Decadal variation of the precipitation relationship between June and August over South China and its mechanism

This paper investigates the causes for the interdecadal change in the relationships between early and late summer precipitation over South China (SC). It is found that the correlations of the precipitation over SC between June and August shift from weak positive in 1979–1995 to significantly negative in 1996–2019. Further analysis demonstrates that the distinct evolution of sea surface temperature (SST) pattern between the two periods accounts for the interdecadal variations of their relationships. Although the warming of the tropical western Indian Ocean (WIO) in June favors increasing precipitation over the SC via enhancing the Northwest Pacific Subtropical High (NWPSH) during the whole period, the associated SST anomalies in August are rather different between the two periods. Specifically, the WIO warming in June corresponds to slightly positive anomalies over the tropical central-eastern Pacific in August during the 1979–1995, which has weak impact on the NWPSH and results in a weak precipitation correlation between June and August. However, the WIO warming in June corresponds to La Niña’s rapid development in August during the 1996–2019, which favors the enhancement of the NWPSH via increasing the regional Hadley circulation. Due to the climatologically northward movement of NWPSH from June to August, the enhanced NWPSH in August acts to decrease the precipitation over SC, causing a significantly negative correlation between precipitation in June and August. Overall, the distinct evolution of tropical SST pattern is the key factor inducing the change of the relationships between June and August precipitation in the two periods.


Introduction
South China (SC) is in the East Asian (EA) monsoon region with significant intraseasonal and interdecadal variations of summer precipitation (Su et al. 2014;Ha et al. 2016Ha et al. , 2019Ding et al. 2008;Wu et al. 2010b). The burst of the SC summer monsoon around mid-May marks the beginning of the EA summer monsoon. SC is located to the northwest side of the Northwest Pacific subtropical high (NWPSH), with significant southwest water vapor transportation, precipitation increasing and reaching the maximum precipitation in June (Ding and Chan 2005;Karori et al 2013). In mid-June, convection in the SC Sea intensifies, the NWPSH jumps north to control SC, which indicates the end of the rainy season in SC and the start of the Meiyu season over Yangtze-Huaihe River (Su and Xue 2010;Su et al. 2017;Feng et al 2016).
Due to the different mechanisms of precipitation formation before and after June, the rainy season in SC (i.e. May to August) can be divided into two different phases. The first phase runs from May to June (refers to as the "early rainy season"), and the second phase lasts from July to August (refers to as the "late rainy season") (Yuan et al. 2010;Yuan and Chen 2013). In the early rainy season, the precipitation over SC is mostly associated with the monsoon systems (Chi et al. 2005;Chang et al. 2006). Along with the monsoon systems moving to northward to the Yangtze-Huaihe River, the tropical depressions and cyclones become dominant contributors to the precipitation over SC in the late rainy season (Ren et al. 2002;Lee et al. 2010). This is also consistent with the recent findings (Yuan et al. 2019;Su et al. 2014) that the precipitation in July-August over SC is significantly correlated with the local cyclone circulation, while the precipitation in May-June over SC is accompanied by both the local cyclone circulation and the strong anticyclone circulation anomalies over the Northwest Pacific (NWP).
In terms of the decadal variations of precipitation over SC in the rainy season, the sea surface temperature (SST) anomalies in the tropical Pacific Ocean and the Indian Ocean have been found as a major modulator. For instance, many studies have found that the summer (June to August) precipitation over SC shows a significant increasing trend in the early 1990s and a significant decreasing trend in the early twenty-first century, respectively (Ding et al. 2008;Wang et al. 2009;Wu et al. 2010b;Ha et al. 2016Ha et al. , 2019. The former change is related to the Walker circulation and Hadley circulation anomalies caused by the continuous warming of the tropical Indian Ocean and the Pacific Ocean in summer, which leads to the increase of snow cover in winter and spring over the Tibetan Plateau and the strengthening of the East Asian monsoon trough (Wu et al. 2010a;Zhu et al. 2014). The latter change is attributed to the strengthening of the sinking branch of circulation over SC caused by the tropical Pacific SSTA (Ha et al. 2016). The NWPSH also experienced a significant decadal shifting around the late 1990s, which is supported by an Empirical Orthogonal Analysis (EOF) for 1979-2016 (Huang et al. 2018). The change is modulated by the variation of SST mode in the Pacific from canonical El Niño-Southern Oscillation (ENSO) to "Modoki" central ENSO (Huang et al. 2018;Funk and Hoell 2015;Li et al. 2015). In addition, when the tropical Pacific is positive SST anomaly, the NWP anticyclone is weak and the cyclone anomaly over SC; while the tropical Pacific is developing phase of La Niña, the NWP anticyclone (NWPSH) is strong and SC is controlled by anticyclone Kosaka et al. 2012;Wang et al. 2013;Qiao et al. 2021). However, the NWPSH is mainly related to the development of La Niña in the same period after 2000 (Huang et al. 2018).
Overall, previous studies have shown that the precipitation feature of May-June is different from the July-August and the interdecadal variations of summer precipitation over SC. Under different interdecadal backgrounds, the summer precipitation in SC, the NWPSH related to precipitation and the SST anomalies all have significant interdecadal variations around 1990 (Li et al. 2015;Gong et al. 2016;Ha et al. 2016Ha et al. , 2019. It is necessary to further discuss if the precipitation relationships have changed between different months in summer around the 1990s, and whether there are interdecadal changes in the circulation field and key areas of SST related to variable precipitation correlation. Comprehending these scientific issues will promote understanding of intraseasonal variations in precipitation and improve short-term climate prediction. The rest of the article is organized as follows. Second section briefly describes the data and methods. Third section presents the analyses of the relationships between the precipitation in June and August over SC to demonstrate the significant interdecadal shifting in the correlations of precipitation around late 1990s. The characteristics of the circulation and SST before and after the late 1990s are also presented. Fourth section investigates the links between the SSTA in the tropical Central Pacific (CP), West Indian Ocean (WIO), Maritime Continent (MC) and the decadal shifting of the correlations between the precipitation in June and August. Fifth section discusses the physical mechanism associated with the rapid development of La Niña. The conclusion and discussion are presented in sixth section.

Datasets
The precipitation data at 295 stations in SC were obtained from the national daily precipitation dataset developed by the National Meteorological Information Center, China Meteorological Administration (CMA). The analysis focused on the period from June to August of 1979-2019 to characterize the interdecadal variations of the relationships between early (June) and late (August) summer precipitation. The precipitation data during the period of 1979-2019 were also extracted from Version 2 Global Precipitation Climatology Project (GPCP) (Adler et al. 2003) to replicate the analog exercises performed with the CMA in-site precipitation observations. The GPCP product includes monthly land-surface precipitation based on ~ 80,000 rain gauge stations across the globe.
The monthly mean meteorological fields during the period of 1979-2019 were retrieved from the National Centers for Environment Prediction-Department of Energy (NCEP-DOE) Atmospheric Model Intercomparison Project-II reanalysis dataset (Kanamitsu et al. 2002) to analyze the atmospheric circulation and surface flux anomalies associated with the summer precipitation over SC. Two sets of variables were considered in this study: (1) three-dimensional upper-air quantities on a 2.5° × 2.5° global grid, including geopotential height, horizontal and vertical wind components, and relative humidity, and (2) two-dimensional surface variables on a T62 Gaussian grid, including 10-m wind, 2-m specific humidity, skin temperature, latent heat net flux, sensible heat net flux, downward shortwave radiation flux, downward longwave radiation flux, upward solar radiation flux and upward longwave radiation flux.
In addition, the downward net heat flux, downward shortwave and upward longwave radiative flux, upward latent and sensible heat fluxes, and near-surface wind speed, temperature, humidity information of the sea surface/near-surface air provided by the OAFlux dataset are also adopted in the calculation (1.0° × 1.0° resolution; Yu and Weller 2007). The monthly mean subsurface temperature, oceanic current and vertical velocity from 1980 to 2019 are obtained from the NCEP Global Ocean Data Assimilation System (GODAS, 1.0° × 0.33°, with 40 vertical levels, Saha et al. 2006). The ocean data are from the Hadley Centre Sea Ice and Sea Surface Temperature (HadISST) dataset (Rayner et al. 2003), and the Extended Reconstructed SST (ERSST) version 4 (ERSST.v4, Huang et al. 2015;Liu et al. 2015). The results presented in third and fourth section are based on the analyses for CMA, NCEP2 and ERSST.v4. However, the same conclusions can be drawn if we performed the analog exercises using GPCP, OAFlux and HadISST (Table 1).

Methods
The SC region selected for this study encloses [21°N-29°N; 110°E-121°E] (Fig. 1, Yim et al. 2014). Within this region, continuous precipitation observation data at 295 stations from June to August in 1979-2019 can be obtained after the quality control process. For the convenience of discussion, we defined a reverse index of precipitation, Index = Pre 6 − Pre 8 , to characterize the variation of the relationship between June and August precipitation in SC. Here, Pre 6 and Pre 8 are the normalized precipitation values.
The SST indices for three key regions (the rectangular boxes in Fig (1) Here, the radiative fluxes (i.e. SWF and LWF) are obtained from the satellite observations, and the surface latent heat fluxes can be calculated using a bulk algorithm (Yu and Weller 2007): where, ρ and L e denote the air density and the latent heat of evaporation, respectively. C e represents the turbulent exchange coefficient for latent heat fluxes. U is the wind speed at the near sea surface level; Δq is the difference of specific humidity between the sea surface and near-surface air. The term UΔq is determined by both atmospheric and oceanic states, and UΔq can be decomposed using the following formula Similarly, the latent heat fluxes can be decomposed and written as We can obtain where the first and second terms are affected by the difference of specific humidity at the interface of the air-sea interaction and the atmospheric wind anomalies at the nearsurface, respectively. The third term denotes a nonlinear interaction term, which can be ignored as the magnitude of this term is usually small. In this study, all the linear trends and climatic mean values are removed before the calculation using Eq. (5).

Circulation and SST associated with reverse index
A 21-year sliding correlation is carried out for any two months of precipitation with a moving step of 1 year. As shown in Fig. 2a, there is a weak positive correlation between the June and July precipitation, and a strong positive correlation between the July and August precipitation. The latter shows a weakening trend starting from 2007. In contrast, the negative correlations (referred to as "inverse relationship") between the June and August are seen in Fig. 2a, indicating the reversed variations of precipitation between these two months. In addition, the inverse relationship between June and August precipitation continues to strengthen, especially in the recent 20 years. Figure 2b shows the interannual variations and correlation coefficients (CC) of precipitation between June and August over SC. The mean precipitation CC are 0.25 in 1979-1995 and −0.45 (passing the 95% significance test) in 1996-2019. The spatial patterns between the reverse index and the precipitation in June and August also show significant interdecadal changes. Precipitation in June and August during the prior period has a poor correlation with the reverse index, and the significant area is smaller and located northerly (Fig. 3a, b). In the latter period, the correlations between precipitation and reverse index in June (positive) and August (negative) are significantly enhanced, and the significantly related area extends from SC to the east near 130°E (Fig. 3c, d). We note that our results are consistent with the findings in existing studies (Zhu et al. 2011;Leung et al. 2020).
To understand the reasons for the reverse change of precipitation relationship between June and August. The water vapor fluxes (WVF) and their convergence/divergence in Southeast Asia are regressed onto the precipitation reverse change index. During the prior period, the significant southwest water vapor transportation over SC is shown in June (Fig. 4a), providing the favorite condition for the formation of precipitation. In contrast, there is no significant water vapor transportation, and local water vapor divergence is weak in August (Fig. 4b). During the latter period, the strong local water vapor convergence, and an increase of the southwest water vapor transportation over SC are significant in June, while the local water vapor divergence associated with the anticyclone circulation over the SC is dominant in August (Fig. 4d). As a result, the precipitation in August is significantly less than that in the prior period. A significant negative correlation between the June and August precipitation is observed. In addition, the water vapor over SC in June mainly comes from the west, north and south borders and is exported from the east border. On the contrary, the water vapor over SC in August is mainly imported from the east border and exported from the north and south borders. However, the net water vapor input (output) in June (August) of the latter period (1996-2019) are larger (Table 2).
Because the anticyclone in the NWP region is closely related to the change of the NWPSH (Lu 2002;Yang and Sun 2003), the 500 hPa geopotential height in June and August during the two periods are also regressed onto the precipitation reverse index. During the prior period, the significant negative geopotential height anomalies over the subtropical region are evidenced in June (Fig. 5a), which corresponds to the cyclonic circulation in Fig. 4a. In August, there is no significant geopotential height anomalies in East Asia (Fig. 5b), which corresponds to no apparent vapor transportation in Fig. 4b. During the latter period, the positive geopotential height anomalies in the south of 20°N and the significant negative geopotential height anomalies in East China are evidenced in June (Fig. 5c). These circulation anomalies are corresponding to cyclone and anticyclone WVF anomalies over the north and south of SC in Fig. 4c, providing beneficial conditions to the precipitation over SC. The positive geopotential height anomalies gradually move northward from June to August. In August, the center of maximum geopotential height anomalies is located on SC as shown in Fig. 5d, which is coincident with the anticyclone and water vapor divergence over SC in Fig. 4d.   Fig. 4 The water vapor flux regressed by inverse index (vectors,10e 2 kg∕(m * s) , blue vectors indicate passing 90% significance test) and its convergence/divergence (shadings, 10e −5 kg∕m 2 * s , pass-ing 90% significance test) at 1000-300 hPa in June (a) and August (b) during 1979-1995. (c) and (d)  In the positive anomaly years (years with positive inverse index) during the latter period, the 5880-geopotential height contour (referred to as "588 line") is westward in June and August, indicating that the NWPSH is strong. In June, SC is located on the north side of 588 line accompanied by southwest water vapor transportation to SC (Fig. 4c), which is conducive to precipitation. In August, the 588 line and SC are located on the same latitude, so the precipitation in SC is suppressed, which is conducive to the formation of a reverse change relationship. In the negative anomaly years (years with negative inverse index) during the latter period, the 588 line continues to weaken and to the east, contrary to positive anomalies (Fig. 5c, d). However, there is no such changing regularity of NWPSH in positive and negative abnormal years of the prior period (Fig. 5a, b).
The 850 hPa wind (relative vorticity) fields in June of the two periods are very similar, with cyclone (positive relative vorticity) anomalies in SC and anticyclone (negative relative vorticity) anomalies in the south of SC (Fig. 6a, c). In August of the latter period, there is a significant anticyclone (negative relative vorticity) in SC, and its intensity is much larger than that of the prior period (Fig. 6b, d). The significant positive vertical velocity (upward) anomalies in June and the significant negative vertical velocity (downward) anomalies in August are evidenced in SC region during the two periods. However, the vertical velocity anomalies in the later periods are much stronger than those in the prior period (Fig. 7).

Physical mechanism of precipitation reverse change
To further understand the reasons of why the evolution of the local circulation from June to August in SC and NWPSH change significantly after 1996, the SST in Indian and pacific are regressed onto the inverse index during two periods. During the prior period, the regressed SST shows the positive SSTA over the tropical WIO and the El Niño over tropical CP in June (Fig. 8a), and only the El Niño over tropical CP in august (Fig. 8b). During the latter period, there are also positive SSTA over the tropical WIO in June, but the La Niña over tropical CP (Fig. 8c). In August, the intensity of La Niña is further enhanced, with significantly positive SSTA over MC and negative SSTA over tropical CP. (Fig. 8d). In other words, the significant change of precipitation relationship between June and August during the two periods may be related to the phase shift of tropical pacific SST pattern. According to the 21-year sliding regression of reverse index and tropical WIO (MCCP), the correlation between reverse index and tropical WIO is always very high, while the correlation between reverse index and MCCP increases significantly for nearly 20 years. Previous studies pointed out that the IO warming excites a Matsuno-Gill-type response in tropospheric temperature (TT), with the Kelvin wave wedges penetrating the equatorial western Pacific, resulting in the enhancement of NWP anticyclone and NWPSH (Terao and Kubota 2005;Xie et al. 2009;Wu et al. 2009). Especially in June, the warm tropical Indian Ocean can only affect the precipitation in the south of the Yangtze River (Yuan et al. 2019;Zheng and Wang 2021). The dipole SST of the MC and the tropical CP also have independent and coordinated effects on the NWPSH (Gill 1980;Wu and Liu 1992;Lu et al. 2006;Sui et al. 2007;Wu et al. 2010a;Chung et al. 2011;Wang et al. 2013;Xiang et al. 2013). The El Niño (negative SSTA over MC and positive SSTA over tropical CP) is conducive to the generation of cyclones over the NWP, while the La Niña (positive SSTA over MC and negative SSTA over tropical CP) is conducive to the generation of anticyclones over the NWP (Tao et al. 2017;Tang et al. 2021;Wang et al. 2013;Yuan et al. 2019;Qiao et al. 2021).
During the prior period, the pattern of the water vapor transportation and tropospheric temperature regressed against the tropical WIO SST in June verifies the previous research results. SC is located to the northwest of the anticyclone, and the southwest water vapor transportation is conducive to the increase of precipitation (Fig. 9a). During the latter period, the results of the June tropical WIO SST regression are consistent with those of the prior period (Fig. 9b). Different from the El Niño in the prior period, the latter period is the rapid development of La Niña (Fig. 8b,  d). Therefore, the Hadley circulation and the Intertropical Convergence Zone (ITCZ) with stronger intensity are more northward in the latter period (Figs. 10d, 11d). Compared with the prior period, the ITCZ near the maritime continent extends northward by about 5° in August, which causing the ascending and descending branch of the Hadley circulation both northward. In the prior period, the descending branch of Hadley circulation is located near 11°N, and South China is the weak upward movement (Fig. 11b). But in the latter period, the intensity of the descending branch of Hadley circulation increases significantly and moves northward to SC (Fig. 11d). Therefore, there is a significant anticyclone over SC in the latter period, which does not exist in the prior  (Fig. 9d, f). Finally, the precipitation relationship between June and August over SC changes from the weak positive correlation to the significant negative correlation.
To further clarify the influence of La Niña on NWPSH and NWP anticyclone, we remove the linearly related parts between MC and CP, and then we study the impacts of them on the NWP anticyclone respectively. From the regression results, these two key areas of SSTA can only affect the local circulation anomalies (Figures are omitted). Considering the joint influence of two key areas of SSTA at the same time, the significant easterly anomalies are produced Fig. 6 The 850 hPa wind (vectors, m∕s , blue vectors indicate passing 90% significance test) and relative vorticity (shadings, 10e −6 s −1 , passing 90% significance test) regressed by inverse index in June (a) and August (b) during 1979-1995. (c) and (d) is as (a) and (b), but for 1996-2019 in the tropical central Pacific and the significant anomalous anticyclone shows up in the NWP (Fig. 9e, f).
Considering the weak correlation between June and August precipitation in the prior period, we use the June and August precipitation to regress the East Asian circulation fields and SST respectively (the results are basically consistent with the results of inverse index regression).
The regression results are similar in June and August, that are, the convergence of water vapor caused by local cyclone over SC, and the water vapor transported by NWP anticyclone (Fig. 12a-d), resulting in a weak positive correlation between June and August precipitation during this period (CC is 0.25). It has been analyzed in this paper that the interdecadal change of precipitation relationship Fig. 7 Same as in Fig. 6, but for the 500 hPa vertical velocity ( 10e −2 Pascal * s −1 ) between June and August is mainly caused by the Pacific dipole anomaly in August (The SST patterns of the tropical Indian Ocean in June are consistent in two periods). The SST pattern of the tropical Pacific in this period is El Niño (Fig. 8b, d), which leads to the southerly position and weak intensity of the meridional Hadley circulation and NWP anticyclone (Figs. 9d, 10b).

The physical mechanism of the rapid development of La Niña
Based on the analyses in the previous section, we note that the rapid development of La Niña from June to August plays a crucial role in the interdecadal variation of precipitation relationship between June and August over SC. The question then arises what causes the La Niña grow from June to August? Following this question, this section further discusses the role of local air-sea interactions in the interseason variation of the La Niña.
First, we found that there are consistent positive feedbacks of air-sea interaction in the two periods. However, the phases of SST Pattern in the prior and latte period are opposite (Fig. 8b, d), which leads to the completely opposite results. Here, we present the analysis results for the latter period to demonstrate the role of air-sea interactions in the gradual development of La Niña. Figure 13 presents the evolution of the latitude-vertical velocity anomalies regressed against the La Niña. The development of La Niña forms a closed circulation circle, with the significant easterly and westerly winds in the low and high level and the significant upward and downward movements in the MC and the tropical CP, respectively, which lead to the intensification of convective activities in the MC and the enhancement of the meridional Hadley circulation (Fig. 11c, d).
In June and August, the easterly winds are dominated in the tropical CP at the surface (Fig. 14a, b), with weak monthly variations. In contrast, large monthly variations of the wind direction are evidenced in MC, with southeast and south winds in June and southwest and south winds in The average air temperature at 850-250 hPa (shadings, K , passing 90% significance test) and 850 hPa wind (vectors, m * s −1 , blue vectors indicate passing 90% significance test) in June during the prior (a) and latter (b) period regressed onto the tropical WIO. The 850 hPa wind (vectors, m * s −1 , blue vectors indicate passing 90% significance test) and its convergence/divergence (shading, s −1 , purple counters indicate passing 90% significance test) and 500 hPa geopotential height (red counters, m , passing 90% significance test) in June (c) and August (d) during 1979-1995 regressed onto the MCCP. (e) and (f) is as (c) and (d), but for 1996-2019 August (Fig. 14a, b). The results of 10 m wind regressed by La Niña show that the abnormal northeast winds in the MC and the abnormal easterly winds in tropical CP are presented in June (Fig. 14a), which are superimposed on the climatic wind fields and make the local wind speed decrease in MC and increase in CP (Fig. 14c). Similar results are evidenced in August, but the strength of wind anomalies is enhanced (Fig. 14d). The significantly abnormal easterly winds cause the warmer ocean currents flowing westward from the tropical CP to the MC, and the colder deep ocean currents flowing eastward from the MC to the tropical CP (Fig. 14e, f). Thereby forms a closed ocean current and further strengthens the intensity of the La Niña (Xie 1999;Kug and Kang 2006;Ham et al. 2007).
Equatorial easterly wind anomalies triggered by development of La Niña not only affect the ocean currents, but also cause significant changes in the net radiation flux, latent heat flux and the first, second and third items of latent heat flux (Fig. 15). The regression results show that the significant increase of the net radiation fluxes over the MC in June and August is mainly due to the contribution of the decrease of latent heat release (Fig. 15a). The second term of the latent heat flux related to the sea surface wind speed plays an absolute role in the significant decrease of the latent heat flux release (Fig. 15b). This is because the climatic winds of the MC blow towards northeast and northwest, but the development of La Niña will cause easterly and northeasterly wind anomalies (Figs. 13a, b, 14c, d). The net radiant fluxes and latent heat flux only pass the 90% significance test   The 850 hPa wind (vectors, m∕s , blue vectors indicate passing 90% significance test) and relative vorticity (shadings, 10e −6 s −1 , passing 90% significance test) regressed by SC precipitation in June (a) and August (b) during 1979-1995. (c) and (d) is as (a) and (b), but for the water vapor flux (vectors, 10e 2 kg∕m * s , blue vectors indi-cate passing 90% significance test) and its convergence/divergence (shadings, 10e −5 kg∕m 2 * s , passing 90% significance test) at 1000-300 hPa. (e) and (f) is as (a) and (b), but for the SST (K, dotted areas pass 90% significance test) in August over tropical CP (Fig. 15c), which may be due to the cooling of the tropical CP mainly comes from the upturn of the cold water (Fig. 14e, f). The latent heat flux and the second term of latent heat flux related to wind speed are also the main reasons for the decrease of net radiation flux over CP (Fig. 15c, d).
In summary, the development of La Niña will enhance the equatorial easterly winds, promote the warmer ocean currents westward to the MC and reduce the latent heat release, resulting in the warming of SST. However, it will increase the release of latent heat over the CP and the upturn of the colder water, resulting in the cooling of SST. Finally, the strength of the La Niña is further strengthened, forming a positive feedback mechanism.

Summary and discussion
This study uses China's daily precipitation data and reanalysis products from June to August during 1979-2019 to reveal the interdecadal variation of precipitation relationship between June and August over SC around mid-1990s and its physical mechanism. With a 21-year sliding correlation analysis for the precipitation data, we find that the relationship between June and August precipitation over SC changes from a weak positive correlation (+ 0.25) in 1979-1995 to a significant negative correlation (-0.45, passing 95% significance test) in 1996-2019.
The interdecadal evolution of SST pattern from June to August during the two period accounts for the interdecadal variations of the precipitation relationships. Although the warming of the tropical WIO in June is conducive to increasing precipitation over the SC via enhancing the NWPSH and transporting water vapor to SC during the whole period, the associated SST anomalies in August are rather different between the two periods. Specifically, the WIO warming in June corresponds to slightly positive SST anomalies over the CP and negative SST anomalies over the MC in August during the 1979-1995, which has weak impact on the NWPSH and results in a weak precipitation correlation between June and August. However, the WIO warming in June corresponds to La Niña's rapid development (negative SST anomalies over the CP and positive SST anomalies over the MC) in August during the 1996-2019, which favors the enhancement of the NWPSH via enhancing the convection activities in MC and the meridional Hadley circulations. Due to changes in the north-south position of the NWPSH in different months, causing a significantly negative correlation between precipitation in June and August.
This study suggests that the change of SST evolution from June to August is the key factor responsible for the interdecadal variation of precipitation relationship Fig. 13 The average vertical velocity (shadings, 10e −2 Pascal * s −1 , purple counters indicate passing 90% significance test) and latitude-vertical velocity (vectors, m * s −1 , black vectors indicate passing 90% significance test) in 5°S-5°N in June (a) and August (b) during 1995-2019 regressed onto the MCCP in mid-1990s, but why the WIO SST anomalies in June corresponding to different SST anomalies over the tropical Pacific in August before and after mid-1990s remains a puzzle, which needs to be investigated in the future. Additionally, apart from the impact of the WIO SST on the precipitation over SC in June, it seems that La Niña's development also plays a role in contributing to enhance precipitation over SC during the 1996-2019 (Figs. 9, 11). That suggests the enhanced impact of the ENSO developing phase on the precipitation over SC after mid-1990s for both June and August, which will be further clarified in our future work.