Water resources are an essential part of the ecosystem in the extremely arid northwestern part of China. Previous studies revealed a dry-to-wet climate change since the late 1980s in this region, which suggested a relief from the drought condition. However, the analysis in this study using the updated data shows that the arid situation has continued and even intensified in the past decade. This is reflected by the fact that the low-level air relative humidity and deep soil relative humidity have decreased in the past decade. Examination of the standardized precipitation evapotranspiration index (SPEI) and self-calibrating Palmer drought severity index (sc-PDSI) indicates that the severity and spatial extent of aridity and drought have increased substantially in northwestern China in the most recent decade. It is shown that the drought intensification in northwestern China is mainly caused by the increase of evaporation that results from the continuous rise in temperature, which will pose a continuous threat to the ecosystem and economic development in this region, especially under the background of global warming.
Because of its location in the center of the Eurasian continent, far from any ocean, northwestern China is characterized by an extremely arid climate. The annual rainfall is less than 150 mm in most stations. Some stations even witness annual rainfall less than 50 mm. Meanwhile, this region is blockaded by the Tibetan Plateau in the south, the Pamir Plateau in the west, and the Altay and Tianshan Mountains in the north. The dry wind coming off the lee slopes of the mountain ranges and plateaus makes this region even more arid, including the Taklamakan Desert in the Tarim Basin, Gurbantunggut Desert in the Junggar Basin, and Qaidam Desert in the Qaidam Basin.
However, the climate in extremely arid northwestern China is suggested to be changing for the better, as first proposed by Shi et al. (2002, 2003). They analyzed several key indicators and pointed out a strong signal of climate change from a warm/dry to a warm/wet climate since 1987. Their indicators include a 22% and 33% increase of precipitation in northern and southern Xinjiang Province, respectively (for 1987–2000 compared to 1961–86), an 84.2% increase of the annual meltwater of Glacier No.1 (42°30′N, 86°26′E) at the source area of the Ürümqi River (1985–2001 compared to 1958–85), a 7% increase of the annual total runoff of the Xinjiang area (1987–2000 compared to 1956–86), water level rise, and area expansion of the inland lakes (e.g., Bosten Lake and Ebinur Lake). Other indicators include the increasing frequency of flood disasters, increase of vegetation cover, and reduction of sandstorm and duststorm days. Meanwhile, by comparison analysis of the paleoclimatology and a simulation under the double carbon dioxide scenario, despite large uncertainties, they proposed a prospect of warm and wet climate in the twenty-first century for this region (Shi et al. 2007). Their comprehensive studies attracted the attention of both the scientific community and the public as water resources are of essential importance to the ecosystem and economics in this extremely arid region.
Their analysis was confirmed by several studies (Bi et al. 2007; Chen et al. 2009; Jiang et al. 2009; Shi and Xu 2007; Shi et al. 2007); for example, Li et al. (2010) found a significant glacial runoff increase of the Ürümqi River after 1987. Chen et al. (2009) showed a significant monotonic increasing trend and 10.5% increase of the runoff of the Aksu River since 1990. Zuo et al. (2005) showed an increasing trend of the relative humidity in western China, with extrema of 2% decade−1 during the past 50 years. Jiang et al. (2009) investigated the variation of drought and wet conditions in Xinjiang province with a self-calibrating Palmer drought severity index (sc-PDSI) during 1961–2003 and revealed that both annual and seasonal mean climates are prone to become wetter, mainly as a result of precipitation increase. Zhai et al. (2010) revealed a downward dry trend and an upward wet trend for the northwestern China by analyzing the time series of the average annual Palmer drought severity index (PDSI) and the standardized precipitation index (SPI) for 483 meteorological stations in China using monthly data from 1961 to 2005.
The above studies are mainly based on the data before 2000; however, there are large climate changes in the recent decade. For example, 9 of the 10 warmest years since the instrumental observations are mostly in the past decade, making the past 10 years significant. On the other hand, there is still some argument about the likely water resource change under the background of global warming. Dai et al. (2004) estimated that warming would enhance the water-holding capacity and increase the potential surface evapotranspiration, enhancing drying near the surface and increasing the risk of drought. Labat et al. (2004), however, estimated that the continental precipitation will increase under the global warming. Dry/wet change in an area is affected not only by precipitation but also by temperature, because of its effect on potential evaporation. It has been pointed out that the strongest warming due to the global warming during the last 100 years has occurred in the interior region of Asia (Bueh et al. 2003; Hu et al. 2000; Hu et al. 2003; Hulme et al. 2006; Solomon et al. 2007). This may change the previous estimation and a new evaluation is needed, especially under the background of strong global warming.
The purpose of this work is to examine the dry/wet change in extremely arid northwestern China. Section 2 describes the data used in this study and presents the approach to quantify the extent of drought. Section 3 discusses the dry/wet change in the most recent decade. Section 4 presents the change of drought indices. We found that northwestern China experienced intensified drought in the past decades. A discussion of our results and conclusions is provided in section 5.
2. Data and methodology
The primary datasets employed in this study are the daily relative humidity, precipitation, mean temperature, maximum temperature, and minimum temperature of 730 stations over China derived from the China Meteorological Administration. Although the earliest observation of these datasets at some stations is from 1 January 1951, the number of stations throughout China is low until the mid-1950s. We adopted the beginning date as 1 January 1960 to achieve enough station coverage and also have a time period long enough to conduct trend analysis. The missing values were screened and stations with too many missing values were dropped. Finally, we used 47 stations in northwestern China region (see the black dots in Fig. 1). These stations are distributed mainly in the oasis and surrounded by arid desert regions, ranging from a height of 34.6 m (Turpan, 42°56′N, 89°12′E) to 3504.4 m (Sai-K'o-Lo-T'e-Ma, 40°31′N, 75°24′E), averaging at 1149.3 m. We also used the soil relative humidity data from the surface agriculture dataset, which were compiled by the China Meteorological Administration. The earliest soil relative humidity data started from 1991 in this region; however, missing data are common in the early period. Therefore, we only adopted 13 stations with sufficient observation (see the open squares in Fig. 1). In addition, we used the gridded monthly mean reanalysis datasets (Kalnay et al. 1996; Kistler et al. 2001) from the National Centers for Environmental Prediction (NCEP) and National Center for Atmospheric Research (NCAR) for the same time period as the station dataset. To calculate the drought indices, the latest release of Climatic Research Unit (CRU) time series (TS) datasets (http://www.cru.uea.ac.uk/cru/data/), namely CRU TS3.2, is also utilized (Harris et al. 2013; Mitchell and Jones 2005).
The degree of drought is represented using the standardized precipitation evapotranspiration index (SPEI). SPEI is a new drought index devised by Vicente-Serrano et al. (2010), which is based on both precipitation and temperature data and take advantages of the previously widely used climatic drought indices PDSI by Palmer (1965) and SPI by McKee et al. (1993). It combines the sensitivity of PDSI to changes in evaporation demand and the multitemporal nature of SPI. The detailed procedure of computing SPEI can be found in Vicente-Serrano et al. (2010). For comparison, the calculation of SPI is also performed.
To validate previous efforts in drought analysis by using SPEI, another famous drought index, namely PDSI, and a station-based dataset are used here to analyze decadal variation of dryness or wetness in northwestern China. PDSI is a landmark in the development of drought indices. The PDSI is created by Palmer (1965), based on supply and demand in a two-stage “bucket” model of soil. The prominent advantage of PDSI is its sensitivity to changes in evapotranspiration demand caused by temperature fluctuations and trends. However, it is widely recognized that the PDSI has numerous deficiencies (Alley 1984). A great improvement to PDSI is the self-calibrating PDSI devised by Wells et al. (2004) and this solves most of the problems of calibration of PDSI. We used the sc-PDSI instead of the original PDSI in this study. As for the necessary input quantity of water holding capacity in calculation of PDSI, the water holding capacity in each site is obtained from a soil texture–based dataset from Webb et al. (1993).
The approach to obtain potential evapotranspiration (PET) follows Thornthwaite (1948) and the generalized logistic distribution is utilized to fit the aggregated series. There are concerns that the Thornthwaite method may overestimate the PET (Dai 2011; Hobbins et al. 2008; van der Schrier et al. 2011), especially in the tropics. The overestimation is associated with the exclusion of cloud cover and vapor pressure deficit in the parameterization of Thornthwaite. The Penman–Monteith method (Burke et al. 2006; Shuttleworth 1992) recommended by the Food and Agriculture Organization (FAO) is thought to be more realistic in estimating PET. However, it requires additional input fields such as solar radiation, wind speed, cloudiness, and humidity, which makes the calculation and evaluation more complicated. Several works have proved that the choice of the method of PET has limited effects on both PDSI and sc-PDSI. For example, van der Schrier et al. (2011) showed that PDSI values based on Thornthwaite and Penman–Monteith parameterizations are very similar in terms of correlation, regional average, and trends. Dai (2011) compared four forms of the PDSI over the global land areas and evaluated them with Gravity Recovery and Climate Experiment (GRACE) satellite-observed soil moisture, streamflow, and water storage. He showed that the choice of parameterization method of PET has small effects on actual evapotranspiration and both PDSI and sc-PDSI. Dai (2011) also pointed out that large uncertainties may exist in data such as surface wind speed, radiation, and humidity. We compared the PET estimated by these two approaches in northwestern China using CRU TS datasets; the results show that the time evolution of PET obtained by the Thornthwaite method (PET_th) is highly correlated with that by Penman–Monteith (PET_pm), that the differences of PET_th and PET_pm are small in northwestern China, and that the regional averaged PET_th and PET_pm have a similar long-term trend in the past two decades (not shown). Therefore, considering the simplicity and also the possible large uncertainties in historical data needed for computing the Penman–Monteith PET, the Thornthwaite method is adopted in this study.
3. The dry/wet change in the recent decades
The studies of the dry/wet change have mostly concentrated on the water resources related to the meteorological elements such as relative humidity, precipitation, temperature, soil moisture, etc. The consistency and confirmation of the multiple elements help to characterize the climate change fully in the above-mentioned region. The time series and their 9-yr running mean of the observed meteorological elements in the past 50 years are plotted in Figs. 2a–f corresponding to precipitation, screen air temperature, PET_th, relative humidity, soil relative humidity (20 cm), and soil relative humidity (100 cm). All the times series are obtained by annual average. The relative humidity, precipitation, and temperature are based on the average of the daily observations of the 47 stations in northwestern China (the dots in Fig. 1). The series of soil relative humidity are obtained by averaging the 13 stations (the open square in Fig. 1). The annual rainfall is about 121 mm, far below the average rainfall in eastern China. The rainfall shows a long-term increase during the past 50 years with an obvious shift around 1987. The average precipitation is about 110 mm from 1960 to 1986, while it increased to about 134 mm from 1987 to 2009, an increase of about 20% between these two periods, which directly led to the dry-to-wet change studies in the recent years as noted in the introduction. An interesting feature of this change is that the change shows an abrupt shift, with the largest rainfall occurring in 1987 in this region. Since then, the annual total rainfall remains at the level of ~130 mm. However, the rainfall of this magnitude is still far from making this arid region wet.
The variation of dryness/wetness is determined by water balance in a region; therefore, studies on the long-term dry/wet change must evaluate the precipitation and evaporation simultaneously. The evaporation, however, is largely influenced by temperature change. Figure 2b illustrates the average temperature of the 47 observation stations, which is characterized by continuous increase since the early 1970s. The average temperature is about 7.33°C during 1960–76, while the average temperature rocketed to about 8.69°C during 1997–2009. The increasing rate is ~0.33°C decade−1 since 1960, ~0.42°C decade−1 since 1970, ~0.46°C decade−1 since 1980, ~0.52°C decade−1 since 1990, and ~0.60°C decade−1 after 2000; all are greater than the global average increasing rate (about 0.177° ± 0.052°C decade−1 from 1981 to 2005) (Solomon et al. 2007). In the extreme years, for example, the temperature is 8.93°C in 2006, 9.24°C in 2007, 8.82°C in 2008, and 8.78°C in 2009, while it is only 6.74°C in 1960, 6.64°C in 1967, 6.74°C in 1969, 6.82°C in 1976, and 6.54°C in 1984. The strong warming in this arid region favors the increase in surface evaporation, as shown in Fig. 2c. The PET increased in the past two decades, consistent with the strong warming in the recent decades. We noted that the PET decreased during the period from the early 1960s to the mid-1970s, which may be the results of the temperature drop during this period.
When we combine the effect of both precipitation and evaporation change, a comprehensive evaluation of the wet/dry change can be made. Figure 2d shows the variation of relative humidity in northwestern China. For relative humidity, the annual mean value is about 50.5%, indicating an arid climate compared to the usual value of ~80% in southeast China and ~75% in east China, and still less than the ~65% in northeast China and ~60% over the Tibetan Plateau. The relative humidity remains low from 1960s to the mid-1980s, averaging at about 49.8% from 1960 to 1986. A strong shift can be observed around 1987 with an average increase in relative humidity to about 51.9% during the period of 1987–2003, amounting to a 4.2% increase compared to the former period. This increase corresponds to the previous studies on the dry-to-wet change. However, a strong decreasing trend is also evident in the most recent decade. Strong low relative humidity values are continuously observed after 2003. The relative humidity is 47.8%, 48.2%, and 47.5% in 2007, 2008, and 2009, respectively. Therefore, arid northwestern China is turning from a relatively wet period (about 20 years from the late 1980s to the early 2000s) back to a dry condition.
This turning back to a dry condition can also be confirmed with soil relative humidity data. Figures 2e and 2f are the regional averaged soil relative humidity at the 20-cm and 100-cm layers, representing the soil moisture at the shallow level and the deep-rooted level. Soil relative humidity is the ratio between soil gravimetric water content rate and field moisture capacity; therefore, it can be used as a measure of the dry/wet extent of soil. A slightly decreasing trend can be observed at the 20-cm level, and a decreasing trend is obvious at the deep-rooted level, indicating a long-term drying trend. As the short-term and interannual signal is smoothed at the deep level, the 100-cm soil relative humidity is a better proxy than the 20-cm one. The average value of the 100-cm soil relative humidity during 1994–2001 is around 76%, while it decreases after 2000 and drops to around 38% during 2008–11, which amounts to only 50% of the value in the middle and late 1990s.
We also calculated the evolution of the above six indices for the summer season, as this region is usually tortured with high temperature and severe drought in summer. As shown in Fig. 3, similar results can be obtained. The rainfall (Fig. 3a) has increased in the late 1980s, while a decreasing trend can be observed in the most recent decade. The average temperature (Fig. 3b) presents a continuous long-term increase, especially in the past 20 years, which will lead to increased PET (Fig. 3c). Therefore, the water balance is biased to be deficient in the recent years, with a decreasing tendency of relative humidity near the surface (Fig. 3d) and decrease of soil moisture content at the deep level (Fig. 3f).
4. The drought indices changes
Figure 4 shows the evolution of zonally (70°–95°E within China, covering northwestern China) averaged sc-PDSI, SPEI, and SPI at the time scale of 12 months; the evolution of the difference between SPEI and SPI is also shown. From the perspective of evolution of sc-PDSI and SPEI, it is evident that there is main drought episode from late 1990s to 2009, especially between 34° and 50°N. In contrast, wetness and dryness appeared alternatively before the late 1990s in northwestern China, indicating there are no prolonged and extensive drought events. As mentioned above, SPEI takes not only precipitation deficits but also temperature anomalies into account, whereas the SPI is solely based on precipitation data. From the evolution of SPEI and SPI from late 1990s to 2009, the two indices manifest different patterns. According to a latitude–time cross section of SPI, it can be concluded that the climate was becoming humid during the late 1990s to 2009 from 30° to 45°N. However, the temperature anomaly plays a large role in the drought conditions. There has been a general temperature increase during the last 50 years, especially in the most recent 10 years, and consequently recent warming has increased atmospheric moisture demand. Therefore, the rapid increase in temperature may offset the increase in precipitation and enhance and prolong drought. The procedures of SPEI and SPI are identical, so the difference between SPEI and SPI illustrates the contribution of temperature anomalies to drought events. From Fig. 4d, recent warming has a much stronger positive contribution to drought and offsets the precipitation increase in northwestern China, especially since the late 1990s in northwestern China.
Furthermore, the temporal evolutions of regional averaged SPEI and SPI are shown in Fig. 5. It is clear that the evolution of regional mean SPEI and SPI corresponds to that in Fig. 4. From Fig. 5, the drought in northwestern China persists in recent years as a result of the positive temperature anomaly, although precipitation has increased in that period as indicated by the temporal evolution of SPI. In addition, before the late 1990s, the temporal evolution of SPEI is similar to that of SPEI, illustrating that the temperature anomaly is not significant.
To validate previous results in drought analysis using CRU TS3.2 dataset, we also calculated the sc-PDSI, SPEI, and SPI using a station-based dataset. Figure 6 shows the change of sc-PDSI, SPEI, and SPI from 1987–97 to 1998–2009 in northwestern China. It is evident that the values of sc-PDSI and SPEI decreased in most parts of northwestern China. In this analysis, the differences of sc-PDSI between 1998–2009 and 1987–97 in 31 sites (accounting for two-thirds) are negative with the minimum value of −6. In contrast, there are only 3 sites with a positive value greater than 2, indicating more dryness in 1998–2009 than that in 1987–97. The differences of SPEI between 1998–2009 and 1987–97, however, have only one positive value station, with the other stations having negative values. Overall, the severity and spatial extent of aridity and drought have increased substantially in northwestern China in the last 10 years. As mentioned above, global warming plays an important role in the recent drying since the high temperature increases the atmospheric demand for moisture.
5. Discussion and conclusions
The precipitation increase in the 1990s inspired an anticipation of a climate change from warm/dry to warm/wet in the extremely arid northwestern area of China. If this anticipation comes true, the ecosystem and environment will definitely be improved, with the lessening of pressures on scarce water resources. However, our analysis of the updated data to the end of 2009 reveals that the drought condition has worsened in the recent two decades. We show that the soil moisture and the near-surface relative humidity have both decreased, and the drought indices indicate that the drought state has deteriorated in northwestern China since the early 1990s.
Decadal circulation changes may have helped intensify the drought condition in recent years. Figure 7 demonstrates the circulation change in East Asia in the most recent two decades (i.e., the average of 1998–2009 minus that of 1987–98). Positive geopotential height change can be observed above northern China and Mongolia, which weakens the Asian low in summer, favors the downward velocity anomaly, and helps to restrain large-scale convection in this region. An anomalous easterly can be observed in northwestern China, which is opposite to the climatology wind direction and weakens the wind speed in this region. The weak wind can reduce the eastward transfer of water vapor from the midlatitude westerly, leading to drier conditions and favoring the occurrence of extremely hot weather. This is further validated by the temperature anomaly in East Asia, which shows a striking warming change (shading of Fig. 7). The maximum temperature change in northwestern China exceeds 0.5°C. This circulation-related regional warming, superimposed on the global warming, inevitably increased the frequency of extremely hot and dry days, and helped to result in the continuous drought condition in recent years, as illustrated by the strong SPEI decrease in Fig. 5.
The water balance in a region is determined not only by the precipitation but also by the evaporation, which is highly influenced by the temperature fluctuation. Although SPEI, SPI, and sc-PDSI can reveal dryness and wetness change over large regions, they show different sensitivities. For the extremely arid northwestern China with large potential evaporation, the sensitivity of the drought index to changes in evaporation demand should be taken into account. Therefore, SPEI and sc-PDSI are more suitable to detect the tendency of dryness and wetness severity in historical and scenario conditions in climate change impact assessments there.
Under the background of global warming and a stronger regional warming in the inner region of the Eurasian continent, the evaporation will inevitably increase in northwestern China. For this extremely arid region with rare precipitation, the evaporation increase dominates the precipitation change and determines the dry/wet condition. The climate change projection into the end of the twenty-first century using multimodel ensemble indicates a stronger warming in central Asia, and the models show strong consistency in this region (Solomon et al. 2007). For example, by the end of the twenty-first century (2080–99), the projected temperature increase in central Asia is more than 3°C for the Special Report on Emissions Scenarios (SRES) A1B scenario, while the precipitation change in this region is small and has large uncertainty. Therefore, temperature-related evaporation increase will be large and very likely surpass the precipitation change in the future. As a result, the drought condition will be serious, which is a great pressure on the ecosystem and economic development in this extremely arid region.
The authors thank two anonymous reviewers and editor Dr. Aiguo Dai for their helpful comments and suggestions. We thank Dr. Debashis Nath for a careful and thorough review of the manuscript. This work was supported by the Special Scientific Research Project for Public Interest under Grant 201006009, the National Basic Research Program of China under Grant 2009CB421405, and the National Natural Science Foundation of China under Grant 41175041.