1 Introduction

Global warming has profoundly changed many regional climates over the twentieth century (IPCC 2013). Surface air temperature averaged over China has increased by 0.25 °C/10 yr. during the past 50 years, at a rate almost twice the global mean rate of 0.13 °C/10 yr. (Ren et al. 2005; Li and Yan 2009; IPCC 2013; Ji et al. 2014). Surface warming and other climate change are known to affect regional water cycle, hydroclimate and water resources (Chen et al. 2007; Zhao et al. 2015; Ljungqvist et al. 2016). Numerous studies have suggested that arid northwestern China are among the most sensitive areas to climate change based on both historical records and future climate projections (Zhang et al. 2004; Li et al. 2012; Yang et al. 2012; Chen et al. 2012, 2014a). Xinjiang covers over 80 % of the northwestern arid area of China; it is featured by complex topography and continental arid climate, which leads to a fragile ecological environment with scarce water supply (Li et al. 2013). Water is the critical factor in Xinjiang in keeping the ecosystems function properly, but the regional water cycle in Xinjiang has been affected greatly by climate change (Zhang and Wang 2007; Chen et al. 2015). Thus, investigations of the regional water budget changes over Xinjiang are extremely important not only for understanding its regional water cycle characteristics, but also helping water resources management and sustainability in this region.

Xinjiang refers to a vast area over 1.6 million km2 generally confined within 35°-50°N and 70°-100°E. The terrain of Xinjiang is dominated by three mountain ranges, namely the Altay, Tianshan and Kunlun Mountains from north to south (Fig. 1). The Junggar Basin, Ili Valley and Tarim Basin are distributed between these massive mountains. Xinjiang also contains the Tarim River and Taklimakan, the largest inland river and desert in China, respectively. Located in the eastern portion is the Turpan Basin, which represents the lowest elevation of the land in China. In addition to the mountains, grasslands comprise the major seasonal land cover in northern Xinjiang (North XJ), but deserts are the dominant land cover in southern Xinjiang (South XJ).

Fig. 1
figure 1

The schematic topography map of the Xinjiang region

Full size image

The hydroclimatic characteristics in Xinjiang are very complicated with strong spatial variations due to heterogeneous landscapes and water exchanges among different regions. Annual temperature and precipitation in North XJ typically range from approximately −4 to 9 °C and 150 to 200 mm/yr., respectively. In South XJ, the annual ranges are from 7 to 14 °C and 25 to 100 mm/yr. Many studies have examined the impacts of the recent climate change on the water cycle in Xinjiang. Observational studies show that both precipitation and temperature in Xinjiang have increased with large regional variations (He et al. 2003; Ma and Fu 2006; Liu et al. 2009; Li et al. 2012; Jiang et al. 2013; Zheng et al. 2013; Chen et al. 2014a), especially in the different slopes of the Tianshan Mountain (Yuan et al. 2003, 2005; Han et al. 2004). An abrupt change in precipitation over Xinjiang occurred around 1987 (Hu et al. 2002; Shi et al. 2003; Chen et al. 2014b), and the water cycle has accelerated significantly since then (Wang et al. 2003; Chen et al. 2009; Wang et al. 2013; Chen et al. 2014a). These studies suggest that Xinjiang has become warmer and wetter during the recent decades.

In Xinjiang, water budget components are very sensitive to climate change. Li et al. (2013) showed that pan evapotranspiration in Xinjiang decreased significantly from 1960 to 1991, but increased significantly from 1992 to 2007 mainly due to wind speed changes. Mountain glaciers are the main source of runoff in mountainous areas of Xinjiang, but their areas decreased by about 12 % during 1963–2000 due to warmer temperatures (Lan et al. 2007), which lead to more melting of snow and ice (Shen et al. 2013). Chen et al. (2009) analyzed the streamflow of the Tarim River in South XJ, and found increasing trends over its three main sub-basins, although only the streamflow at Aksu River showed a significant increasing trend. These findings indicate that recent climate change have had a large impact on regional water cycle in Xinjiang.

However, the majority of previous studies were mainly based upon field observations and reanalysis datasets. It is still unclear about the interactions of the water exchange among the atmosphere, land surface and soil layers in Xinjiang. Moreover, little effort has been devoted to investigate regional long-term hydroclimatic trends over Xinjiang, and few studies have comprehensively analyzed the regional variations of the these trends considering the dynamical links of the water cycle components.

To address these issues, especially the regional differences in hydroclimatic changes between North and South XJ, we applied the Community Land Model version 3.5 (CLM 3.5) in this study with enhanced hydrological processes to simulate the historical water budgets in Xinjiang from 1951 to 2000 based on observation-constrained atmospheric forcing (Liu et al. 2012). The advanced Geographic Information Systems (GIS) platform was used to derive the water divide of the Tianshan Mountain, which separates North and South XJ. Also, a non-parametric Mann-Kendall test (MKT) method was used to analyze the trends of the water budget components to investigate their response recent climate changes in different parts of Xinjiang. The main objectives of this study are to (i) simulate and analyze the water budget components (e.g. precipitation, evapotranspiration, runoff, soil moisture, etc.) in Xinjiang, (ii) study the hydroclimatic trends of Xinjiang from 1951 to 2000 in response to recent climate change, and (iii) compare and investigate the regional differences of the hydroclimatic trends between North and South XJ.

2 Model and experimental design

2.1 CLM model description and enhancement

CLM was developed by the National Center for Atmospheric Research (NCAR) to represent land surface processes in the Community Earth System Model (CESM). The CLM version 3.5 includes a land surface model (LSM) and improved vegetation canopy, snow accumulation, carbon cycle and ET parameterizations (Niu et al. 2005, 2007), which lead to better model performance, particularly at the global scale (Oleson et al. 2008; Wang and Zeng 2011). In addition, enhanced representation of hydrological processes has significantly improved the water cycle simulations (Oleson et al. 2008). To reduce the systematic wet biases in simulated soil moisture, Niu et al. (2011) improved the parameterization with a correction scheme (Eq. (1)) to enhance the water exchanges between soil and underground water. In this study, the soil porosity parameter (f mic ) is also introduced into soil and underground water exchange scheme in CLM 3.5 in order to reinforce the recharging water from soil to aquifer and to weaken the rising water from aquifer to soil using this equation:

$$ \mathrm{Q}=-{K}_{bot}\frac{-{Z}_{\nabla }-\left({f}_{mic}{\varphi}_{bot}-{Z}_{bot}\right)}{Z_{\nabla }-{Z}_{bot}} $$
(1)

where, Q is the water exchange (mm) between the soil and aquifer, K bot is the hydraulic conductivity (mm/s), Z ∇and Z bot are the depths (mm) of the water table and the bottom of the soil layer, respectively, ψ bot is the matric potential (mm). The parameter f mic is closely related to soil texture, ranging from 0 to 1. Specifically, the gravity-caused penetrating water increases as f mic closes to 0, conversely the capillary rise water increases as f mic closes to 1. The soil textures in this study are assumed to be unchanged over time for simplicity (with spatial distribution), and f mic is set to 0.5.

Currently, higher versions of CLM (4.0/4.5) have already offered optimized parameterizations of soil moisture and biogeochemical schemes. However, according to Lawrence et al. (2011), the soil moisture variability still shows negative bias compared with observations. Based upon the previous assessment of the applicability of CLM 3.5 over China (Li and Ma 2015; Li et al. 2016), we opted for CLM 3.5 to simulate the long-term hydroclimatic trends over Xinjiang with enhanced hydrological parameterization in this study.

2.2 Atmospheric forcing data

Atmospheric forcing is another key factor for accurate land surface modeling (Dickinson et al. 2006; Qian et al. 2006). But there are only sparse weather stations in Xinjiang, especially in its deserts and Gobi (Fig. 1). Nevertheless, several widely used reanalysis or land assimilation products (e.g., NCEP, ERA40 and GLDAS) have large biases for Northwest China (Personal communication with Dr. Fei Chen 2012; with Dr. Yang Hong, 2013). Thus, field observations are still needed for improving the quality of the atmospheric forcing data over our study area. Here, 6-hourly data of precipitation, surface air temperature, wind speed, wind direction, air pressure and relative humidity records from 1951 to 2000 were derived from 90 meteorological stations in Xinjiang (Fig. 1) operated by China Meteorological Administration (CMA). We used these data to generate the atmospheric forcing by Kriging spatial interpolation algorithm. The temporal interpolation of the atmospheric forcing was generated by curve fitting method. Due to the lack of observations, the radiation data were obtained from the Princeton meteorological forcing data (hydrology.princeton.edu/data/, Sheffield et al. 2006). Finally, the new observation-constrained atmospheric forcing data were constructed at 0.5° grid spacing, 3-h interval for a 50-year time span from 1951 to 2000. Several studies have already shown that this observation-constrained atmospheric forcing data can significantly improve the land surface simulations over China (Liu et al. 2012; Li and Ma 2015).

2.3 Experimental design

The optimized CLM 3.5 was driven offline with the observation-constrained atmospheric forcing data from 1951 to 2000. The model was first spin-up for 400 years with repeated 50-years forcing to allow soil water parameters to reach a sufficient equilibrium. The final 50 years’ simulation was used to analyze the surface water budgets in North and South XJ. The general water balance equation can be written as:

$$ \frac{dW}{dt}=P-E-R $$
(2)

where (dW/dt) is the terrestrial water storage (TWS) variations, which equals the precipitation (P) minus the actual evapotranspiration (E) and runoff (R).

The rank-based MKT (Mann 1945; Kendall 1975) was used to detect hydroclimatic trends based on a linear regression analysis. The MKT method is a nonparametric and commonly used method that can validate the significant level of monotonic trends in hydro-meteorological time series (Yue and Wang 2002; Geng et al. 2016).

Regional difference analyses in North and South XJ require accurate basin division procedures. Here, the ArcGIS-Desktop 10.1 platform (www.esri.com) and Arc-Hydro extension tools (resources.arcgis.com/en/communities/hydro /) were utilized to extract the Tianshan Mountain water divide, which separates North and South XJ. Arc-Hydro uses a digital elevation model (DEM) with 90 m horizontal spacing derived from the Shuttle Radar Topography Mission (SRTM, lta.cr.usgs.gov/SRTM) (Jarvis et al. 2008). All of the Xinjiang catchments were then delineated and the Tianshan Mountain water divide polylines were used as the boundaries of North and South XJ (Fig. 1).

3 Results and discussion

3.1 Surface water budgets

Figure 2 shows the spatially-averaged annual precipitation and temperature (interpolated from station observations), and CLM-simulated evapotranspiration, runoff, TWS, P-E time series for whole Xinjiang, North XJ and South XJ from 1951 to 2000. The associated linear trends are all summarized in Table 1. Annual precipitation in Xinjiang shows an obvious increase during the second half of the twentieth century. It increased by an average rate of 4.9 mm/10 yr. This increasing rate reaches 4.6 mm/10 yr. in South XJ and 5.7 mm/10 yr. in North XJ. As surface temperature and precipitation increase, the simulated evapotranspiration rate over Xinjiang also increases by 3.9, 2.7 and 4.3 mm/10 yr. for whole Xinjiang, North XJ and South XJ, respectively (Fig. 2b).

Fig. 2
figure 2

Annual a precipitation from observation-constrained forcing; b evapotranspiration; c runoff; d TWS variation; e precipitation minus evapotranspiration (P-E) and f temperature time series Xinjiang (including North and South XJ) simulated by CLM 3.5

Full size image
Table 1 Annual trends from 1951 to 2000 of the water budget components and temperature averaged over Xinjiang
Full size table

Due to the precipitation increases, the simulated runoff in North XJ also increased by 3.0 mm/10 yr., while it remained relatively constant in South XJ. Therefore, runoff changes are negligible for surface water budget changes in South XJ. The whole Xinjiang runoff increased by 0.9/10 yr. during 1951–2000. This runoff increase is due to increased precipitation and warmer temperatures (Fig. 2f), which can cause more melting of snow and glacier ice (Chen et al. 2014a). These factors cause the North XJ runoff rate increased rapidly after the 1987 climate shift when both precipitation and temperature increased substantially over Xinjiang (Fig. 2a-b). In addition, TWS and P-E also exhibited larger inter-annual variations after 1987 than before (Fig. 2d-e).

These results indicate that the water input (P) increased faster in North XJ than in South XJ, but for the water consumption (E), the increasing rate was larger in South XJ than in North XJ due to warmer temperature As a result, the increment of P-E flux in North XJ was as high as 4.3 mm/10 yr., more than 7 times larger than in South XJ (Fig. 2e). Also, the gap of P-E between North XJ and South XJ increased from 1951 to 2000. The results further suggest that North XJ water budget components are more sensitive to climate change than South XJ, where the components are characterized by gradual trends due to the vast desert environment.

3.2 Spatial distributions of trends

Figures 3 show the linear trend maps for the water budget components and temperature over Xinjiang. We can see precipitation increased over nearly entire Xinjiang from 1951 to 2000, with the largest increase (over 15 mm/10 yr) in the Junggar Basin, and decreased in the eastern portion of the Ili Valley and the nearby Karamay city (Fig. 3a). Evapotranspiration also increased over most of Xinjiang and the change patterns generally mimic the precipitation trend distribution, but with smaller magnitudes (Fig. 3b). In the Ili Valley, precipitation increased while evapotranspiration decreased, leading to large runoff increase in this region (Fig. 3a-c). Runoff decreased in portions of South XJ due to the much smaller net water flux (P-E) and rapid infiltration in the desert. In North XJ, the maximum trend value of P-E flux can be up to 14 mm/10 yr., much higher than in South XJ (Fig. 3e). Spatial variations in TWS trends are small over most Xinjiang, with wet conditions causing TWS to increase slightly in portions of the Taklimakan Desert. As seen in Fig. 3, the trends for all of the water budget components are larger in North XJ than in South XJ, with the highest trends distributed in North XJ. Thus, the acceleration of the water cycle in North XJ is much stronger than in South XJ.

Fig. 3
figure 3

Spatial distributions of the annual linear trends from 1951 to 2000 for precipitation (a); CLM 3.5 simulated evapotranspiration (b), runoff (c), TWS variation (d), precipitation minus evapotranspiration (P-E) (e) and temperature (f) using Mann-Kendall algorithm. The symbols represent statistically significant increases (“+”) or decreases (“△”) (0.01 significance level) respectively, and the red dash line represents the boundary of North and South XJ

Full size image

3.3 Soil moisture and water table depth

The total soil depth in CLM 3.5 is 3.433 m, and the soil column is divided into 10 layers with different thicknesses. In this study, the top 7 layers (0.62 m) are considered as “shallow” soil layers, while the rest 3 lower layers (2.813 m) are deemed “deep” soil layers. Figure 4a-b shows the trend distributions of simulated soil water content in the shallow and deep layers. These results show that soil moisture in the shallow layer increased over entire Xinjiang, especially in the Junggar Basin, eastern North XJ and the southern slope of the Tianshan Mountain, with the largest increases at the mountainous areas along Tianshan and Altay Mountains (Fig. 4a). Soil moisture trends are more spatially heterogeneous in the deep layer (Fig. 4b). In North XJ, the deep-layer soil moisture shows strong positive trends, whereas the negative trends are seen in the central Taklimakan Desert and the Turpan Basin in South XJ. The soil moisture trend patterns are closely related to the precipitation and temperature distributions and landscape heterogeneities over the entire region. Overall, North XJ experienced larger precipitation increases, while larger loss of water occurred in South XJ due to larger evapotranspiration trends (Fig. 2b). In general, the large desert area in South XJ exhibited limited ability to store and retains water in deep soils.

Fig. 4
figure 4

CLM 3.5-simulated soil moisture content for the spatial distributions of the annual linear trends in a shallow layers (0–0.62 m) and b deep layers (0.62–3.433 m), the symbols and dash red line are same as Fig. 3; and the temporal anomaly soil moisture series in c shallow and d deep layers

Full size image

Figures 4c-d shows the temporal variations of soil moisture in the shallow and deep layers. The results suggest that the soil moisture in Xinjiang generally increased in both layers from 1951 to 2000. Positive anomalies are particularly common after the climate shift around 1987, which coincides with the precipitation and temperature increases (Fig. 2a-b). The soil moisture response to warm-wet climate trend during 1951–2000 is larger in North XJ than in South. North XJ also displays larger inter-annual variability in soil moisture. Although precipitation increased over entire Xinjiang, the excess water did not infiltrate into the deep soil layers over South XJ. Therefore, the soil moisture in the deep layer of South XJ increased minimally compared with the soil moisture increase in the deep layer in North XJ.

Groundwater is an important factor for land-atmosphere interactions (Niu et al. 2007), and it can even provide water for evapotranspiration in arid regions (Gutowski et al. 2002). In this study, regional groundwater variations were simulated using the water table depth (WTD) as a long term indicator and plotted in Fig. 5. In Fig. 5a we can see the WTD trends are highly heterogeneous over Xinjiang, with significant reduction (i.e., WTD becomes shallower) over North XJ, except in the western Junggar Basin. The Altay and Ili Valleys exhibited the highest rate of increase in the WTD at approximately 0.2 m/10 yr. The WTD trend in South XJ show positive trends (the WTD becomes deeper) and is opposite to the trend in North XJ, except on small portions of the southern slope of the Tianshan Mountain. A clear WTD boundary can be seen in Fig. 5a as the trend shifts from positive to negative. This boundary is coincident with the divide between North and South XJ.

Fig. 5
figure 5

CLM simulated WTD in Xinjiang a spatial trend distributions, the symbols and dash red line are same as Fig. 3; and b temporal variations, the larger negative/positive anomalies indicate the shallower/deeper WTD

Full size image

Figure 5b shows that the WTD changes for North and South XJ are very different. The WTD in North XJ exhibited apparent decadal variations in response to recent climate change, becoming increasingly shallower after 1987. However, South XJ WTD slowly deepened despite the precipitation increases (Fig. 2a). One possible reason is that in arid regions, especially in the deserts, the groundwater is little recharged by precipitation because of the deep WTD and strong evapotranspiration, this would lead to a decoupling between the precipitation and WTD trends in long-term scale in South XJ. The simulated WTD changes over Xinjiang are consistent with previous studies (e.g., Xia et al. 1991), which indicate that the water content stored in soil and underground have different sources in North and South XJ. Xia et al. (1991) showed that the soil and subsurface water in Gurbantünggüt Desert (the main part of the Junggar Basin in North XJ) is derived mainly from precipitation, while in Takilimakan Desert are mainly recharged by ground water (i.e., little infiltration from precipitation). This is the reason why the WTD in North XJ responds to the increased precipitation more closely than in South XJ. Although the climate in Xinjiang showed a warm-wet trend in the last few decades, the groundwater outlook in South XJ remains vulnerable.

4 Summary

Xinjiang is the largest area with arid climate in inland China. Recent climate change, especially the shift change around 1987 has affected Xinjiang’s regional water budget in various aspects. This study used the optimized CLM 3.5 forced with observation-constrained atmospheric forcing to investigate water budget changes and compare the hydroclimatic trends between North and South XJ. The divide of North and South XJ was represented by the water divide formed by the Tianshan Mountain. Trends in water budget components were then analyzed using the MKT method.

Xinjiang experienced warming and wetting from 1951 to 2000. Annual precipitation and evapotranspiration increased across whole Xinjiang. The trends in North XJ were more spatially heterogeneous than in South XJ. Nearly all the largest water budget trends occurred in North XJ, for both increases and decreases. Regionally averaged precipitation increased faster in North XJ than in South XJ (where it has a lower annual precipitation and drier climate than North XJ), whereas evapotranspiration increased faster in South XJ than in North XJ. This worsened net water flux (P-E) gap between North and South XJ during the second half of the twentieth century.

Significant increases in precipitation and snowmelt caused runoff to increase much faster in North XJ than in South XJ, where runoff remained relatively constant. The total water storage exhibited no significant trends across Xinjiang. Due to increased precipitation variations, the inter-annual terrestrial water storage (TWS) variations in North XJ became larger after 1987 than before.

Soil moisture and water table depth (WTD) trends were non-uniform across Xinjiang. The shallow layer (0–0.62 m) soil moisture increased across Xinjiang, especially in mountainous areas. Changes in the deep layer (0.62–3.433 m) soil moisture were heterogeneous, with significant increases in North XJ but with small changes in South XJ. Both the shallow and deep layer soil moisture trends were affected by the 1987 climate shift in North XJ, while the deep layer in South XJ displayed no response to the climate change. Moreover, the WTD trend showed opposite changes in North and South XJ. The WTD in North XJ became shallower after the climate shift in 1987, but the WTD in South XJ became deeper and was not affected by the increases in temperature and precipitation in the recent decades.

These results of this study indicate the water shortage in Xinjiang, especially South XJ, have not improved fundamentally despite the increases in temperature and precipitation during 1951-2000. The TWS variations showed no significant increases and ground water deletion continued over large areas of South XJ, as suggested by the deepening WTD there.

Besides the climate-induced changes investigated here, the water budgets in Xinjiang are also affected directly by human activities in recent years (Chen and Xu 2004; Li et al. 2014). For example, afforestation and irrigation have considerable impacts on evaporation and soil moisture in arid regions like Xinjiang. Moreover, the modeling study presented here has several limitations. One problem is the simulated WTD cannot go deeper in this very arid area due to the relative coarse spatial resolution. And uncertainties exist within the atmospheric forcing data, which were constrained with sparse observations. Moreover, quantitative investigations to distinguish the impact of climate change and other human activities on water budgets and the hydroclimate over Xinjiang are still needed in the future.