The observed precipitation data employed in this study were derived from over 2,944 stations of the National Meteorological Information Center operated by the China Meteorological Administration (CMA). Among the 2,944 stations, data from 347 stations during the period 1979–2019 were extracted for South China. The linear trends of precipitation were removed before performing the composite analysis for better interannual change signals.

The SH dataset of 48 stations that also originated from the CMA with 41 years of continuous observations from January 1979 to December 2019 over the TP used here was provided by Duan et al. (2018); Duan et al. (2022). Less than 7.3% of the entries are missing values. These data have been publicly released and can be freely downloaded (http://data.lasg.ac.cn/TPSHLH/TP_station_SHLH/).

The gridded datasets used in this study were derived from the European Centre for Medium-Range Weather Forecast (ECMWF) Interim Re-Analysis dataset (ERA-Interim; https://apps.ecmwf.int/datasets/). The same time periods mentioned previously were used, with a spatial resolution of 1.0° × 1.0° and a total of 23 vertical levels from 1,000 hPa to 200 hPa (Dee et al., 2011). The monthly mean SST was provided by the Hadley Centre (HadISST; https://www.metoffice.gov.uk/hadobs/hadisst/) with a horizontal resolution of 1.0° × 1.0° and a time span from 1870 to present (Rayner et al., 2003).

The Niño-3 index is calculated as the SST anomaly (SSTA) in the domain of 150°–90°W, 5°S–5°N. We selected the ±0.5 standard deviation as the criterion for the El Niño and La Niña events. The Niño-3 index was derived from https://www.cpc.necp.noaa.gov/data/indexes/sstoi.indexes.

According to the similarity theory (Monin and Obukhov, 1954), the positive algorithm method is used to estimate sensible heat flux (SH) with the bulk transfer equation as follows (Li et al., 2000; Duan et al., 2018): S H = ρ · C p · C H   · u · T g − T a , (1) where u , T _g , and T _a denote the wind speed at 10 m, ground temperature, and the near-surface air temperature at 2 m, respectively. ρ is the air density with the value of 0.8 kg m⁻³ (Yeh and Gao, 1979). C _p (=1005 J kg⁻¹·K⁻¹) denotes the specific heat capacity at a constant pressure. C _H is the bulk transfer coefficient of heat, which is assumed to be identical to the drag coefficient with a constant value of 0.008 (Yeh and Gao, 1979).

Referring to previous studies (Lu et al., 2010; 2015; Tu and Lu, 2020; Hao and Lu, 2021), if the variable Z is affected by multiple factors such as X and Y at the same time, and the relationship between Z and the influencing factors X, Y, etc., is non-linear; it can be written in the general form of Z = Z(X, Y). To explore the relative importance of influencing factors, and to assess whether variable Z is more influenced by variable X or Y, multiple linear regression Z = AX + BY + ε can be used to fit their relationship, in which A = ∂Z/∂X and B = ∂Z/∂Y expressed the mean the change rates of Z with X and Y, respectively. Therefore, considering the interannual change rates and the perturbation scales of X and Y, two equations can be obtained to estimate the contributions of X and Y to the variable Z: S X = ∂ Z / ∂ X · σ X , (2) S Y = ∂ Z / ∂ Y · σ Y , (3) where σ _X (σ _Y ) in the equation represents the standard deviation of the interannual perturbation of X and Y. With the coefficients fitted and the normalizations performed (σ _X = σ _Y = 1), the final equations of contribution can be expressed in the form of S X = A (4) and S Y = B . (5)

Thus, the dominant relative effect of variables X and Y on variable Z can be assessed by comparing the magnitude of the absolute values of A and B.

The index of the SAH is defined as the longitude of the eastern ridge point corresponding to the 1,254 dagpm isoline on the 200 hPa geopotential height of the strongest warm high system that appears in the upper troposphere of the TP and its adjacent areas (Ren et al., 2015; Zeng et al., 2016; Guan et al., 2018). Similarly, the WPSH index in midsummer (July and August) is defined as the western ridge point of the 588 dagpm isoline on the 500 hPa geopotential height field over the Northwest Pacific (Huang and Wang, 1985; Yang et al., 2017; Guan et al., 2018).

A relative zonal position index of the SAH and the WPSH was constructed to determine the relative zonal position between the SAH and the WPSH. This index suggests that the greater (smaller) the absolute difference value is, the farther (closer) the relative zonal position of the two high pressures is.

The ridge point index of the WPSH defined in this study is verified by the 74 Atmospheric Circulation Index sorted by the National Climate Center of China (NCC), as shown in Figure 1. It can be seen that the correlation coefficients between the western ridge point index of the WPSH calculated here for July and August from 1979 to 2019 and the index provided by the NCC have r = 0.82 and 0.93 (p < 0.05) (Figures 1A,B), respectively. To verify the rationality and accuracy of the calculated eastern ridge point index of the SAH, a similar method is applied to calculate the western ridge point index of the SAH in JJA. There is a relatively good correlation with the western ridge point index of the SAH provided by the NCC, with a correlation coefficient of 0.75 (Figure 1C). Accordingly, the ridge point index of the WPSH and the SAH in this article are the western ridge point index of the WPSH and the eastern ridge point index of the SAH obtained by the same calculation method.

The spatial distribution of the 48 observation stations and the climatological variations of the SH over the TP are shown in Figure 2. The TP presents a strong heat source from March to September, in which the SH is much stronger in the spring, especially when it reaches its annual peak in May (Zhao and Chen, 2001b; Liu et al., 2012; Xu and Chen, 2018). In addition, comparing the monthly variation of the standard deviation of the SH, it can be seen that the standard deviation of the SH is relatively larger in May (Qian et al., 2004). Therefore, in some studies, the SH in May was used to characterize the intensity of the spring heat source over the TP (Zhao and Chen, 2001b; Duan et al., 2017; Zhan et al., 2020; Duan et al., 2022).

Figure 3 shows the leading mode of the spatial and temporal distribution of EOF of the SH at 48 stations on the TP in May with its variance contribution of 23.2%. The spatial distribution (Figure 3A) shows that except for a few stations in the northeast TP, most of the TP stations are positive, denoting the SH on the TP mainly showed consistent spatial distribution characteristics throughout the whole region. The change curve of time coefficient series (Figure 3B) indicates the SH on the TP mainly shows a significant reduction trend (Qian et al., 2004; Duan and Wu, 2008; Wang et al., 2012; Duan et al., 2017) in the spring over the past 41 years, with a remarkable interannual variability. Recent studies (Zhu et al., 2017; Wang et al., 2019; Fan et al., 2021) also pointed out that recovery occurs every 2 decades. However, for simplicity, we removed a linear trend from the time series to obtain interannual signals for the following research.

To investigate the relationship of precipitation in midsummer in South China with the spring SH over the TP and the SSTA in Niño-3 in the preceding winter, the correlation coefficient patterns between the precipitation in midsummer in South China and the spring SH over the TP and the Niño-3 index in the preceding winter (December) are presented in Figure 4. In particular, the preceding winter (December) is chosen mainly because of the lagging influence of the ENSO on climate (Wang et al., 2001; Chen et al., 2018). From Figure 4A, it can be seen that a significant negative correlation exists between the spring SH over the TP and the midsummer precipitation in South China, while a remarkable positive correlation exists in the middle and lower reaches of the Yangtze River Basin, which shows a spatial dipole mode in eastern China. Clearly, the SSTA in the Niño-3 region has a similar relationship with the precipitation in eastern China (Figure 4B). As shown in the figures, the anomalous midsummer precipitation in South China (denoted by the rectangular box) is affected by both the spring SH over the TP and the variation of SST in Niño-3 areas, indicating that during in the years of the positive (negative) phase of the spring SH and the ENSO, the midsummer precipitation in South China will be obviously less (more), which is basically consistent with the results of Huang and Wu (1989) and Zong et al. (2010). Both the spring SH over the TP and the SSTA in Niño-3 have significant regulatory effects on midsummer precipitation in South China. Therefore, the research area of this study focuses on South China (110–120 °E, 20–28 °N), which is basically consistent with the scope of South China determined by previous studies (Wang and Fan, 2013; Yang et al., 2014; Wu and Mao, 2016).

To further demonstrate the key areas of the SH over the TP and the SSTA of the Pacific that are closely related to midsummer precipitation in South China, Figure 5 depicts the regression coefficient distributions of the standardized midsummer precipitation in South China regressed to the spring SH over the TP and the SSTA in the Pacific in the preceding winter. The spatial distribution pattern of the regression coefficient of midsummer precipitation in South China with the spring SH (Figure 5A) is identical to the spatial distribution of the first EOF model of the spring SH over the TP (Figure 2A), which indicates when the spring SH is relatively stronger (weaker) over the TP, the amount of midsummer precipitation in South China is relatively less (more), which is consistent with Figure 4A. Figure 5B shows a close relationship between midsummer precipitation in South China and the Pacific SSTA. The most significant correlation areas are in Niño-3 regions, implying that when the index in the preceding winter is on the high (low) side, the midsummer precipitation in South China is relatively less (more), which is in accordance with Figure 4B. In addition, the overall correlation coefficients are −0.42 and −0.34 (passing the 90% significance level test), respectively (in Figure 6).

The correlation and regression analysis results show that the spring SH over the TP and the SSTA in Niño-3 regions in the preceding winter are significantly related to the midsummer precipitation in South China. It can be seen from Figure 6 that when the spring SH and the SSTA in Niño-3 areas in the preceding winter are in a negative (positive) phase, the midsummer precipitation in South China is generally more (less).

In order to explore whether the spring SH over the TP or the ENSO has a greater influence on the midsummer precipitation in South China, the contributions of the spring SH over the TP and the ENSO are calculated with the method mentioned previously in Section 2. Figure 7 shows the distribution of S_SH, S_Niño-3, and their differences.

As shown in Figures 7A,B, the spatial distribution characteristics of the spring SH over the TP and the ENSO contributions to midsummer precipitation in eastern China are consistent with Figure 4. Combined with Figure 4, the spring SH over the TP and the ENSO have significant positive contributions to the Yangtze–Huaihe River Basin and the middle and lower reaches of the Yangtze River, respectively, while for the same region of South China, both show significant negative contributions (absolute values in the figures). When these non-linear relation-impacting factors have concurrent effects, it is difficult to discern their respective contribution to the midsummer precipitation in South China. However, with the difference between S_SH and S_Niño-3, the contribution from the spring SH over the TP and the ENSO can be compared (Figure 7C). There are regional differences in the impact of the SH and the ENSO on precipitation in South China; namely, the precipitation in the eastern coast of South China is mainly dominated by the spring SH over the TP, while the central and southern of South China is dominated by the ENSO.

Additionally, the PDO phase transition can weaken the impact of the ENSO on summer precipitation in southern China (Chan and Zhou, 2005; Wang et al., 2008; Song and Zhou, 2015; Dong, 2016). To clarify the impacts of the spring SH over the TP and the ENSO on the midsummer precipitation in South China under different PDO phases, the sliding t-test has been carried out for the spring SH and midsummer precipitation in South China, as shown in Figure 8A. Over the past 41 years, both the midsummer precipitation in South China and the spring SH over the TP changed significantly in 1996, as shown in Figure 8A (passing the 95% significance level), while ENSO events had no remarkable change. Figure 8B displays the 11-year low pass-filtered PDO in the preceding wintertime. Note that the PDO also experienced a positive-to-negative phase transition around 1996. Hence, for the convenience of research, the period 1979–1996 is recorded as T1, and the 1997–2019 period is T2. In addition, the 11-year low pass-filtered PDO is positively correlated to the sliding correlation between the ENSO and midsummer precipitation in South China, especially during the positive PDO phase (not significant during the negative PDO phase), which indicates that the PDO may play a moderating role on the ENSO and its relationship with midsummer precipitation in South China.

Eight categories are selected to further examine the regulation of the PDO on the effects of the SH over the TP and the ENSO on the midsummer precipitation in South China, as shown in Table 1. Then, composite analyses can be conducted for the different categories to distinguish the relative contributions of different factors to the precipitation in South China.

Figure 9 depicts the composition of midsummer precipitation anomalies in South China in stronger and weaker years of the spring SH over the TP and the preceding ENSO under the positive PDO phase (period T1). On the one hand, the midsummer precipitation in South China significantly decreased (increased) in the years with a strong (weak) spring SH, especially in the southeast coastal areas (Figures 9A,B, respectively), which is consistent with the results obtained in Figure 6; on the other hand, in years of preceding El Niño (La Niña) events, the midsummer precipitation in South China significantly decreased (Figures 9C,D, respectively).

The PDO plays an important role in modulating the impacts of the ENSO (Power et al., 1999; Chen et al., 2013). During the T2 period (1997–2019), when the PDO was in a negative phase, the sliding correlation coefficient between the PDO and the ENSO with the midsummer precipitation in South China was not significant. Compared with the T1 period, the effect of the ENSO on midsummer precipitation in South China was enhanced after being modulated by the PDO (Figures 10C,D). At this time, the impacts of the spring SH over the TP and the ENSO in the preceding wintertime on midsummer precipitation in South China are relatively independent, which is consistent with Figure 4. In addition, it can be seen (Figures 10A,B) that the modulation effect of the PDO makes the influence of the spring SH on precipitation in midsummer in South China relatively symmetrical in the spatial distribution mode.

As shown in Figure 11, during the SH(+)/PDO (+) period (Figure 11A), the intensity of the SAH and the WPSH were consistent with the climatic state, and South China was under the control of low-level anticyclone anomalies at 850 hPa located between the SAH and the WPSH. This results in the convergence and transportation of water vapor to the middle and lower reaches of the Yangtze River in the Yangtze–Huaihe River Basin, leading to less precipitation in South China. Meanwhile, during the SH(−)/PDO (+) period (Figure 11B), as the intensity of both the SAH and the WPSH weakened abnormally, the SAH retreated westward, and the WPSH moved eastward. Under the configuration of the anomalous anticyclone centered in the Yellow Sea and the anomalous cyclone circulation centered in the South China Sea in the lower layers, the water vapor from the eastern coastal areas of South China was continuously transported to central and northern China. Meanwhile, there was weak water vapor convergence in the central and southern parts of South China, which is consistent with the precipitation anomalies in Figure 9B.

Moreover, the WPSH was stronger (weaker) in the El Niño (La Niña) event years under positive PDO phases, extending (retreating) westward (eastward). The whole area in South China is in the center of an anticyclone (cyclone) circulation anomaly with downdraft (updraft) movement, which gives rise to the whole-layer moisture flux diverging (converging) (Figures 11C,D) in the Yangtze–Huaihe River Basin, resulting in less (more) midsummer precipitation in South China. Hence, in the years of ENSO/PDO(+) , midsummer precipitation in South China is reduced (Figures 9C,D).

From the perspective of the circulation field, compared with the pre-mutation T1 period (positive PDO phase), due to the modulation of the PDO, in the SH(+) and El Nino years in T2 (negative PDO period) in Figures 12A, C, the SAH and the WPSH were abnormally enhanced. This manifested as the SAH extending eastward and the WPSH moving westward, resulting in their being very close to each other. Under this anomalous circulation configuration, the lower levels of South China were dominated by anomalous anticyclonic circulation. The water vapor was transported and converged to the Yangtze River Basin, leading to decreased midsummer precipitation in South China, which is identical to the result of Figures 10A,C. In the SH(−) and La Nina years in T2 shown in Figures 12B, D, the intensity variation of the SAH and the WPSH were relatively weak. South China was under the control of anomalous cyclonic circulation, accompanied by stronger ascending motion and water vapor convergence and transport; hence, midsummer precipitation increased abnormally (Figures 10B,D).

What is more, in the remarkable positive (negative) phase years of the spring SH and the ENSO, the SAH over South Asia strengthened (weakened) and extended (retreated) eastward (westward), and the WPSH reinforced (weakened) while moving west (east). There is a trend of “moving toward each other or moving away from each other” (Figure 12) (Tao and Chu, 1964) between the two high-pressure areas.

Whether the effects of the relative position changes of the SAH and the WPSH on the precipitation in South China are independent or synergistic is an important question. To analyze the influence of the zonal relative position variation of the SAH and the WPSH on the midsummer precipitation in South China, Figure 13 shows the index of the west ridge of the 588 dagpm isoline on the 500 hPa geopotential field and the east ridge of the 1,254 dagpm isoline on the 200 hPa geopotential field of the SAH (Figure 13A) in midsummer from 1979 to 2019. The normalized zonal relative position variation of the ridge point of the two high pressures is shown in Figure 13B, from which we can learn that in the past 41 years (1979–2019), compared to the climatological mean of the east ridge of the SAH and the west ridge of the WPSH, the position of the two high-pressure system ridges has changed westward 55% of the time and eastward 45% of the time. The analysis found that the relative motion of the SAH and the WPSH accounted for 70%, of which the motion of moving toward and away from each other was 37.5% and 32.5%, respectively (Table 2).

In summer, under the effect of heating over the TP, a strong anticyclone is formed in the upper level of South Asia, and the airflow diverging from the TP to the east forms a horizontal convergence in the middle levels of the Pacific Ocean, thus playing an important role in the formation and maintenance of the WPSH. The two high pressures at different heights in the subtropical summer zone can be regarded as a whole, and they tend to move in opposite directions and apart from each other. According to ±1.0 σ of zonal relative position change in Figure 13B, 9 and 5 years of significant moving toward and away from each other in the distance between the SAH and the WPSH were selected and recorded as type 1 and type 2, respectively, in Table 3. In terms of the interannual variation scale, the upper SAH oscillates from east to west, and the middle and lower WPSH also change significantly at the same time, affecting the regional climate in Asia. Studies have found that the SAH and WPSH overlap areas mainly correspond to an arid climate (Guo and Zhi, 2008).

Figure 14 depicts the spatial distribution of midsummer precipitation anomaly in South China according to the zonal relative position of the SAH and the WPSH. When the SAH and the WPSH noticeably moved away from each other, that is, the SAH retreated to the west, and the WPSH retreated to the east, the midsummer precipitation in all South China areas increased significantly (Figure 14A), while the midsummer precipitation in South China reduced markedly when the two high pressures moved toward each other, namely, the SAH moved eastward, and the WPSH extended westward, and the two even overlapped (Figure 14B). The synergistic effect of the zonal relative position anomaly of the two high pressures on the midsummer precipitation in South China is primarily located in the southeast coastal areas, which passed the 90% significance level test.

In summation, what can be seen is that the synergistic influence of the zonal relative position of the SAH and the WPSH on midsummer precipitation in South China lies in when the two high pressures move toward (away from) each other, corresponding to the positive (negative) phase of the standardized ridge position index, the precipitation increases (decreases) significantly.

To discuss the relationship between the zonal relative position anomalies of the SAH and the WPSH and the precipitation in midsummer in South China, the anomalies of the atmospheric circulation field, the water vapor flux, and the water vapor divergence of the SAH and WPSH zonal relative positions were studied using composite analysis. As shown in Figure 15, when the SAH and the WPSH were far apart from each other (Figure 15A), the east extension ridge point of the 1,254 dagpm isoline on the 200 hPa geopotential field was located at 102.3°E, which is about 6 degrees of longitude westward compared with the climatic mean state. At the same time, the WPSH has also retreated to the sea, with the west ridge located at 138.2°E, which is about 16 degrees of longitude eastward from the mean state. In this modal, the whole South China region was under the cyclone circulation anomaly center with an updraft motion of the atmosphere, and the water vapor from both the middle and lower reaches of the Yangtze–Huaihe River Basin and the ocean converged to South China, which is conducive to the abnormal increase of midsummer precipitation in South China (Figure 11A). In contrast to the depiction in Figure 15A, South China was under the control of the anticyclone circulation anomaly shown in Figure 15B, accompanied by a downdraft motion of the atmosphere when the SAH and WPSH approached or overlapped. This is conducive to the convergence of water vapor to the Yangtze–Huaihe River Basin, decreasing the water vapor in South China significantly, and leading to decreased midsummer precipitation in South China.