LW10 are weaker than those in TP10 (see also Fig. S3). During HR18 (Figs. 8e -h), fluctuations with 25 -90 day periods are strong and are associated with the development of the high-pressure system to the southeast of Japan (Fig. 8g). In the upper troposphere, intraseasonal time scale variability is evident for TP10 and HR18 and is accompanied by a wave train from northern Eurasia to Japan (Fig. 9), which contributes to the enhancement of the anomalous anticyclone to the east of Japan. These results indicate the importance of RWPPs along the PFJ to anomalous circulation around Japan, including surface high-pressure systems to the southeast of Japan. Convective activity around the Philippines is also expected to contribute to the development of surface high-pressure systems to the southeast of Japan (Nitta 1987;Kosaka and Nakamura 2010). The composite map of OLR anomalies filtered using the 25 -90 day BPF reveals enhanced convective activity around the Philippines, which is associated with a northward migration of the Boreal summer interseasonal oscillation ( Fig. 9g; see also Fig. S4). Such convective activity is not evident using the 8 day HPF, the 8 -25 day BPF, or the 90 day LPF (Figs. 9e, f, h). However, during HR18 (Fig. 9o), enhanced convective activity associated with the BSISO is located to the south of the Philippines, which is far from the surface high-pressure system to the southeast of Japan. Therefore, it is likely that the direct contribution of convective activity around the Philippines to the development of the surface high-pressure system to the southeast of Japan during HR18 is smaller than in TP10.
Then, we focus on surface low-pressure anomalies from the ECS to Japan filtered using the 8 -25 day BPF (Fig. 8f). In the composite analysis for TP10, they are statistically significant (Fig. 8c), which are related to the development of the upper-tropospheric trough around the KP (Fig. 9b). The development of the upper-tropospheric trough is also seen during HR18 (Fig. 9j); however, it is centered over the Sea of Japan and shifts north-eastward compared with that in TP10. Meanwhile, surface low-pressure anomalies over the ECS during HR18 are clearer than those in TP10 (compare Figs. 8f and 8c). This feature cannot be explained by the development of the upper-tropospheric trough only. Takemura et al. (2019) mentioned the importance of lower-tropospheric cyclonic circulation anomalies over the ECS to the southerly moisture flux toward western Japan using a potential vorticity (PV) budget analysis of HR18 and discussed that diabatic heating associated with active convection over the ECS acts to maintain lower-tropospheric cyclonic circulation anomalies. In the present study, we investigate the time evolution of surface low-pressure anomalies over the ECS from late June to early July 2018 (Fig. 10). On 26 June 2018 (Fig. 10a), negative OLR anomalies at 20°N, 140°E are observed and are associated with enhanced convection in this region and weak low-pressure anomalies in the western part of the region. Both the active convection and the low-pressure anomalies intensify and propagate northwestward toward the ECS until the beginning of July 2018 (Figs. 10b -d) before moving into Japan in early July (Figs. 10e, f). Although the low-pressure anomalies partly correspond to the track of typhoon Prapiroon, which rapidly moved northward to the north of western Japan on 4 July (not shown), the enhanced convection and low-pressure anomalies persisted over the ECS in early July. Using a forecast experiment, Enomoto (2019) showed the role of Prapiroon in the intensification of the Baiu frontal zone during HR18. We argue that the persistence of the low-pressure system over the ECS also played an important role in maintaining the lower-tropospheric southerly moisture flux during HR18. These results indicate that the development of active convection at 20°N, 140°E in late June 2018 is closely related to both the persistence of the low-pressure system over the ECS and the formation of typhoon Parpiroon. A possible mechanism for the development of this active convection is discussed in Section 4.
Then, we focus on the upper-tropospheric trough over the KP during HR18, which is much sharper than that in TP10. Composite maps of the time-filtered 360-K PV anomalies (left panels of Fig. 11) indicate that the contribution of lower-frequency variability to the development of the trough over the KP is larger than that of higher frequency variability. The amplitude of the positive PV anomalies filtered using the 25 day LPF (~ 1.25 PVU) is ~ 1.5 times larger than those filtered using the 8 -25 day BPF. However, positive PV anomalies filtered using the 25 day LPF is not   evident over the KP during HR18 (right panels of Fig.  11), which indicates the importance of variabilities with periods shorter than 25 days. Particularly, during HR18, positive PV anomalies filtered using the 8 day HPF are much larger than those in TP10 (compare Figs. 11a and 11d). Comparing the maximum positive PV anomalies over the KP during HR18 with those of the other heavy rainfall events over western Japan that exceed the 95th percentile (Fig. 12), it is clear that PV anomalies filtered using the 8 day HPF over the KP during HR18 are particularly high. This result indicates that the predominance of higher frequency variability caused the development of the sharp upper-tropospheric trough over the KP associated with the concentration of strong mid-tropospheric upwelling anomalies over western Japan during HR18. This is discussed further in Section 4.

Additional effects of the strong RWPP along the STJ during June 2018
We have shown that the remarkable RWPP observed in late June 2018 strengthened the upper-tropospheric ridge to the east of Japan and consequently intensified the surface high-pressure system to the southeast of Japan. We discuss additional effects of this strong RWPP in this section. The strong RWPP along the STJ caused a wave breaking around the Date Line and the consequential evident penetration of positive PV anomalies toward the subtropical region to the south of Japan (Fig. 13a). These positive PV anomalies were accompanied by negative potential temperature anomalies at the dynamical tropopause (Fig. 13b). Vertical and longitudinal distribution of the square of Brunt-Vaisälä frequency anomaly shown in Figs. 13c and 13d indicate that the upper-tropospheric cold temperature leads to thermodynamically unstable atmospheric conditions and activates convection around 20°N, 140°E (Fig. 13e), which propagated from the east. Although we calculate the divergence of 500 hPa Q-vectors over this region to examine if dynamically induced ascent due to positive PV intrusion exists, a clear convergence of the Q-vectors is not found over the activated convection (not shown). These results indicate that thermodynamic instability made the primary contribution to the further development of convective activity around 20°N, 140°E during HR18. As discussed in Section 3.2, active convection propagated northwestward toward the ECS during HR18 and was accompanied by a surface low-pressure system that  remained over the ECS and played an important role in the persistent southwesterly moisture flux in the lower troposphere. The remarkable RWPP along the STJ in late June is thus one of the essential factors for the occurrence of HR18.
4.2 Possible mechanism for the formation of a sharp upper-tropospheric trough over the KP In Section 3.2, we found that PV anomalies filtered using an 8 day HPF over the KP during HR18, which are much larger than those in TP10, contributed to the development of a sharp upper-tropospheric trough over the KP. In this section, we describe the development of this upper-tropospheric trough over the KP and discuss a possible mechanism for the higher frequency variability in this region.
The time evolution of 360 K PV anomalies around Japan from 12 UTC 5 July, 2018, to 06 UTC 7 July, 2018, is shown in Fig. 14. At 12 UTC 5 July, 2018 (Fig. 14a), southward penetration of positive PV anomalies toward KP is found in association with weak RWPP along the STJ (see also Fig. 7d). The longitudinal horizontal scale of the positive PV anomalies gradually decreases over KP from 12 UTC 5 July to 06 UTC 7 July . In other words, high PV anomalies are stagnant over the KP; conversely, low PV anomalies over northern China gradually move eastward. Regarding the 90 day LPF 200 hPa zonal wind field as a basic flow (Fig. 15a), the STJ is located over northern China and is accompanied by regional maximum zonal winds > 30 m s −1 from 80°E to 100°E and slower winds (< 20 m s −1 ) in the region 100°E to 120°E. The zonal winds are even weaker over the KP. Thus, the region from northern China to the KP can be considered as one of the exit regions of the STJ, where Rossby waves tend to be stagnant and amplified (Shutts 1983;Nakamura and Huang 2017). Therefore, we compare the longitudinal gradient of 90 day LPF 200 hPa zonal winds in the region among the heavy rainfall events over western Japan (Fig. 15b). We find that the deceleration of zonal wind during HR18 is larger around 115°E than it is during other events. These results indicate that the basic flow in the STJ exit region during HR18 leads to the stagnation and amplification of Rossby waves. As discussed in Section 3.1, during summer 2018, the seasonal mean upper-tropospheric geopotential height anomalies are positive in the mid-latitudes of the Northern Hemisphere, particularly over northern China in association with the several extreme heat events (Kobayashi and Ishikawa 2019). Such seasonally scaled positive upper-tropospheric geopotential height anomalies    over northern China can contribute to the enhanced diffluence and deceleration of the basic flow near the STJ exit region. This indicates that there exists the possibility of relationships between the seasonally scaled anomalous circulation over north China and HR18. This issue should be further investigated.

Summary and conclusions
In order to examine statistical large-scale atmospheric characteristics during the past heavy rainfall events that occurred widely in western Japan since 1979, we conducted a composite analysis of atmospheric fields. The results show that during these heavy rainfall events, the atmospheric fields are characterized by the upper-tropospheric trough over the KP, the upper-tropospheric ridge to the east of Japan, the surface high-pressure system to the southeast of Japan, and lower-tropospheric southwesterly moisture flux. Results of the composite analysis also indicate that clear RWPP along the PFJ over Siberia tends to occur just before the stronger heavy rainfall events, such as those in TP10, and contributes to the enhanced upper-tropospheric trough and ridge around Japan.
Further analysis considering various time scale variabilities reveals that surface high-pressure anomalies to the southeast of Japan, which are generally enhanced by the RWPP along the PFJ, are dominated by variability with a 25 -90 day period. These are also likely enhanced by convective activity around the Philippines associated with the northward migration of the active phase of the BSISO. However, variability with an 8 -25 day period dominates lower-pressure anomalies over the ECS in relation to the development of the upper-tropospheric trough around the KP.
We also investigated the atmospheric fields during HR18. The atmospheric features during HR18 are generally similar to those found among other heavy rainfall events. The RWPP along the PFJ enhances the surface high-pressure system to the southeast of Japan for both HR18 and TP10. Notably, the surface high-pressure systems to the southeast of Japan were dominated by 25 -90 day period variabilities for both HR18 and TP10. During HR18, in addition to the RWPP along the PFJ, a remarkable RWPP occurred along the STJ in late June 2018 that intensified the surface high-pressure system to the southeast of Japan.
We further discussed another effect of this remarkable RWPP in late June along the STJ. Results of our analysis indicate that the low-pressure system with an 8 -25 day period develops to the south of Japan in association with wave breaking induced by the remarkable RWPP in late June along the STJ. This wave breaking leads to the southward penetration of positive PV anomalies accompanied by negative potential temperature anomalies in the tropopause. This leads to thermodynamically unstable atmospheric conditions and activates convection around 20°N, 140°E, which then propagates northwestward toward the ECS accompanied by the surface low-pressure system just before HR18. Consequently, the simultaneous development of both the high-pressure system to the southeast of Japan and the low-pressure system over the ECS contributed to the extreme southerly moisture flux into western Japan.
During HR18, the sharp upper-tropospheric trough was observed over the KP. We found that high frequency variability with a period shorter than 8 days is predominant in this trough. We discussed the mechanism for the predominance of higher frequency variability over the KP, comparing the longitudinal gradient of 200 hPa zonal winds filtered using a 90 day LPF from northern China to the KP among the heavy rainfall events over western Japan. We found that during HR18, the significant deceleration of the basic flow around 115°E compared with those in other events contributed to the stagnation and amplification of Rossby waves.
Finally, as described in Section 3.2, we found that the direct contribution of the BSISO to the development of the surface high-pressure system to the southeast of Japan during HR18 is less than during the other analyzed events because the active phase of the BSISO is situated south of 10°N and far from the surface high-pressure system to the southeast of Japan. However, in mid-June 2018, the northward migration of the amplified active phase of the BSISO was clearly observed (not shown). The role of intraseasonal variability in the excitation of quasi-stationary Rossby waves that propagate in the mid-latitudes of the Northern Hemisphere should be further investigated. Additionally, it is still unclear how such a remarkable RWPP along the STJ in late June was excited. The mechanisms driving extreme events, including heavy rainfall and heat waves, around Japan also require further consideration.

Supplements
Scatter diagrams of regionally averaged (31.25 -35°N, 130 -135°E) 3 day mean anomaly fields of water vapor flux divergence in the lower troposphere for days representing 3 day precipitation peaks during the heavy rainfall events over western Japan that exceed the 95th percentile are shown in Fig. S1. The decomposition of water vapor flux is based on Eq. (2) in Sekizawa et al. (2019). The red and orange circles represent HR18 and TP10, respectively. Figure S2 is as in Fig. S1 but for regionally averaged (30 -45°N, 60 -120°E) zonal components of 200 hPa WAF 8 days before 3 day precipitation peaks during heavy rainfall events over western Japan that exceed the 95th percentile. Figure S3 is as in the left panels of Fig. 8 but for (left) MD10 and (right) LW10. Figure S4 is as in Figs. 9g and 9o but for from 12 days to 3 days before 3 day precipitation peaks in TP10 (a -d) and HR18 (e -h). The contour interval is 5 W m −2 .