The impact of climate changes on water level of Qinghai Lake in China over the past 50 years

Previous studies have compared Qinghai Lake annual level trends with annual river runoff trends and climatic data; these studies have tried to investigate the impacting factors of the Qinghai Lake level (Tang et al. 1992; Ma 1996; Liu 2004; Li et al. 2005; Li et al. 2007; Liu et al. 2008; Shi et al. 2010; Zhang et al. 2011a; Jin et al. 2013). Most studies have concluded that annual lake level variations before the 2000s were primarily driven by climate and runoff (Ma 1996; Li et al. 2007; Zhang et al. 2011a; Jin et al. 2013). However, Qinghai Lake lies in a critical transitional zone (Figure 1), where the Southeast Asian monsoon (SEAM), the westerly circulation (WC) and the Qinghai-Tibet Plateau monsoon meet. These forces impact the lake differently in different time periods (LZBCAS 1994), making the distribution and moisture source of the precipitation highly complex and unusual (Wang et al. 2004, 2008; Xu et al. 2007; Henderson et al. 2010; An et al. 2012; Cui & Li 2015b). Due to the lake watershed being closed, precipitation is the sole source of river runoff, and the runoff lags behind the precipitation (Cui et al. 2011). Therefore, due to using and comparing annual data of lake level, runoff, and precipitation, the reason for Qinghai Lake level variation is not fully explained. Nor is the contribution of each factor on lake level variation yet fully understood.

This study analyzed monthly lake level, climate factors, and river runoff data (1960–2012) to investigate the reasons for lake level variation, using principal component analysis (PCA) as an analytical method. The objectives of this study were: (1) to investigate the monthly and annual variations of Qinghai Lake level; (2) to identify the reasons for lake level variation; and (3) to discuss future possible variation of lake level. It is hoped that study results will provide insight into the hydrological processes of the Qinghai Lake Basin, and inform water resource management in the Qinghai Lake Basin and the northeastern Qinghai-Tibet Plateau.

Qinghai Lake (36 °32′–37 °15′N, 99 °36′–100 °47′E), a brackish lake with a water surface area of 4,400 km² and a water volume of 71.6 × 10⁹ m³ at an altitude of ∼3,193 m above sea level (a.s.l.), is the largest lake in China (Figure 1). The lake is a closed basin with a watershed area of approximately 29,661 km² with no surface water outflow. It lies in the cold and semiarid region of China's NE Qinghai-Tibet Plateau. The average annual air temperature in the lake area is ∼1.2 °C. The average annual precipitation is ∼357 mm, with more than 65% occurring in summer. Evaporation is three to four times higher than precipitation (Li et al. 2007). The climate of Qinghai Lake is influenced by two air masses: SEAM and WC. The SEAM reaches this region in summer, while the WC dominates in winter, resulting in a clear seasonality of precipitation (An et al. 2012; Cui & Li 2015b).

The lake water has a salinity of 15.5 g/L and pH of 9. Its water chemistry is similar to many saline lakes in the Tibetan Plateau; the cations of lake water is in the order: Na⁺ > Mg²⁺ > K⁺ > Ca²⁺, and anions is in the order: Cl⁻ > SO₄²⁻ > HCO₃⁻. More than 50 rivers or streams flow into Qinghai Lake (LZBCAS 1994). The rivers are asymmetrically distributed, with most of them located north and northwest of the lake (Figure 1; Yan & Jia 2003). Most of the rivers are seasonal, with 85% of the annual discharge occurring between June and September. The main rivers are Buha River, Shaliu River, Haergai River, Quanji River, and Heima River (LZBCAS 1994; Li et al. 2007), with a total discharge of 1.43 × 10⁹ m³/yr. This river water accounts for 80.3% of the total surface water volume flowing into Qinghai Lake (Table 1).

For this study, the runoff of Buha River was chosen to represent the hydrological characteristics of rivers flowing into Qinghai Lake. The Buha River is the largest river in the Qinghai Lake Basin (LZBCAS 1994; Li et al. 2007), with a catchment area of approximately 14,932 km², which is about 50% of the Qinghai Lake Basin. Its discharge is 8.09 × 10⁸ m³/yr, which accounts for approximately 50% of the total volume of surface water flowing into Qinghai Lake (Cui & Li 2015a). Furthermore, from a study methods perspective, only two rivers (Buha River and Shaliu River) have hydrological observation stations at their estuaries, needed to quantitatively estimate runoff. Based on these factors, the Buha River was used as a representative river to represent the hydrological characteristics of all rivers flowing into Qinghai Lake.

Figure 1 shows the locations of meteorological and hydrological stations in the Qinghai Lake Basin. Monthly lake level records and river runoff records from 1961 to 2012 were collected from Xiashe hydrological station and Buha River hydrological station, respectively (Figure 1). There are only two weather stations (Tianjun and Gangcha weather stations) in the Qinghai Lake Basin (Figure 1). Precipitation and temperature, for example, observed at the Tianjun weather station were consistent with those observed at the Gangcha weather station (Cui et al. 2011; Cui & Li 2015b); the Gangcha weather station is approximately 13 km north of the lake and is the nearest weather station to the lake (Figure 1). Therefore, monthly climate-related factors (precipitation, temperature, evaporation, and wind speed) of Gangcha weather station were used to analyze the impact of climate changes on lake level variation. Annual data were derived from monthly data.

In this study, correlation analysis (CA) and PCA were used to study the impacts of climate and runoff on lake level (Astel et al. 2007; Olsen et al. 2012). Calculations were done using SPSS 13.0 software. The PCA is a multivariate technique where observations about the data set are described using several inter-correlated quantitative dependent variables (Manly 2000; Davis 2002; Shaw 2003). A ‘percent explained variance’ approach is often used, where the first few principal components (PCs) explain a large fraction of the total variance (e.g., greater than 60–70%). A given variable is considered to be an important contributor if its absolute loading exceeds approximately 0.75 (Olsen et al. 2012). The analytical goal is to extract the most important information from the data set, to represent it as a set of new orthogonal variables called ‘PCs’, and to display similar patterns across observations and variables as points in maps (Shaw 2003).

From 1961 to 2012, Buha River runoff fluctuated obviously, ranging from 1.99 × 10⁸ m³ to 19.47 × 10⁸ m³ (Figure 4), but with no clear overall increase or decrease over time. River runoff was relatively high in the 1980s and 2000s and relatively low in the 1970s and 1990s (Table 2). Changes in runoff can be divided into three stages (Figure 4): (1) fluctuating and decreasing stage (1961–1980) with a decreasing rate of 0.34 × 10⁸ m³/yr; (2) fluctuating stage (1981–1995); and (3) fluctuating and increasing stage (1996–2012) with an increasing rate of 0.44 × 10⁸ m³/yr. Monthly runoff was relatively high in the summer (June to August) and relatively low in the winter (December to February) (Figure 3).

According to Figure 4, at the annual scale, high evaporation peaks generally corresponded to low precipitation and low humidity, but also to relatively high temperature and high wind speed (e.g., 1969, 1979, 1995, and 2001). The opposite also holds true: low evaporation peaks generally corresponded to high precipitation and high humidity, but also to relatively low temperature and low wind speed (e.g., 1967, 1983, 1989, 2005, and 2012). CA results (Table 3) showed that the evaporation was highly negatively correlated to precipitation and humidity. Statistically, the absolute values of correlation coefficients were all above 0.35, and all were significant at a p value of <0.05. These results suggested that high evaporation was closely related to dry climate, and low evaporation was closely related to wet climate. Humidity was positively correlated with precipitation (Table 3), namely, the precipitation had a positive effect on humidity. Due to the high wind speed despite low temperatures, monthly evaporation was relatively high in March and April (Figure 3).

At the annual scale, high runoff peaks generally corresponded to high precipitation and low evaporation (e.g., 1967, 1983, 1989, 1999, and 2005); conversely, significant drops in runoff generally corresponded to low precipitation and high evaporation (e.g., 1969, 1973, 1979, 1995, and 2001). Precipitation changes affected the generation of surface runoff directly, whereas temperature changes affected the evaporation of surface runoff (Cui & Li 2015a). These results indicated that drops in river runoff were closely related to a dry and relatively warm climate, whereas increases in river runoff were closely related to wet and relatively cold climates. Precipitation had a positive effect on runoff, while potential evaporation due to temperature and wind speed played a negative role (Li et al. 2007; Zhang et al. 2011a). Pearson correlation coefficients between river runoff and climatic factors showed that runoff was positively correlated with precipitation and humidity, but negatively correlated with evaporation and wind speed; all were significant at a p value of <0.01 (Table 3). However, the correlation coefficient between runoff and temperature was not statistically significant (Table 3). These results demonstrated the river runoff in the Qinghai Lake Basin was primarily influenced by precipitation.

According to Table 2 and Figure 4, lake level variation showed a similar trend as precipitation and runoff, but revealed an inverse relationship with evaporation when considering data by decade (relative low in the 1970s and 1990s, relative high in the 1980s and 2000s). Peaks in lake level increases corresponded to relatively high precipitation, humidity, and runoff, but also to relatively low evaporation and wind speed (e.g., 1967, 1983, 1989, 1999, 2005, and 2012; solid line in Figure 4). Drops in lake level corresponded to relatively low precipitation, humidity, and runoff, but also to relatively high evaporation and wind speed (e.g., 1969, 1973, 1979, 1995, and 2001; dotted line in Figure 4). Pearson correlation coefficients showed that lake level variation was highly positively correlated to precipitation, humidity, and river runoff, and negatively to evaporation and wind speed. All correlations were significant at a p value of <0.001 (Table 3). However, the correlation coefficient between lake level variation and temperature was 0.05, with a non-significant relationship. These results suggested that lake level increase was closely related to wet climate; lake level decline was closely related to dry climate. For monthly lake level variation, the maximum drops were in November and December and were closely related to the freezing of Qinghai Lake's surface from November to March (Che et al. 2009). The maximum increases in lake level were in July and August and were closely related to high river runoff and high precipitation in summer season (Cui et al. 2011).

Pearson correlation coefficients showed that during SEAM or WC periods, lake level was highly positively correlated to precipitation, humidity, and river runoff, and negatively correlated to evaporation and wind speed (Table 4). However, the correlation coefficient between lake level variation and temperature was below 0.05 and the relationship was non-significant.

PCA results (Tables 5 and 6) showed that the cumulative variance for the first five PCs were relatively high and reached almost 85.4% (>85%), and the eigenvalue of PC5 was 1.1 (>1). In order to distinguish the contribution of potential influencing factors during two periods (SEAM and WC), the cumulative variance was selected above 95%, so PCs would be analyzed from PC1 to PC8 (Table 5). PC1 (variance of 38.0%) was strongly positively correlated with humidity and was negatively correlated with evaporation of the SEAM period, with each absolute loading exceeding 0.84 (Table 6). PC2 (variance of 18.0%) was strongly positively correlated with wind speed and was negatively correlated with temperature of the SEAM and WC periods, with each absolute loading exceeding 0.65. PC3 (variance of 11.8%) was strongly positively correlated with precipitation and river runoff of the WC period, with each absolute loading exceeding 0.66. PC4 (variance of 9.5%) was strongly positively correlated with wind speed of the SEAM period, with loading exceeding 0.56 (Table 6). PC5 (variance of 8.1%) was strongly positively correlated with evaporation of the WC period, with loading exceeding 0.53. PC6 (variance of 4.9%) was strongly positively correlated with temperature of the WC period, with loading exceeding 0.45. PC7 (variance of 3.0%) was strongly positively correlated with humidity of the WC period, with loading exceeding 0.37. PC8 (variance of 2.3%) was strongly positively correlated with river runoff and precipitation of the SEAM period, with loading exceeding 0.3 (Table 6). In all, PC1, PC4, and PC8 represented the comprehensive contribution of influencing factors (climate and river runoff) of the SEAM period to annual lake level variation, with a sum variance of approximately 49.8%; PC3, PC5, PC6, and PC7 represented the comprehensive contribution of influencing factors of the WC period to annual lake level variation, with a sum variance of approximately 27.8%; and the comprehensive contribution of influencing factors of both the SEAM period and WC period was approximately 18.0% (PC2).

Owing to the closed basin of the Qinghai Lake, precipitation was the sole source of water inputting into the basin, meaning that river water was mainly fed by precipitation in the basin (Cui & Li 2005a). Further, precipitation directly affected humidity (Table 3), thus contributions of river runoff and humidity on lake level variation were merged into the precipitation's contribution. Therefore, the relative contributions of temperature, precipitation, evaporation, and wind speed to Qinghai Lake level variation were approximately 13.8%, 36.3%, 27.1%, and 18.4%, respectively (Tables 5 and 6). These all indicated that annual lake level variation was primarily influenced by precipitation and evaporation of the SEAM period.

The results show that the changes of lake level were divided into two stages: decreasing stage (1961–2004) and increasing stage (2004–2012) (Figure 2). Pearson correlation coefficients between lake level variation and influencing factors suggest that: the lake level during the decreasing stage (1961–2004) was controlled by the comprehensive effect of climate and river runoff; and the lake level during the increasing stage (2004–2012) was only sensitive to river runoff (Table 7). Increase of river runoff would be dependent upon the combined effects of changing rainfall patterns (increase in intensity of individual summer rainfall) and improving grass vegetation within the basin as a result of global warming (Jin et al. 2013). Furthermore, thawing permafrost (due to climate warming) would also increase river runoff in the alpine catchment (Ye et al. 2003; Walvoord & Striegl 2007; Connon et al. 2014), which contributes more surface water flowing into the lake, making the lake level increase. However, annual precipitation did not clearly increase during this period of 2004–2012. Therefore, the contributions of melt water from glacier shrinkage and permafrost degradation, under climate warming on the TP, to river runoff cannot be ignored in any study of lake level variation on the TP.

Evaporated water vapor from Qinghai Lake strongly influenced local precipitation, creating a distinct regional water cycle. The annual contribution of lake evaporation to the basin's total precipitation was approximately 23%, or 26.7 × 10⁸ m³, concentrated mostly in the summer (Cui & Li 2015b). Annual evaporation from the Qinghai Lake was found to be approximately 39 × 10⁸ m³ (Liu 2004; Shi et al. 2010), meaning net output water of evaporation was approximately 12.3 × 10⁸ m³. The large number of lakes on the Tibetan Plateau, including more than 1,500 lakes covering a total lake area of 44,993 km² (Xu et al. 2011; Song et al. 2014a), clearly results in a significant contribution of evaporation from lake surfaces into atmospheric vapor. The contribution of evaporation on lake level variation could not be ignored.

A tree ring-width chronology of the northeastern Tibetan Plateau, spanning 4,500 years, suggested that large-scale climate warming might be associated with an even greater moisture supply in this region (Yang et al. 2014). If so, the lake's level would rise, due to increasing precipitation and runoff (Christensen et al. 2014; Song et al. 2014b). Consequently, the rising water level in Qinghai Lake since 2004 may reflect the comprehensive effect of increased precipitation and runoff and changing rainfall patterns, resulting from global warming. If warming trends continue, the lake level of Qinghai Lake may continue to rise in the future decades. Accordingly, a better ecological environment and richer biodiversity could be expected in the Qinghai Lake Basin after years with continuous rising water level (Li et al. 1999; Yang et al. 2003; Kong et al. 2011).

Overall, Qinghai Lake water level was sensitive to climate and river runoff. Precipitation, river runoff, and evaporation had direct effects on lake volume, while temperature, humidity, and wind speed had indirect effects on lake volume. Precipitation and runoff had directly positive effects on the lake level, evaporation and wind speed negatively impacted lake level. The relative contributions of influencing factors in the SEAM and WC periods to annual lake level variations were approximately 49.8% and 27.8%, respectively. The relative contributions of temperature, precipitation, evaporation, and wind speed on lake level variation were approximately 13.8%, 36.3%, 27.1%, and 18.4%, respectively. In general, annual level variation of Qinghai Lake was primarily influenced by precipitation and evaporation of the SEAM period.

The study was supported by the National Science Foundation of China (NSFC41201039, NSFC41130640); the National Key Technology R & D Program (2012BAH31B03) and the Fundamental Research Funds for the Central Universities, PCSIRT (No. IRT1108).