Figure 1 shows the relative contribution rates of the seven moisture sources to the climatological summer precipitation over China. The results are then further summarized in Figure 2 as a pie chart. In addition, the moisture trajectories of air parcels 10 (8, 6, 4, and 2) days before reaching each subregion is presented in Figure 3. It should be noted that a fair amount of the air parcels move at high latitudes, whereas the evaporation and release of moisture are mainly cycled at relatively lower latitudes. Broadly speaking, most air parcels arriving in the three northern subregions (i.e., Northwest, North and Northeast China) originate from the CN, MLW, SibArc and OKS sources, whereas air parcels that arrive in the southern subregions (i.e., Southwest, South and Southeast China) are governed predominantly by the tropical ocean and CN sources. Specifically, the precipitation center of the MLW source is located in Northwest and Northeast China, with area-averaged contribution rates of approximately 29.4 and 14.4%, respectively (Figures 1A, 2A,C). The precipitation of the SibArc and OKS sources contributes little to summertime precipitation over the whole of China (around 1–7%), and the target region is spatially confined to Northwest and Northeast China, which is located in the vicinity of the source origins (Figures 1B,C). The moisture trajectory of the MLW, SibArc and OKS sources can also be seen from the evolution of air parcel locations 10–2 days before reaching Northwest and Northeast China (Figures 3A,C). Regarding the tropical ocean sources, the IO source precipitation plays a decisive role in the precipitation of southwestern China (48.4%), with the magnitude then decreasing northeastward radically from Southwest to Northwest China, the YRV, and South China (Figures 1D, 2A,C–E, 3D–F). The SCS and PO sources precipitations have similar impacts on East China (Figures 1E,F, 3D–F). The main difference is that the former makes a more prominent contribution to the precipitation of South China, with the contribution rate decreasing with latitude (Figures 1E, 2F, 3F), while the latter mainly affects the precipitation of Northeast China and the contribution rate weakens westwards from coastal to inland regions (Figures 1, 2C, 3D–F).

In addition to the six remote sources, most air parcels arriving in China are primarily dominated by the local CN source (Figures 1G, 2, 3). Notably, considering the local evaporation, the contribution rates of the CN source precipitation differ considerably among the different regions—roughly between 25.8 and 68.3%. The center of China, including the North–Northwest China, YRV and Southwest China subregions, is overwhelmingly influenced by moisture transport from surrounding areas (∼80%), which arguably reflects the strong feedback of local evaporation. This finding is consistent with those of Numaguti (1999) and Sun and Wang (2014b), who attributed continental evaporation as the primary origin for continental precipitation. Technically, the contribution of the CN source appears to be overestimated, partially due to the 10-day tracing method, which considered only one process of evaporated moisture uptake, transport and release and therefore omitted any distant transport beyond 10 days. To sum up, these seven sources cumulatively account for greater than 95% of the total summertime precipitation, which may provide valuable precursors for synoptic-scale rainfall forecasts. The remaining contributions will come from residual sources across the globe, especially for the precipitation of Northwest China (Figure 1H).

Figure 4 shows the variations of total precipitation and relevant sources of precipitation in the different subregions of China. For each subregion, summertime precipitation shows distinct interdecadal and interannual variations. Broadly speaking, the interannual components of precipitation explain larger amounts of the total variance than the interdecadal components (Table 1). The precipitation of Northwest China and its CN source display a similar increasing trend from 1979, superimposed with clear interannual variations (Figure 4A), and their correlation coefficient is high at 0.92 (Table 2). Their interdecadal components explain approximately 18% of the total variance, with the remaining 82% from interannual timescales. The MLW and IO sources (i.e., the 2nd and 3rd) show clear interannual variations, accounting for 88 and 97% of the total variance, respectively. In comparison, the summertime precipitation in North–Northeast China, the YRV, and South China show similar interdecadal variations superimposed with interannual variation, which is characterized by a consecutively southward retreat of summertime precipitation on interdecadal timescales (Figures 4B–E). In particular, the interdecadal component of YRV precipitation accounts for 44% of the total variance. Specifically, there is increased rainfall in Northeast China peaking in the mid-1980s, a southward displacement to North China peaking in the early 1990s, a rapid southward shift to the middle and lower YRV peaking at 1998, and then a further extension to South China peaking in the early 2000s. The decadal evolution of rain belts is consistent with the findings of previous studies (Ding et al., 2009; Zhu et al., 2011; Chu et al., 2019), which reported similar features in terms of a southward retreat. One common trait they share is the crucial role of moisture from the CN source in terms of total precipitation. For North China, around 68.3% of the total precipitation is determined by the CN moisture source, while 11.8% is determined by the PO moisture source (Figures 2B, 3B). Similarly, the CN and PO moisture transport pathways account for 52.7 and 19% of the rainfall in Northeast China, respectively, and show noticeable interdecadal variations (explained variances of 24 and 12%, respectively), whilst the MLW moisture transport pathway displays evident interannual variability (explained variance of 86%). For the YRV and South China, the SCS moisture source plays a secondary and a dominant role in total rainfall, respectively, featuring similar interdecadal variations with an explained total variance of roughly 46%. In addition, the IO source makes interdecadal contributions to the rainfall in the YRV and South China (explained variance of 29 and 22%, respectively), and the PO source makes interannual contributions (93 and 99%, respectively). As for Southwest China, the interdecadal change in precipitation is distinct from the other subregions, being characterized by decreasing rainfall from 1979 and increasing rainfall since the early 2000s, and the explained variance of the interdecadal component is 15% (Figure 4F). Almost identical interdecadal changes are found for the SCS source, with an explained variance of 19%, whilst similar interannual variations exist for the CN and IO sources, both with an explained variance of 95%. These results suggest that the variations of summertime precipitation differ among the six subregions of China, which is fundamentally due to the significant variations in moistures source and their relative contributions.

To better understand the causes and mechanisms involved in the variations of summertime precipitation in China, we analyzed the large-scale atmospheric circulation anomalies associated with the moisture sources. The key circulation patterns were identified to investigate their impacts on summertime precipitation in the different subregions. Figures 5–10 show the water vapor flux anomalies accumulated in the lower–middle troposphere and 500-hPa geopotential height anomalies regressed on the precipitation source, and the correlation of the precipitation source anomalies with the key atmospheric circulation pattern indices.

For Northwest China (Figure 5), the CN, MLW, IO and SibArc sources are the four leading ones, accounting for 52.3, 29.4, 15.3 and 0.9% of the total precipitation, respectively. The circulation pattern associated with the MLW source shows an evident Eurasian wave train, with anomalously high pressure over a region covering Siberia to the Kara Sea and anomalously low pressure over the regions of North Europe and North Asia to Japan (Figure 5B). Such a high-latitude wave train is also apparent in the circulation pattern associated with the SibArc source (Figure 5D). Note that the circulation anomalies over the region from the Arctic to the North Atlantic project largely to the negative phase of the North Atlantic Oscillation, consistent with Hua et al. (2017). However, for the CN source, there is a fairly weak high-latitude wave train pattern over the Eurasian continent, and the circulation anomalies over China are insignificant (Figure 5A), suggesting an important contribution from local evapotranspiration. For the IO source, a similar but slightly southward-shifted wave train is identifiable over the midlatitudes extending from Europe to China and resembling the East Atlantic/West Russia pattern. Meanwhile, enhanced easterly winds are apparent over the northern India, representing an enhanced ISM (Figure 5C). Here, we define a high-latitude and a midlatitude wave train index, as described in Data and Methods. The commonality of the high-latitude and midlatitude wave trains is the cyclonic anomaly over northern China and the anticyclonic anomaly over southern China associated with the enhanced ISM. The cyclonic circulation anomaly facilitates cold-air outbreaks from high latitudes, and the anticyclonic tropospheric anomaly favors moisture transport from the tropical oceans. Therefore, the northwesterly and southwesterly branches of moisture pathways converge in Northwest China, resulting in excessive precipitation over northern and southern Northwest China, respectively (Figures 5E–H).

For North China (Figure 6), the CN, PO, MLW and SCS sources tend to be the major contributors, with contribution rates of 66.3, 11.8, 7.8 and 7.1%, respectively. The critical role of the CN source is in agreement with previous studies reporting that evaporation from the land area over and south of North China is the most important contributor (around 61.1%) to the summertime rainfall in North China (Sun and Wang, 2014; Chu et al., 2019). The circulation patterns related to the CN and MLW sources are a similar high-latitude wave train, and the anomalously low center over Northeast China will certainly lead to a uniform increase in summertime precipitation over North China (Figures 6A,C). In particular, the MLW source gives rise to a rainfall belt across North China (Figure 6G). Besides the CN source, the PO is the most significant source of moisture and dynamics for precipitation in North China. An apparent meridional wave train that closely resembles the positive phase of the PJ teleconnection pattern acts as an atmospheric bridge linking the PO moisture source to rainfall in North China, accompanied by cyclonic (anticyclonic) anomalies in the tropical (midlatitude) Northwest Pacific (Figure 6B). Such anomalously strong convective activity over the tropics tends to the promotion of a Rossby wave teleconnection in the poleward direction (Kosaka and Nakamura, 2010). The East Asian–western North Pacific monsoon (Wang et al., 2008) provides an additional source of moisture from the PO for the abundant rainfall over eastern North China (figure not shown). In addition, the SCS source is associated with an enhanced WNPSH, which brings abundant water vapor along the western margin of the subtropical high (Figure 6D). This finding is consistent with those of Ding et al. (2009) and Jiang et al. (2017) who reported that the southwesterly moisture transport associated with the PO and SCS sources is the dominant moisture pathway for variations of North China precipitation. In summary, the mid–high-latitude wave train, subtropical high, and PJ teleconnection are dominant circulation systems for North China precipitation.

Figure 7 shows the major circulation patterns that govern the variations of Northeast China precipitation. The CN, PO, MLW and OKS sources explain 52.7, 19, 14.4 and 7% of total rainfall, respectively. A common feature for the circulation anomalies is the prominent cyclonic anomaly over Northeast China, which can be regarded as the Northeast China vortex—the classic climate system affecting northern China. This pattern is especially evident in the circulation patterns associated with the CN, MLW and OKS sources (Figures 7A,C,D). It is reasonable to assume that the Northeast China vortex converges together the water vapor branches from the west (MLW source), south (CN source), and east (OKS source), and thus produces heavy rainfall over vast regions of North and Northeast China (Figures 7E,G,H). However, an anticyclonic circulation system associated with the PO source is identifiable over the Okhotsk Sea, reflecting the remote influence of the PJ teleconnection (Figure 7B). The southwesterly moisture transmission branch transports moisture from the PO along the western margin of the anticyclone and yields excessive rainfall over East China (Figure 7F). In addition, the precipitation over Northeast China might be influenced by the East Asian summer monsoon, which transports water vapor from the SCS source (Sun et al., 2017)—something that can also be seen in the OKS-related circulation pattern (Figure 7D). To sum up, summertime precipitation in northern China, including the Northwest, North and Northeast China subregions, is primarily affected by atmospheric regimes in in mid–high-latitudes, such as mid–high-latitude wave trains, the Northeast China vortex, and the PJ teleconnection. The exception is that the subtropical high has additional impacts on North China precipitation and the ISM has extra effects on Northwest China precipitation.

Considering the complex regime of atmospheric water vapor transport over northern and southern China (Zhou and Yu, 2005), the atmospheric sources of the southern subregions could be different from those of the northern subregions. Aside from North China, the YRV and South China subregions are significantly affected by the East Asian summer monsoon and subtropical high. Therefore, there will be intrinsic linkages among the summertime precipitation in these three subregions. For instance, the anomalously stronger East Asian summer monsoon brings excessive rainfall to North and South China and insufficient rainfall to the YRV. Such an out-of-phase relationship in rainfall is linked to the variability of the East Asian monsoon (Ding et al., 2009). Summertime precipitation in the YRV is predominantly controlled by the CN (52.8%), SCS (21.4%), IO (14.6%), and PO (10%) sources, which together account for approximately 98% of total precipitation (Figure 2E). These relative contributions of moisture sources are slightly distinct from those reported by Chu et al. (2019), partly due to the different methods used for classifying water vapor sources. In the present study, the CN precipitation source for the YRV is controlled by southwesterly and northerly moisture branches, which transport moisture from southern China and North–Northeast China, respectively (Figure 8A). This suggests that, similar to North China rainfall (Sun and Wang, 2014), the YRV rainfall largely originates from the evaporation taking place over land areas, rather than oceanic factors (Wei et al., 2012). In addition, a strengthened subtropical high, along with a weakened East Asian summer monsoon (figure not shown), is the key contributor to the southwesterly moisture pathway of YRV rainfall, which carries water vapor from the SCS and IO sources (Figures 8B,C), consistent with the findings of Fan et al. (2008) and Wang and Chen (2012). Simultaneously, a weakened ISM is apparent alongside the strengthened subtropical high (Figure 8C), which corresponds to more rainfall over the YRV, consistent with the conclusion in Zhang (2001). The PJ teleconnection appears to be closely linked with the PO precipitation source (Figures 8D,H). The aforementioned out-of-phase relationship with total precipitation is reproduced in the CN and IO precipitation sources (Figures 8E,G). These results indicate that the subtropical high and PJ pattern provide uniformly excessive summertime rainfall over the whole of East China, while the evaporation from land areas and the ISM provide additional information for the uneven distribution of East China rainfall.

In comparison, the precipitation in South China is dominated by the SCS, IO, CN and PO sources (Figure 9), with contribution rates of 36.5, 27.4, 25.8 and 10%, respectively. Note that CN is no longer the primary source for South China precipitation owing to the increased influence from the tropical oceans. Like the YRV, the precipitation in South China is affected primarily by enhanced southerly winds from the SCS source, corresponding to an intensified subtropical high (Figure 9A). On the contrary, a weakened subtropical high and enhanced East Asian–western North Pacific monsoon can also bring water vapor from land evaporation over the north of South China (Figure 9C), resulting in less rainfall over the YRV and, in turn, heavy rainfall over South China (Figure 9G). In addition, a weakened ISM transports water vapor from the Indian Ocean through the westerly branch of moisture pathways over northern India (Figure 9B), leading to excessive IO-source precipitation over western South China (Figure 9F). The PO source further causes spatially restricted rainfall over South China through the PJ teleconnection and the enhanced monsoon (Figures 9D,H). These results suggest that the moisture transport from the SCS and IO sources dominates the variation in South China precipitation (Sun and Wang, 2014; Chu et al., 2019) through a strengthened subtropical high and weakened ISM. Conversely, land evaporation from CN becomes less important, particularly relative to YRV and North China precipitation, with moisture contributions mainly from the YRV region or the PO through a weakened subtropical high.

The relative contributions to precipitation in Southwest China are distinct from those of other subregions in that the IO (48.4%), CN (41.5%) and SCS (8.8%) are the primary sources for Southwest China precipitation, which together explain 99.1% of the total precipitation (Figure 10). The amplified subtropical high, along with weakened ISM, brings water vapor from the Bay of Bengal and the Indian Ocean via a southwesterly pathway (Figure 10A). These two branches of moisture pathways result in strong convergence and hence heavy rainfall over Southwest China (Figure 10D). Similarly, the intensified subtropical high associated with the SCS source also transports moisture from the PO through the southeasterly moisture branch (Figure 10C), resulting in a belt of heavy rain over the whole of southern China and southern North China (Figure 10F). Conversely, the CN source is associated with a weakened subtropical high and enhanced East Asian–Western North Pacific monsoon, albeit the pattern is more zonally orientated, with the easterly winds bringing moisture transport from the evaporation over land areas (Figure 10B). This indicates that the heavy rainfall over South China can provide further evaporation sources for Southwest China precipitation (Figure 10E). To sum up, different from northern China, the three southern China subregions are primarily controlled by atmospheric circulation regimes associated with tropical ocean origins south of the mainland, such as the subtropical high, East Asian summer monsoon, and ISM.