Spatial distributions and seasonal variations of tropospheric water vapor over the Tibetan Plateau and the surrounding areas are explored by means of water vapor products from the high-resolution Atmospheric Infrared Sounder (AIRS) on board the Aqua satellite and the NASA Water Vapor Project (NVAP). Because NVAP has a serious temporal inhomogeneity issue found in previous studies, the AIRS retrieval product is primarily applied here, though similar seasonal variations can be derived in both datasets. Intense horizontal gradients appear along the edges of the plateau in the lower-tropospheric (500–700 hPa) water vapor and columnar precipitable water, in particular over the regions along the southeastern boundary. Rich horizontal structures are also seen within the plateau, but with a weaker gradient. In the mid- to upper troposphere (300–500 hPa), horizontal gradients are relatively weak. It is shown that there is always a deep layer of high water vapor content over the plateau with a peak around 500 hPa, which can extend from the surface to roughly 300 hPa and even to 100 hPa at some locations. This layer of high water vapor content has consistent influence on precipitating processes in the downstream regions such as the valleys of the Yellow and Yangtze Rivers. Estimated vertically integrated water vapor flux and moisture divergence in the two layers (500–700 and 300–500 hPa) further confirm the effect of the Tibetan Plateau on the downstream regions. In particular, the mid- to upper-layer water vapor (300–500 hPa) tends to play an essential role during both the warm and cold seasons, confirmed by the spatial distribution of seasonal-mean precipitation.
In addition to its dynamic and thermodynamic effects on weather and climate in East Asia and likely the entire Northern Hemisphere, the Tibetan Plateau can significantly influence spatial and temporal variations of water vapor in the East Asian monsoon region as well (Tao and Yi 1999). The plateau itself is often considered as a crucial moisture source or a “transfer station” for East Asia (Xu et al. 2002). In boreal summer, abundant moisture originated from the plateau region feeds heavy rainfall bands in the Yangtze River Valley. Recent studies also indicated that the Tibetan Plateau might be a main pathway for cross-tropopause water vapor transport (e.g., Tian et al. 2011); hence, the hydration of the global stratosphere could be especially sensitive to the changes of water vapor over the Tibetan Plateau (Fu et al. 2006; Smith et al. 2000). Past studies on water vapor over the plateau and the surrounding areas were based on limited ground station observations, reanalysis products, and limited satellite retrievals or any kind of combinations (Ji et al. 1989; Yang et al. 1992; Zhuo et al. 2002; Wang et al. 2006; Liang et al. 2006). Long-record radiosonde data are available for many locations but are limited over the plateau region (e.g., Zhang et al. 2012). Furthermore, the inhomogeneity issue has been discovered to be widely existing in the raw radiosondes (e.g., Trenberth et al. 2005), though enormous efforts have been made to identify and further correct errors and biases (e.g., Dai et al. 2011). Corrected radiosondes have been applied in the East Asian region to investigate variations of tropospheric water vapor on various time scales (e.g., Xie et al. 2011; Zhou et al. 2012; Zhao et al. 2012). Nevertheless, there is no doubt that a further improved understanding of tropospheric water vapor over the Tibetan Plateau including spatial (horizontal and vertical) distributions and temporal variations would still be necessary. This could be accomplished using the satellite-based retrievals for their continuous spatial and temporal coverage. Trenberth et al. (2005) demonstrated the reliability of the Special Sensor Microwave Imager (SSM/I) columnar water vapor product in exploring long-term variations of tropospheric water vapor following surface temperature changes during the past decades. However, the SSM/I-based water vapor products are only available over global oceans and are in general vertically integrated with no vertical information.
In the early 1990s, a water vapor product from the National Aeronautics and Space Administration (NASA) Water Vapor Project (NVAP) became available by merging retrievals from various satellites, such as the Television and Infrared Observation Satellite (TIROS) Operational Vertical Sounder (TOVS), SSM/I, and radiosonde observations. Although the dataset only extends to 2001, it has been applied in various aspects of atmospheric and climatic studies (Simpson et al. 2001; vonder Haar et al. 2003). Another important water vapor product is also available, which is derived from the Atmospheric Infrared Sounder (AIRS) on board the Aqua satellite launched in May 2002. This dataset provides a means for high-quality vertical profiles of both water vapor and temperature with a continuous spatial and temporal coverage (Divakarla et al. 2006; Tobin et al. 2006). Tian et al. (2011) used AIRS water vapor data to explore aspects of stratosphere–troposphere exchange near the Tibetan Plateau for the period of 2003–08. Interesting features of seasonal variations in upper-tropospheric–lower-stratospheric water vapor have been discovered, which are related to deep convective activities.
In the first part of our work on the water vapor over the Tibetan Plateau, Zhang et al. (2012) recently explored the applicability of the gridded (level 3) monthly tropospheric water vapor (version 5) retrievals from the AIRS instrument and the Advanced Microwave Sounding Unit (AMSU) over the Tibetan Plateau by comparing with carefully processed radiosonde data. It is shown that the monthly gridded AIRS/AMSU water vapor retrievals could in general provide a very good account of spatial patterns and temporal variations of tropospheric water vapor content over the Tibetan Plateau and the surrounding areas, specifically below 250 hPa where most atmospheric water vapor content is located.
The objectives of this study are thus to further explore the spatial distribution and seasonal variations of tropospheric water vapor over the plateau region and to examine how the atmospheric component of the water cycle functions over the plateau and more importantly how it may relate to rainfall variations in the downstream regions. To have a relatively longer time span, both NVAP and AIRS water vapor products are used and compared.
2. Data description
In addition to AIRS, two microwave sounders are also on board the Aqua satellite: AMSU and the Humidity Sounder for Brazil (HSB; Aumann et al. 2003; Lambrigtsen et al. 2003). Because the HSB instrument failed after a short period, the data used here are a combination of AIRS and AMSU [i.e., the version 5 monthly gridded (level 3) AIRS/AMSU product without HSB with a spatial resolution of 1° × 1°]. The original monthly products are simply the arithmetic averages weighted by the counts of the days at each grid, and actually include two files corresponding to the ascending and descending orbits of the satellite, respectively. Ascent is from the Southern Hemisphere (SH) to the Northern Hemisphere (NH) with an equatorial crossing time at 1330 local time (LT), while descent is from NH to SH with an equatorial crossing time at 0130 LT. Thus, the final monthly data used here are the averages of these two portions. The monthly AIRS/AMSU retrieval data are available from September 2002 to the present. Here, we focus on the period of January 2003–December 2010.
The NVAP dataset is designed to be model independent and relies mainly on satellite measurements. Early versions of NVAP (1988–99) included layered and total column water vapor on a 1° × 1° grid, combining water vapor measurements from radiosondes, TOVS, and SSM/I (vonder Haar et al. 1995; Randel et al. 1996). The so-called next-generation (NG) dataset, NVAP-NG (2000–01), included retrievals from SSM/I, AMSU, Advanced TIROS Operational Vertical Sounder (ATOVS), Special Sensor Microwave Temperature Profiler 2 (SSM/T2), Tropical Rainfall Measuring Mission (TRMM) Microwave Imager (TMI), and TOVS Pathfinder, on a 0.5° × 0.5° grid (Forsytheet al. 2003). Thus, the NVAP data can roughly cover the time period of 1988–2001.
To estimate water vapor flux and divergence fields for the AIRS/AMSU period (2003–10), zonal and meridional monthly wind components from the European Centre for Medium-Range Weather Forecasts (ECMWF) Interim Re-Analysis (ERA-Interim) are applied (http://www.ecmwf.int). The ERA-Interim fields have a 0.75° spatial resolution and are regridded to the level 3 AIRS grids. It is necessary to be cautious when any reanalysis product is used over the plateau because of very limited in situ observations (e.g., Zhang et al. 2012). Using independent sounding observations, Bao and Zhang (2013) recently evaluated the performance of several reanalysis products over the Tibetan Plateau, including the National Centers for Environmental Prediction–National Center for Atmospheric Research Reanalysis I (NCEP–NCAR-I), Climate Forecast System Reanalysis (CFSR), 40-yr ECMWF Re-Analysis (ERA-40), and ERA-Interim. It is found that ERA-Interim and CFSR tend to have a smaller root-mean-square (RMS) error and bias than NCEP–NCAR-I and ERA-40.
Also, the monthly precipitation data (3B43) from TRMM are used to estimate seasonal-mean rainfall in the region. The 3B43 is a standard TRMM product and merged from various rain-rate products including both passive and active microwave and infrared retrievals, and rain gauge information over land (Huffman et al. 2007). However, it is necessary to mention that reliable rain gauge information, critical for calibrating satellite retrievals, are also very limited in the plateau region. Wu (2012) assessed two satellite-based precipitation data, TRMM 3B42 (a 3-h temporal-resolution version of the monthly 3B43) and National Oceanic and Atmospheric Administration (NOAA) Climate Prediction Center (CPC) morphing technique (CMORPH) (based on various available passive microwave retrievals), over the Tibetan Plateau using surface rain gauges, and concluded that TRMM 3B42 tends to be more consistent with rain gauge data regarding both precipitation amount and frequency.
3. Temporal variation of water vapor over the Tibetan Plateau
Combined together, NVAP and NVAP-NG products and AIRS/AMSU data can last about 22 years. However, the consistency between these two distinct kinds of products has to be checked first. We focus on the domain of 27°–40°N, 74°–104°E, which roughly covers the entire plateau.
Figure 1 depicts both the monthly and annual-mean time series of total precipitable water content (PWC). Evident seasonal variations appear in these three datasets. However, annual-mean PWC in general does not vary much. Nevertheless, there is a decreasing trend and possibly an artificial jump around 1997/98 in NVAP, clearly indicating the existence of a temporal inhomogeneity issue, which is likely caused by merging various data resources. Obviously, one should be very cautious if applying NVAP and NVAP-NG for examining variations on longer-than-seasonal time scales. Relative to NVAP and NVAP-NG, the AIRS/AMSU product tends to provide a consistent description of temporal variations of water vapor (Fig. 1).
To further examine seasonal variations of water vapor, annual cycles in the domain-mean time series for these three products are computed and compared (Fig. 2), even though NVAP-NG lasts for only two years. Their seasonal cycles are very similar, although the NVAP data have a serious temporal inhomogeneity issue. In general, the NVAP and NVAP-NG products tend to have greater amplitudes than AIRS/AMSU during all months. In particular, during boreal summer (June–August), NVAP is much larger than AIRS/AMSU possibly because of extremely large values during the early period (1988–91).
Water vapor reaches its peak in boreal summer, resulting from the seasonal march of the South Asian summer monsoon, which brings warm and wet air from lower-latitude regions. As Jiang and Fan (2002) and Zhou et al. (2005) pointed out, during boreal summer, the primary water vapor sources over the plateau are the Bay of Bengal and Arabian Sea, and the water vapor transported to the plateau is hence strongly modulated by the low-level southerly monsoon flow; in boreal winter, westerly dry flows from the midlatitudes become dominant over the plateau region so that tropospheric water vapor content reaches its minimum. Thus, the sources of water vapor over the Tibetan Plateau strongly vary with season, following an evident seasonal shift in the large-scale circulation. Distinct seasonal features including horizontal distributions and the vertical structure of water vapor will be further examined in the next sections.
4. Spatial distribution of water vapor over the Tibetan Plateau
As shown in Fig. 1, NVAP has temporal inhomogeneity issues. We will thus primarily focus on the AIRS/AMSU data hereafter, even though their annual cycles are similar to each other. Figure 3 illustrates the spatial distributions of seasonal-mean columnar water vapor from AIRS/AMSU data during 2003–10 for the warm (April–September) and cold (October–March) seasons, respectively. Spatial structures of seasonal-mean water vapor from NVAP and NVAP-NG are also computed for comparison (Fig. 4).
Water vapor content in general decreases from the southeast to the northwest over the Tibetan Plateau, with intense horizontal gradients along the south-southeastern boundary (Fig. 3). It is of interest to note that a regional maximum appears over the southeastern tip of the plateau. The horizontal gradient within the plateau is relatively weak specifically over the central and northern part of the plateau, though regional minima of water vapor are readily seen, partly manifesting regional elevations. Much more water vapor is observed during the warm than the cold season, following evident seasonal variations as shown in Fig. 2, while the seasonal differences in spatial distributions of water vapor are not evident. Abundant water vapor and associated large horizontal gradients over the southern part of the plateau during the summer season obviously result from the northward transport of water vapor associated with the low-level monsoon flow and also the steep orographic slopes blocking its farther north-northwestward intrusion. Along the southern edges are the lower reaches of Yarlung Zangbo River usually called “the Southern Tibet Valley,” the major water vapor entrance from the lower latitudes (Gao 2008). Elevations are relatively lower to the east than to the west, forming a southeast–northwest orientation of horizontal water vapor gradients. On the other hand, the northern edges of the plateau, because of lower elevations than in the central part of the plateau, are actually not the driest regions. The regions with lower topographical heights tend to have more water vapor content in that more water vapor tends to be in the lower layers of atmosphere. Within the plateau, on the other hand, rich horizontal structures of water vapor tend to manifest the complexity of topography as well. During the winter season, as stated above the mean flow is in general controlled by the midlatitude westerlies so that the atmosphere over the plateau is dry. Dry westerlies in winter and wet southerly monsoon flows in summer are thus the reasons for much more summertime water vapor over the plateau than in the cold season, suggesting that the dominant control of atmospheric water vapor content by the large-scale circulation shifts, though seasonal changes in surface conditions are also very important by affecting local evaporation. Furthermore, complex and steep orography tends to shape intense horizontal gradient along the edges of the Tibetan Plateau.
Similar features can also be found in NVAP and NVAP-NG (Fig. 4). Nevertheless, detailed discrepancies can readily be seen. The horizontal gradient along the southern edges in NVAP and NVAP-NG is much weaker than in AIRS/AMSU data. Furthermore, there is no local maximum in the southeastern part of the plateau during both seasons (Figs. 3 and 4). Within the plateau, evident differences exist in local minima as well. These differences may need to be further clarified, though we tend to have more confidence in the AIRS/AMSU product.
Columnar water vapor content can illustrate overall horizontal distribution of water vapor, but with no vertical information. To examine vertical distributions of water vapor over the Tibetan Plateau, layer-averaged precipitable water is further applied. As above, we focus on the AIRS/AMSU data, while the NVAP and NVAP-NG product is applied for comparison. Because the NVAP and NVAP-NG layered precipitable water content is available only below 300 and 200 hPa, respectively, we use the two layers of 500–700 and 300–500 hPa to represent the lower and upper troposphere, respectively.
Figure 5 depicts horizontal distributions of layered water vapor during warm and cold seasons. Spatial structures for lower-tropospheric water vapor generally follow that of columnar water vapor especially along the south-southeastern edges where high horizontal gradients are located, indicating the dominance of lower-tropospheric water vapor. Large horizontal gradients are also seen in the upper-tropospheric water vapor. However, gradients in the upper layer are not as steep as in the lower layer. The fact that much richer horizontal structures appear in the lower-tropospheric water vapor than at the upper layer suggests a direct impact from the surface as we discussed above. It is further noted that the maximum/minimum zones at the two layers are not exactly matched vertically. Also, the directions of the horizontal gradient are different between the two layers. Water vapor content decreases from the south to the north at the upper-tropospheric layer, while at the lower layer the gradient direction is primarily from the outer boundary to the central especially along the southeastern edge where the gradient is oriented from the southeast to the northwest, confirming that orography strongly impacts the lower layer while its impact becomes relatively weak at the upper layer. Obviously, these features occur during both seasons. Nevertheless, detailed, seasonal differences in spatial distributions of water vapor can be found, again because of seasonal variations in the large-scale circulation and local surface conditions.
The differences between the lower- and upper-tropospheric layers suggest a frontal-like structure of tropospheric water vapor over the plateau, which is roughly oriented from the northwest to the southeast. The upper-level westerly dry air from the midlatitudes tends to be overlying the low-level southerly moist flow specifically during boreal summer. This frontal-like vertical structure might have characteristics of a warm front during boreal summer and characteristics of a cold front during boreal winter depending on seasonal changes in the large-scale monsoon circulation and the midlatitude westerly flows.
In general, similar features appear in NVAP and NVAP-NG. However, large differences exist especially in the upper layer. Overall NVAP and NVAP-NG tend to have larger values than AIRS/AMSU, and show relatively weak seasonal changes. Large discrepancies are also observed in the direction of the horizontal gradient, particularly for the upper-tropospheric layer during the cold season.
Figure 5 provides a gross account of the three-dimensional frontal-like structure of tropospheric water vapor over the plateau with possible, distinct seasonal characteristics. To get a more detailed picture of the three-dimensional structure of tropospheric water vapor, the monthly AIRS/AMSU water vapor data with full vertical resolution are further applied in next section. Because our aim is to explore how the plateau may affect spatial distributions and temporal variations of tropospheric water vapor not only over the plateau but also in the surrounding regions, the analyses will thus extend to cover a relatively larger area (Figs. 6 and 7).
5. Water vapor transport from the Tibetan Plateau
a. Vertical cross sections of seasonal-mean water vapor
To examine zonal and meridional variations of water vapor, relative departures [D(i, j, k)] from corresponding zonal averages are estimated for each season:
Here, i, j, and k denote longitude, latitude, and pressure levels, respectively. Here, q(i, j, k) (g kg−1) represents the water vapor mixing ratio and (g kg−1) is the corresponding zonal average. Various vertical cross sections of seasonal-mean relative departures (i.e., D) are then constructed. We intend to provide a detailed account of three-dimensional structures of tropospheric water vapor over the plateau for both the warm and cold seasons.
Figures 6 and 7 illustrate vertical cross sections of the seasonal-mean D at several specific longitudes and latitudes during warm and cold seasons. From the longitudinal–vertical (pressure) cross sections, we can readily find that positive D peaks around 500 hPa over the plateau at all four specific latitudes and during both seasons, forming a positive D zone extending roughly from 700 to 300 hPa. This is further confirmed by the latitudinal–vertical (pressure) cross sections at 84.5° and 93.5°E during both seasons. This high, positive D zone covers the entire plateau indicating much higher water vapor content over the plateau relative to the surrounding areas. Note that to the west of the plateau, D tends to be negative and is thus basically smaller than to the east of the plateau where positive D values are always located between 300 and 500 hPa. However, the opposite appears at the lower troposphere (below 700 hPa) along 28.5°, 32.5°, and 35.5°N specifically during the warm season. In addition, above 300 hPa and near the surface, negative D values occur in the downstream regions of the plateau and form a three-layer vertical structure of water vapor. A layer with high positive D values over the plateau peaking around 500 hPa can extend eastward to about 120°E.
Water vapor transport from the south is evident on the latitudinal–vertical (pressure) cross-section plots at 84.5° and 93.5°E. A thick layer of high positive D can extend from the surface to roughly 200 hPa. It is also interesting to note that in contrast to positive D values to the south of the plateau, albeit weak, D is in general negative to the north of the plateau and centered around 300 hPa. The meridional variations at the upper troposphere are associated with the intrusions of midlatitude dry westerly flows.
Above 300 hPa, D decreases northward for all areas and during both seasons. Also, D decreases from the east to the west in the warm season as mentioned above. Over the region (28°–33°N, 85°–95°E), D keeps positive even to 100 hPa for both seasons along 28.5° and 32.5°N,which might confirm that this is the major moisture source region for the stratosphere (e.g., Tian et al. 2011).
In summary, D is positive over the Tibetan Plateau from the surface to about 300 hPa and even to 100 hPa at some locations, manifesting a deep layer of high positive D with a peak at around 500 hPa. The maxima of D could reach 1.5 and 1.2 during warm and cold seasons, respectively. For the surrounding areas, larger D values are found in the south than in the north as expected. Also, D tends to be larger at the west side of the plateau below 700 hPa during the warm season, while it is opposite above this level suggesting the mid- to upper-tropospheric air's midlatitude origin especially to the west of the plateau. These zonal and meridional differences tend to be smaller during the cold season when the low-level southerly monsoon flow disappears. The evident high positive D between 300 and 700 hPa suggests the importance of water vapor located at this layer. Zhai and Eskridge (1997) estimated that the water vapor content at this deep layer might actually contribute to about 80%–90% of PWC over the plateau. Thus, we are focused more on the water vapor within this layer in section 5b.
b. Characteristics of water vapor transport
The existence of a deep layer of high positive D with a maximum around 500 hPa over the Tibetan Plateau illustrated in Figs. 6 and 7 suggests that the plateau may not only affect stratospheric water vapor balance but also effectively influence the tropospheric water vapor budget in the downstream regions. To examine this effect, vertically integrated water vapor flux Q and divergence of water vapor flux Qdiv are estimated for the two layers of 500–700 and 300–500 hPa:
Here, p1 = 700 and p2 = 500 hPa for the lower layer; p1 = 500 and p2 = 300 hPa for the upper layer. The wind fields are derived from the ERA-Interim products.
Seasonal-mean water vapor flux (i.e., Q), divergence of moisture flux (i.e., Qdiv) for both layers, and vertical velocity ω at 500 hPa are shown in Figs. 8 and 9 . For moisture flux fields (i.e., Q), larger differences between these two layers are observed during the warm than the cold season. During the warm season (Fig. 8), the magnitudes of moisture flux at the lower-tropospheric layer (500–700 hPa) appear weaker over the plateau than in the surrounding regions except along the southeastern edge where the intense southwesterly fluxes of water vapor occur. Along the northern edge, the west-northwesterly water vapor fluxes appear, which are stronger than over the plateau. In the upper-tropospheric layer (300–500 hPa), the moisture fluxes are basically zonal over the plateau except along the southern edge where the southwesterly flows are dominant, and there is no minimum in moisture fluxes over the plateau. The water vapor flux appears strong to the east of the plateau and in the downstream regions of the Yellow and Yangtze Rivers, while it is relatively weak to the southwest of the plateau. There is an increasing tendency of water vapor flux from the west to the east especially between 27° and 42°N specifically for the upper layer, roughly the latitudinal range of the plateau. Comparing the structures of water vapor fluxes between the two layers indicates that water vapor transport in the upper-tropospheric layer from or via the plateau plays a leading role for the downstream regions during the warm season. Moisture flux divergence provides another useful means to examine water vapor sources/sinks. At the lower layer (500–700 hPa) moisture convergence is in general seen over the plateau except along its south-southeastern edge, while weak divergence or convergence tends to occur at the upper layer (300–500 hPa). During the warm season, upward motions are dominant over the plateau, specifically along its southeastern and northwestern edge. It can thus be deduced that the plateau helps to collect moisture from the lower levels, mainly benefited from westerly and southwesterly water vapor transport, and transport them upward to higher levels, manifesting the typical overturning circulation feature. In 300–500 hPa, two zones of negative values indicating moisture convergence appear to the east of the plateau: one is between 102° and 109°E and another is over the downstream region of the Yangtze River roughly matching the high values of water vapor flux. The latter convergence zone is also found in 500–700 hPa with centers shifted southwestward. It is noted that moisture divergence with relatively high values is mainly located along the southern and the western edges of the plateau at both layers, where moisture flux vectors turn to the northeast and begin to strengthen. In particular, the flows along the downstream valley area of Yarlung Zangbo River with a strong moisture flux and divergence in 500–700 hPa are the major contributors carrying water vapor upslope into the plateau (Yang et al. 1987). West of the plateau, there is another large zone of moisture divergence for both layers, indicating the effect of dry westerly flows, consistent with Wang et al. (2009). For moisture convergence zones in the upper layer specifically to the east of the plateau, they are mostly located over the regions of high water vapor flux. Thus, they may be considered beneficiaries of water vapor transport from the same layer over the plateau. The convergence of water vapor at the upper layer between 102° and 108°E may also be partly affected by dry midlatitude westerly winds. The convergence zone located downstream of the Yangtze River is somewhat weaker but appears in both layers. Upper-troposphere water vapor from the Tibetan Plateau meets with the westerly wind perturbations from the northwest mid- and downstream of the Yangtze River (Jiang and Li 2009). This convergence of water vapor flux certainly has an important influence on precipitating systems in these regions.
During the cold season (Fig. 9), both water vapor flux and divergence are weaker than during the warm season and the differences between the two layers tend to be smaller. The streamlines for water vapor flux become more zonal over the plateau with only weak perturbations. The oceanic, southerly transport of water vapor along the southern edges of the plateau, a major contributor of water vapor during the warm season, becomes much weaker at 500–700 hPa and in general disappears at the upper layer (300–500 hPa).Westerly winds and associated perturbations are dominant in the cold season. Observed water vapor flux between 27° and 40°N during the cold season is similar to that in the warm season. In both layers the northwesterly flows in northeastern China have intensified in the 38°–45°N, 110°–120°E area, though the magnitude of the moisture flux is relatively low. A relatively strong moisture flux appears in the lower layer in southern China, indicating direct transport of oceanic moist air from lower latitudes; and at the upper troposphere the highest water vapor flux zone has also shifted southeastward to the coastal regions of the Zhejiang and Fujian Provinces (27°–33°N, 117°–122°E), though the corresponding moisture convergence becomes much weaker. Furthermore, centers of high moisture divergence are discovered moving to the south and southwestern edges of the plateau for both layers. Note that moisture convergence in 500–700 hPa becomes weaker and only covers the area to the north of the plateau, however, it extends to the west, where a divergence zone is discovered in 300–500 hPa, exactly corresponding to strong upward motion at 500 hPa. The upward flow in the southern area in the warm season becomes weaker during the cold season. However, a large area of intense vertical motions appears west of the plateau, which corresponds to the midlatitude-origin precipitating systems and associated moisture transport during the cold season. The divergence zone at 500–700 hPa to the north of the plateau during the warm season disappears during the cold season possibly because of weakened convergence over the plateau, while the convergence zones over the plateau and just to the east are still seen during the cold season, albeit weaker. This quasi-stationary convergence zone might indicate a unique climate feature in this area. There is another convergence region located in the Tarim basin (37°–42°N, 77°–88°E), to the northwest of the plateau, occurring for both layers only during the cold season possibly because of weakened convergence over the plateau as well. Major convergence zones could also be deduced from the 32.5°, 35.5°, and 38.5°N, and the 102.5°E cross sections in Figs. 6 and 7.Therefore, most of the moisture divergence and convergence zones appearing in the summer season can still be seen during the cold season, though their preferred locations may have shifted and amplitudes become smaller. Hence, water vapor over the Tibetan Plateau specifically between 300 and 500 hPa including that pumped from lower levels may be considered as a consistent contributor to water vapor in the downstream regions such as the valleys of the Yellow and Yangtze Rivers. Thus, the southerly transport of water vapor from the lower latitudes that can only be seen during the warm season may just act as an enhancer of transporting water vapor from the plateau to the downstream areas in the warm season, and it may also explain the seasonal difference of the magnitudes of tropospheric water vapor over the plateau and the regions to the east.
Seasonal-mean precipitation during 2003–10 is further shown in Fig. 10 using the TRMM monthly rainfall product 3B43. There are evident seasonal differences in precipitation for the entire region. During the warm season, intense rainfall occurs along the southern edges of the plateau because of the summer monsoon; seasonal rainfall also covers large areas to the east of the plateau. During the cold season, seasonal-mean precipitation seen to the west of the plateau is associated with the midlatitude systems, but is very weak along the southern edge of the plateau. East of the plateau, the major precipitation zone generally shrinks to the southeastern corner of the domain. The precipitation fields are to some extent consistent with moisture flux and divergence fields in the lower levels.
To further understand the relationships between precipitation and moisture flux transport, seasonal-mean water vapor flux and divergence in 700–1000 hPa are depicted in Fig. 11. Complicated features can be seen in the divergence field of 700–1000 hPa along the boundaries of the plateau especially along the southern edge likely caused by complex, steep orography. In general, major rainfall zones to the south of the plateau and along its southern edge, as well as in southeastern China during the warm season, correspond to intense moisture flux in 700–1000 and 500–700 hPa. Specifically, along the southern edge of the plateau, the precipitation zone tends to be located over the low-altitude regions and is in general corresponding to low-level (700–1000 hPa) moisture convergence. However, it should be further emphasized that the correspondence between precipitation and low-level moisture convergence seems to be heavily modulated by local orographic circulation along the southern edge of the plateau. Also, along the eastern edge of the plateau and in the downstream regions of the Yellow and Yangtze Rivers, the relationships between precipitation and water vapor flux and convergence in 700–1000 hPa tend to be weak particularly during the cold season. This tends to suggest that to the west and east of the plateau, and in southern China, precipitation during the cold season could be contributed to primarily by moisture at the upper layers.
6. Summary and conclusions
The high-resolution AIRS/AMSU retrievals and NVAP and NVAP-NG products are applied to investigate the spatial distribution and seasonal variation of tropospheric water vapor over the Tibetan Plateau and the surrounding areas. The NVAP is shown to have a serious temporal inhomogeneity issue, though it has similar seasonal variations as AIRS/AMSU. Thus, the AIRS/AMSU retrieval data are focused on here. The NVAP and NVAP-NG product is applied only for comparison.
Columnar water vapor content decreases from the southeast to the northeast over the Tibetan Plateau with intense horizontal gradients along the boundaries of the plateau during both the warm and cold seasons, in particular along the southeastern edges. Complex, steep orography is the major reason through the blocking of upslope low-level winds. Seasonal variations are evident in columnar water vapor over the plateau with much more water vapor being seen during the warm season than during the cold season. Nevertheless, the seasonal differences in spatial distributions of water vapor are in general weak. Also, spatial distributions and seasonal variations of columnar water vapor are basically dominated by the lower-tropospheric-layered (500–700 hPa) water vapor. On the other hand, because of the relatively weak direct impact from the surface, the upper-tropospheric-layered (300–500 hPa) water vapor manifests different features than the lower layer of water vapor and columnar water vapor. These vertical differences suggest a three-dimensional frontal-like structure of tropospheric water vapor over the Tibetan Plateau. This frontal-like structure may have distinct seasonal characteristics caused by seasonal variations evident in the large-scale circulations. Similar results can be derived from the NVAP and NVAP-NG product, though it has a weaker seasonal cycle and in general shows weaker horizontal gradients than the AIRS/AMSU data.
Detailed spatial (horizontal and vertical) structures of water vapor are further explored by estimating relative departures (i.e., D) from corresponding zonal averages during both the warm and cold seasons using the AIRS/AMSU retrieval data. Various longitudinal and latitudinal vertical cross sections during both seasons indicate that D tends to be positive over the Tibetan Plateau from the surface to about 300 hPa and even to 100 hPa in some locations with a peak around 500 hPa. This deep layer tends to have much larger positive D values than over surrounding areas. In the surrounding areas, D tends to decrease northward as anticipated. An east–west difference also exists in D. Below 700 hPa, D tends to be larger to the west of the plateau, while it is opposite above this level. These zonal and meridional differences in general become weaker during the cold season likely because of the disappearance of the low-level southerly monsoon flows.
Vertically integrated water vapor flux and moisture divergence are further estimated for the layers of 500–700 and 300–500 hPa to explore what the major sources of water vapor would be over the plateau and how various water vapor transports may vary seasonally and affect the downstream regions. Water vapor convergence occurs at 500–700 hPa over the plateau, while divergence tends to appear at 300–500 hPa, suggesting an overturning circulation pumping moisture from the surface/lower levels to the upper levels. The tropospheric layer of 300–500 hPa over the Tibetan Plateau seems to be a constant contributor for water vapor in the downstream regions including the valleys of Yellow and Yangtze Rivers. In other words, water vapor in the mid- to upper troposphere over the plateau has an essential effect upon precipitation processes in these downstream regions during both seasons. This is generally confirmed by the spatial distributions of seasonal-mean precipitation. To further explore the role of the plateau, variations of tropospheric water vapor on subseasonal and interannual time scales will be focused in future work.
This research is jointly supported by the State Key Basic Research Development Program (Grant 2012CB417204), the R&D Special Fund for Public Welfare Industry (Meteorology) by the Ministry of Finance and the Ministry of Science and Technology (Grants GYHY200806007, GYHY201006014, and GYHY201206039), and the National Natural Science Foundation (Grants 41175064, 41175080, 40875022, and 40633016). We would thank three anonymous reviewers for their comments and suggestions. Detailed suggestions from the editor, Dr. Aiguo Dai, greatly improved the manuscript.