Previous studies on precipitation have mostly been based on ground observations and, as a consequence, the results from these studies lack adequate spatial representation. Ground observation stations in this region are located primarily in plain river basins and the eastern TP; there are few observations in the western TP (Chen et al., 2012). In contrast, the TRMM data cover the entire TP, possess high temporal and spatial resolution, and are highly reliable (Shen et al., 2010). Therefore, the TRMM 3B43 and 3A12 datasets were used in this study.
The majority of TP rainfall occurs in summer. Figure 1 presents the summer precipitation (June, July and August) over the TP during 1998-2010. The precipitation amount was less than 100 mm (rate of 1 mm d-1) over the western TP, northwestern TP and Qaidam Basin, but was more than 800 mm over the southeastern TP. The precipitation amount decreased from the southeast to the northwest. The distribution of precipitation based on the TRMM data is basically consistent with that of ground observations (Fu et al., 2006). However, it is clear that the TRMM data possess advantages over ground observations in depicting the spatial distribution of precipitation (Fujinami et al., 2005; Fu et al., 2006).
The EOF method was used to assess the temporal and spatial characteristics of summer precipitation. The accumulated variance of the first three eigenvectors of precipitation exceeded 73% (Table 1). Thus, the first three eigenvectors are analyzed in this paper.
Figure 2 shows that spatial patterns depicted by the first three eigenvectors are significantly different to each other. Figure 2a shows that the first eigenvector has an explained variance of 34.9% and that the summer precipitation annual anomaly over the south and north of the Tanggula Mountains changes inversely. This reflects the effects of the terrain on precipitation processes.
The second eigenvector of summer precipitation has an explained variance of 23.7% (Fig. 2b). The precipitation anomaly over the eastern and western areas of the TP changes inversely, which is consistent with observations (Fujinami et al., 2005; Fu et al., 2006). Water vapor can barely reach the western regions, which makes the western TP a dry region. This indicates that the precipitation annual anomaly changes inversely between the western TP and the Hengduan Mountains in the east of the Tanggula Mountains. The vapor that moves from the Hengduan Mountains onto the TP only reaches the 90°E region (the source of three rivers). This pattern reflects the transport border for vapor that comes from the Bay of Bengal onto the TP through the valleys of Nujiang River, Lancang River and Jinsha River, the Yangtze River. The third eigenvector accounts for up to 15% of the total variance. This pattern illustrates that the annual anomaly of TP precipitation changes inversely between the central TP (between Mount Kunlun and Gangdis\e Range) and the surrounding TP. The central TP is characterized by a dry region that is rich in permafrost, so the precipitation may come from regional water cycles of ice and snow and the permafrost active layer (Duan and Wu, 2005; Wang and Guo, 2012). Based on the above analysis, the TP was divided into five parts (Fig. 3) to investigate the characteristics of precipitation and hydrometeors.
Figure 4 shows the monthly variation of precipitation in the different regions, which illustrates that precipitation over the TP possesses monsoonal features, i.e., rainfall is concentrated during summer, and that large differences exist among the five sub-regions. Region I is far from the influence of the summer monsoon over the Indian peninsula. The precipitation in Region I is much less compared to the other regions, showing little seasonal variation, possibly because of its terrain (Guo and Wang, 2014). The precipitation in the other four regions peaks in summer. Additionally, the amount of precipitation in Region IV and Region V is significantly greater than that in the other three regions, which can be attributed to the water vapor advected by the South Asian summer monsoon (Goswami, 2005). Specifically, Region V shows the greatest monthly accumulated precipitation over the entire TP, due to the abundant vapor transported by the monsoon. The precipitation amount in Region I and Region III is much lower than in the other three regions because both are influenced the least by the South Asian monsoon (Yeh and Gao, 1979).
The accumulated monthly precipitation is highly unevenly distributed over the TP. The following analyses may explain why the precipitation has such spatial and seasonal variations. First, we analyzed the seasonal CWC (the sum of the CIC and CLWC, Fig. 5), then the precipitable water (PW) (the sum of PLWC and PIC, Fig. 6) and, finally, the LH profiles (Fig. 7), with the aim to elucidate the possible transformation among phases of hydrometeors.
Figure 5 illustrates the seasonal variations of the CWC profiles in the five regions. It is important to note that each diagram in Figs. 5, 6 and 7 has its own starting height, to reflect the average elevation of the region. Region I (Fig. 5a) shows double peaks of CWC in all seasons except winter. The first peak occurs near the lowest plotting height at 4 km, which reflects evaporation from the surface, and the second peak occurs at ∼9 km, which represents pure ice water content (Gao et al., 2003). Apparently, the CWC in winter is the lowest among the seasons, and this is either due to a lack of advection toward this region or large retrieval uncertainties over ice/snow-covered surfaces.
Region II (Fig. 5b) has the lowest CWC over the entire year compared to the surrounding regions. Region III (Fig. 5c) also has very low CWC, but we can see the LWC has a peak at around 3 km (near-surface). Region IV and V (Figs. 5d and e) have very similar vertical profiles to each other, and the CWCs in both regions show double peaks, with much higher CWC in Region V than Region IV.
From Fig. 5, it is clear that the maximum CWC happened in different seasons in different regions; specifically, spring in Regions I and V, and autumn in Regions II, III and IV. In Region I, summer shows a very similar CWC vertical distribution as spring, except for lower values at 4 km. Winter shows the lowest CWC in all regions except Region IV, and summer shows the lowest CWC in Region IV. Compared to the lower peaks in all five regions, Region V has the largest CWC and Region II has the lowest CWC. Region I shows a high peak in three seasons and Regions IV and V show very clear high peaks in all four seasons, with the highest in autumn in Region V.
Figure 6 shows the corresponding PW profiles, which is the sum of the PLWC and PIC. The PW decreases with height, which is the same as CWC and changes with height. However, the maximum PW shows different seasonal variations in different regions. In Regions I and II, PW is largest in spring and summer and lowest in winter. In Regions III and IV, PW in spring and autumn is larger than that in summer and winter, and the PW minimum occurs in summer in region IV. In Region V, PW is largest in spring and is lowest in winter, which is similar to the seasonal trend of CLWC. In total, PW in Region V is the largest over the entire TP.
To further analyze the vertical distribution and relationship of CWC and PW, Fig. 7 illustrates the profiles of LH. Similar to CWC and PW, LH decreases with height. The LH is high at near-surface heights; however, unlike the CWC, there is a peak at 9 km in Regions IV and V. It reflects the LH release during the induced condensation of water vapor, and also reveals that the condensation process is more intense than the former. As a result, LH strengthens the convection and, thereby, the LH can reach a greater height, at which the CLWC is smaller .At this height, the liquid water condenses, almost turning into ice crystals. In ice-phase cloud, the LH becomes smaller. In Regions II and III, CWC and PW are at a minimum, which also corresponds to a minimum LH.
Figure 7 shows that LH is relatively large below 8 km. This suggests that water vapor condensation over the TP mainly occurs near the surface, which partly explains the strong convection over the TP caused by diabatic heat near the surface. This is wholly in agreement with (Fu et al., 2007). The LH profile of the TP reaches a maximum at around 6-7 km. The LH in Region V is largest over the TP, and is seven to eight times larger than that in Regions II and III. The LH in Regions I and II is nearly the same. The LH maximum is in autumn in Regions III and IV, but in summer in Region V. In all regions except Region IV, the LH in winter is the lowest of the year.
As shown in Figs. 5, 6 and 7, the seasonal variations of PW and CWC are almost the same in all regions over the TP, which means that PW and CWC are closely related to each other. The hydrometeors over the eastern and southeastern TP are more abundant, while they are less over the central and northwestern TP. In seasonal terms, the CWC maximum occurs in summer in the western and central TP. Over the eastern TP, the CWC in summer is smaller than that in spring and autumn, especially over the northeastern TP; hydrometeors in summer are less than in winter. This can be attributed to the differences in moisture transportation and weather systems in different seasons. In Region III, where westerly winds prevail, water vapor mainly comes from the westward stream which originated from southern China flows. Considering that the Qaidam Basin lacks abundant moisture, little precipitation in Region III can be expected.
The 2A12 algorithm cannot provide accurate CLWC and CIC in the mixed-phase region of cloud and precipitation; therefore, the boundary of the CLWC and CIC is unclear. However, regarding the LH profiles in Fig. 7, the LH is larger in Regions IV and V than in the other regions. During the summer half of the year, compared to the winter half, uplift in Regions IV and V is stronger than in the other regions. Therefore, the profile distribution of PW and CWC may be rational, except for some disparity in values.
In general, the CLWC and PW should not equal precipitation. To discuss the relationship between hydrometeors and precipitation, the correlation coefficients between precipitation and a selection of variables——CIC, CLWC, CWC, PIC, PLWC and PW——were calculated. The hydrometeor content was defined as the summation of all hydrometeors in a unit area:
\begin{equation} \label{eq1} p=\sum_{i=1}^{14}{w_i H_i} . \end{equation}
Here, p is the hydrometeor path (g m-2), wi is the content of each hydrometeor in a specific level (g m-3), and Hi is the thickness of each level.
There is a complex link between summer hydrometeor and precipitation values (Table 2). Precipitation is rare in Regions I and III. In Region I in particular, the precipitation amount is small, and monthly precipitation changes are small within a year, i.e., monsoonal characteristicsare less obvious.Thus, the correlation between hydrometeors and precipitation is poor. Region III is located in the northeastern TP and is controlled by westerly winds rather than the monsoon, meaning that the correlation here is also weak. This could be attributed to the fact that there is little water vapor advected in Regions I and III, where precipitation mainly comes from the regional recycling of vapor (Wang and Guo, 2012). In addition, the satellite data resolutions are low, meaning they are unable to capture showery precipitation. In the other three regions, the correlation between CIC and precipitation is significant; however, the CLWC is only significantly correlated with precipitation in Regions II and V. In the eastern TP, the PWC and PIC are both correlated with precipitation significantly. A key feature is that, south of the Three Rivers (Yangtze River, Yellow River and Lantsang River) Source, precipitation is closely related to the CIC and PW, and the effect of orographic lift is notable. In Region IV (the Tanggula Mountains), no significant correlation exists between precipitation and CLWC or CWC. The cloud type over this region is mainly high cloud, in which ice prevails. As a result, the correlation between precipitation and CIC is significant. In the western and northeastern TP, where vapor is not readily available, the precipitation amount is small and related poorly to hydrometeors. This may be caused by two reasons: (2) the accuracy of the retrieving algorithm is limited and (2) precipitation is mainly derived from regional water recycling.
Additionally, not all CWC results in precipitation, and different types of cloud have various water amounts. It should be noted that the formation of precipitation from clouds is a dynamic process, so hydrometeors can only partly reflect precipitation, which results in a poor correlation between precipitation and hydrometeors.
To further discuss the spatial and temporal distributions of each hydrometeor type and the corresponding influencing factors, the zonal and meridional distributions of hydrometeors are presented in Figs. 8 and 9.
Figure 8 illustrates the anomalies of the zonal vertical distribution of hydrometeors, LH and vertical circulations along 30°-37.5°N and 5°S-5°N. In the northern mid-latitudes, there is a large-value region of hydrometeors (positive anomaly) within 120°-160°E (Fig. 8a). In this area, there also exists a maximum center of CIC, and the center corresponds to the upward branch of vertical circulation over the western Pacific. To the east and west of the TP, the heights of CWC are apparently different to that over the TP. In the equatorial region, two maximum centers exist within 60°-100°E and 120°-160°E, respectively, which both correspond to the upward branch of Walker Circulation (Fig. 8d). There is a minimum center within 160°-80°W, and this corresponds to the downward branch of Walker Circulation. It can also be seen that there is Walker-like circulation in the midlatitudes of the Northern Hemisphere.
In Figures 8b and e, over the midlatitudes in the Northern Hemisphere, PW in the Western Hemisphere is notably larger than that in the Eastern Hemisphere. There is a positive anomaly center of PW over the western Pacific, and PW over the TP is larger than that of other regions within the same latitude. Over the west of the TP, there is a weak negative anomaly center. Similarly, a positive anomaly center of PW is found on the upward branch of the Walker circulation. Comparatively, in the equatorial region, the zonal negative anomaly center of the PW and CWC is also found in the east of the Pacific.
Studies show that, during the summer, the TP can also be a heat source (Flohn, 1957; Yeh et al., 1957), which creates convection and a large amount of cloud due to large-scale upward motion. There is a positive anomaly of LH in the upper troposphere over the central Pacific. The center of the positive anomaly is at the top of the upward motion. At the same levels, there are negative anomalies of LH centers in other regions. However, in the lower level over the TP, there is a negative anomaly center of LH; the temperature is lower compared to that in other regions in the same latitude. In the upper troposphere over the TP, there is a large positive anomaly center of LH, which corresponds to the PW positive anomaly center, as well as to the upward motion. The negative and positive anomaly of LH separates at the boundary between CIC and the CLWC, which is with the drag of vertical motion. The ascending vapor condenses and heats the atmosphere. In the anomaly centers of PW, CLWC and CIC are consistent in two latitude cross profiles. However, these are different from the situation over the TP and its adjacent regions. Over the western Pacific, PW is mainly from the CLWC, but over the eastern Pacific, it is mainly from the higher CLWC transforming into CIC.
In order to analyze the special characteristics of hydrometeors in temporal and spatial distributions over the TP, Fig. 9 illustrates the meridional distribution anomalies and vertical circulations of hydrometeors and LH between 40°S and 40°N. In Fig. 9a, the TP causes an extension of the upward branch of the Hadley Circulation stretching to the south of the TP. In the upward branch of the Hadley Circulation, the meridional positive anomaly center of CLWC and CIC is within 10°-20°N. Similarly, negative anomalies of hydrometeor centers are found in the downward motion area of the Hadley Circulation in the Southern Hemisphere. Over the southern slope of the TP, there is an upward motion area of monsoon circulation, and there is another maximum center of CWC. It should be noted that this center is mainly vapor condensation caused by the orographic lifting effect; hence, it is not the CIC center anomaly. In a similar way, there is a negative anomaly center of hydrometeors over the northern slope of the TP——an arid atmosphere center in north slope of the TP, which is caused by the water vapor condensing and dropping in south slope because of terrain lifting. This is consistent with the spatial structure of the CLWC and CIC over Regions IV and V. Figure 9b presents the PW spatial structure. Apparently, positive and negative anomalies of hydrometeors are centered in the upward and downward motion areas of the Hadley Circulation, respectively. Because of the existence of the TP, the positive anomaly center of hydrometeors in the Northern Hemisphere is found to be mainly located over the southern TP, while the negative anomaly center is over the northern TP. The positive anomaly center of hydrometeorsis consistent with the CWC values in Fig. 8a. The LH has the same spatial structure as the hydrometeors. In the hydrometeor profiles within 10°-20°N, a positive anomaly of LH is centered on 400 hPa, where the boundary between CLWC and CIC is located. The LH over the southern TP reflects the lifting force that condenses water vapor. The negative anomaly center of LH over the northernTP is related to surface evaporation.
By analyzing Figs. 8 and 9, it can be seen that the lifting and condensation effect makes the southern and eastern TP two positive anomaly centers of hydrometeors. On the one hand, the hydrometeor center is located on the ascending branch of the Hadley and Walker Circulation, because there is abundant vapor in the monsoon trough in the Bay of Bengal, which is located south of the TP, and the hydrometeors are the product of water vapor and upward motion; while on the other hand, the vapor condenses, releasing LH to reinforce the upward motion of the Hadley and Walker Circulation, which is more significant over the southern and southeastern TP. Additionally, the regional water cycling induced by evaporation is also an important process in the formation of cloud and precipitation.