Roles of water vapor sources and transport in the intraseasonal and interannual variation in the peak monsoon rainfall over East China


 ERA-Interim reanalysis data from 1979–2017 are utilized to analyze the intraseasonal and interannual variabilities in the moisture transport associated with the summer rainfall over East China. Three key issues of moisture sources for East China summer precipitation are discussed in this study: the identification of the primary moisture sources of summer precipitation, determination of their individual contributions to the intraseasonal variation in summer precipitation, and determination of the extent to which each source affects the interannual variation in precipitation during the peak monsoon rainfall period (PMRP). Through the Lagrangian method, the water vapor is divided into six sources, namely, the land area source (LD source), East China source (EC source, except for the target region), Indian Ocean source (IO source), Pacific Ocean source (PO source), South China Sea source (SCS source), and regional evapotranspiration sources. This study proposes that the rainfall over South China (SC) during the PMRP is mainly influenced by the IO source water vapor transport (52.4%). For the middle and lower Yangtze River valley (YRV), the rainfall contribution from southwesterly moisture transport accounts for approximately 50% of the total rainfall during the PMRP and determines the intraseasonal and interannual variation in the summer rainfall. The water vapor from terrestrial evaporation, which includes EC source water vapor (38.9%) and NC source water vapor (21.8%), is the most important moisture source for rainfall over North China (NC) during the PMRP.


Introduction
Climatically, the summer monsoon system over China has a signi cant effect on water vapor transport, which can have a fundamentally dominant effect on the summer rainfall. Hence, the intraseasonal and interannual variations in summer rainfall are determined by the summer monsoon. Previous studies have revealed that water vapor arrives in East China mainly through three branches during the boreal summer (Zhai and Eskridge 1997;Simmonds et al. 1999;Xue et al. 2003;Zhou and Yu 2005;Chan and Zhou 2005;Fan 2006;Fan et al. 2008;Wang and Chen 2012;Li et al. 2013;Chu et al. 2015). The Indian summer monsoon transports water vapor from the Indian Ocean. The water vapor over the Paci c is moved by southeasterly water vapor transport. The water vapor is transported via the western margin of the western Paci c subtropical high. The terrestrial source rainfall originates from Eurasia, which is fundamentally carried by midlatitude westerlies. Moreover, for East China, the effect of regional evapotranspiration on summer monsoon rainfall cannot be ignored. The abovementioned water vapor transport makes the water vapor sources of the summer rainfall over East China more complex than expected (Ma et al. 2004;Pathak et al. 2014).
Based on the station data provided by the China Meteorological Administration (CMA), the rainfall during the boreal summer constitutes approximately 40% of the annual total rainfall over East China. The anomalous precipitation events can cause severe oods over East China. Hence, the intraseasonal and interannual variation in summer rainfall has been a focus of meteorologists (Ding and Chan 2005;Zhou and Yu 2005;Zhao et al. 2007;Li et al. 2013;Gong et al, 2017Gong et al, , 2019Li et al. 2019). Ding and Chan (2005) found that water vapor transport from the Sino-Indian Peninsula and South China Sea has a considerable in uence on the variation in the summer monsoon rainfall over East China. Zhao et al. (2007) suggested that the southwesterly water vapor transport over East China moves consistently northward during the ood season, as does the associated vertical upward water vapor transport, along with the shift in the main rain band from South China (SC) to North China (NC) via the middle and lower Yangtze River valley (YRV). In addition, Li et al. (2019) found that an anomalous South Asian anticyclone in April is closely associated with an earlier Meiyu onset in June. Moreover, the variations in water vapor transport have a considerable effect on both the interannual variation and intraseasonal variation in the summer monsoon rainfall over East China. Zhou and Yu (2005) proposed that the sources of water vapor associated with the typical anomalous precipitation of China are different from those connected with the mean state during the boreal summer. Li et al. (2013) revealed that the interannual variation in the summer monsoon rainfall over SC is primarily affected by water vapor transport from the Indian Ocean and Bay of Bengal. Although the intraseasonal and interannual variations in the summer rainfall over East China have been widely studied, the corresponding mechanisms have not been clearly illuminated.
Thus, to improve the prediction methods for summer rainfall, the water vapor transport processes of summer rainfall should be studied further.
The Eulerian method was extensively used in the abovementioned studies. The Eulerian method can calculate the temporal variations in meteorological elements over a speci ed region. However, it is di cult to distinguish the contribution from each water vapor source to the summer rainfall based only on the Eulerian method because the water vapor transport processes are not independent of each other. Chen et al. 2011;Jiang et al. 2017). For instance, the summer monsoon rainfall in uenced by the southwesterly or southeasterly water vapor transport over East China cannot be accurately distinguished by the Eulerian method because the transport processes converge over the South China Sea before arriving in East China. Considering the de ciencies of the Eulerian method, the hybrid single-particle Lagrangian integrated trajectory (HYSPLIT) model Hess 1997, 1998) has been extensively utilized to analyze moisture transport systems (Brimelow and Reuter 2005;Perry et al. 2007;Sodemann and Stohl 2009;Van der Ent et al. 2010;Drumond et al. 2011;Jiang et al. 2013Jiang et al. , 2017Sun and Wang 2014a, b;Pathak et al. 2017). The major advantage of the Lagrangian particle dispersion model is that it can calculate the atmospheric water cycle based on the variation in the water vapor content within the air parcel along its motion trajectory (Sun and Wang 2014a, b).
Based on three-dimensional positioning and meteorological reanalysis data, the HYSPLIT model can calculate the motion trajectory of a speci c air parcel. Through the Lagrangian method, Jiang et al. (2013) studied the characteristics of moisture contributions during the boreal summer over the YRV. Chu et al. (2019a) focused on the effect of water vapor transport processes on the variations in the seasonal mean rainfall over East China. Chen and Luo (2018) used the Lagrangian model to explore the paths and sources of the water vapor carried to SC during the pre ood season. Moreover, based on a Lagrangian model, Sun and Wang (2014b) quantitatively calculated the water vapor transport from every water vapor source to East China for the 2000-2009 period.
Climatically, the role of the prime moisture source in the variations in summer rainfall over East China is di cult to identify. Many studies have used the Lagrangian method to reconstruct the source-receptor relationship for the mean climate state (Chu et al 2019a), while few of the studies analyzed the intraseasonal variations in the water vapor transport processes and the relationships with rainfall. Chu et al. (2019b) found that because of the intraseasonal changes in the moisture transport, obvious discrepancies occurred between the interdecadal variabilities in the total rainfall and peak monsoon rainfall over SC during the pre ood season. In fact, although the rainfall over SC during the pre ood season is mainly in uenced by the southwesterly moisture transport, the moisture source is quite different in each month. During the May, the southwesterly water vapor released over SC is carried from the bay of bengal. While, during the June, the rainfall over SC is mainly in uenced by the southwesterly water vapor transport, which carried the water vapor from the tropical Indian Ocean. It means that the pivotal water vapor source of the peak monsoon rainfall is di cult to be clearly identi ed through the source-receptor relationship for the mean climate state. It should be noted that the peak monsoon rainfall period (PMRP) is the most important rainfall period over East China during boreal summer. Anomalous persistent torrential rain occurs frequently during the PMRP and causes severe ooding in East China. Hence, it is necessary to distinguish the role of each water vapor source in the peak monsoon rainfall.
In this study, the interannual and intraseasonal variations in the moisture transport from each source and their relationships with the peak monsoon rainfall over East China during 1979-2017 are quantitatively calculated by the HYSPLIT model. The structure of this study is arranged as follows. Section 2 introduces the reanalysis data and Lagrangian method employed in the paper. The intraseasonal variation in the moisture sources and transport in East China are investigated in Sect. 3. The most in uential water vapor transport for the interannual variation in precipitation during the PMRPs is discussed in Sect. 4. Finally, Sect. 5 provides the conclusions.

Data
In this paper, the output from the HYSPLIT model is utilized to determine the main moisture sources of East China. Based on the results proposed by Trenberth et al. (2011), the ERA-Interim reanalysis data are the most suitable reanalysis data for capturing water vapor transport. Hence, ERA-Interim 6-hourly reanalysis data are employed as the HYSPLIT model input data in our study. Through these data, the water vapor released in the target region can be backtracked. Then, the water vapor sources of the target region can be found. The output from the HYSPLIT model, which is utilized to describe the trajectories, includes the three-dimensional positions of air parcels (longitude, latitude, and altitude). Moreover, we utilize the ERA-Interim reanalysis data, available at 1° × 1° resolution, to quantitatively calculate the contributions of the water vapor transported from every target source (Emil et al. 2016). The abovementioned calculations utilize a 6-hourly mean evaporation and rainfall dataset from 1979-2017.

Methods
During boreal summer, there are three PMRPs over East China: the pre ood season for SC, the Meiyu season for the YRV, and the rainy season for NC (Coching 1934). Essentially, the PMRPs that occur over East China are fundamentally controlled by the East Asian summer monsoon, which gradually moves northward during boreal summer. (Ding et al. 2008;Si et al. 2009). Note that the interannual variations in the summer rainfall over SC, the YRV, and NC are signi cantly different. Previous studies indicated that strong southeasterly water vapor transport can lead to more summer rainfall over NC and SC and less summer rainfall over the YRV, and vice versa (Zhou and Yu 2005;Ding et al. 2009). The out-of-phase relationship in the summer rainfall over East China is generally led by the western Paci c subtropical high. Hence, the water vapor sources of the summer rainfall over East China should be subdivided. Based on previous studies, Chu et al. (2019a)  A Lagrangian particle dispersion model, namely, HYSPLIT 4.9, which was proposed by Hess (1997, 1998), is utilized to calculate the trajectories of a large number of particles representing the air parcels over East China in the period of 1979-2017. Li et al. (2010) indicated that the water vapor is mainly transported below the upper troposphere. Thus, we select to release the particles from the 500 m level to the 3500 m level, with a 500 m interval (seven levels are selected). The particles are released over these regions with a 1° × 1° resolution. Moreover, based on the conclusion of Trenberth et al. (2011), the mean time of one atmospheric water cycle process, which ranges from evaporation to precipitation, is 10 days. Thus, the air parcels over East China are backtracked for 10 days to ensure that the trajectories are of acceptable accuracy. The particles are traced backed every 6 hours during the rainy season in SC, the YRV, and NC (1 May to 31 August, 123 days).
The HYSPLIT 4.9 model is utilized to calculate the water vapor sources of the summer monsoon rainfall over East China. For these three subregions of East China, more than 1,000,000 particles representing air parcels are backtracked by the Lagrangian method. To utilize a systematic distribution of the trajectories to concisely introduce the water vapor sources of the summer rainfall over East China, Chu et al. 2019a used the distribution of air parcels 10 days before reaching the target region to represent the real trajectories.
To reasonably estimate the role of each moisture source in the variation in rainfall in a target region, it is necessary to calculate the variations in the water vapor maintained in the air parcels along the trajectories. Considering the in uences of evapotranspiration and rainfall along the motion trajectories, Sodemann and Stohl (2009) calculated the amount of water vapor transported from the speci c evapotranspiration source locations to the target locations. To improve the applicability of the abovementioned method, Sun and Wang (2014a) presented the "areal source-receptor attribution method" to compute the amount of water vapor transported from the examined evapotranspiration sources to a target region. Chu et al. (2019a) revealed that the abovementioned method based on the Lagrangian model is valid and credible and is suitable for studies over East China during the ood season.
The areal source-receptor attribution method proposed by Sun and Wang (2014a, b) mainly includes seven steps for computing the moisture uptake and release along the trajectories from an areal source region to an areal target region. The core of this method is to calculate the moisture uptake and release along the target-bound trajectories of each air parcel. Speci cally, the moisture uptake over an examined source region can be obtained by calculating the net moisture gained over the source region. The en route loss can be obtained by proportionally subtracting the moisture released en route from the original moisture uptake region over the examined source region. The contribution to the precipitation in the target domain is thus obtained by calculating the ratio of the moisture released from the examined source region to the total moisture release over the target domain.

Roles Of Water Vapor Sources In The Intraseasonal Variation In The Summer Monsoon Rainfall
As shown in Fig. 2, the southwesterly moisture transport trajectories account for 54.6% of the total trajectories arriving in SC, representing the primary branch of moisture transmission in SC (Fig. 2a).
Furthermore, 35.4% of the air parcels along the southeasterly moisture transport trajectories arrive in SC. For the YRV, 51.4% of the air parcels originate from the southwesterly moisture transport trajectories, and 31.7% of the air parcels come from the southeasterly moisture transport trajectories (Fig. 2b). In addition, 62.9% of the air parcels that arrive in NC are controlled by midlatitude westerlies. The air parcels coming from the southern region account for 16.8% of the total trajectories arriving in NC (Fig. 2c). However, the trajectories over NC exhibit signi cant intraseasonal variations. During the NC rainy season (from mid-July to mid-August), which is the main precipitation period, the air parcels come from the southern region, accounting for more than 25% of the total trajectories. However, note that a fair number of the air parcels move at higher levels, whereas the evaporation and release of the moisture is mainly cycled at relatively lower levels. In addition, because the average residence time for moisture in the atmosphere is 10 days, this difference may limit the moisture transport from a given region to a target region. Therefore, the prime source of air parcels may not be the most important moisture source of precipitation over a target region.
Notably, to analyze the effects of water vapor sources on the variations in the peak monsoon rainfall, the local evaporation and precipitation along the air parcel trajectories should be the focus. Notably, the water vapor carried over the South China Sea and East China should be calculated independently because it is in uenced by southwesterly and southeasterly moisture transport. Moreover, considering recycled precipitation, the target region should also be considered an independent moisture source. Accordingly, Chu et al. (2019a) divided the water vapor sources of summer rainfall over East China into six categories. As shown in Fig. 2d, these six moisture sources include the land area source (LD source), East China source (EC source, except for the target region), Indian Ocean source (IO source), Paci c Ocean source (PO source), South China Sea source (SCS source), and target regional evapotranspiration source (SC source, YRV source, and NC source). Then, the mean rainfall contribution originating from every water vapor source that contributes to East China is quantitatively calculated.
For SC and the YRV, the water vapor carried over the IO source and SCS source, which is controlled by the Indian summer monsoon, has the primary effect on the summer rainfall (shown in Fig. 3d and 3e). The summer rainfall originating from the above-mentioned sources occupies more than 50% of the total summer rainfall over SC (Fig. 4a). Considering the moisture lost en route, a fair amount of the IO source and SCS source rainfall are lost during the travel of the air parcels. During the boreal summer, there is signi cant northward water vapor transport over East China. Hence, as shown in Fig. 3b and 3c, the EC source precipitation and recycled precipitation contribute much less rainfall over SC than over other areas. However, for the YRV and NC, the recycled rainfall accounts for a fair amount of the summer rainfall. The water vapor from the EC source also has a considerable effect on the summer rainfall over the YRV and NC. Moreover, the East-Asia-summer-monsoon-controlled water vapor transport mainly releases water vapor over the southeastern YRV and eastern SC (Fig. 3f). The in uence of the LD source moisture transport controlled by the midlatitude westerlies is mainly limited to NC (Fig. 3a). Here, note that signi cant intraseasonal variations in the primary water vapor transport channel occur over East China during the ood season. Chu et al. (2019b) found that because of the intraseasonal variabilities in the water vapor transport, there is a signi cant difference between the interdecadal variabilities of the total rainfall and peak monsoon rainfall over SC during the pre ood season. Therefore, the effects of water vapor transport on the intraseasonal variations in summer rainfall are discussed in the following.
The intraseasonal variations in the moisture contribution from each terrestrial and oceanic source to the daily rainfall are shown in Fig. 4. For SC, the YRV, and NC, there are obvious differences in the primary moisture sources of the summer monsoon rainfall. The contribution from the IO source to SC is described by a bell-shaped curve with a maximum contribution during June (Fig. 4a). For the peak IO source contributions, IO source rainfall constitutes more than 50% of the total rainfall over SC. Additionally, the PO source water vapor transport decreases signi cantly, led by the enhanced Indian-summer-monsooncontrolled water vapor transport. The water vapor contribution by the SCS source to SC is in uenced by both the southwesterly and southeasterly water vapor transport. The southwesterly water vapor transport is the primary moisture source of SC at the onset of the Indian summer monsoon. The southeasterly water vapor transport plays the most important role in the rainfall over SC as the Indian summer monsoon abates. Thus, the curve of the SCS source precipitation over SC is at during the boreal summer (Fig. 4a). In addition, because the vertically integrated moisture transport over SC is directed southward during the ood season, there is a signi cant decrease in the moisture carried from terrestrial sources (LD source, EC source, and SC source).
For the YRV rainfall during the PMRP, there are two peaks of summer precipitation. Because of the enhancement in the southwesterly moisture transport, the moisture from the IO source, SCS source, YRV source, and EC source is transported northward. The moisture from the SCS source, YRV source, and EC source rainfall reach and is released over the YRV earlier than IO source moisture, which leads to the rst peak (Fig. 4b). Then, there is a signi cant increase in the IO source rainfall during the second peak. At the same time, the SCS source, YRV source, and EC source rainfall decrease because a fair amount of the moisture from these sources is transported to NC by the enhanced southwesterly moisture transport (Fig. 4b). In addition, Fig. 4b shows that the water vapor carried from the PO source and LD source is reduced during these two peaks. Thus, for the YRV, the Indian-summer-monsoon-controlled moisture transport is the primary moisture transport contributing to rainfall during the PMRP.
The intraseasonal variations in the water vapor transport processes for NC are shown in Fig. 4c. As shown in Fig. 4c, the intraseasonal variations in the summer rainfall are dominated by the water vapor transport processes from the EC source and NC source, which account for more than 50% of the total rainfall over NC. During the PMRP, except at the beginning, the precipitation curve is at, as observed in Fig. 4c, which means that only a small percentage of the data changes occur because of the water vapor transport from the LD source. In addition, it is observed that the IO source precipitation over NC decreases during the PMRP, and a further decrease is observed when the contribution from the PO source increases. Note that the peak monsoon rainfall over NC occurs between the end of July and the beginning of August, which is slightly later than the occurrence of the peak of IO source precipitation and slightly earlier than the largest contribution from the PO source. Hence, the major moisture transport channel of precipitation over NC during the PMRP needs to be further discussed.
To illuminate the major water vapor transport during the PMRPs, three time series of a duration of 15 days, the peak of the summer precipitation over SC (P1: 6-20 June), the peak of the summer precipitation over the YRV (P2: 23 June-7 July), and the peak of the summer precipitation over NC (P3: 17-31 July), are selected for further studies. The intraseasonal variations in each water vapor transport process are shown in Table 1. For SC, 4.48 mm/d and 1.96 mm/d water vapor are carried from the IO source and SCS source, respectively, through southwesterly water vapor transport during P1, thus dominating the precipitation during the PMRP (Table 1). Then, during P2, the average values of water vapor carried from the IO source and SCS source to the peak monsoon rainfall over the YRV increased by 1.21 mm/d and 0.29 mm/d, respectively. Furthermore, the variations in the water vapor carried from other sources are minimal (Table 1). For NC, the water vapor transport processes originating from the EC source and NC source have the foremost in uence on the intraseasonal variation in the peak monsoon rainfall (Table 1). However, for P3, the water vapor contributions from the IO source and SCS source increased by only 0.17 mm/d and 0.19 mm/d, respectively (Table 1). Compared to those in P2, there is an approximately 50% increase in the rainfall contributions from the IO source and SCS source during P3. During the same period, the moisture contributions from the PO source increase by 0.20 mm/d. This result means that the EC source rainfall over NC is dominated by both the Indian-summer-monsooncontrolled and East-Asian-summer-monsoon-controlled water vapor transport, which carry water vapor from the YRV and SC. To illuminate the effects of different water vapor sources on the intraseasonal variations in the peak monsoon rainfalls over East China, we analyze the spatial variations in the water vapor carried from the abovementioned water vapor sources and released over the target regions (shown in Fig. 5, Fig. 6, and Fig. 7). Moreover, the PMRPs denote the foremost consistent rainfall period over East China, which is connected with strongly converged water vapor transport. The daily mean anomalies in the vertically integrated moisture transport and vertically integrated moisture ux divergence during the PMRPs based on composite analysis are shown in Fig. 8, and an anomalous anticyclone shift northwestward over the western North Paci c can be observed during the boreal summer (Fig. 8).
During P1, because of the enhancement in the Indian-summer-monsoon-controlled water vapor transport, there is a positive anomaly of westerly water vapor transport over SC (Fig. 8a). For SC, the rainfall increases a fair amount, led by the increase in the IO source water vapor transport (Fig. 5e1). Meanwhile, the meridional water vapor transport from the SCS source increases signi cantly, being primarily affected by the western Paci c subtropical high (Fig. 8a). There is an obvious increase in the peak monsoon rainfall over eastern SC, and the main increase in the water vapor originates from the SCS source ( Fig. 5d1). Furthermore, for the southeastern YRV, the signi cant increase in rainfall is primarily in uenced by the Indian-summer-monsoon-controlled water vapor transport, which carries water vapor from the IO source, SCS source, and SC source (Fig. 6c1, 6d1, and 6e1). During this period, the oceanic source moisture transport has minimal in uence on the precipitation over NC.
During P2, a signi cant enhancement in the meridional water vapor transport over southeastern China occurs, induced by the shift in the western Paci c subtropical high (Fig. 8b). Meanwhile, there is a positive anomaly of water vapor ux divergence over the YRV region. Because of the anomalous anticyclone, it is di cult for the water vapor carried from the PO source to be directly transported to East China. Therefore, during the PMRP, a signi cant reduction occurs in the PO source water vapor transport to East China ( Fig. 5f2 and Fig. 6f2). Both the Indian-summer-monsoon-controlled and East-Asiansummer-monsoon-controlled water vapor is transported via the South China Sea before arriving in East China. Hence, during PMRPs, the meridional water vapor motion from the SCS source enhances signi cantly. The Indian-summer-monsoon-controlled water vapor transport has the foremost effect on the peak monsoon rainfall over the YRV, which leads to a signi cant increase in the rainfall over the whole YRV (Fig. 6e2). Furthermore, the increase in the rainfall over the eastern YRV is mainly affected by the SCS source and SC source water vapor transport ( Fig. 6c2 and 6d2). Therefore, the increase in SC source water vapor transport is primarily in uenced by oceanic water vapor transport, which carries water vapor from southern China. In addition, during P2, the peak monsoon rainfall over NC is primarily controlled by terrestrial source moisture transport. The increase in the LD source, SC source, and EC source water vapor transport leads to a signi cant increase in the rainfall over northeastern NC (Fig. 7a2,   7b2, and 7c2).
During P3, because of the shift in the western Paci c subtropical high, the northwest Paci c anticyclone moves northwestward. The westerly moisture transport anomalies, via the southern bound of the northwest Paci c anticyclone, decrease the zonal moisture transport over southeastern China. There is a positive anomaly of water vapor ux divergence over NC (Fig. 8c). Meanwhile, a signi cant anomaly of southerly moisture transport over NC is observed. Therefore, the water vapor contribution from oceanic water vapor transport sources to the rainfall over NC increase signi cantly (Fig. 7d3, 7e3, and 7f3). However, the water vapor carried from terrestrial sources, including the EC source and SC source, has important effects on the peak monsoon rainfall over NC during P3 ( Fig. 7b3 and 7c3). Note that the anomalous center of the westerlies-controlled rainfall is over northern NC (Fig. 7a3). The anomalous center of the SC source rainfall is over southern NC (Fig. 7c3). The abovementioned results mean that a fair amount of the land evapotranspiration water vapor released over NC is primarily controlled by southwesterly and southeasterly water vapor transport, which carries the water vapor from southeastern China.
In summary, the intraseasonal variations in the peak monsoon rainfall over the YRV and SC are principally affected by the Indian-summer-monsoon-controlled water vapor transport. The intraseasonal variation in the peak monsoon rainfall over NC is in uenced by both the Indian-summer-monsooncontrolled and East-Asian-summer-monsoon-controlled water vapor transport processes, which carry water vapor from southeastern China.

Roles Of The Water Vapor Sources In The Interannual Variation In The Peak Monsoon Rainfall
The persistent rainfalls over East China during PMRPs determine the effect on the spatiotemporal characteristics of the total rainfall during the boreal summer (Chu et al. 2019b). Anomalous persistent torrential rain can often cause severe oods (Shen et al. 2007;Chu et al. 2016). Considering the differences in the water vapor transport channels over East China, the related mechanisms controlling the climate dynamics of these moisture transport channels are complex and different. Simplifying the moisture transport system over East China bene ts the forecasting of the summer rainfall and even the shift in the summer rain belt. Based on the conclusions in Sect. 3, the peak monsoon rainfall is induced by moisture transport from individual moisture sources (Fig. 4). Hence, it is necessary to study the in uence of the primary water vapor sources and channels during the PMRPs.
Climatically, the IO source rainfall determines the amount of monsoon rainfall over the YRV and SC during the PMRPs (Fig. 4). The rainfall over NC is signi cantly affected by the EC source moisture transport. However, the interannual variations in the PMRPs should be considered in the study of the primary moisture source in the peak monsoon rainfall. Hence, based on a 15-day running mean method, the PMRPs, which have the most precipitation during the rainy season, are picked out from the pre ood season, mei-yu season, and rainy season period provided by CMA for each year. Then, to illustrate the role of the primary moisture source in the interannual variation of peak monsoon rainfall, the linear trends of the daily mean rainfall and moisture contribution from the primary moisture source were removed before the aforementioned linear regression. The time series of the daily mean rainfall and moisture contribution from the primary moisture source during PMRPs are shown in Fig. 9. The correlation coe cients for the detrended time series of the daily mean rainfall and moisture contribution from the primary moisture source are signi cant at the 99% con dence level. Based on Fig. 9, 10 years that exhibit the strongest (weakest) rainfall during PMRPs are selected as the strong (weak) rainfall years. For PMRPs in the strong and weak rainfall years, the amount of water vapor carried from every water vapor source to the target regions is calculated (Table 2). For SC, the difference in the IO source rainfall between the strong and weak rainfall years is 3.19 mm/d (Table 2). For the strong rainfall years, the average rainfall contribution from the IO source to the YRV region increases by 2.84 mm/d ( Table 2). As Table 2 shows, for strong rainfall years, the average water vapor transport from the EC source to the NC region increases by 1.33 mm/d. Furthermore, note that the LD source rainfall, which is controlled by the moisture transport due to the midlatitude westerlies, increases only by approximately 0.2 mm/d. The differences in the water vapor transport processes from six water vapor sources to the rainfall over East China during PMRPs are shown in Fig. 10, Fig. 11, and Fig. 12. For the peak monsoon rainfall over SC, an obvious increase in rainfall occurs over SC because SC receives a fair amount of atmospheric water vapor from the IO source (Fig. 10e). For the strong rainfall years, the distribution of the peak monsoon rainfall over the YRV is determined by the IO source water vapor transport (Fig. 11e). During the same period, the PO source and LD source rainfall, which are controlled by the moisture transport from the southeast and midlatitude westerlies, respectively, only slightly change ( Fig. 11a and 11f). Therefore, the increase in the EC source, YRV source, and SCS source water vapor transport is mainly in uenced by the southwesterly moisture transport (Fig. 11b, 11c, and 11d). As shown in Fig. 12c, for the peak monsoon rainfall over NC, the interannual variation is foremost affected by the EC source water vapor transport. The anomalous EC source moisture, which originates from the YRV and SC, is mainly released over western NC. Furthermore, a signi cant increase occurs in the IO source, SCS source, and PO source water vapor transport (Fig. 12d, 12e, and 12f). This means that the interannual variation in the peak monsoon rainfall over NC is affected by southwesterly and southeasterly water vapor transport.
The results above suggest that the rainfall contribution from the primary moisture source dominates the spatiotemporal characteristics of the rainfall during PMRPs. Hence, the precipitation contribution from the primary moisture source can be used instead of the total precipitation, which is bene cial for simplifying the related mechanisms controlling the climate dynamics of summer precipitation over East China. As shown in Fig. 13 and Fig. 14, the detrended time series of the peak monsoon rainfall over SC, the YRV, and NC is used to regress the anomalies of evaporation and vertically integrated water vapor transport.
A previous study indicated that East-Asian-summer-monsoon-controlled water vapor transport mainly manifests as meridional water vapor transport, while Indian-summer-monsoon-controlled water vapor transport mainly manifests as zonal water vapor transport during the boreal summer (Huang et al., 1998). Zhang (2001) found that strong meridional water vapor transport always corresponds to weak zonal water vapor transport over southern China. The opposite trend is observed when the zonal moisture transport is enhanced. Figure 13a shows the differences in the water vapor transport over SC during P1.
The westerly moisture transport anomalies over southeastern China, the southerly moisture transport anomalies over the South China Sea, and the northerly water vapor transport anomalies over the western North Paci c manifest a clockwise gyre pattern, meaning that the climatic clockwise moisture transport over the South China Sea connected with the subtropical high is strengthened during the strong rainfall years of SC. At the same time, there is an obviously positive anomaly in the Indian-summer-monsooncontrolled water vapor transport, which primarily carries the water vapor from the IO source through the Arabian Sea-India-Bay of Bengal-Indochina Peninsula-Southwest China-SC moisture transport branch.
The enhanced IO source water vapor transport leads to a signi cant positive anomaly in the rainfall over SC during strong precipitation years. Figure 14a shows that the IO source water vapor that arrives and releases over SC primarily originates from the Bay of Bengal, South India, and the Arabian Sea. Figure 13b shows the differences between the P2 moisture transport of the strong and weak precipitation years of the YRV. Climatically, during P2, the edge of the anomalous anticyclone over the western North Paci c gradually moves northwestward, resulting in southerly moisture transport anomalies reaching the YRV via the western bound of the western Paci c subtropical high. During the strong rainfall years of the YRV, the zonal water vapor transport over the YRV and the meridional water vapor transport over the South China Sea and SC are enhanced signi cantly (Fig. 13b). This enhanced water vapor transports were mainly in uenced by the enhancement of the subtropical high and Indian summer monsoon. Hence, through the Bay of Bengal-Indochina Peninsula-Southwest China-SC-the YRV moisture transport branch, a fair amount of water vapor from the IO source arrives and releases over the YRV. The increase in the rainfall carried by the IO source that is released over the YRV is signi cantly in uenced by evaporation over the Arabian Sea (Fig. 14b). Figure 13c shows the differences between the P3 moisture transport of the strong and weak precipitation years of NC, which exhibit a clockwise gyre pattern of moisture transport anomalies over the western North Paci c, meaning that the climatic clockwise moisture transport over the western North Paci c connected with the subtropical high increases during the strong rainfall years in comparison to that in the weak rainfall years. Meanwhile, the zonal water vapor transport over southeastern China weakens signi cantly, which leads by the weakening of the Indian summer monsoon and the strengthening of the subtropical high. Accordingly, there is an obviously positive anomaly of southerly moisture transport over East China. Therefore, for strong rainfall years in NC, the strong meridional water vapor transport, which carries water vapor from the SC source, leads to an obvious increase in rainfall. SC and the southern YRV are the primary evaporation sources of rainfall over NC (Fig. 14c).
In summary, for East China, the intensity of the subtropical high and Indian summer monsoon has an important effect on Indian-summer-monsoon-controlled water vapor transport, which is the foremost water vapor source of rainfall during PMRPs. The spatial distribution of the IO source, PO source, and EC source rainfall is signi cantly affected by the position of the subtropical high. Hence, the subtropical high and Indian summer monsoon in uence and may even control the amount and distribution of the summer precipitation over East China. In addition, note that the anomalous anticyclone in the background circulation during summer provided unfavorable climatic conditions for PO source water vapor transport.
Thus, the PO source water vapor transport hardly in uences the interannual variation in the rainfall over SC and the YRV during PMRPs.

Conclusions And Discussion
This study uses the Lagrangian method and areal source-receptor attribution method to illustrate the effects of water vapor transport on the intraseasonal and interannual variations in the summer monsoon rainfall over East China. In this paper, three key issues of the moisture sources of East China summer precipitation are discussed: the identi cation of the primary moisture sources of summer precipitation, determination of their individual contributions to the intraseasonal variation in summer precipitation, and determination of the extent to which each source affects the interannual variation in the precipitation during the PMRP. The main ndings of this study focus on the following aspects: Based on the water vapor branches debated by previous studies, the water vapor sources for East China are de ned by six categories. The southwesterly water vapor transport, which includes the IO source and SCS source water vapor transport, is the primary source of the summer monsoon rainfall over SC and the YRV. Considering the water vapor lost en route, the amount of Indian-summer-monsoon-controlled water vapor decreases during transport. The southeasterly moisture transport mainly releases moisture over the southeastern YRV and eastern SC. The in uence of the LD source moisture transport is mainly limited to NC. The amount of EC source precipitation and recycled precipitation over SC remains low, whereas there is a large amount of EC source precipitation and recycled rainfall over the YRV and NC.
The intraseasonal variations in the summer monsoon rainfall over the YRV and SC are fundamentally in uenced by the Indian-summer-monsoon-controlled water vapor transport. The intraseasonal variation in the summer monsoon rainfall over NC is fundamentally in uenced by both the Indian-summermonsoon-controlled and East-Asian-summer-monsoon-controlled water vapor transport, which mainly carries the water vapor from the YRV and SC. The amount and distribution of summer precipitation during PMRPs are signi cantly affected by the persistent rainfall over East China. The rainfall contribution from the primary moisture source dominates the spatiotemporal characteristics of the rainfall during PMRPs. Thus, the IO source moisture transport determines the interannual variation in the peak monsoon rainfall over the YRV and SC. The interannual variation in the summer monsoon rainfall over NC is signi cantly affected by the EC source moisture transport, which is mainly affected by the Indian-summer-monsoon-controlled water vapor transport.
The intensity of the subtropical high and Indian summer monsoon has an important effect on the Indiansummer-monsoon-controlled water vapor transport, which is the foremost water vapor source of rainfall during PMRPs. The spatial characteristics of the IO source and EC source rainfall are signi cantly affected by the position of the subtropical high. Hence, the subtropical high and Indian summer monsoon in uence and may even control the spatiotemporal characteristics of summer monsoon rainfall over East China. Moreover, the PO source water vapor transport plays an important role in the interannual variation in the peak monsoon rainfall over NC.
In this study, the effects of water vapor transport on the intraseasonal and interannual variations in rainfall over East China during PMRPs are discussed. During PMRPs, oceanic moisture transport has the most important impact on the rainfall over East China. The relationships between the low-frequency modes and the precipitation originating from each moisture source should be studied further.
Furthermore, although the PO source moisture transport has minimal in uence on the peak monsoon rainfall over East China, the water vapor contribution from the PO source during the period after P3 greatly increases. Hence, the PO source moisture transport may need to be analyzed in future research, which will require more calculations and discussions. Note that because of the disadvantages of the Lagrangian model, only one "evaporation (uptake)-precipitation (release)" cycle is considered in the current study. The analysis of multicycle processes may be necessary for future research, which will necessitate more computations and discussions.
Moreover, the intraseasonal variations in the moisture contributions from target regional evapotranspiration sources are quite different. For NC, during the PMRP, the increase in the NC source precipitation is signi cantly affected by the increase in the EC source precipitation. However, for the YRV and SC, the YRV source and SC source precipitation are constant during PMRPs. Because of the southerly water vapor transport over East China during PMRPs, a fair amount of water vapor originated from the YRV source and SC source transport released over NC. This means that the EC source rainfall over NC is fundamentally affected by the evapotranspiration over the YRV and SC (Fig. 14c). Considering the delaying effect of summer rainfall (Ma et al. 2004;Pathak et al. 2014), the evapotranspiration over the YRV and SC during P3 (P2) is signi cantly affected by the precipitation over the YRV and SC during P2 (P1). Therefore, the relationship between the precipitation over SC during P2 and the precipitation over NC during P3 is worth studying. Declarations