1 Introduction

The Tibetan Plateau (TP) , often referredto as the world's third pole, is of great importance to the global climate. Covering about a quarter of the Chinese land area and elevated up into the mid troposphere, the TP exerts huge impact on atmospheric circulation, precipitation, and ecologicalconditions. In China, summer precipitation over the Yangtze River valley (YRV) attracts considerable attention and bears enormous importance to peopleresiding in and around this area. Given the large population and economic activity of the region, people in the YRV suffer considerably from droughts and floods in summer (Ding and Hu, 2003; Arndt et al., 2014) .

The climatesystem is highly sensitive to hydrological processes (Webster, 1994) . Water vapor transport and its anomaliesare closely related to rainfall conditions. In the East Asian monsoon region, the summer moisture fluxes from the Somali jet, the Bay of Bengal, the South ChinaSea, and the westernPacific subtropical high bring water vapor from the ocean toEast China. Analysisof these transport featureshas been quiteextensive (e.g., Tao and Chen, 1987;Ninomiya and Kobayashi, 1999; Ding and Sun, 2001; Ding and Hu, 2003) . Meanwhile, it is also very important to take into account the TP, owing to its considerable physical influence in the region.A large amount of water is preserved in the TP, which serves as a “global water tower” with respect to attracting and deflecting water vapor on the globalscale (Xu et al., 2008a) . In terms of general circulation, the East Asian monsoon, which predominately takes charge of the moisture transport to the YRV, is affectedby the TP's thermal and mechanicalforcing (Wu and Zhang, 1998; Wu et al., 2007, 2012) . Interacting with the tropical oceans, the TP plays a role as “transfer platform”of water vapor transport between East Asia and the Indianmonsoon area (Xu et al., 2002) . Diverted by the plateau, the water vapor transport accounts for pro-longed rainy weather in the vicinity of East China in summer (Zhou et al., 2005;Shi and Shi, 2008) . Around the TP, thecomponents of the earth systemshow com-plex interactions with one another, and affect regional rainfall (Lau and Li, 1984; Liu and Yin, 2001; Sato and Kimura, 2007; Lau, 2016) .

The main purpose of this study is to investigate the role of moisture flux around the TP in modulating the YRV summer rainfall. The moisture flux is separated in two parts: rotational (nondivergent) and divergent (nonrotational) . As indicated by the watervapor budget, the local hydrological balance (between evaporation and precipitation) is only affected by the divergent moisture flux. On the other hand, it is the rotational moisture transport that supplies water vapor to the rainy region from a distance.Previous studies of water vapor transport have mainly treated the situation as a whole (Simmonds et al., 1999; Zhou and Yu, 2005) , i.e., withoutseparating the two roles. Therefore, systematically studying the moisture flux around the TP in ways that distinguish these two components is a worthy avenue of research.

In this study, reanalysis data are used to explore in detail the features of water vapor transport around the TP and its effecton the precipitation over the YRV. Both the climatological and anomalous states are examined, and the relative importanceof the moisture supply over the southeastern corner of the TP is analyzed and discussed.

2 Data and methodology 2.1 Data

The data used for calculating water vapor trans-portare from ERA-Interim (Dee et al., 2011; http://apps.ecmwf.int/datasets/data/interim-full-daily/) , and the monthly average verticallyintegrated water vapor flux is derived from the dailymean flux calculated by the ECMWF. For the mid troposphere, 500-hPa wind and specific humidity fields are collected at 0200, 0800, 1400, and 2000 BT (Beijing Time) . The dataset is believed to represent a significant improvement comparedwith older generation reanalysis data, such as ERA-40 and NCEP-NCAR, especially for a number of parameters over the TP with respect to interannual variability (Wang and Zeng, 2012; Bao and Zhang, 2013; Lin et al., 2014) . Moreover, the water vapor in ERAInterim shows nearrealistic features to that observed by satellite and radiosonde between 50flS and 50flN (Kishore et al., 2011) . The Precipitation Reconstruction over Land (PREC/L) product, which is produced by the Climate Prediction Center from gauge observations (Chen et al., 2004) , is also used on the monthly scale for the summer season (June-August) . This dataset is provided by NOAA/OAR/ESRLPSD, Boulder, Colorado, USA (http://www.esrl.noaa.gov/psd/) . Both datasets are availableat a resolution of 1fl (latitude) ×1° (longitude) for the period 1985-2014. The YRV is represented by the grid area of (27.5fl-32.5flN, 110fl-22.5°E) .

2.2 Methodology

A vertically integrated moisturebalance equation depicts the atmospheric branch of the hydrological cycle,

where , E is the evaporation, q the specific humidity, g the gravity, P the precipitation, p the pressure, subscripts “t” and “s” mean top and surface, and V the velocity field. Accordingto Chen (1985) , the water vapor flux can be separated into rotationaland divergent components, which can be expressed in forms of streamfunction (ψQ) and velocity potential function (χQ) :

The water-balance equation can then be written as

or

As indicated by the equation, only the divergent part of water vapor flux modulates the balance of local precipitation and evaporation, and the rotational component accounts for supplyingthe water vapor that is essential to persistent rainfall.

3 Climatology of water vapor transport around the TP 3.1 Total vertically integrated flux

To identify and verify the featuresof water vapor transport duringseasonal transition, the climatological vertically integratedwater vapor transport (hereafter total moisture flux) in Apriland July is firstdepicted in Fig. 1. The water vapor flux rotates around both sides of the TP, whose main body lies in the westerlies. The flux magnitudeover the plateau itself is relatively weak due to its elevation and the resulting low specific humidity of air. In April (Fig. 1a) , the westernPacific subtropical high is domi nant over the South China Sea, resulting in comparatively heavy water vapor flux towards South China.The Somali jet has not gained its power, and the water vapor transport over the ArabianSea and the Bay of Bengal isquite weak in April.The water vapor flux in the eastfront of the troughsouth of the Himalaya over Burma (Yin, 1949) is comparatively strong. An anticyclonic pattern controls the Arabian Sea. In July, flows in the mid and low latitudes get changeddrastically (Fig. 1b) . The transport on the wholein tensifies. It is clear that a strong water vapor band, referred to as the “great moistureriver” by He et al. (2007) , originatesfrom the SouthernHemisphere, rushes into the Northern Hemisphere via the Somali jet, and then consecutively flows over the ArabianSea, the southern Indian Peninsula, and the Bay of Bengal. Another moisture band stems from the South China Sea and western tropical Pacific Ocean.The two branches compose the “big triangle”, whose dynamic and thermodynamic signalssignificantly affect Asian hydrology (Xu et al., 2002) . A cyclonic circulation appears south of the Himalaya, corresponding to the Indian low vortex (He et al., 2007) . The TP's influence is clearly reflectedby the deflection of southerly flow from the Bay of Bengal. After circling aroundthe southeastern corner of the TP, the “big river” mergeswith the flux from the western Pacific subtropical high over the South China Sea, turning around almost orthogonally into the East Asian monsoon region, and going as far as 40°N or even farther north.The distinction between the South and East Asian monsoon is evident (Huang et al., 1998) ; and since most water vapor is concentrated in the lowest 2-3 km of the atmosphere, lowlevel circulation (figure omitted) dictates the direction and magnitude of columnintegratedwater vapor transport.

3.2 Rotational and divergent part of the vertically integrated flux

Viewed separately, the averagerotational component of verticallyintegrated water vapor flux is generally similar to that of the whole part (Fig. 2) . The magnitude of the rotational part is a little less than the total flux, indicating that it composes the majority of the total moisture flux. From the rotational part's perspective, the TP's influenceon the vorticity of moistair motion is more representative. In July, when the northward flux over the Bay of Bengal encounters the steep orography of the Himalaya, it deflects in two directions (Fig. 2b) . The westward branch hikes againstthe mountain range, resultingin a cyclonic path over the south of the TP (increase in relative vorticity in response to conservation of potential vorticity) . The eastward branch flows through relatively lower mountain ranges, allowing it to continuously propagate through the southeastern TP and farther into East Asia.

While the total transport over the west body of the plateaucarries water eastward, moving downstream towards East China, the rotational flux brings water vapor westward out of the plateau. Rotational transport can only supplywater vapor, and it is the divergent part that affects the local source and sink of water vapor. Therefore, the contrast in transport direction between total and rotationalflux indicates that the flux over the western plateau only modulates the local precipitation and evaporation budget (where divergent flux plays a dominant role) . These aspects can be seen in Fig. 3 more clearly:over the central and western TP, the rotational current flows in the direction opposite to that of the total moistureflux. As can be inferred from the waterbalance equation, the divergent part thus flows eastward with stronger incal conditions. As indicatedin the following section, the rotational part not only transports water vapor, butalso affects the YRV summer rainfall significantly.

The divergent flux (Fig. 4) is generally one order of magnitude less than the rotational part, and strengthens from spring to summer. Indicative of the sourceand sink of water vapor, the divergent flux broadly reflects the landseaconfiguration: the tropical oceans are the major evaporative supply, while the maritime and monsoonal continents act as water vapor sinks. Unlike the total moisture flux pattern, the divergent part's intensity over the TP is comparable to that of the low latitudes. In July, the southeastern corner of the TP, togetherwith the YRV and subtropical western Pacific, is the location of the convergent center, indicative of a water vapor sink. The seasonal transition is mainly demonstrated by the convergence center's shifting from the westerntropical Pacific to East China and its adjacent ocean, which, as pointed out by Chen et al. (1988) , exhibits a 30-50-day oscillation associated with the Meiyu rainy season.

3.3 The rotational and divergent parts of water vapor flux in the mid troposphere

While the verticallyintegrated water vapor flux denotesthe total effecton the balance between column precipitation and evaporation, it is also of interest to investigate the major flow and convergencenear the surfaceof the TP and their influence on downstream rainfall. Xu et al. (2003, 2008b) showed that the flux over the western boundary of the YRV, which largely comes from the TP in the mid troposphere, is also significant to downstream rainfall.In this section, 500 hPa is chosen as a representative level for the mid troposphere to investigatethe condition and effect of water vapor transport over the TP and its surroundings. The integral interval is unit thickness (1 hPa) with the assumption of uniformity near this layer.

The climatological summer pattern (JuneAugust; the same below) of the rotational component southof 30°N at 500 hPa is to a large extent similar tothat of its verticallyintegrated counterpart (Fig. 5a) . The major differences take place over the TP and to its north.This is consistent with the fact that the synoptic systems in the tropics are mainly barotropic, and the baroclinicity develops primarilyin the mid and high latitudes. The TP and the area to its north witness eastward water vapor flux comparable to the magnitude of that in the low latitudes. The eastward moisture transport over the YRV consists of streams not only from the “big triangle”, but also from the westerlies (Wu and Zhang, 1998) . A vortex (Lin, 2015) predominates over the central TP.

Unlike the verticallyintegrated pattern, the divergent part at 500 hPa over the TP is not well distributed spatially (Fig. 5b) . The convergencezone still prevails along the subtropical western Pacific. Along the periphery of the southernHimalaya, the divergent flux reaches its maximum. Intensive convergence covers central and western parts of the TP near theelevated surface, while significant midlevel divergenceprevails over northern India and south of the Himalaya. Over the TP, the total moistureflux in the mid troposphere exhibits a similar patternas the divergent part (figure omitted) . The maximum against the southern periphery can possibly be traced back to the distinctive pool of water vapor maximum right over the TP (Xu et al., 2008b) . It is also of interest that while the difference between the vertically integrated rotational and divergent parts is one orderof magnitude, the two components at 500 hPa are roughly the same. It can be inferred that the inhomogeneity of specific humidity mainly contributes to the divergent component.

The Asian monsoon exhibitsgreat variability in termsof precipitation, which is a vital factor for the socioeconomicsof the region (Lau and Li, 1984; Tao and Chen, 1987) . In the following two sections, the linear trend is removed along the annually averaged summer data for both rainfall and moisture flux. By detrending, the interdecadal variation is taken out from the dataset, leaving only the interannual changes.The role and significance of moisture flux over the TP on the YRV summerrainfall on interannual timescale are discussed.

4 Summer water vapor transport aroundthe TP in YRV flood/drought periods

By using the PREC/L dataset, the YRV flood/drought summersare selected based on the criterion that the detrended summerprecipitation anomaly for the total YRV shouldbe larger than one standard deviationin extreme years. In this way, 1985, 1990, 2001, and 2013 are chosen as YRV drought summers, while 1993, 1996, 1997, 1998, and 1999 are selectedas YRV flood summers.

In flood summers, anomalous water vapor rushes into the YRV from the western boundary (Fig. 6a) . An anticyclonic anomaly prevails over theSouth China Sea with its northern periphery carryingabundant water to the YRV, which is related to the PacificJapan teleconnection pattern that developsin the southwesterly monsoon flow (Kosaka and Nakamura, 2006) . Among the contributions to this anomaly, there are three dominant branches. The supply from the westerntropical Pacific is the most significant one. The other two branches, originating from the former, are all related to the TP. One flows over the Bay of Bengaland turns aroundjust south of the Himalaya. The other goes farther into the Arabian Sea and finallycircles anticyclonically towards the south of the TP. All three branches move through ocean areas, travelling all the way along the anomalous transport path and turning around south of the TP to the YRV. The supplysouth of the TP is proven to be important to the downstream precipitation.

When the YRV suffersdeficient summerrainfall, the pictureexhibits different characteristics (Fig. 6b) . The intensity of the transport anomalyis, on the whole, approximately the same as in YRV flood sum-mers. Two abnormal vortexes control the north of the South China Sea and the northeast of the Bay of Ben gal, respectively. The Indian monsoon low is therefore much stronger, especially along the southern periph ery of the TP. An anticyclonic anomaly is situated over the Yellow Sea, bringing strong abnormal vapor from the ocean east of the YRV, and from there going on to carry it towards North China. Unlike the pattern in flood summers, a broad and much more intense cy clonic anomaly prevails north of the TP in drought summers, which brings a substantial amount of water vapor to Northeast China.

Owing to the directlink with precipitation, the divergent component anomalouspatterns in extreme summers are more clearcut (Fig. 7) . In flood summers, there is a convergence band extending from the YRV to North India and south of the Himalaya. In drysummers, however, a broad divergence bandis situated from the YRV to easternand central parts of the TP.However, only the abnormal divergent flux south of the YRV and in the southeastern TP exceeds the 90% confidence level in both situations. From the perspective of divergent moisture transport, the TP and YRV are closely related to one another.

5 Significance of the water vapor supplyfrom the southeastern TP for YRV summer rain-fall 5.1 Correlation between rotational water vapor flux and YRV summer rainfall

Figure 8 shows the correlation distribution between the summer rotational component of vertically integrated zonal water vapor flux and totalsummer rainfall in the YRV. The moisture flux of this vast areain the lowlatitude Pacific is negatively correlated to the YRV precipitation. This is corresponding to the cyclonic/anticyclonic anomaly pattern mentioned above related to the PacificJapan teleconnection. On the other hand, the zonal water vapor supply by the rotational part southeastof the TP is significantly in phase with the YRV rainfall, besides the YRV surrounding region. This further confirms that the water vapor transport at the southern peripheryof the Himalaya is closelylinked with the YRV precipitation. It is safe to say that the southeastern corner of the TPis a sensitive area with respect to the downstream rainfall.

5.2 Effect of the water vapor flux anomaly over the southeastern corner of the TP

In this subsection, the southeastern cornerof the TP is represented by the area (20°-30°N, 90°-105°E) , in which we sum the zonal rotational vertically inte-grated water vapor. The summers are chosen as the eastward/westward wate r vapor supply extremesusing the same method as when filtering the YRV flood/drought years, mentionedabove. The composite anomalies show a similarpattern as the YRV flood/drought summers, respectively. When water vapor supply is intense over the southeastern corner of the TP, more water vapor reaches the YRV from its western boundary (Fig. 9a) . In these circumstances, the East Asian monsoon andSouth Asian monsoon are moreclosely related throughthe top of the “big triangle.” When the southeastern TP suffers weak water vapor flux, the YRV's precipitation is supplied dominantly from its southernboundary, and the East Asian monsoon pushes farther north (Fig. 9b) . The south eastern cornerof the TP serves as a bridge between the two components of the Asian monsoon, which display inverse proportions to one anotherin terms of moisture supply to the YRV (Zhang, 2001) .

The precipitation departure percentages in the eastward/westward water vapor anomalous summers are shown in Fig. 10. The pattern corresponds well with the moisture transport anomaly in the YRV. The YRV and central China gain abovenormalrainfall when stronger eastward water vapor supply dominates the southeastern TP. However, drought takes place inth e region when the elevated water vapor transportdecreases. South China and the southern North China display opposite change trends with regard to the YRV, partlydue to the different pace and progress of the East Asian monsoon in the two circumstances. A tripole alsoemerges in the South Asianmonsoon region. The Himalayan foothillsand southern India both covary with the YRV, while central India exhibits an opposite anomalytrend. The variation of moisture transport over the southeastern TP is believed to also be responsible for the rainfalloscillation in India, possibly through the cycloni c circulation south of the TP. Thesephenomena and proposed mechanism are de-scribed in Day et al. (2015) .Due to its aridity, Northwest China shows large variability with respect to the anomalous percentage. The anomaly of divergent water vapor flux in the weak southeastern TP water supply summers is characterized by a strong convergence center over the South China Sea, and a divergence center over the main bodies of the TP, the East China Sea, and Japan (figure omitted) .

6 Conclusions and discussion

In this study, based on the ERAInterim data for the period 1985-2014, the characteristics of the rotational and divergent components of water vapor flux around the TP are investigated, in the climatology and in composite analysis. Their impacts on the YRV rainfall are also discussed. The following conclusions are drawn.

In the climatology, the rotational part of vertically integrated water vapor flux around the TP contributes to the majority of total moisture flux. Therefore, the pattern of rotationalwater vapor flux is generally similar to that of the total.However, the opposite situationis true over the western TP, implying that the flux there mainly consistsof atmospheric divergent flow just over the groundsurface, modulating local rainfall, yet rarely providingwater vapor downstream. The movement of the divergent part's center isindicative of the seasonaltransition.

The TP also exerts significant influence near its surface. At 500 hPa, the whole plateauacts as a water vapor supplier for the YRV, witha parallel magnitude to that in the low latitudes. The divergent component of moisture flux reaches its maximum valuesand gradient over the southernTP, along the Himalayan mountains.

Composite vertically integrated water vapor flux in YRV flood/drought summers shows that the water vapor supplychanges most strikingly in the South China Sea and the south of the TP. The role of the TP as a transfer platform strengthens/weakens in YRV flood/drought summers.From the perspective of the divergent part, the anomaly in the YRV and the eastern TP is closelyrelated to one another.

Correlation analysisverifies the importanceof the water vapor supply from the southeastern corner of the TP. It acts as a bridge between two subsystems of the Asian monsoon, and exerts influence on both atmospheric circulation and precipitation.

The separation of water vapor flux into rotational and divergent parts helps us to understand the different roles that they play. In both climatological and extreme states, the southern TP plays an important rolein the hydrological cycle of Asia. The Asianmonsoon consistsof various components, and the interaction among them is complex.In this study, the TP shows its power both at the mid level and throughout the whole column. The hydrological cycle is closely related to, and imposes influence upon, atmospheric heating, and hence the thermal forcingof the TP. Upward transport supplyingthe wet pool of moisture over the TP and the onset order of Asian monsoon rainfall in early summer can both be traced back to the elevated heating structureof the TP (Yanai et al., 1992; Wu and Zhang, 1998; Xu et al., 2008a) . Thus, it is necessary to investigatethe relationship between thermalforcing and the two ingredients of moisture flux. Furthermore, how the “atmospheric water tower” (Xu et al., 2008a) and its amplifying effect (Liu et al., 2003) in association with other cli-matecomponents will change is of great importance and interest with regards to the YRV rainfall, and much more.

Acknowledgments: We would like to thank the two anonymous reviewers for their insightful and detailed comments and suggestions. We also extend our gratitude to the editorsfor their efforts in improving this paper.