Interactions of Multiple Atmospheric Circulation Drive the Drought in Tarim River Basin

Global warming is likely to cause overall drying of land surfaces and aridity increasing leading to expansion of dry climate zones. There is an increased risk of extremely arid environment and large deserts developed progressively in the central Asia. However, the key factors causing the drying in mid-Asia remain inconclusive. Here, we analyzed the relationship among precipitation, water vapor transportation in Tarim River Basin (TRB) and Multiple Atmospheric Circulation (MAC) to explore the mechanism of MAC driving the drying in TRB, through comparing MAC between abundant and scarce precipitation years. We found that Westerly Circulation (WC) and Asian Summer Monsoon (ASM) are likely to promote the precipitation respectively. Whereas, they not only have their own influence but also restrict each other and facilitate the forming of peculiar water vapor transport channel for TRB, which is probably to restrain the precipitation and its distribution pattern and accelerate the drying in this region. Our results enrich the findings on mechanisms of wet places becoming wetter while dry areas getting drier under the global warming.

mountain precipitations were resulted from the forest pump and the accelerating of local water recycling, and the decrease in plain precipitations, the mainly reason of the drying in TRB, is related with MAC.
In this paper, we define MAC as the air movement to and from or around something in a closed/quasi-closed system with different spatial and temporal scales, such as general circulation on planetary scale, u wind (zonal wind), Westerly Circulation, Asian Summer Monsoon and so on. And Westerly Circulation (WC) is defined as the prevailing west-to-east winds in westerly belt located in the planetary wind belt between subtropical high pressure zone and subpolar low pressure belt (roughly between latitudes 30~60°N). Since the main factor behind the drying in TRB is the decreasing of plain precipitations 16 , we examined the relationships between any two of MAC, plain precipitations and water vapor flux, analyzed the impact of MAC on them, compared abundant and scare precipitation years, and revealed physical mechanism of MAC driving drought in TRB.

Results
Drying and seasonality of precipitation and water vapor flux in TRB. The scarce or zero precipitation is one of the key factors leading to drought. From later 1970s to the beginning of 21st century global warming was significant, although precipitation in mountains increased rapidly owing to the accelerating local water recycling 17 , precipitation in plains showed an smoothly decreasing trend, as well as the trend of the average annual precipitation in total TRB. Besides, comparing Fig. 1a,b, the variations in u wind in 600 hPa over Equatorial eastern Pacific was more consistent with precipitation in plains than mountains. It is implied that the decreasing precipitation in plains of TRB is the main factor behind the local drying, the MAC may be the culprit, and water vapor transportation, the joint between precipitation and MAC, should be considered.
Our examinations revealed that the changes of precipitation in TRB was complex and has closed linking with water vapor transportation. Firstly, the spatial distribution of average annual precipitation in TRB presented a pattern of decreasing gradually from northwest to southeast with a seasonality changes, and this was consistent with the direction of the source of water vapor transportation. The seasonality character could be reflected as follows: the most of precipitation occurred in the west-northwest of the TRB in springs, and the water vapor transportation basically came from the west and northwest. In summers, the center of precipitation moved to the north TRB with the entire distribution pattern decreasing from north to south, and the same change took in the water vapor transportation, plenty of water vapor entered into the basin across the Tianshan Mountains from the north and east boundaries. In autumns, the patterns of precipitation and water vapor transportation returned to that in springs; The direction of winter precipitation distribution from strong to weak and direction of water vapor transportation have shifted into west-east. Secondly, the highest precipitation reached 160 mm in Tianshan Mountains in summer, whereas the secondary maxima is only 60 mm in spring. And the water vapor entered into TRB was also mainly from the northern and eastern boundary in summer. Finally, the net input of water vapor from the western boundary is positive throughout the year, while for other boundaries, it is positive in summer and negative in winter. However, the maximum water vapor net input is in summer from east and north. In addition, seasonal changes in the water vapor transportation from western boundary is much less than the other borders. The seasonal characteristics indicated that the channel of TRB water vapor transportation determines the spatial-temporal distribution of precipitation in the region, this implies that both of precipitation and water vapor transportation are closely related to WC and ASM (Fig. 2).

The coordination of atmospheric circulation in different levels controls precipitation in TRB.
To further explore the impact of MAC on TRB precipitation, we characterized the summer atmospheric circulation and its divergence field over TRB in upper (300 hPa for the vector and divergence fields, 200 hPa for u wind), middle (600 hPa) and lower (850 hPa) troposphere, and analyzed their correlation with TRB precipitation by comparing the difference between abundant and rare precipitation years. The results showed that TRB was at the water vapor convergence area in 300 hPa and 850 hPa (Fig. 3), in both abundant and rare precipitation years. However, TRB is in front of the westerly trough at 600 hPa in abundant precipitation year, while in a flat westerly circulation in dry year. Compared with the dry years, the water vapor entering into TRB is mainly from the northern and eastern boundary in abundant precipitation years (Fig. 4). It is indicated that the precipitation of TRB is related to the circulation configuration of different levels.
Correlation analysis between TRB precipitation and zonal wind showed that there were four significantly high correlation regions (P > 0.05), which located at I (65°E, 55°N), II (65°E, 45°N), III (135°E, 0°N) and IV region (110°W, 10°N) at 200 hPa (Fig. 5). In area I, which locates at the north to the TRB, there is a significant negative correlation between u wind and TRB precipitation. It implies that the weakness of WC there is beneficial to TRB precipitation. Area II is at the entrance of WC to the TRB, and there is a notable positive correlation between u wind and TRB precipitation there, which suggests that more water vapor can transport from the Atlantic to the TRB region when the WC is strengthened. Regions III and IV situate respectively at the western equatorial Pacific warm pool and subtropical Eastern Pacific, where El Ni no and the aerial support of Walker Circulation occur. The positive correlation between u wind and precipitation in TRB over the western Pacific suggests that the weakening of WC over tropical eastern Pacific Ocean might lead to the reducing of precipitation in TRB. Consequently, it is concluded that the significantly correlated areas over equatorial Pacific Ocean reflects the impact of the ASM and associated sea surface temperature on precipitation of TRB.
Comparative analysis shows that the correlation between u wind and TRB precipitation was weaker on 600 hPa and 850 hPa than 300 hPa in I and II zones. While in eastern Mongolia and Qingzang-Tibet Plateau, the correlation is strengthened, with a shift from negative (positive) to positive (negative) in northern (southern) parts. It suggests that the strengthening of upper WC in TRB upstream region or the weakening of WC in middle and lower downstream are benefit to the precipitation of this region. Moreover, in IV region, correlation coefficient is negative in upper troposphere but positive in the middle and lower. It is likely that their different situations in the upper and lower branch of walker circulation cause the opposite correlation with TRB precipitation (Fig. 5).
The restriction between WC and ASM drives the drying in TRB. The relationship between the situation of precipitation and water vapor flux implies that the decreasing precipitation and drying in TRB is linked to the interaction between WC and ASM. Correlation analysis in Table 1 shows that there is significant positive correlation (P > 0.001) between the WC index in middle and whole layer and net input amount of water vapor flowing through west boundary into TRB. While the water vapor net input from east has negative correlation with the lower WC index (P > 0.005). In a word, the correlation of WC index with zonal water vapor net input is closer than that with meridional water vapor net input (Table 1). We can conclude that strengthened WC will enrich the water vapor flowing into TRB across west boundary, but weaken that from east boundary.
Due to convective activity, the link between ASM and ENSO is close 18 . ENSO event has a significant impact on China precipitation and associated atmospheric circulation 19,20 . Syed et al. pointed out that ENSO also affect local climate in the central region of Southwest Asia and found positive precipitation anomalies accompanying with warm ENSO 21 . Plenty of studies insist ENSO takes some impact on China and regional precipitation. In this paper, the relationship between ENSO and the precipitation in TRB was comparatively analyzed, using Multivariate ENSO Index (MEI) established by Wolter and Timlin 22 . Statistics analysis showed that MEI has a positive significant correlation with the water vapor net input cross western (W_WVNI), but negative notable correlations with that through eastern (E_WVNI), northern (N_WVNI) boundaries of the TRB, and the meridional water vapor net input (M_WVNI) (Fig. 6 and Table 2). Since there was a significant negative correlation between ASM with the same period ENSO after 1980 23 , ASM is not conducive to water vapor entering the TRB from western border but eastern and northern borders.
In summary, strengthened WC is likely to increase water vapor net input to TRB from west boundary, and the reinforced ASM is benefit to moisture flowing into TRB cross east boundary. However, their redoubling simultaneously will restrict each other. In fact, western water vapor net input is obviously weakened during the period of ASM outbreak and development from March to June, but rebound when ASM starts to retreat in August (Fig. 2). Therefore, the competition between WC and ASM will reduce the precipitation and induce drying in the area.  1974, 1981, 1987, 1996 and 2002, and years with scarce precipitation are 1961, 1970, 1980, 1985 and 1986. We created the vector and divergence fields of general circulation using Grid Analysis and Display System Version 2 (GrADS2.0) and GrADS has been implemented worldwide on a variety of commonly used operating systems and is freely distributed over the Internet (http://badc.nerc.ac.uk/help/software/grads/).

Discussion
Empirical results revealed that the central Asia is undergoing an increased risk of extremely arid environment. However, the underlying causes of this phenomenon are far from being well understood. To reveal a mechanism of it, we take TRB as a typical example to analyze how precipitation, water vapor transportation and MAC have influence on drying. It is found that interactions of WC and ASM decrease the amount of precipitation and thus promote the emergence the drying. Our findings well enrich the mechanisms of wet places becoming wetter while dry areas getting drier under the environment of global warming [24][25][26] .
It should be noted that our results are obtained based on data analysis. However, for a long time prediction on climate change, data analysis is not very effective due to requiring large amounts of data. In this case, it may provide useful information by constructing mathematical models to explain the phenomenon observation or make the forecast of climate change. As a result, we need to investigate climate change based on both mathematical analysis and data processing in the future work.
In this paper, we just focus our attention on the region of TRB. However, it is believed that the findings may be also widely applicable in other regions. Two main reasons may support our conclusion. Firstly, from a geographical point of view, some regions have similar geographical environments with TRB, such as Sahara Desert 27 . Secondly, from the data point of view, the data in some areas exhibit same features with TRB 28,29 . We hope that our efforts will provide a new starting point for the study of drying causes in modeling 30,31 .

Method
The computation of water vapor transportation and water vapor net input (WVNI). A rectangular box was chosen to investigate the budget for the water vapor flux. The vertically integrated water vapor flux Q can be expressed as 17 .
where g is the gravitational acceleration, P s is the surface pressure, P t is pressure at the top of the air column taken as 300 hPa, q is the specific humidity, V is the wind velocity, and p is the atmospheric pressure. The zonal and meridional components of Q denoted by Q x and Q y , are expressed respectively as:  where u and v are zonal and meridional wind respectively. When Q x of western boundary or Q y of southern boundary is positive, or Q x of eastern boundary or Q y of northern boundary is negative, it will be indicated that there is water vapor transported into TRB. Otherwise, there is water vapor transported out of TRB. The zonal  Estimate of the air divergence. The physical meaning of air divergence is the air quality which converges in or diverges out of per unit volume during per unit time. The horizontal air divergence A is: where, u and v are zonal and meridional wind velocity respectively. When A > 0, it is divergence; when A < 0, it is convergence. The unit is 1/s (inverse seconds). Besides, the R/S 36 , Mann Kendall (MK) 37 mutation analysis, and linear trend analysis are used to analyze the mutation time and changing trends of net input of water vapor, precipitation and other relevant factors.