During the boreal summer, a large amount of water vapour is brought to the Asian continent by the monsoonal flows and contributes largely to pre-summer precipitation in Southern China and summer Meiyu rainfall in the mid-lower of the Yangtze River Basin (YRB). The droughts and floods related with the monsoon rainfall have enormous social and economic impacts in China and the other Asian summer monsoon regions, in particular for the reprehensive region of mid-lower YRB, a large segment of the world's population (Tao and Chen, 1987; Ding, 1994; Ding and Chan, 2005; He et al., 2007).
In this context, the identification of the physical mechanism of water moisture transport in the atmosphere from evaporation sources to Asian summer monsoon regions and the geographical source/sink of moisture associated with the maintenance of monsoon rainfall are significant for the understanding of the regional hydrological cycle and for advancing our knowledge on the monsoon rainfall variations (Simmonds et al., 1999). Moreover, as might be expected, the knowledge of the source/sink of moisture can improve the predictive skill of numerical models and also help design the observational network in the upstream region, which could in turn be of important operational value for heavy rainfall forecasts (Xu et al., 2008).
A variety of previous studies has been conducted to identify the moisture transport mechanisms and the associated sources/sinks of moisture based on various methodologies and input data with significantly different results among authors. For example, Tao and Chen (1987) suggested that the Somali jet and the east flow from the south of the western Pacific subtropical high are the key moisture transport passages for heavy rainfall over eastern China. In another case study, however, three different low-level monsoonal streams are regarded as the main processes governing the moisture transport for summer precipitation over China, i.e. the southwesterly flow towards the Indian peninsula and the Bay of Bengal connecting with the Indian summer monsoon, the southerly flow in the South China Sea and the southeasterly flow following the southern flange of the Northwestern Pacific subtropical high (e.g. Ninomiya and Kobayashi, 1999; Ding and Sun, 2001; Ding and Hu, 2003). The studies by Simmonds et al. (1999) and Zhang (2001) also indicated that the inter-annual anomalous precipitation that occurred over China appears to be more closely related to the anomalous water vapour transport by the easterly stream, basically due to the anomalous strength of the Northwestern Pacific subtropical high.
There has been no consensus on the exact location of moisture sources and their relative contributions to summer rainfall over east China so far. The South China Sea is usually regarded as the main moisture source for summer precipitation (Ding, 1994; Huang et al., 1998; Simmonds et al., 1999). Lau et al. (2002) and Chow et al. (2008) also suggested that the South China Sea acted as a leading moisture source for the torrential rain over the YRB in summer 1998. In contrast to these findings, other studies suggest that the westerly flow from the Indian peninsula and Bay of Bengal is the dominant water vapour transport routes for the summer or Meiyu precipitation over China (e.g. Chen, 1994; Li, 1999; Xu et al., 2003a, 2003b; Qian et al., 2004). The moisture sources and transport processes for the YRB region were also explored by Xu et al. (2003a, 2003b, 2008), who found that the main moisture sources for the mid-lower YRB cover the large low latitudinal areas in the Indian Ocean, South China Sea and Northwest Pacific.
The moisture source discrepancy can partly attribute to the fact that direct identification of the moisture source regions is considerably difficult over Asian summer monsoon regions. On one hand, Asian summer monsoon hydrological cycle itself involves different comprehensive climatic and synoptic processes at a wide range of temporal and spatial scales. As suggested by Gimeno et al. (2010a), the Indian-Asian continent can receive moisture from six different major oceanic source regions. On the other hand, previous studies were restricted to the diagnosis methods adopted, which applied the Eulerian approaches. The information on moisture sources and transport conditions in the previous studies is often derived indirectly from the common analysis methods based on general circulation models with tagged water tracers or integrating the atmospheric moisture transport fluxes. Using general circulation modelling, an evaluation of the major evaporating regions that contribute to water precipitating requires the predetermined source regions and the results are dependent on the quality of the modelling data; the moisture fluxes across the region boundaries cannot provide any information on the geographical sources of moisture due to the fast transition of wind fields (Sodemann et al., 2008b).
More sophisticated methods have recently been developed to determine the moisture origin/sink by backward/forward trajectories of air parcels residing over the target region with Lagrangian models. As a pioneering work, Stohl and James (2004, 2005) developed a Lagrangian model and used it to determine the moisture source for a region through the net changes in the quantity of water along a large number of backward or forward trajectories in space and time. The main characteristics of the different Lagrangian analysis methods for moisture transport are described in several studies with the discussions of the advantages and disadvantages of both Eulerian and Lagrangian methods (Nieto et al., 2008; Sodemann and Zubler, 2010). The Lagrangian methodology has been successfully applied in climatological studies on sources of moisture in other regions in the Sahel (Nieto et al., 2006), Iceland (Nieto et al., 2010a), the Orinoco River basin (Nieto et al., 2008), the south American monsoon area (Drumond et al., 2008), northeastern Brazil (Drumond et al., 2010), central America (Duran-Quesada et al., 2010) and the section around Mediterranean Sea (Nieto et al., 2010b). Recently, Gimeno et al. (2010b) carried out an analysis of the seasonal contribution of the principal global sources of moisture to precipitation in continental regions. Sodemann et al. (2008a, 2008b) also examined the net water vapour changes along the back trajectories to infer the moisture sources for precipitation in a target region.
Believing that the Lagrangian method can provide more precise details concerning the transport pathways of air masses and the moisture variability in comparison to the traditional Eulerian techniques, we herein attempt to present a quantitative and spatially detailed climatology of moisture sources and transport to the YRB area using the Lagrangian approach. Hereby, this study aims to identify the leading moisture sources affecting the YRB water vapour, in particular, the distant moisture transport following the low-level streams of Asian summer monsoons and finally to quantify the seasonal and inter-annual variability of moisture sources and transport to the YRB.
The target region for this study was the YRB, where the rainy period of east Asian summer monsoon is the noted Meiyu or Plum rains caused by the convergence of several atmospheric streams, such as the downstream of Tibetan Plateau, the southwest monsoon winds from Indian Ocean across Indian peninsula and Bay of Bengal and Arabian Sea and summer monsoon winds form south China Sea. The influence of these wind streams on the YRB is different because of the complex topography, large ocean–land contrast. Therefore, the knowledge on the details of moisture sources and related transport processes could benefit the understanding the precipitation regimes for Asian summer monsoon system. Furthermore, the high frequency of flood/drought events occurred in summer over YRB region is also a reason to subject this region to the Lagrangian analysis.
The article is organized as follows: The data sets and methods applied in the study are described in Section 2. Section 3 contains the Lagrangian analysis results, together with some discussions. The conclusions are presented in Section 4.
2. Data and methods
For this study, we tracked the air column masses overlying the YRB backward in time to establish the air moisture budgets using a three-dimensional Lagrangian transport model FLEXPART developed by Stohl and James (2004, 2005). At the start of the simulation, the column atmosphere over the extended Asian monsoon region (60°N–0°N, 40°E–140°E) was divided into 2.2 million particles filling the atmosphere homogeneously. One particle load has a constant air mass of around 7.0 × 1011 kg. Then, the particles are advected by the meteorological fields from the National Centers for Environmental Prediction Final (NCEP FNL) Operational Global Analysis data with a resolution of 1° × 1° and 26 levels vertically extending from surface to 10 hPa (see http://dss.ucar.edu/datasets/ds083.2/data/), as well as driven by the superimposed stochastic turbulent and convective motions. The modelling period is from 00:00 UTC of 15 June to 00:00 UTC of 15 September for the years 2004–2009 with the integration time step 600 s. Within this context, the simple physical concept of a particle was used, which consisted of a tiny portion of atmospheric matter with negligible internal motion and unique thermodynamic properties. The modelled particle position (latitude, longitude and altitude), and the interpolated meteorological parameters from the input analyses (temperature, specific humidity, air density, atmospheric boundary layer height and the tropopause height) were output every 6 h.
For a single air particle, the relation between evaporation (e) and precipitation (p)e − p = m(dq/dt) can be determined by the temporal moisture (q) change dq/dt and the particle mass m along the trajectory. The sums of evaporation e and precipitation p from all the particles in the atmospheric column over an area could be, respectively, the evaporation and precipitation rates E and P per unit area. The method can also track (E − P) from any studied region backward or forward in time along the trajectories, diagnosing the relationships between net moisture source and net moisture sink regions. The full details of method and its limitations are described by Stohl and James (2004, 2005) and Sodemann and Zubler (2010).
To search the moisture sources for the YRB, all the particles entering the middle and lower YRB (26.5°N–33.5°N, 110°E–122.5°E) during the boreal summertime from 1 June to 31 August of 2004–2009 were backtracked in time. Similar to the previous work, we traced the E − P backward from the YRB region with the 10-day transport time, which is the average residence time of water vapour in the atmosphere (Numaguti, 1999). Thus, a 6-h database of trajectories entering into YRB region was constructed with the six-summer E − P values in the 1° × 1° grids, and then averaged to summer seasonal composite maps. The back trajectory E − P values of the previous day number n are denoted with (E − P)n. For example, (E − P)1 represents the total gain or loss of moisture on the previous day along the back trajectories. Accordingly, (E − P)1−n indicates the integrated value of (E − P) from 1 to n days in back time.
To catch a glimpse of the underlying transport mechanisms, we also used a totally independent procedure to calculate the vertically integrated moisture fluxes (VIMFs) and its divergence based on the entire atmospheric mass. The VIMF is defined as , where g is the acceleration of gravity, q is the specific humidity, Ps is the surface pressure, and v is horizontal wind vector. The NCEP/NCAR reanalysis data on the 2.5° × 2.5° grids were used to derive the VIMFs and the flux divergences, together with a flux correction for air mass balance, and then averaged to the summer seasonal composite maps.
3. Results and discussions
3.1. Seasonal moisture sources and sinks in summer
Figure 1 shows the summer mean (E − P)n fields on the second, fourth, sixth, eighth and tenth day and the 10-day integrated (E − P)1−10. Despite the relative complexity of the backtracking method, the interpretations of the patterns in Figure 1 are simple and straightforward for the source and sink regions of YRB moisture. In the regions characterized by (E − P) > 0, the evaporation exceeded the precipitation, which indicates that air particles were destined for the YRB with the net gain moisture in the vertical column; such regions were therefore designated as moisture sources. In contrast, the air mass transit to the YRB had a net loss of moisture over the regions of (E − P) < 0, which were regarded as moisture sink regions. Thus, Figure 1 can be interpreted as the moisture source YRB region during the summer season projected onto the corresponding evaporation sources.
Since the main purpose of this work is to identify the remote moisture sources, the (E − P)1 pattern reflecting the process of recycling and nearby sources was not shown here. As shown in Figure 1, one can see that the spatial extent of the source and sink regions expand and become less intense with the days backward elapsed. The patterns of (E − P) > 0 on the (E − P)2 for the second day and the (E − P)4 for the fourth day display that the East China Sea, South China Sea and Philippine Seas are the major moisture source regions in a shorter distance affecting the YRB. These results are consistent with the pattern of the VIMT and its divergence (Figure 2). The source of South China Sea connects the prevailing northeastward wind directly from this region passing over the subtropical region, while the East China Sea source is associated with the northwestward wind. These two wind streams combined before they arrived at the YRB region. The spatial pattern of (E − P)6 at the sixth day indicates a presence of moisture sources in the tropical ocean covering from the equator to 20°N with the highest values centred at Indian peninsula and Bay of Bengal. The relevance of this region is more evident with longer backward time, particularly for (E − P)8 at the eighth day and (E − P)10 at the tenth day, when the source expanded southward with the air flows crossing the equator region. This indicates that the water vapour budget and precipitation over the YRB could be partly influenced by the cross-equatorial transport. This finding is in agreement with the previous studies (Fan, 2006).
The corresponding moisture sink regions were characterized by the negative values of E − P. Considering the topography and the VIMT divergence distribution (Figure 2), it can be found that the sink regions with considerable loss of moisture can be categorized into two kinds: one highlighted the effect of topography hindrance for the most sinks and the other closely related to the regions of air flow convergence. It could be deduced from Figure 2 that the air masses originating in the Bay of Bengal and Arabian Sea region tend to lose considerable moisture over Indian peninsula, southwest China, and the west part of Indian continent and Indo-China peninsula. As the moisture sinks for precipitation, the largest loss of moisture occurred over south slope of Tibetan Plateau, and the moisture reduction was resulted from the convergence of air flow over the southwest corner of Tibetan Plateau and Sichuan Basin in China. However, for the air masses originating in the South China Sea, only a small proportion of moisture was lost before arrival at the YRB region due to no strong topography hinder. More details about the moisture transport processes are provided in Section 3.3.
The distribution of (E − P)1−10 shown in Figure 2(f) corresponds to the integrated moisture of all particles from day 1 to 10, indicating the averages of the moisture gains and losses over the previous 10 days. In addition to the minor sources (E − P > 0) located near the YRB itself and over the south part of China, the aggregated perspective of Figure 2(f) further confirms the significant moisture sources: the short-distance sources at the South China Sea and the part of East China Sea, and long-distance sources over the Indian peninsula, Bay of Bengal and Arabian Sea. These patterns imply from a general point of view that the air masses are dried or moistened along their journey to YRB region, affected by long-distance transport as a function of meteorological conditions.
3.2. Quantifying the relative contributions of major sources
To quantitatively evaluate the relative contribution of the different moisture sources to the atmosphere overlying the YRB, we selected four representative moisture source regions: (1) South China Sea including large part of Indo-China peninsula (100–120°E, 0–20°N), (2) Indian peninsula and Bay of Bengal (75–100°E, 0–20°N), (3) East China Sea including part of Japanese Sea (121–140°E, 200–40°N) and (4) Arabian Sea (50–75°E, 0–20°N). The four source domains and the distribution of topography elevation are presented in Figure 3.
Taking the advantage of the ‘domain filling’ technique used here, we calculated the evaporative moisture mass by multiplying the air parcel mass (7.0 × 1011 kg) and the change in specific humidity along their trajectories. In Lagrangian frame, since the changes in humidity for air parcels are mainly caused by the local hydrological cycle rather than the advection from the other region, the changes in evaporative moisture mass calculated herein thereby mainly reflect the local evaporation processes. Integrated over the area of source region, the evaporated moisture mass backward in time over the four source regions with the temporal accumulations are shown in Figure 4(a) and (b). It is obvious that the evaporated moisture mass for the source regions have the peaks but with the different transport timescales before reaching the YRB region. Figure 4 largely quantifies the total evaporation for all source regions. The evaporation is closely related to the land or sea surface temperature.
The net uptake of source moisture is of more significance in water vapour contribution to the YRB region. We could also obtain the net uptake moisture mass by multiplying the mass of all air parcels and the absolute change in specific humidity. Figure 5(a) and (b) presents the temporal evolutions of the net uptake of moisture mass of the summertime averaged (E − P) and the accumulations over four source regions, which are similar to the work by Gimeno et al. (2010a, 2010b) and Duran-Quesada et al. (2010). The completely different variation patterns are shown in Figures 4 and 5. The source region over South China Sea had the dominant net uptake of moisture for the YRB during the period of 1–5 days backward in time with a peak at day 3 and a very significant contribution during the 2–4 days backward in time (Figure 5(a)). The moisture contributions of the South China Sea region were greater than those of the source regions of Bay of Bengal and Arabian Sea over backward 1–5 days.
The contribution of the Bay of Bengal sources had a negative net uptake of summer moisture in the period of backward 2–5 days. The negative net uptake existed in the Arabian Sea region but with 3–6 days. This negative value could imply that there are significant moisture losses along the trajectories before leaving the source region. This phenomenon is more obvious by the effects of hinder over the west part of Arabian peninsula and Indian-China peninsular coast during the air parcel transport, the areas highlighted in blue in Figure 1. However, the Bay of Bengal and Arabian Sea regions contributed more moisture than South China Sea region, respectively, backward after 5.5 and 7 days.
The moisture peaks for the Bay of Bengal and Arabian Sea regions after 6 days of backtracking implies the existence of persisted moisture supply practically from these two source regions. The total summer contribution of moisture mass from the South China Sea for days 2–10 backward was 22 × 1013 kg year−1, whereas the corresponding total summer supply from the Bay of Bengal for days 6–10 backward was 17.5 × 1013 kg year−1, following by the Arabian Sea with a moisture mass 13 × 1013 kg year−1 for 7–10 days backward (Figure 5(b)). On the basis of the quantitative estimations, the South China Sea made a largest contribution of 26.6% to the YRB summer moisture, following by the Bay of Bengal (20.5%), East China Sea (17.5%) and Arabian Sea (13.6%). It is easily found that these four regions only cover parts of Asian summer monsoon regions, thus besides these contributions, the contributions of 21.8% left might be made by other regions. Their different transport timescales depended on the distances to the YRB and the atmospheric circulations over the transport pathways.
3.3. Moisture transport pathways
All previous studies on the YRB moisture and air mass transport processes are limited to a short time period. In this section, we depicted the moisture transport pathways from the sources to the YRB region more qualitatively with a climatological view. To achieve this, all particles reaching the target region of YRB during the six summers of 2004–2009 were tracked backward in time, and their spatial evolution of air parcel locations can be approximately served as the transport pathways. Figure 6 shows the seasonal mean density field as a percentage of particles reaching the YRB in the previous 2, 3, 4, 5, 6 and 7 days over the six summers on 1° × 1° grid. One can see that the distributions of air parcels reaching eventually the YRB region varied significantly with time, and several dominant transport pathways are roughly identified.
The air mass transport mostly came from the south and west side of YRB in backward 2 days controlled by Asian summer monsoon. The minor transport made by the northeast and northwest side of YRB was associated with Tibetan Plateau and East China Sea, which was consistent with the source distribution identified in Figure 1. The strong particle transport started from the north of South China Sea and the Indian-China peninsular at backward 3 or 4 days. This change indicated that most of the particles passed over the regions with a destination for the YRB. Another striking finding is that the relative large value of particle transport exhibited a V-shape (Figure 6(c), (d) and (e)). This V-shape distribution indicated that despite the southwesterly flows over the regions, not all the particles come directly from the South China Sea and Indian-China peninsular regions. Even though the analysis of the mechanism of formation of transport pathways is beyond the scope of this study, the key role of topography of the Tibetan Plateau in the regional water cycle within the Asian monsoon regimes can be seen easily. This topic has been investigated by a larger number of studies (Yanai et al., 1992; Wu et al., 2006; Liu and Ding, 2008; Xu et al., 2008; Bothe et al., 2010, 2011).
We thereby can get an overall picture of moisture transport pathways. The pathways for the moisture coming from the sources with relative short transport timescales (e.g. East and South China Sea) are controlled by the dominant air flow associated with Asian summer monsoon regimes. However, the transport pathways that originated from long-distance regions, particularly over Bay of Bengal and Arabian Sea, follow the southwest flow from the tropical ocean and then are separated into two branches in the southwest of China, as a result of the effects of large topography, reaching the target region of YRB lastly. On combining Figure 6 with Figure 1, we can deduce that the large parts of the moisture from the sources of Bay of Bengal and Arabian Sea cannot reach the YRB, as a result of the loss of moisture for precipitation due to the landing and the presence of large topography.
3.4. Subseasonal variations
Following the analysis on seasonal mean basis in the previous sections, we further examined the subseasonal variations or monthly changes in moisture sources and transport within summer. Figure 7 shows the 10-day integrated values of (E − P) for the months of June, July and August, respectively. Not surprisingly, during three summer months, the patterns of moisture sources (E − P > 0) were overlapped in most of the regions suggesting the strongest influences of the oceanic moisture sources within southeast Asia on the YRB summer monsoons. The sink regions (E − P < 0) located in the south slope of Tibetan Plateau and the landing coast areas of summer monsoon streams.
However, a closer look at Figure 7 revealed that all moisture source regions shifted slightly for each month. For example, compared with the monthly distributions of June and August, the South China Sea source was extended southward slightly in July and covered the region of Indonesian. Similarly, the sources of Arabian Sea and Bay of Bengal in the high summer (July) covered the largest areas over the Arabian Sea and Bay of Bengal during the three summer months (Figure 7(a), (b) and (c)). The large East China Sea source region, including some parts of north and east in June, was drastically reduced with its negligible contribution to the YRB in July, the source region was expanded somewhat (Figure 7(c)). These differences in the source distributions have the substantial monthly impacts on the water vapour transport from southeast Asia to northeast Asia.
To assess the underlying physical processes in the subseasonal evolution of moisture sources, we analysed the seasonal VIMF patterns and the divergence derived from the wind differences between July and June (Figure 8(a)), as well as those from the differences between July and August (Figure 8(b)). It can be found that two additional air flows arrived the YRB regions between July and August: one being southwest wind streams from the Bay of Bengal and Arabian Sea and the other being the air streams of the anticyclone circulation over the tropical west Pacific Ocean and the north part of South China Sea. These two distinctive air flows of east Asian summer monsoon are converged over the YRB region resulting in a corresponding convergence of VIMF for the summer monsoon precipitation. The stronger southwest Indian monsoon may provide a greater air mass transport to the YRB in July than in June or August. It could also be confirmed by the convergences in the VIM that the contribution of moisture from the Bay of Bengal and Arabian Sea to the YRB region was enhanced in July. It is not difficult to discriminate that the additional circulations are associated with the evolution of Asian summer monsoon, which matured in July. The evolution of the Asian monsoon circulation is an important factor in regulating the relative contribution of moisture sources.
For a further quantitative characterization of the subseasonal variation, Figure 9 presents the monthly variations of integrated mass of net uptake moisture of 1–10 days for four moisture sources in June, July and August. The regions of South China Sea, Bay of Bengal and Arabian Sea maintain their status as the moisture sources for YRB in June, July and August. However, there were the most evident differences in the contribution of moisture sources to the YRB between July and June (or August). In July, the moisture mass supply for the South China Sea from day 2 to day 10 backward was 10.23 × 1013 kg·year−1, for the Bay of Bengal from day 3 to day 10 backward was 8.9 × 1013 kg·year−1 and for the Arabian Sea from day 6 to day 10 backward was 10.34 × 1013 kg·year−1. The East China Sea region was characterized by a net sink of moisture in July. In June, the South China Sea and Arabian Sea region supplied a moisture mass of 5.7 × 1013 kg·year−1, whereas the Bay of Bengal region supplied only 2.05 × 1013 kg·year−1. A striking feature of moisture source in June, which is not detected by previous studies, is the contribution from East China Sea, which supplied 4.34 × 1013 kg·year−1; that is, the East China Sea supplied approximately 2.0 times more moisture in June than did the Bay of Bengal region. In August, the contrast was not as marked as June, with the South China Sea region supplying slightly more amounts of moisture (8.02 × 1013 kg·year−1), whereas Bay of Bengal and Arabian Sea regions supplying comparable amounts of moisture (5.01 × 1013 and 5.13 × 1013 kg·year−1, respectively).
3.5. Inter-annual variability
The inter-annual variability and anomalies in the moisture transport could lead to the extreme events, such as drought or flooding. Figure 10 presents the inter-annual variability of the oceanic moisture source contributions, again with the mean CPC Merged Analysis of Precipitation (CMAP) (Xie and Arkin, 1997) and station-based precipitation, respectively, over YRB region in the same figure as a reference. One can find pronounced inter-annual variability of main moisture sources to YRB region during the summer. Apparently, except for the East China Sea region, other three main sources of moisture show a same year-to-year variability trend in line with the precipitation, particularly for Bay of Bengal and Arabian Sea. For example, during the summer of the years 2004, 2006 and 2007, moisture source contribution appears to shift from a minimum to a relatively larger one in the following year (2005, 2007 and 2008), which is accompanied by a precipitation shift from a minimum to a maximum. These point again towards a leading role of South China Sea, Bay of Bengal and Arabian Sea moisture with respect to the variability of YRB summer precipitation. This finding also reinforced our credibility of the moisture sources diagnosed herein. However, it should be noted that the moisture source shifts over South China Sea are not accompanied by obvious changes in precipitation amount, especially for the year 2008, which implies that variability in moisture origin over this region is not necessarily totally associated with variations in precipitation amount.
To analyse the role of all moisture source contributions in inter-annual variability precipitation variation, the contribution to YRB moisture of the main four sources under dry or wet summer conditions in the YRB was also assessed. Precipitation over the YRB during the relatively short study period was characterized by two extreme summers: a relative drought in 2004 in YRB and, in contrast, an intense precipitation over the YRB in summer of 2008. These extremely different summer conditions enabled comparison of the different moisture supplies and extension of the moisture sources for these distinct periods, providing a better understanding of the role of the main moisture sources in the summer precipitation in the YRB.
Figure 11 shows the 0.015 mm day−1 contours for 1- to 10-day integrated (E − P)1–10 during both summer. The blue lines correspond to the wet period (2008) and the red line corresponds to the dry period (2004). It can be found that, compared with the dry summer (2004), the 0.015 mm day−1 contours covered larger areas in tropical and subtropical regions and extended southward significantly over the South China Sea in wet summer (2008). Moreover, the intensity of Arabian Sea and Bay of Bengal source was much greater for the wet summer, with the area in this period characterized by values > 0.015 mm day−1 covering most regions of Arabian Sea and Bay of Bengal. In contrast, during the dry year, similarly high values were completely absent from the entire Arabian Sea and the regions likelihoods, whereas the high values were located mainly over the East China Sea and north of South China Sea. These results provide additional evidence that the combined effect of high/low moisture supply from the Arabian Sea and Bay of Bengal regions and low/high moisture supply from the East China Sea are responsible for extreme summer precipitation in the YRB. Interestingly, during the wet summer, the Tibetan Plateau and the Sichuan Basin appeared to supply some moisture to the YRB, but this did not occur during the dry summer.
The weak but detectable inter-annual variability prompts questions about the influence of large-scale climate modes on shifts in the YRB moisture sources. However, an analysis of 6-year time series cannot reveal any significant correlations between moisture source variability and either the western Pacific Ocean high or the El Niño/Southern Oscillation. On the basis of the longer time series, this aspect would nevertheless be valid.
4. Summary and conclusions
In this study, we used an objective Lagrangian model FLEXPART combined with a robust and recently developed dynamic three-dimensional backtracking algorithm to detect the major long-distance moisture sources and the relative importance for the atmosphere over the YRB. The method applied in the study allowed not only the identification of moisture sources for the YRB but also the quantification of moisture transport originating from the sources. The YRB was chosen due to the large spatial and temporal variability in precipitation of east summer monsoon over the region, which is caused by the convergent zones of atmospheric moisture transport from the South China Sea, Bay of Bengal, Arabian Sea and the East China Sea.
Because of the high storage and large time consumption, the long-time simulation configured with high spatial and temporal resolutions as done in this work is not easy to conduct; thus, we focused the study on a six-summer period (2004–2009) for the FLEXPART simulations. Although this period cannot be considered as a standard period at the climate scale, and the Lagrangian method has some limitations, the robustness of the results throughout the six-summer period strongly provided us the significant details of moisture sources and transport to the target YRB region. The main findings are summarized as follows:
1.The moisture source regions for the YRB covered the vast areas and presented the moisture transport patterns during summer, governed by advections driven by summer monsoon flows from tropical and subtropical source regions. We identified the four important moisture source regions affecting the YRB region: the South China and Philippine Seas (26.6%), the South Asian subcontinent and Bay of Bengal (20.5%), East China Sea (17.5%) and Arabian Sea (13.6%). All these moisture sources are consistent with the previous results derived from Eulerian approaches, except the Northwest Pacific Ocean. The YRB area was characterized by complicated transport structure from major distant moisture sources and the complex nature of summer monsoon precipitation.
2.The contribution of South China Sea moisture to the YRB region was dominant throughout the 10-day transport period with a clear prominent maximum between 2 and 5 day before arrival at the YRB. Owing to a longer distance of transport, the moisture supply from the Bay of Bengal and Arabian Sea were characterized by more transport time, approximately 4–8 and 7–10 days, respectively. Lagrangian backward projections of the mean source region location confirm that the Tibetan Plateau is an effective moisture transport barrier. A large part of the moisture transported from the Arabian Sea and Bay of Bengal could not reach the YRB, as a result of the loss of moisture in the progresses of landing and from the presence of Tibetan Plateau.
3.All moisture sources for the YRB exhibited an obvious inter-seasonal variability for each summer and subseasonal variations in importance of moisture sources within summer. The South China Sea is the dominant summer moisture source for the YRB supplying approximately 20 × 1013 kg water vapour during summer, with a peak in July. With a slightly little in June, the variation of Bay of Bengal source presented a similar properties as Arabian Sea source during summer, with the strongest moisture contributions equalling to these from South China Sea in July, and decreasing somewhat in August. The East China Sea as moisture source appears only in early summer (June), and the contribution of the YRB in July and August was almost negligible. In this regard, this significant finding was the identification the region over the northeast of East China Sea and its longitudinal extension to the Japan as a moisture source in early summer due to its short trajectories with the moisture contribution to the YRB, which have not been apparent in the moisture source distributions derived from the classical analyses (e.g. Tao and Chen, 1987; Ding, 1994, Huang et al., 1998; Simmonds et al., 1999; Xu et al., 2003a, 2003b; Chow et al., 2008), and this represents a major advantage of the Lagrangian approach. The subseasonal marching of the Asian monsoon circulation is an important element in regulating the relative importance of different sources. Its northward movements and its relative intensity during boreal summer could help explain the observed subseasonal differences in the contribution of the moisture sources to YRB. For example, for the South China Sea region, the strength and movement of south Asian monsoon is a determining factor in the strong contribution of the South China Sea to YRB. In middle summer, the strongest northward atmospheric transport by the south Asian monsoon combined with the Indian monsoon system implies that more water vapour can reach the target region. For this reason, the contribution of South China Sea, Arabian Sea and Bay of Bengal to YRB is more compared with that of the time during early or later summer.
4.Lastly, the relationship between moisture source variations and precipitation is examined in the inter-annual scale. In general terms, a good agreement has been obtained between the amounts of moisture originating in the South China Sea, Bay of Bengal and Arabian Sea regions and the amount of precipitation over the YRB region, which in turn confirmed the confidence of our results. We also evaluated the importance of moisture sources to the YRB in a wet and a dry summer. The moisture sources over the Arabian Sea and Bay of Bengal and the southward latitudinal extension of the South China Sea were considerably larger during the wet summer (2008) than during the dry summer (2004), while the intensity of the northwest sources was remarkably higher during dry summer with the large areas of higher values from South China Sea to East China Sea. These results based on the analysis of only one relatively wet and one dry year are required to evaluate the significance from the longer data records.
The results discussed herein relate only to 6 years and to the whole of the YRB, thereby providing a short ‘climatological’ overview of this source of moisture. This relatively short time frame can nevertheless be considered to be a standard period of time at the scale of the global climate, because of the lack of clear extremes in the major modes of climate variability, such as El Niño-Southern Oscillation (ENSO) or North Atlantic Oscillation. During the period of analysis, one negative NAO event was observed, as well as one instance of La Niña and two of El Niño, all of which were weak or moderate in intensity. Nevertheless, we are currently exploring an extension to the FLEXPART data set as far back as 1979, allowing us to explore the inter-annual variability of the source of moisture and the role of the main modes of climate variability such as ENSO. Recent changes in atmospheric circulation patterns are partially responsible for the trend of declining precipitation over the YRB (Ding et al., 2010; Zhang et al., 2010; Li et al., 2011) and the occurrence of major drought episodes (Zhang et al., 2009). So the inter-annual variability of these processes of moisture transport should be also assessed to understand their influence at different timescales. To undertake these research tasks, a longer period of data will be analysed and the results of this will form part of a further work.
It also should be stressed that the objectives of this study were focused concentrated on the identification of the moisture sources from the YRB. As an interesting research topic, effects of continental areas (e.g. the Tibetan Plateau) on the moisture transport could not be ignored and will be investigated in future work.
This research was jointly funded by the National Natural Science Foundation of China (Grant Nos. 41105027 and 41130960), the China Postdoctoral Science Foundation (Grant No. 20110490488) and the social commonweal profession research program of Ministry of Science and Technology of the People's Republic of China (GYHY201006009 and GYHY201006053).