Long-term precipitation changes in the Baiu and 2 Akisame seasons in Japan over the past 120 years 3 (1901–2020)

30 Long-term variations in precipitation during the major rainy periods in Japan—the Baiu 31 (June–July) and Akisame (September–October) seasons—are investigated using 32 precipitation records from 44 weather stations in western to eastern Japan over the past 33 120 years (1901–2020). The total amount of Baiu precipitation has increased over the 34 1901–2020 period, mainly during the mid–late stages of the season (late June–July) over 35 regions on the Sea of Japan side of the country. In contrast, the precipitation amount 36 during the Akisame season has decreased, mainly during the mid-stage (late September– 37 early October) over all regions. The frequency and intensity of heavy precipitation have 38 generally increased in both seasons, but the trends are much stronger for the Baiu season 39 compared to those for the Akisame season. A prominent positive trend, 23.5% per 100 40 years (18.1% per ℃ ), which is much higher than the Clausius–Clapeyron rate 41 (approximately 7% per ℃ ), is observed for the Sea of Japan side of western Japan for 42 the seasonal maximum 1-day precipitation total during the Baiu season. It may be


50
There are two major rainy periods in Japan. The first one appears in early summer (typically 51 June-July) and is called the "Baiu" season, which is identified as the period of peak 52 precipitation and minimum sunshine duration and is especially evident in southwestern 53 Japan (Ninomiya and Murakami 1987). The Baiu season is associated with the seasonal 54 northward migration of the Meiyu-Baiu rainband, which extends zonally from eastern China 55 to southern Japan and is accompanied by a frontal zone (Ninomiya and Murakami 1987;56 Wang and LinHo 2002). The second one is recognized in early autumn (typically 57 September-October) and called the "Akisame" or "Shurin" season (referred to hereafter 58 "Akisame"), which is associated with the southward migration of a frontal zone (Matsumoto 59 1988). Akisame precipitation is generally weaker and more intermittent than that during the 60 Baiu season; however, the total precipitation amount is greater in eastern Japan than that 61 during the Baiu season (Sekiguchi and Tamiya 1968;Kato 1997). Most of the annual 62 precipitation falls during the two rainy seasons, making these seasons vital for water 63 resource management; however, heavy precipitation events frequently occur during the 64 rainy seasons, resulting in disasters such as floods and mudslides. Therefore, their long-65 term variations and changes are one of the major concerns. 66 On a broader perspective, the Baiu season is a phenomenon associated with the East 67 Asian monsoon, which is one component of the Asian monsoon system (Wang and LinHo 68   2002). The Meiyu-Baiu rainband that is responsible for Baiu precipitation is maintained by 69 3 moisture transport with low-level southerly monsoonal flows, driven by the zonal pressure 70 difference between the warmer Asian continent and the relatively cooler Pacific Ocean 71 (Kodama 1993;Kawamura and Murakami 1998;Wang and LinHo 2002). The rainband is 72 accompanied by a quasi-stationary front that is characterized by the strong meridional  Ongoing global warming has influenced spatial and temporal precipitation distribution. As 91 atmospheric moisture content increases with temperature, precipitation intensity can be 92 expected to have increased at short-time scales. In Japan, the intensity and frequency of . Nevertheless, it is known that the spatial pattern for changes in total precipitation on 99 a broad scale roughly follows a "wet-gets-wetter" pattern (Held and Soden 2006) owing to 100 the thermodynamical change. Given this rough approximation, precipitation amount during 101 the rainy seasons is anticipated to increase as the climate warms.

102
Long-term variations and trends of Baiu precipitation have been investigated using 103 precipitation observations. Misumi (1994) found inter-decadal variations of Baiu precipitation 104 (June-July), with a wetter period in 1924-1944 and a drier period in 1952-1972, especially 105 in southwestern Japan. Endo (2011) analyzed precipitation records for western and eastern 106 Japan from 1901 to 2009 and showed that precipitation over the Sea of Japan side of the 107 country had decreased significantly in the early phase (early-mid-June), while it had 108 increased significantly in the late phase (mid-late July) with amplification of year-to-year 109 5 variability. Otani and Kato (2015) reported a recent decrease in precipitation in northwestern 110 Kyushu in late June by comparing precipitation data for 1971-2000 with data for 2001-2010. 111 They attributed the decline in rainfall to a reduction in heavy precipitation events (>50 mm/d). and showed a significant increase in the frequency of heavy rainfall events (>130 mm/3hr) 118 during the Baiu season (June-July), especially in July. Based on precipitation estimates from 119 satellite radar observations, Takahashi and Fujinami (2021) showed that the frequency of 120 heavy precipitation events (>10 mm/h) during the Baiu season (mid-June-mid-July) 121 increased by 24% between 1998-2008 and 2009-2019. 122 In contrast to these insights into the Baiu precipitation, long-term precipitation variations 123 during the Akisame season are poorly understood. The existing literature is limited to the 124 studies of Oguchi and Fujibe (2012) and Duan et al. (2015), who examined precipitation 125 data from 1901 at seasonal and regional scales in Japan. They identified decreasing trends 126 in precipitation amount during fall (September to November), with a statistical significance 127 at some weather stations. Therefore, it is necessary to investigate the long-term variations 128 for the Akisame season. Comparison of the results with those in Baiu precipitation could 129 6 help us further understand the characteristics of the two rainy seasons.

130
This study investigates long-term variations in precipitation for the Baiu and Akisame 131 seasons, as an extension of the work by Endo (2011), by extending the analysis period to 132 recent years and adding an analysis for extreme precipitation and for the Akisame season, 133 and focuses on exploring similarities and differences in the long-term changes between the 134 two rainy seasons. Seasonal variation in precipitation during the warm season in Japan is 135 characterized by two rainy periods that are separated by a break spell, which are caused by 136 the migration of the Baiu/Akisame frontal zones and development of the WPSH, in 137 association with seasonal evolution of the East Asian monsoon, resulting in a well-defined 138 wet/dry cycle in the climatology (Chen et al. 2004;Ding 2007;Inoue and Matsumoto 2007).

139
Considering the climatological feature, our analysis includes the beak spell (August), which 140 could provide some insight on the mechanism of the long-term variations. analysis was based on daily and 10-day or 11-day accumulated data (e.g., 1-10 July, 11-148 20 July, and 21-31 July), which were compiled and digitalized by the JMA. Few precipitation 149 7 records were missing for this period (e.g., Fujibe et al. 2006;Oguchi and Fujibe 2012). 150 In the analysis, the period from June to October was classified into three seasons: the Baiu 151 season (June-July) and the Akisame season (September-October) as the wet spell, and 152 the high summer season (August) as the relatively dry spell. The Baiu and Akisame seasons 153 are further divided into three stages: the early stage, the mid-stage, and the late stage, as 154 shown in Table 2. In addition, the mid-late stage is defined for the Baiu season by combining 155 the mid-stage and the late stage.  Table 1. We also defined five combined regions: eastern Japan (EJEP) comprises 162 EJ and EP, western Japan (WJWP) comprises WJ and WP; Sea of Japan side of eastern 163 and western Japan (EJWJ) comprises EJ and WJ; Pacific side of eastern and western Japan 164 (EPWP) comprises EP and WP; and eastern and western Japan (ALL) comprises EJ, EP, 165 WJ, and WP. Regional mean data were calculated where more than 80% of the station data 166 had a quality flag of normal or quasi-normal (JMA 2021).

167
Long-term changes were evaluated in two ways. In the first way, the climatology between 168 the first half of the 20th century      Next, long-term changes are investigated in another way. A linear trend was calculated for 204 1901-2020 using the least squares method. The results are summarized in Table 2, which   205 show the changes as mm (Table 2a) and % per 100 years (Table 2b) August, which led to shortening of the dry spell between the first and second rainy periods 225 ("Changma" and "second Changma", respectively). Interestingly, this observed feature in 226 Korea is consistent with the long-term trends for WJ and EJ in our study ( Fig. 2 and Table   227 2). Lee et al. (2017)

Heavy precipitation 231
This section examines long-term changes in heavy precipitation, where it is defined as 232 precipitation over 100 mm per day (R100mm). Figure 3 compare the frequency of heavy 233 precipitation between the two epochs (1901-1950 and 2001-2020). For the Baiu season, 234 there is a great increase in the frequency of R100mm events for all the regions except EP.

235
The most prominent increase is found for WJ, where the frequency increases significantly 236 in early-mid July and the timing of the peak shifts from late June to early July. No significant 237 change is observed during the high summer season except late August for WJ. As for the In this section, we focus on the differences between the long-term changes identified for 260 the Baiu and Akisame seasons. Table 4 summarizes regionally averaged long-term trends 261 for surface-air temperature (SAT), precipitation amount, and the intensity and frequency of 262 heavy precipitation. Here, in addition to the frequency of R100mm events described in the 263 previous section, the intensity of heavy precipitation is evaluated using the seasonal 264 maximum 1-day precipitation total (Rx1d), which is defined as the maximum 1-day 265 precipitation total during the Baiu season (June to July) or the Akisame season (September 266 to October) for each year. The long-term SAT trends were calculated using data from 267 selected weather stations to avoid the effects of urbanization, following JMA (2022), as 268 shown in Table 1. Two stations, Miyazaki (47830) and Iida (47637), were relocated in May

272
The long-term SAT trends show significant increases for both seasons for all regions, with 273 a slightly larger increase in the Akisame season (Table 4). On the other hand, the long-term 274 trends for precipitation amount in the Baiu and Akisame seasons are opposing, with large 275 regional differences. For example, the trend for ALL is +5.2%/century (+4.2%/℃) for the Baiu 276 season, but it is −6.3%/century (−4.3%/℃) for the Akisame season. The seasonal contrast 277 is the most prominent for WJ, where the trend is +12.0%/century (+9.2%/℃) for the Baiu 278 season and −14.1%/century (−9.6%/℃) for the Akisame season. Negative trends prevail for 279 EP for both seasons, but the statistical significances of these are low.

280
Rx1d and R100mm frequency have increased overall for both seasons, but the increase 281 rates are much higher for the Baiu season than those for the Akisame season. For example, 282 the trend for Rx1d for ALL is +10.4%/century (+8.4%/℃) for the Baiu season, while it is 283 +6.4%/century (+4.4%/℃) for the Akisame season. The former rate is close to the increase 284 expected from the Clausius-Clapeyron (C-C) relationship (approximately +7%/℃). There 285 are substantial regional variations, including in the high rates for Rx1d: +23.5%/century 286 (+18.1%/℃) for WJ and +13.2%/century (+12.7%/℃) for EJ during the Baiu season, and 287 +15.5%/century (+12.0%/℃) in EP during the Akisame season. These rates of increase 288 greatly exceed the C-C rate, suggesting that there is some dynamical enhancement. This pattern may be caused by an enhancement of south-westerly moisture flows associated 294 with the East Asian monsoon and its interaction with topography. TC activity may also 295 contribute to the pattern for the high summer season, as noted in Section 3. However, the 296 spatial pattern for the Akisame season is markedly different from those of the previous two 297 seasons, suggesting that different mechanisms may be important. This matter will be 298 discussed in Section 6. It is also noted that the patterns for precipitation amount and Rx1d as also found by Misumi (1994). Precipitation amount in the Akisame season has gradually 310 decreased and its year-to-year variability has increased, resulting in a more frequent 311 occurrence of small precipitation years in the recent decades.

312
The time series for Rx1d (Fig. 6) and R100mm frequency (Fig. 7) also show seasonal and 313 regional differences. The interannual variability for the Akisame season tends to be greater

319
There is an interdecadal shift in the magnitude of the interannual variability for the Akisame 320 season, including considerably more variability in WJ during the first half of the 20th century 321 and in WP and EP during recent decades relative to other periods. 322 323 6. Summary and discussion 324 We investigated long-term variations in precipitation during the major rainy periods in Japan  This matter should be pursued in a further study.

357
It would be meaningful to discuss whether these observed trends are influenced by         Oguchi, S., and F. Fujibe, 2012: Seasonal and regional features of long-term precipitation   Table 1. List of weather stations whose data were used for precipitation analysis. The   Table 3. As Table 2, but for the number of days with precipitation over 100 mm (R100mm).