1 Introduction

The East Asia monsoon rain belt is the most significant manifestation of the East Asia summer monsoon (EASM) and shifts with it. From June to July, it extends from the Yangtze-Huai River Valley of China to south-central Japan and southern South Korea in a quasi-stationary state (Fig. 1), which is called Meiyu rain belt in China. The location, range, and stagnant period of the Meiyu rain belt significantly influence the precipitation pattern of the East Asian monsoon region in summer, thus affecting agricultural production, residents’ lives, and the occurrence of drought and flood in these regions (Si et al. 2009; Murata 1987; Lee et al. 2010; Lee et al. 2017). Therefore, changes to the Meiyu rain belt are one of the core aspects of studies on East Asian monsoon climate and have important scientific value and practical significance.

Fig. 1
figure 1

Location map of the Meiyu belt (Ding et al. 2007), representative stations of Meiyu, and relevant data sites in central Hunan Province (top left)

Full size image

From 1970 on, the Meiyu rain belt underwent prominent interdecadal shifts. In the late 1970s, the precipitation of Meiyu rain belt obviously increased (Xu 2007). The Meiyu rain belt center appeared in the Yangtze-Huai River valleys in the 1980s and moved south to the Yangtze River and its southern region in the early 1990s (Xu and Jiang 2004; Si et al. 2009) (Fig. 3e). Since 2000, the precipitation of Meiyu rain belt weakened, with the rain belt center moving north to the HRV (Hu and Ding 2009; Si et al. 2009; Zhu et al. 2011). (Fig. 3f). Correspondingly, precipitation pattern in summer over East Asia has changed significantly: the main rain belt of eastern China moved from the Yellow River Valley to the Yangtze River Valley in the late 1970s (Gong and Ho 2002), afterwards transformed to the area between the Yellow River and the Huai River in the beginning of the twenty-first century (Zhao et al. 2008; Wei et al. 2012; Sun and Wang 2015). Coincidentally, the high precipitation region in summer which appeared in southern Japan from the 1980s to 1990s shifted to the Korean Peninsula after 2000 (Si and Ding 2013; Choi et al. 2020).

Due to instrument-measured data limitations, current studies have focused only on Meiyu belt activities over the last 50–60 years, so that long-term studies of rain belt shift have been severely limited. There are more researches to be done on the changes and its driving mechanisms of the Meiyu rain belt from the interdecadal to the centennial.

Existing, rich documentation in East Asia can effectively expand data sources, so that Meiyu variations can be discussed on a longer time scale. Zhang and Wang (1991) discussed the reconstruction and changing trends of the Meiyu in the eighteenth century in the lower reaches of the YRV using the daily rain record called “Qing-Yu-Lu” (Clear and Rain Records) kept by the Qing court. Ge et al. (2008) reconstructed the Meiyu sequence in the YRV from 1736 to 2000 using the rain record called “Yu-Xue-Fen-Cun” (Rain and Snowfall Archives) reported by local government of the Qing Dynasty, and shifts of the summer main rain belt in eastern China were discussed accordingly. Man et al. (2007) reconstructed the Meiyu sequence of Wuhan and Changsha from 1867 to 1872 using diary-recorded data. These studies have extended the study period and reconstructed the variation in Meiyu characteristics in the typical Meiyu area by mining high-resolution weather information from ancient documents. In addition to China, Japanese scholars also used diary-recorded weather data from ancient times to reconstruct the Meiyu in Japan (Murata 1987, 1993). Wang et al. (2007) used the “Chukwookee” data to discuss the changes in summer precipitation in Seoul over the past 200 years. However, there has been little discussion on the interdecadal-centennial shift of the Meiyu belt.

Changsha (112.6°E, 28.1°N) of Hunan Province is located at the southern margin of the Meiyu area in the Yangtze-Huai River Valley, where there is a precipitation concentration period in early summer. It is a debatable whether this precipitation period can be regarded as Meiyu (Liao and Zhang 2018), because in some years, the precipitation period persists long (very short) like (not like) Meiyu. Whether there is a typical Meiyu in this area would be a significant indicator for a southern margin of the Meiyu belt. Xiao et al. (2008) used to reconstruct the early summer precipitation of this region and speculated on the location of the Meiyu belt in the second half of the nineteenth century. Therefore, this paper attempts to use multiple diary-recorded data and the modern instrument measured data to reconstruct the Meiyu changes in Changsha since 1861. By discussing whether there is a typical Meiyu on the interdecadal period, the north-south shifts in the southern margin of the Meiyu belt can be inferred. The change in location of the Meiyu belt in eastern China since the mid-nineteenth century and its relationships with the evolution of a summer precipitation pattern in eastern China were discussed along with different time scales in combination with variation in Meiyu characteristics in the YRV and HRV.

2 Data and methods

2.1 Data

In this study, 2 types of precipitation data were used, including continuous diary-recorded data from 1861 to 1929 and continuous instrument-measured data after 1911. The period from 1911 to 1929 is therefore covered by both diary-recorded data and modern instrument-measured data.

The significance of diaries for high-resolution climate research has been proven by relevant studies (Zhu 2004; Man et al. 2007). Seven diaries centered on Changsha shown in Table 1 screened in this study (see 2.2.2 for the screening method). When Changsha data was insufficient, diary-recorded weather data of Pingjiang, Xiangtan, Xiangxiang, Shuangfeng, Xiangyin, Hengyang, and other sites were used as a supplement, as shown in Fig. 1. Comparing the daily precipitation of these stations with that of Changsha from May to August (1970–2000), the days with the same precipitation accounted for more than 70% of the total number of days.

Table 1 An overview of the diary-recorded data in Changsha and its nearby regions from 1861 to 1929
Full size table

Modern weather data before 1950 was of 2 types. One is the Hydrological data of the Yangtze River Valley (1880–1949) (Changjiang Water Resources Commission of the Ministry of Water Resources 1961) and the other is the Daily Precipitation and Evaporation of Hunan Province before Liberation (1933–1950) (Bureau 1961), which includes data from the Agricultural Weather Station of Hunan Department of Construction during the Republic of China. Modern instrument-measured data after 1951 includes Changsha and Hengyang precipitation data, which are from the China surface climate daily dataset of the International Exchange Station of the CMA National Meteorological Center.

PDO of SST index was obtained from https://www.ncdc.noaa.gov/teleconnections/pdo/, and AMO data was obtained from http://www.esrl.noaa.gov/psd/data/timeseries/AMO/.

2.2 Methods

2.2.1 Criteria for identification of the Meiyu period

Presently, reconstructing the characteristics of the Meiyu from historical and literature data is mainly based on the rainy days in early summer, and the criteria is slightly different according to varying data resolution (Zhang and Wang 1991; Ge et al. 2008). Murata (1987) used diaries to reconstruct the Meiyu sequence in Japanese history, which fully considered the daytime precipitation characteristics elaborately recorded in the diary. Since nighttime precipitation was often ignored (by comparing the weather information in different diaries on the same day and at the same place), the number of daytime precipitation days has been used to determine the starting and ending date of the Meiyu period. For modern precipitation observational records, daytime precipitation is bounded by 8:00 a.m. and 8:00 p.m.

The identification of Meiyu period is as follows: (1) from May 20, if there are 5 days of precipitation (at least 3 days of precipitation >1 mm) or 4 days of precipitation (daily precipitation >1 mm) in 10 consecutive days, it can be considered as a precipitation concentration period. Multiple precipitation concentration periods are allowed. The first (last) rainy day of the first (last) precipitation concentration period is regarded as the starting (ending) of Meiyu period. (2) Referring to other research (Ge et al. 2008), the precipitation concentration period which ends before June 10 or starts after July 10 is not classified as Meiyu Period. (3) If a precipitation concentration matching the above conditions starts in May and there are two or more days without precipitation before June, the date after the rainless days should be set as the starting of Meiyu period. (4) If there is no precipitation concentration period that meets the requirements, it will be considered as no Meiyu period (empty Meiyu).

As the instrument-measured precipitation data from 1930 to 1948 was collected during the day and night, the criteria for identification of a Meiyu period was adjusted, and set as precipitation for at least 5 days within 10 days, of which precipitation should be >1 mm.

2.2.2 Processing of diary-recorded data and identification of a Meiyu period

After screening, it can be found that six diaries, such as Guo Songtao’s Diary, had complete records of daily precipitation >0.1 mm, which met the criteria set in Section 2.2.1 above. However, for the period 1892 to 1906, the Xiangqilou Diary was chosen due to the lack of data.

Daily diary precipitation data levels were parameterized as follows: “rain,” “heavy rain,” and “thunderstorm” were categorized as 2, corresponding to precipitation above 1 mm; “drizzle,” “shower,” “light rain,” and “sprinkle” were 1, corresponding to the precipitation below 1 mm; and “cloudy,” “sunny,” and “light drizzle” were 0. The Meiyu period of the daily precipitation level data in each diary was divided according to the criteria discussed in Section 2.2.1. For the period using the Xiangqilou Diary, 3 days of “2” within 10 days would be sufficient. The number of daytime precipitation days from the third pentad of May to the sixth pentad of July was calculated according to the parameterized daily precipitation information. The correlation coefficient of the mean values of daytime precipitation days of every pentad during 1861–1933 and 1933–2017 was as high as 0.92, exceeding the significance level of 0.01.The example of parameterized dairy records and identified Meiyu period is shown in the table (see online resource 1).

2.3 Verification of methods

Zhao et al. (2017) used the data of 20 sites in north and central Hunan including Changsha to establish the starting and ending sequence of the Meiyu in Hunan from 1960 to 2014 (see online resource 2). Compared with Meiyu starting and ending dates identified by methods above, except 14 years of empty Meiyu, there were 31 (75.6%) and 28 (68.2%) years, respectively, with a difference within 5 days in the starting and ending dates. The Meiyu period duration of the two sequences was significantly correlated with a significance level of 0.01. Among 10 years with a difference of more than 5 days on the Meiyu starting dates, 9 of 10 yeas of Changsha were earlier, while amang 12 years ending dates,10 of 12 years earlier, which is consistent with Changsha being on the southern margin of the Meiyu area in the middle and north of Hunan and indicates that the selected method is scientifically to identify Meiyu characteristics.

There were two types of climatological data from 1911 to 1929: instrument-measured data and diary-recorded data. Comparing the Meiyu periods identified by the two types of data, it can be found that during the 19 years, 3 years were defined as empty Meiyu (see online resource 3). In the remaining 16 years, the difference between the starting and ending days of Meiyu within 3 days was 81.2% and 62.5%, while the difference between the starting and ending days of Meiyu within 5 days was 93.8% and 81.2%. This shows that the diary-recorded data is sufficient to meet the needs of the research.

3 Meiyu characteristic value and the shift of the southern margin of the Meiyu belt in Changsha from 1861 to 2017

3.1 Meiyu period in Changsha from 1861 to 2017

As mentioned in Section 2.2, the starting and ending dates and the length of the Meiyu sequences in Changsha from 1861 to 2017 were reconstructed (Fig. 2). Meiyu started on May 22 at the earliest and July 10 at the latest, with an average date of June 11. It ended on June 11 at the earliest and July 21 at the latest, with an average of June 1. The longest Meiyu period was 55 days, and the shortest was 4 days (excluding years of empty Meiyu), with an average of 17.2 days. There were significant discrepancies between the earliest and the latest days in both Meiyu starting and ending days; at the same time, there was a huge difference between the longest and the shortest Meiyu periods.

Fig. 2
figure 2

Meiyu characteristics from 1861 to 2017. (a) Starting dates, (b) ending dates, and (c) length of Meiyu; heavy line: 9 years running mean value

Full size image

Changsha Meiyu had significant interdecadal variability, starting earlier from 1861 to 1890, 1921 to 1940, and 1981 to 2017 and later from 1891 to 1820, and 1941 to 1980. It ended earlier before 1910 and after 1971 and later from 1911 to 1970. The Meiyu period was longer from 1861 to 1870, 1881 to 1890, 1901 to 1950, and 1991 to 2000 and shorter from 1871 to 1880, 1891 to 1900, 1951 to 1990, and 2000 to 2017.

According to Meiyu sequence in the YRV statistics from 1736 to 2000 reconstructed by Ge et al. (2008), the minimum average Meiyu period in any 20 years from 1861 to 2000 was 17.1 days. For the entire period, the years when Meiyu length was greater than or equal to 17 accounted for 70%. Therefore, 17 days was the lowest value of typical Meiyu and adopted as the standard value to judge whether a typical Meiyu period occurred in Changsha from 1861 to 2017. The average Meiyu length in Changsha was 17.2 days, which was higher than the minimum standard of 17 days, confirming that a typical Meiyu period occurred in Changsha on a century scale.

3.2 Variations in the southern margin of the Meiyu belt from 1861 to 2017

The length of the Meiyu period in Changsha is used to determine if it was a typical Meiyu or not and to discuss the north-south shift of the southern margin of the Meiyu Belt. Here R are used to indicate the Meiyu intensity. The annual length of Meiyu (A) in Changsha is compared with the minimum standard value of typical Meiyu (17 days). When A-17 ≥ 0, R is set as “+ 1,” and when A-17 < 0, R is “− 1.”

Year of absence was evaluated according to the description of summer precipitation in Changsha in A Compendium of Chinese Meteorological Records of the Last 3000 Years (Zhang 2004) (the year of 1863), The Chronicle of Famine in Modern China, and Continuation of the Chronicle of Famine in Modern China (Li et al. 1990; Li et al. 1993) (the year of 1880,1904, 1945,1949), and “− 1” was added in case of drought, and “+ 1” was added in case of summer flood. If the information was insufficient, or there was conflicting information, absence was reserved. There were 4 years of absence in total (1865, 1866, 1879, 1909). When ∑ R ≥ 0 in a period, it indicates a typical Meiyu; when ∑R < 0 in the period, it implies that there is no typical Meiyu. Except for the last period, the minimum for other periods was not less than 9 years. When a period in Changsha has typical characteristics of Meiyu, it is believed that the southern margin of Meiyu belt is in the south of Changsha; otherwise, it is in the north. The interdecadal variation of typical and atypical Meiyu is shown in Table 2.

Table 2 Typical and atypical signals of interdecadal Meiyu in Changsha from 1861 to 2017
Full size table

Typical and atypical Meiyu in Changsha appeared alternately on the interdecadal scale from 1861 to 2017, indicating an interdecadal north-south shift in the southern margin of the Meiyu rain belt (Table 2). From the multidecadal change perspective, the southern margin of the rain belt was mainly south of Changsha from 1871 to 1915, in the north from 1916 to 1987, and back in the south from 1988 to 2017.

Based on an analysis of instrument-measured data from Hengyang to Changsha since 1933 (see online resource 4), we found that typical Meiyu signals in both areas were identical. In view of the latitudinal difference between Changsha and Hengyang, it is believed that the southern margin of the rain belt stretched from north of Changsha to South of Hengyang in central Hunan (up to 2 degrees of latitude).

3.3 South-north shift of Meiyu rain belt from 1933 to 2017

To further determine the relationship between the variation of the southern margin of the rain belt and the shift of the entire Meiyu belt, the rain belt was divided into three sections from north to south, namely, HRV (31–34 °N), YRV (28.5–32 °N), and the Changsha-Hengyang area (CSHY, 26.5–29.5 °N) (Fig. 3). Meiyu length was used to indicate its intensity, thus reflecting shift of the whole rain belt.

Fig. 3
figure 3

Meiyu rain belt change maps from 1933 to 2017. Dark gray, light gray, and blank fields, respectively, represent the center, the general typical, and the atypical Meiyu area of the Meiyu belt

Full size image

Anomalies in the length of Meiyu (R) are used to represent Meiyu intensity in YRV and HRV, where a positive value is set as “+1,” a negative value as “− 1,” and a missing record as a “0.” The periodic Meiyu intensity is judged by ∑R, which is stronger when ∑R≥0 as indicated by “+,” and weak when ∑R <0, as indicated by “−.” Data on the Meiyu period in the YRV were from the research of Ge et al. (2008), and the sequence was extended until 2017. Data on the Meiyu period in the HRV were from the study of Xu and Zhang (2007), where the HRV sequence was extended forward to 1933 and backwards to 2017 using the identification method of Ge et al. (2008) based on single-site precipitation data of Dongtai station (the reconstructed characteristic values from 1953 to 2004 are consistent with 92.3% of the original sequence anomaly symbols of Xu and Zhang’s research). The Meiyu mean length in the YRV (HRV) is 24 days (26 days) according to Ge et al. (2008) and Xu and Zhang (2007). The periodic variation of Meiyu intensity in the three regions are shown in Table 3.

Table 3 Periodic Meiyu intensity signal in the CSHY, the YRV, and the HRV from 1861 to 2017
Full size table

To further understand the north-south shift of the rain belt, the relationship between the location of the rain belt center and the rain belt shift should be further considered. If the periodic Meiyu intensity signal is “+” (“−”) in the YRV or HRV region, then the region is considered to be (not be) the precipitation center. If the mean length of Meiyu in a period in CSHY is equal to or exceeds the YRV’s multi-year average value (24d), it would also be regarded as the rain belt center. The rain belt center is considered to be northward when it is in the HRV, southward when in the YRV or in the YRV and CSHY (the rain belt center sometimes could cover both two areas), and abnormally southward when in the CSHY.

The shifts of the Meiyu rain belt center and the southern margin are used to present the north-south movement of the Meiyu belt. The rain belt center is represented by dark gray, and the general precipitation area of typical Meiyu area is represented by light gray. Meiyu belt shifts are shown in Fig. 3. From 1933 to 1943, the Meiyu in the Yangtze-Huai River Valley was weak on the whole, and the CSHY Meiyu period mean length was higher than 24. Thus, the rain belt center was abnormally southward, and the southern margin of the rain belt was south of Hengyang (Fig. 3a). From 1944 to 1954, the Meiyu in the HRV was strong, but weak in the YRV, indicating that the rain belt center was northward, while the Meiyu in CSHY was atypical, meaning the southern margin of the rain belt was north of Changsha (Fig. 3b). From 1955 to 1966, the Meiyu in Yangtze-Huai River Valley was weak overall, and the southern margin of the rain belt was in the north of Changsha (Fig. 3c). From 1967 to 1978, the rain belt center was in the south, and the southern margin of the rain belt was north of Changsha (Fig. 3d). From 1979 to 1990, the Meiyu in the Yangtze-Huai River Valley was heavy on the whole, and the southern margin of the rain belt was north of Changsha (Fig. 3e). From 1991 to 1999, the rain belt center was in the south, and the southern margin of the rain belt was south of Hengyang (Fig. 3f). From 2000 to 2009, the rain belt center was in the north, and the southern margin of the rain belt was north of Changsha (Fig. 3g), which were consistent with the conclusions of existing studies (Hu and Ding 2009; Si et al. 2009). From 2010 to 2017, The Meiyu became strong in the YRV and CSHY, and weak in the HRV, indicating that both the rain belt center and its southern margin moved south (Fig. 3g).

The combinations of Meiyu signals in the YRV, HRV, and CSHY have altered in 8 times from 1933 to 2017. The number of “+” represents the width of the rain belt occupied by the heavier Meiyu, indicating the overall intensity of the Meiyu belt rainfall, and then the rain belt has 6 modes of “−−−,” “+−−,” “−−+,” “−+−,” “++−,” and “−++” from light to heavy. The first four are lighter, and the latter two are heavier. The rain belt is “overall weak,” “weak and abnormal southward,” “weak and northward,” “weak and southward,” “strong and southward,” and “overall strong,” respectively. In these eight periods, the typical Meiyu signals in CSHY are opposite to those in the HRV, indicating that variation in the position of the southern margin of the Meiyu belt is basically consistent with the variation of the whole Meiyu rain belt on the centennial scale. If the Meiyu intensity in the HRV from 1861 to 1932 is estimated based on the 6 signal modes from 1933 to 2017, variation in the belts position can be inferred. Thus, the rain belt was strong and southward from 1861 to 1870, 1881 to 1891, and 1901 to 1911; it was weak and northward or weak overall from 1871 to 1880, 1892 to 1900, and 1922 to 1932 and its position uncertain from 1912 to 1921.

From a multidecadal perspective, the position of the Meiyu belt can be divided into three stages from 1861 to 2017. From 1861 to 1921, both of the southern margin and the center of the rain belt were southward, and the belt was basically stronger and southward. From 1922 to 1978, the southern margin of the rain belt was basically northward, and its center frequently changed, either to the south or north, abnormally south, or was non-existent which indicates that the whole Meiyu belt was weak and that the main rain belt of eastern China deviated from the Yangtze-Huaihe Valley. From 1979 to 2017, the center was mainly southward, and after 1991, the southern edge was mainly southward, which indicates the whole rain belt was stronger and southward. It appears that the northward shift of the rain belt after 2000 may not be a long-term trend.

4 Discussion

4.1 Rain belt shift and variation of East Asian summer monsoon

Previous studies have shown that EASM intensity will significantly influence the position of the summer monsoon rain belt in eastern China (Guo et al. 2004; Lu et al. 2006). Therefore, we further analyzed the relationship between EASM and Meiyu belt variation (Fig. 4).

Fig. 4
figure 4

The spatial and temporal distribution characteristics of the Meiyu belts and the indices of AMO, PDO, and EASM. (a) AMO indices (1861–2017) (the bold line illustrates the 9-year moving averages), (b) PDO indices, (c) EASM indices (1873–2000, data from Guo et al. 2004), (d) the spatial and temporal distribution characteristics. For the YRV and HRV, years with solid lines (without lines) indicate that those were (were not) Meiyu belt centers, while for CSHY, years with solid lines (without lines) indicate that the southern margin of Meiyu belt were in south (north) of the region

Full size image

EASM intensity is closely related to whether the Meiyu precipitation center was in the YRV. Before 1920, EASM was weak, and the rain belt center was in the YRV from 1880 to 1921; from the 1920s to the end of the 1960s, EASM was strong, and the rain belt center was not in the YRV from 1922 to 1966; but since the 1970s, the summer monsoon has weakened, and the rain belt center was mainly in the YRV from 1967 to 2017. The correlation between EASM intensity and variation of Meiyu intensity at the southern margin of the rain belt is very high on a multidecadal scale. From 1880 to 1911, EASM was generally weak, and the southern margin of the rain belt was generally southward; from 1920 to 1970, EASM was strong, and the southern margin of the rain belt shifted rapidly northward; and from 1991 to 2017, EASM was weak, and the southern margin was southward. However, the rain belt southern margin showed frequent variability on an interdecadal scale, and its response to changes in EASM intensity seemed out of sync. When EASM strengthens (weakens), the southern margin of the rain belt moved rapidly (slowly) north (south). The relationship between EASM and the Meiyu intensity in the HRV was very unstable.

Moreover, EASM affects the north-south variation of China’s summer monsoon rain belt by affecting the YRV Meiyu intensity and is the main determinant for the variation in the position of the Meiyu belt. When EASM is weak, monsoon rain belt generally appears in the Yangtze-Huai River Valley (Ge et al. 2008).

To sum up, the characteristics of multidecadal-centennial variation of the Meiyu belt intensity and the southern margin of the belt are mainly controlled by the summer monsoon. However, the Meiyu belt variation on an interdecadal scale, especially the northern margin rain belt location, indicates that there are more influential factors determining rain belt shifts.

4.2 Relationship with PDO and AMO

Interdecadal precipitation pattern variations in eastern China are closely related to interdecadal variations in sea surface temperature (SST) (Wei et al. 2012; Gu et al. 2009; Qian et al. 2009). Among them, the Pacific Decadal Oscillation (PDO) and Atlantic Multidecadal Oscillation (AMO) have the most obvious effects. However, the interdecadal phase of PDO and AMO changed in the late 1970s (Deser et al. 2004; Sutton and Dong 2012), which plays the main role in summer precipitation pattern shifts in East Asia (Wang et al. 2009; Si et al. 2009; Si and Ding 2016; Zhu et al. 2011). We discuss the influence of these two SST models on Meiyu belt activities.

Studies have shown that when PDO or AMO is combined in different phases, their effects will strengthen or weaken each other, thus jointly affecting the summer precipitation pattern (Zhu et al. 2016; Zhang et al. 2018). We discuss the influence of SST combination on the spatial activity characteristics of the Meiyu belt and its mechanism by placing the north-south shift of the Meiyu belt under the combination of the SST models of PDO+/AMO+, PDO−/AMO−, PDO+/AMO−, and PDO−/AMO+ (Fig. 4). Since the relationship between SST and the Meiyu belt in summer is the focus, the June, July, and August average values were adopted for PDO and AMO indices following the methods of Zhang et al. (2018).

PDO−/AMO+: the Meiyu rain belt was weak and northward from 1944 to 1954 and 2000 to 2009 (Fig. 3 b and g). It was weak overall in the entire Yangtze-Huai River Valley (Fig. 3c), and the main rain belt of eastern China was in the Yellow River Basin from 1955 to 1966 (Gu et al. 2005; Zhao et al. 2008). The rain belt was weak and northward or weak overall from 1871 to 1879 and from 1892 to 1900.

PDO+/AMO−: the rain belt rainfall was strong and southward from 1901 to 1911 and 1991 to 1999 (Fig. 3f); and it was strong overall in the entire Yangtze-Huai River Valley from 1979 to 1990 (Fig. 3e).

PDO−/AMO−: although the Meiyu in the YRV was strong from 1967 to 1978 (Fig. 3d), it was weak in the entire Yangtze-Huai River Valley, and the main rain belt of eastern China was in the Yellow River Basin (Gu et al. 2005; Zhao et al. 2008). The Meiyu rain belt was weak and northward or weak overall from 1922 to 1932.

PDO+/AMO+: the rain belt was abnormally southward and weak from 1932 to 1943 (Fig. 3a), and it was strong and southward from 1861 to 1870 and 1881 to 1891.

In summary, the rain belt was southward and stronger under the influence of the PDO+/AMO− and PDO+/AMO+ models; but it was northward and weak or weak overall in PDO−/AMO+ and PDO−/AMO− models.

From the correspondence analysis, the multidecadal SST model AMO mainly affected the variation in intensity of the Meiyu in the YRV. The negative phase of AMO, which was remarkable from 1890 to 2010, usually corresponds to the strong Meiyu in the YRV, a similar finding to Zhu et al. (2016). The influence of AMO on the Meiyu belt was very similar to that of EASM on the Meiyu belt at a multidecadal scale. This similarity may stem from the impact of AMO on EASM, as studies have shown that AMO has a significant modulation effect on the decadal variation of the East Asian monsoon climate. The warm-phase AMO strengthens the thermal difference between land and sea by heating the middle and upper troposphere of Eurasia and finally strengthens the EASM (Wang et al. 2009).

PDO plays a decisive role in the north-south shift of the rain belt on an interdecadal scale. The rain belt is southward for PDO+ and northward for PDO−, and even the main rain belt of eastern China deviated from the Yangtze-Huai River Valley and appeared in the Yellow River Basin.

In a positive PDO phase, the SST of the central North Pacific is colder, which leads to significant negative anomalies of the North Pacific in upper-level atmospheric circulations and low pressure from Japan to the YRV in low-level atmospheric circulations and thus weakening the subtropical high and strengthening the westerly East Asian Jet Stream (EAJS) and moving it to the equator (Zhang et al. 2018; Zhu et al. 2016). Changes in the subtropical high and EAJS have an important influence on the horizontal and vertical movement of water vapor. From a horizontal transportation perspective, the water vapor in the Yangtze-Huai River Valley mainly comes from the South China Sea (SCS) and the Bay of Bengal (BOB) during the Meiyu period. The water vapor from SCS mainly affects precipitation in the YRV, and the amount of water vapor from the BOB is positively and negatively related to the amount of water vapor entering the HRV and the YRV, respectively. When the Pacific Ocean subtropical high is weak, more water vapor from SCS enters the YRV, resulting in precipitation enhancement. The water vapor from BOB is weakened due to the lower transportation capacity of the subtropical high, so that water vapor to the north of the YRV is weakened (Gao and Sun 2016; Gao et al. 2016). From the vertical motion of water vapor, EASJ shifts towards the equator and is strengthened, which causes updraft of the YRV on the south side of the jet stream and downdraft of the HRV on the north side. Furthermore, the secondary circulation caused by vertical meridional circulation strengthens the East Asian northern wind in the lower atmosphere. So when the summer monsoon weakens, the northward transport of water vapor will also be weakened (Zhu et al. 2015, 2016). These two reasons make the Meiyu Belt move southward. When PDO is in a positive phase, AMO phase has a certain influence on the north and south of the Meiyu rain belt (Zhang et al. 2018). There are zonal teleconnection waves in the middle latitude of Eurasia when PDO is in a negative phase, which can strengthen PDO influence on the Meiyu belt. When AMO is in a positive phase, teleconnection waves propagate along the great circular path, and AMO promotes Kelvin waves by affecting the Indian monsoon, thus resulting in a rainfall pattern of southern floods and northern drought in eastern China (Zhang et al. 2018). From 1933 to 1943, the Meiyu belt was abnormally southward, which may be related to the positive AMO phase.

PDO influence in the negative phase on the Meiyu belt is almost the opposite of the positive phase, so the rain belt shifts northward. However, in the case of negative phase, PDO will be affected by the AMO and IOBM (Indian Ocean Basin Mode) phases. If AMO is in a positive phase and IOBM is in a cold phase at this time, the subtropical high would be further strengthened and shifted northward, thus leading to further northward uplift of the rain belt and the occurrence of the high-precipitation zone in north China (Zhang et al. 2018), which was the case in the period from 1955 to 1966. This is one of the reasons why the variation in intensity of the northern margin of the rain belt is so complicated. As the HRV is in the transitional area between the Yellow River Basin and the YRV, the variation in intensity of the HRV is likely to be more similar to the YRV or the Yellow River Basin. For example, a strong Meiyu appeared in the HRV from 1979 to 1990, which was the same as in the YRV, which is another reason for the complicated variation of the northern margin of the rain belt.

4.3 The north-south shift of the Meiyu belt and change in the main rain belt of eastern China

In the past, most scholars have focused on the relationship between the Meiyu belt intensity and the location of the main rain belt of eastern China in summer (Sun et al. 2005), with less discussion on the relationship between the north-south shift of the Meiyu belt and the location of the main rain belt. We consider the latter here.

Lü et al. (2014) analyzed the interdecadal variation of monsoon rain patterns in eastern China from 1900 to 2010 by dividing the time into 10 periods according to the time of summer that precipitation patterns changed in North China, the YRV, and South China and summarized the rainfall distribution for each period. The results are basically consistent with the results of several studies on the variation of summer precipitation pattern and the main rain belt (Ding et al. 2008; Wei et al. 2012; Zhao et al. 2008). Therefore, the signals of rain-pattern intensity identified in Lü et al. (2014) were compared with the interdecadal signals of the Meiyu belt in different regions in Table 4, to address the relationship between Meiyu rain belt shifts and the position of the main rain belt.

Table 4 Interdecadal characteristics of Meiyu intensity in the CSHY, the YRV, and the HRV and summer precipitation intensity in North and South China (Lü et al. 2014)
Full size table

In the seven periods shown in Table 4, the intensity of the southern margin of the Meiyu belt is basically opposite to that of north China. When the rainfall is atypical in the southern margin of the Meiyu belt under normal circumstances, the summer precipitation in north China is relatively heavy. Conversely, when rainfall in the southern margin of the Meiyu belt is heavy, summer precipitation in North China is slight. The corresponding relationship only deviated during the periods from 1979 to 1990 and from 2001 to 2009. However, the overall intensity of the Meiyu belt was basically opposite to the summer precipitation in North China, except for the anomaly that occurred from 2001 to 2009, indicating that precipitation in the southern margin of the Meiyu belt is relatively heavy. Therefore, when the Meiyu belt is southward or abnormally southward, the precipitation in North China is relatively light. When the Meiyu belt is northward or weak overall, precipitation in North China is generally heavier. This means that the location of the Meiyu belt may play a role in the prediction of summer precipitation in North China.

The location of the Meiyu belt is obviously related to the intensity of the summer precipitation period in South China. When the Meiyu belt is southward or abnormally southward, the precipitation in South China may be heavier. However, when the Meiyu belt is northward, the relationship with the precipitation intensity in south China is uncertain.

It is noteworthy that the north-south shift of the Meiyu belt may be related to interdecadal variation of the main rain belt. Studies have shown that the main rain belt has retreated twice since 1951–2000; it was in the Yellow River basin from 1951 to 1977 and in the YRV from 1978 to 1992 and extended southward to south China from 1993 to 2004 (Ding et al. 2008, 2009; Sun and Wang 2015). In the first phase of southward retreat from 1978 to 1992, precipitation in the Yellow River Basin decreased, and the Meiyu was heavier in the YRV and HRV, while precipitation on the southern margin of the Meiyu belt was still light. With further southward extension of the main rain belt to south China from 1991 to 2000 (Ding et al. 2008; Sun and Wang 2015), the Meiyu in the HRV weakened, and its southern margin became stronger. This change may be related to the continuous weakening of the summer monsoon on a multidecadal scale (Ding et al. 2008).

5 Conclusions

  1. 1.

    The characteristic sequence of the Meiyu was reconstructed in Changsha of Hunan province from 1861 to 1916 by processing historical data from various diaries, and the sequence was extended to 2017 using modern instrument-measured data. It can be found that the typical/atypical Meiyu appears alternately on an interdecadal scale in the Changsha area, while the typical Meiyu exists at a centennial scale.

  2. 2.

    There are interdecadal and multidecadal north-south shifts on the southern margin of the Meiyu belt, which may reach 2 degrees of latitude. From a multidecadal change perspective, the southern margin of the rain belt was mainly in the south of Hengyang from 1871 to 1915, in the north of Changsha from 1916 to 1987, and in the south of Hengyang from 1988 to 2017.

  3. 3.

    This study investigated the entire north-south shift of the Meiyu rain belt based Meiyu variation in the YRV and HRV. There were obvious interdecadal and multidecadal variations in the location of the Meiyu rain belt over 157 years, and the trend was basically consistent with the north-south shift of its southern margin. At a multidecadal scale, the rain belt was stronger and southward from 1861 to 1921. From 1922 to 1978, the rain belt was northward, southward, or abnormally southward and overall weak due to the main rain belt of eastern China leaving the Yangtze-Huai River Valley. From 1979 to 2017, the rain belt was southward and stronger. The rain belt was northward after 2000, which may be an interdecadal fluctuation on a long-term scale, not a long-term trend.

  4. 4.

    The multidecadal-centennial scale characteristics of the Meiyu belt activity are mainly controlled by the summer monsoon. If the summer monsoon is strong (weak), the Meiyu will be light (heavy). Meanwhile, the summer monsoon dominates the southern margin of the rain belt, especially when the summer monsoon is strong, the southern margin is generally northward. AMO affects the thermal difference between land and sea through heating the middle and upper Eurasian troposphere, thus affecting the intensity of the East Asian summer monsoon. Since AMO has a positive correlation with the strength of summer monsoon, it showed a similar effect on the Meiyu belt.

  5. 5.

    PDO determines interdecadal variation in the rain belt. This is mainly because the PDO phase determines the SST of the North Pacific, which has an important effect on the intensity of the West Pacific subtropical high and the East Asian westerly jet stream. The intensity of the West Pacific subtropical high and the position and intensity of the East Asian westerly jet affects the horizontal transport and vertical upward movement of water vapor in the YRV and HRV, respectively, thus affecting the position of the Meiyu belt. The PDO positive phase usually corresponds to the rain belt which is southward and strong, while the negative phase corresponds to the rain belt which is northward and weak, or the rain belt is weak overall. However, the influence of PDO is modulated by AMO and IBOM. Namely, if the IBOM is cold,  in a negative PDO and positive AMO phase subtropical height will be greatly enhanced, so that the rain belt will further uplift to the north, resulting in a shift of main rain belt of eastern China from the HRV to North China.

  6. 6.

    The Meiyu belt activity is closely related to the shift of the main rain belt and its interdecadal shift in eastern China. When precipitation in the southern margin of the Meiyu belt is relatively heavy, precipitation in North China is generally lighter, and when the Meiyu belt is northward or weak overall, precipitation in North China is generally heavier. This means that the location of the Meiyu belt may play a role in the prediction of summer precipitation in North China. When the main rain belt shifts from north to south, the location of the Meiyu belt often shows a similar shift.