1. Introduction
Ongoing global climatic change is expected to enhance the global hydrologic cycle, which will affect streamflow and water availability and thereby may disturb the discharge regime of rivers (Barnett et al. 2005; Huntington 2006; IPCC 2007; Oki and Kanae 2006). To provide further evidence for warming-induced hydrological cycle intensification, there has been increasing interest in the linkage of climatic variability/change to hydrological processes and water resources across various temporal and spatial scales in the past decades (Christensen et al. 2004; Immerzeel et al. 2012; Karl and Riebsame 1989; Labat et al. 2004; Milliman et al. 2008; Nijssen et al. 2001; Nohara et al. 2006). Many simulation results have reported an overall increase in global streamflow, which closely parallels to climate warming (Arnell and Gosling 2013; Groisman et al. 2004; Labat 2008, 2010; Milliman et al. 2008). However, the feedback of climate change/variability on the hydrological cycle is highly nonlinear, extremely complex, and less certain, and it furthermore differs substantially within and among different hydrometeorological regions (Arnell and Gosling 2013; Immerzeel et al. 2010; Labat et al. 2004, 2005; Milliman et al. 2008; Milly et al. 2005; Xu et al. 2010). Marked decreases as well as no significant trends in streamflow have also been discerned through historical records or predicted by hydroclimate models in some regions of the globe (Arnell and Gosling 2013; Labat et al. 2004; Milly et al. 2005; Nijssen et al. 2001). Because substantial uncertainty regarding trends in regional streamflow remains (Huntington 2006; Milly et al. 2005), further investigation on regional patterns of climate-induced changes in streamflow is still indispensable to deepen our understanding of the multiscale climate–streamflow cascade.
The Qinghai–Tibet Plateau plays a pivotal role in both regional and global climate because of its high elevation (Pan and Li 1996; Qiu 2008). Growing observational evidence has shown that a striking and accelerating climate warming occurred across the whole plateau in the past decades (Ding and Zhang 2008; Duan et al. 2006; Gautam et al. 2013; Guo and Wang 2012; Liu and Chen 2000; Wang et al. 2008). Liu and Chen (2000) reported a statistically significant warming trend in most areas of the Qinghai–Tibet Plateau since the mid-1950s, with a rise in annual mean temperature of ~0.16°C decade−1 during the period from 1955 to 1996, and the strongest warming in winter among all seasons, with a rate of increase of ~0.32°C decade−1. Based on the daily mean air temperature data during the period from 1961 to 2003, Duan et al. (2006) updated the rates of rise in annual and winter mean temperature to approximately 0.28° and 0.44°C decade−1, respectively. You et al. (2010) examined the monthly mean temperature of 71 stations selected from the China homogenized historical temperature datasets (1951–2004) and estimated the warming rates in annual and winter mean temperature of 0.25° and 0.40°C decade−1, respectively. In addition, the recent works by Wang et al. (2008) and Li et al. (2010) reported a rise in annual mean temperature of up to 0.36°–0.37°C decade−1 for the period from 1961 to 2007, more than twice the warming rate in the period from 1955 to 1996 (Liu and Chen 2000).
As the source of Asia’s major large rivers, the Plateau’s high climatic sensitivity and absence from significant anthropogenic disturbances make it a uniquely desirable environment for examining hydrological response to climate change (Hannah et al. 2007). Its rapid climate warming inevitably exerts profound effects on the global and regional hydrological cycle (Cao et al. 2006; Immerzeel et al. 2010; Lai 1996). Following climate warming, some parts of the Plateau witnessed increasing precipitation and streamflow (Gautam et al. 2013). However, no significant annual precipitation trend as well as a statistically insignificant downward streamflow trend has generally prevailed in the upstream regions of some Chinese large rivers over the plateau since the mid-1960s (Cao et al. 2006; Cuo et al. 2014; Ding and Zhang 2008; Lai 1996; Qin et al. 2010; Zhang et al. 2011; Zhang et al. 2012; Y.-Y. Zhang et al. 2013). The direction and magnitude of trends in precipitation and streamflow varies with basins (Cao et al. 2006; Cuo et al. 2014). Despite a multitude of previous studies, however, it remains unclear as to the linkages among air temperature, precipitation, and streamflow over upstream regions for these large rivers (Cuo et al. 2014). In particular, the differences in the contributions of air temperature and precipitation to streamflow among basins are virtually unknown.
Originating in the Qinghai–Tibet Plateau, the Lancang–Mekong River (LMR) and Nu–Salween River (NSR) are important international rivers across China and Southeast Asia (Fig. 1). Their streamflow fluctuations induced by climate warming could not only affect the downstream water availability, the ecological environment, and food security but could also pose enormous challenges to transboundary water utilization and allocation (Immerzeel et al. 2010). In particular, large-scale hydroprojects from planning to commission along the trunks of both rivers are highly sensitive to potential climate-induced streamflow changes in the upstream regions. Statistically insignificant increasing streamflow trends before 2000 were observed in the upstream regions of both rivers (Cao et al. 2006; Yao et al. 2012). However, this view was challenged by a more recent effort (Zhang et al. 2012), which reported a statistically insignificant decreasing trend in the upstream regions of the LMR during 1958–2000. Therefore, reexamining long-term trends of air temperature, precipitation, and streamflow and relating streamflow to air temperature and precipitation in the upstream regions of the LMR and NSR is of urgent importance for developing a better understanding on the spatiotemporal trends of streamflow over the Qinghai–Tibet Plateau and evaluating the potential impacts of climate change on the water resources and hydropower in the two river basins. The objective of this study is to examine the nature of trends in air temperature and precipitation over the upstream regions of the LMR and NSR, how changing climatic conditions affect streamflow trends, and subsequently to test whether a linkage of the variability of temperature and precipitation to observed streamflow trends exists.
Topography of the upstream regions of the LMR and NSR and the location of meteorological and hydrologic stations used in this study, with box-and-whisker plots of multiyear monthly air temperature and precipitation from the 1950s to 2010.
Citation: Journal of Hydrometeorology 16, 5; 10.1175/JHM-D-14-0238.1
2. Study area
In this study, the upstream regions of the LMR and NSR were delimited as the catchment areas above the Jiuzhou (25°47′N, 99°13′E) and Gongshan (27°44′N, 98°41′E) hydrologic stations, respectively (Fig. 1). The regions are located in the alpine belt of the Qinghai–Tibet Plateau and Hengduan Mountains, which has complex terrain, including glaciers, plateaus, mountains, and canyons. The altitude ranges from approximately 1300 to 6968 m MSL. The annual mean temperature varies from 12°C at lower elevations to less than −3°C at higher elevations. The summer months (June–August) have a high monthly mean air temperature of 5°–20°C, and the winter months (December–February) have a low monthly mean air temperature from −20° to 5°C (Fig. 1). Because of the high altitudes, highly rugged terrain, and a particularly cold climate, direct anthropogenic disturbances are very absent over the most remote and least developed areas in southwest China. Therefore, trends in streamflow can factually reflect climatic variability/change.
The upstream regions of the LMR and NSR cover 10.6% (~84 220 km2) and 39.1% (~106 480 km2) of the whole basin areas and contribute 6.1% and 18.5% of the annual water discharge into the South China Sea and the Indian Ocean, respectively (Campbell 2009; Guo 1985; Robinson et al. 2007). The upstream regions of the LMR exist in the eastern Qinghai–Tibet Plateau and are dominated by the East Asian monsoon in summer (June–August) and midlatitude westerlies in winter (December–February), whereas the upstream regions of the NSR exist on the southern Qinghai–Tibet Plateau and are predominantly affected by the Indian monsoon in summer and midlatitude westerlies in winter (Cuo et al. 2014; Guo 1985). The monsoon precipitation plays an important role over both upstream regions and brings about a heavily uneven spatial and temporal distribution of precipitation. The annual precipitation ranges from 250 to 1000 mm. More than 60%–90% of the annual precipitation occurs in the wet season (May–October), and the remainder occurs in the dry season (November–April; Fig. 1). Accordingly, the monsoon precipitation also contributes more than 70% of the annual streamflow (Cuo et al. 2014; L. Zhang et al. 2013). Monthly mean streamflow from the upstream regions of both rivers fluctuates greatly, with distinct flood and nonflood seasons (Fig. 2). High streamflow with pronounced interannual variation occurs during the wet season, whereas low streamflow with small interannual fluctuation occurs in the dry season. In comparison with the LMR, the upstream region of the NSR has a higher streamflow with greater interannual variation in the wet season.
Box-and-whisker plots of multiyear monthly streamflow observed at the (a) Jiuzhou and (b) Gongshan hydrologic stations from the 1950s to 2010.
Citation: Journal of Hydrometeorology 16, 5; 10.1175/JHM-D-14-0238.1
3. Data and methods
a. Data sources and preprocessing
Meteorological data such as monthly precipitation totals, and monthly mean, maximum, and minimum air temperature recorded from the 1950s to 2010 at 16 meteorological stations are used in this study. These data are derived from the Monthly Surface Climate Dataset in China catalog (SURF_CLI_CHN_MUL_MON), issued by the National Meteorological Information Center of the China Meteorological Administration [Chinese National Meteorological Information Centre (CNMIC) 2005]. These examined meteorological stations have operated since the 1950s, with the exception of some observational records collected since the late 1970s and early 1980s (Fig. 1). To guarantee the consistency and completeness of the meteorological dataset, three years (1958–59 and 1969) with missing data at Dingqing and Nangqian stations were excluded prior to analyses. These very few and nonconsecutive missing observations should have a minimal impact on the results.
Monthly mean streamflow data recorded at the Jiuzhou and Gongshan hydrologic stations from the 1950s to 2010 were also employed. Because the upstream regions above both stations are generally less disturbed by large-scale human development, these long time series of flow data are believed to be indicative of climate change. At the Jiuzhou hydrologic station, monthly observational streamflow records are available from 1956 to 2010. Following the closure of Gongguoqiao Dam, located in the middle reach of the Lancang River, in 2011, the Jiuzhou hydrologic station was fully inundated and hence stopped operating (Fan et al. 2015). The Gongshan hydrologic station has been in operation since 1979. Its monthly streamflow data for the period 1954–78 have to be extrapolated from the empirical relationship of its monthly streamflow data to those recorded at the Daojieba hydrologic station during the period 1979–2010. The Pearson correlation coefficients between the monthly mean streamflow at the two stations are 0.63–0.97, with a p value of <0.01. The Daojieba hydrologic station is about 390 km downstream from the Gongshan hydrologic station (Fig. 1) and has the longest period of record for streamflow among all hydrologic stations on the mainstream of the NSR.
b. Methods
1) Trend analysis and change-point detection
The Mann–Kendall test is one of the nonparametric tests widely used for trend analysis in hydrometeorological time series (Hamed 2009; Novotny and Stefan 2007; Oguntunde et al. 2011; Yue et al. 2002). However, meteorological and hydrologic data always tend to exhibit significant serial correlation, which has adverse effects on the results of trend tests. To remove the serial correlation from the examined data, a trend-free prewhitening (TFPW) procedure is needed prior to applying a trend test (Yue et al. 2002). In the process, the slope of the trend is computed using the Theil–Sen regression estimator (Sen 1968; Theil 1992). A more detailed description of the TFPW Mann–Kendall (TFPW-MK) test was provided by Yue et al. (2002). In this study, the TFPW-MK test was employed to examine trend significance in hydrometeorological time series and to estimate trend slope at the annual and seasonal scales. Statistical significance of trends is considered at a confidence level of 95% or higher. On account of the inconsistent observational periods at the examined meteorological stations, trends in air temperature and precipitation were tested using a moving window for periods of study with varying start years (Adam et al. 2007). Therefore, trends for periods covering 1955–85, 1960–90, 1965–95, 1970–2000, 1975–2005, and 1980–2010 were analyzed. Furthermore, field significance of the observed trends in air temperature and precipitation was evaluated using Monte Carlo simulations (Elmore et al. 2006; Livezey and Chen 1983).
The Breaks For Additive Season and Trend (BFAST) method proposed by Verbesselt et al. (2010) was used to detect abrupt changes in monthly meteorological and hydrologic time series. It is a generic change detection method integrating the iterative decomposition of time series into trend, seasonal, and noise components with change-point detection methods. An additive decomposition model is adopted to iteratively fit a piecewise linear trend and a seasonal cycle. For each iteration, the ordinary least squares residuals-based moving sum test was used to detect whether abrupt changes are occurring in the time series. If the tested change is statistically significant (p value < 0.05), the breakpoints are determined by the Bayesian information criterion, and then the number of breaks, the occurrence time, and the confidence interval of the occurrence time for each break are estimated. The preceding procedure is iterated until the number and position of the breakpoints are unchanged. More details on this procedure can be found in Verbesselt et al. (2010).
2) Random forest regression
A correlation of streamflow to meteorological variables, especially to precipitation, can be expected in the relatively less disturbed regions (Hu et al. 2011; Novotny and Stefan 2007). To investigate the relationships between streamflow and climate variables and to identify the explanatory power of meteorological variables for streamflow, the random forest regression proposed by Breiman (2001a,b) was used in this study. Random forest regression does not overfit and is less sensitive to noise in the bootstrap training sample such that it is suitable to handle high-dimensional data, high-order interactions, and nonlinearities among variables (Breiman 2001a,b; Liaw and Wiener 2002). In comparison with other machine learning algorithms and standard data models such as linear or logistic regression, random forest regression can produce better predictive accuracy and provide more reliable information about the underlying mechanism (Breiman 2001a,b). In essence, random forest regression involves collections of regression trees formed by growing unpruned trees depending on a random vector. For each regression tree in the forest, approximately one-third of the observations are left out of the bootstrap training sample and are used as a test set to obtain the out-of-bag (OOB) estimate of the generalization error. The importance of a variable is the average increase in squared OOB residuals when the variable is permuted (Liaw and Wiener 2002). The “mean of squared residuals” is computed as 
4. Results
a. Temperature and precipitation trends
1) Trends in annual and seasonal air temperature
As illustrated in Fig. 3a, mean annual air temperature is generally increasing from the mid-1950s to 2010 in the upstream regions of the LMR and NSR. Consistent and statistically significant upward trends for significance level 99% occur between 1980 and 2010. The magnitude of trends in the mean annual air temperature has a median of 0.52°C decade−1 across the study area. It ranges from 0.31 to 0.67°C decade−1 with a median of 0.56°C decade−1 and from 0.38 to 0.69°C decade−1 with a median of 0.47°C decade−1 in the upstream regions of the LMR and NSR, respectively (Fig. 4a). Increasing positive trends over time are mostly observed for the stations with longer records. Upward trends in annual maximum air temperature are predominant between the mid-1950s and 2010 in the upstream regions of both rivers (Fig. 3b). However, the trends for most of the examined stations are statistically insignificant at the 95% level. The change rate of annual maximum air temperature for the period 1980–2010 has a median of 0.50°C decade−1 over the two upstream regions. It varies from 0.25° to 0.85°C decade−1 (median = 0.42°C decade−1) and from −0.26° to 0.75°C decade−1 (median = 0.53°C decade−1) for the LMR and NSR, respectively (Fig. 4a). Statistically significant positive trends are mainly observed at the stations with lower elevation (Figs. 1, 3b). Insignificant increasing tendency for the annual minimum air temperature is predominantly seen for the upstream regions of both rivers during the periods of study (Fig. 3c). Its magnitude over the common period 1980–2010 is −0.14° to 1.31°C decade−1 (median = 0.73°C decade−1) and 0.59° to 1.70°C decade−1 (median = 0.85°C decade−1) in the upstream regions of the LMR and NSR, respectively (Fig. 4a). Therefore, the median rise rate of the annual minimum air temperature reaches 0.83°C decade−1 over the two upstream regions. Nevertheless, statistically significant trends occur mainly in the upstream regions of the NSR (Fig. 3c).
Trends in (a) annual mean, (b) maximum, and (c) minimum air temperature, and (d) annual precipitation totals. Blue triangles and asterisks denote statistically significant trends at a confidence level of 95% and 99%, respectively.
Citation: Journal of Hydrometeorology 16, 5; 10.1175/JHM-D-14-0238.1
Box-and-whisker plots of (a) air temperature and (b) precipitation trends from 1980 to 2010.
Citation: Journal of Hydrometeorology 16, 5; 10.1175/JHM-D-14-0238.1
A significant seasonal air temperature rise is also prevalent in the upstream regions of both rivers over the periods of study (Fig. 5), especially for the period 1980–2010. However, seasonal differences are apparent. Winter is the season with the strongest observed air temperature warming (Fig. 5d). Its temperature rise rate between 1980 and 2010 is 0.38°–1.04°C decade−1 (median = 0.76°C decade−1) and 0.55°–0.98°C decade−1 (median = 0.68°C decade−1) in the upstream regions of the LMR and NSR, respectively (Figs. 4a, 5d). The resulting median over the two upstream regions reaches 0.69°C decade−1. Moreover, the trends for all stations exclusive of the Anduo station are statistically significant. Warming of the summer air temperature is also seen over the upstream regions of the LMR and NSR, with a median rise rate of 0.35°C decade−1 from 1980 to 2010 (Figs. 4a, 5b). The magnitude of trends is 0.16°–0.65°C decade−1 (median = 0.44°C decade−1) and 0.14°–0.47°C decade−1 (median = 0.34°C decade−1), respectively. However, upward trends are statistically insignificant at five of the examined sixteen stations, of which four stations fall within the regions with lower elevation. Pronounced air temperature warming also occurs in spring (March–May) and autumn (September–November; Figs. 5a,c). The median rise rate between 1980 and 2010 is 0.44° and 0.45°C decade−1 in spring and autumn, respectively (Fig. 4a). Meanwhile, a statistically significant increasing tendency in spring air temperature spreads over all stations, with a rise magnitude of 0.31°–0.71°C decade−1 (median = 0.46°C decade−1) and 0.31°–0.67°C decade−1 (median = 0.42°C decade−1) in the upstream regions of the LMR and NSR, respectively (Figs. 4a, 5a). Regarding autumn, the temperature increase rate reaches 0.26°–0.67°C decade−1 (median = 0.48°C decade−1) and 0.30°–0.60°C decade−1 (median = 0.40°C decade−1) in the upstream regions of the LMR and NSR, respectively (Figs. 4a, 5c). Autumn air temperature warming is statistically significant at all stations exclusive of the Changdu station. Figure 5 also shows persistent positive trends in seasonal air temperature for most of the stations with longer records. However, the magnitude of trends over different observed periods varies among seasons and stations.
Trends in seasonal means of monthly mean air temperature in (a) spring, (b) summer, (c) autumn, and (d) winter. Blue triangles and asterisks denote statistically significant trends at a confidence level of 95% and 99%, respectively.
Citation: Journal of Hydrometeorology 16, 5; 10.1175/JHM-D-14-0238.1
With respect to abrupt changes in the trends of annual and seasonal air temperature over the upstream regions of both rivers, there are only a few stations with a consistent change-point time for the examined hydrometeorological variables (Table 1). Across the upstream regions of the LMR and NSR, except for the mean annual air temperature with the predominant abrupt changes generally occurring in the late 1990s, sudden changes of the other annual and seasonal air temperature are observed at only a few stations, and their occurrence times generally disagree.
Change-point analyses on hydrometeorological variables observed at meteorological and hydrologic stations in the upstream regions of the LMR and NSR. The stations that recorded an abrupt change are listed and the occurrence time of the change point is shown in parentheses.
2) Trends in annual and seasonal precipitation
Compared with the trends in annual air temperature, less consistent increasing trends in annual precipitation are observed over the investigated periods in the two upstream regions, especially in the upstream region of the LMR (Fig. 3d). However, positive trends in annual precipitation are still prevalent over the period 1980–2010, although only a few stations show statistically significant trends. Statistically significant positive trends and statistically insignificant negative trends are primarily seen in the upstream region of the LMR. Therefore, the median trend rate between 1980 and 2010 is 14.54 mm decade−1, ranging from −21.83 to 71.60 mm decade−1, and 13.65 mm decade−1, ranging from −1.12 to 37.00 mm decade−1, in the upstream regions of the LMR and NSR, respectively (Figs. 3d, 4b).
Unlike the trends in the seasonal air temperature, there are few stations with a consistent or persistent trend in seasonal precipitation (Fig. 6). In comparison, the trends in spring precipitation are more consistent and more persistent in the two upstream regions. Increasing spring precipitation trend between 1980 and 2010 occurs at all stations exclusive of the Weixi station, of which only a few stations have a statistically significant tendency (Fig. 6a). The change rate of spring precipitation is from −24.58 to 30.40 mm decade−1 (median = 8.31 mm decade−1) and 2.54 to 13.33 mm decade−1 (median = 8.75 mm decade−1) in the upstream regions of the LMR and NSR, respectively (Figs. 4b, 6a). Negative trends in precipitation spread over about half of the stations in the remaining three seasons (Figs. 6b–d). However, a few statistically significant trends are found. Positive but few statistically significant trends are still slightly dominant across the study area, especially for the period 1980–2010. Increasing precipitation trends between 1980 and 2010 predominate in summer for each of the two upstream regions, whereas they mainly occur in autumn for the upstream region of the LMR and in winter for the upstream region of the NSR. Only significant trends for summer precipitation emerge at the Mangkang and Deqin stations, with an upward rate of 41.86 and 26.67 mm decade−1, respectively, whereas a significant decrease trend of autumn precipitation occurs at the Zaduo station, with a downward rate of 11.17 mm decade−1 (Figs. 6b–d).
As in Fig. 5, but for seasonal precipitation totals.
Citation: Journal of Hydrometeorology 16, 5; 10.1175/JHM-D-14-0238.1
Very few stations are observed to abruptly change their trends of annual and seasonal precipitation in the upstream regions of the LMR and NSR (Table 1). Only the Zaduo station undergoes a sudden change in winter precipitation occurring at the early 1990s.
b. Streamflow trends and their climatic linkages
1) Streamflow variability and trends
As illustrated in Fig. 7, the annual streamflow at the Jiuzhou and Gongshan hydrologic stations is insignificantly increasing between the mid-1950s and 2010, with a rise rate of 18.22 and 23.21 m3 s−1 decade−1, respectively. A substantial increase in decadal streamflow has occurred at both stations since the 1970s. From the 1950s to 1970s, decadal streamflow first increased and then decreased.
Time series plots for annual and seasonal streamflow at the (a) Jiuzhou and (b) Gongshan hydrologic stations from the 1950s to 2010. Significant differences were accepted at a confidence level of 95%.
Citation: Journal of Hydrometeorology 16, 5; 10.1175/JHM-D-14-0238.1
Similar to the seasonal patterns and trends of precipitation, streamflow and its interannual variation are the largest in summer, the smallest in winter, and moderate in the other two seasons (Fig. 7). Increasing streamflow in all seasons of the year is seen at both stations, with a trend slope ranging from 3.44 to 38.87 m3 s−1 per decade. With the exception of a statistically significant upward trend of spring streamflow at the Jiuzhou hydrologic station, the other streamflow trends are of no significance. Larger positive autumn streamflow trends are observed at the Gongshan hydrologic station, whereas the magnitude of upward streamflow trends in the remaining seasons is roughly equivalent at both stations. At the decadal scale, the trends in seasonal streamflow are more variable (Fig. 7). The trends of decadal spring streamflow oscillate at both stations. The variations of the decadal summer and autumn streamflow are similar to that of the annual streamflow, with a consistent increasing trend since the 1970s. The trends of decadal winter streamflow are found first to decrease and then to increase, with the time of the turning point occurring in the 1970s and the 1980s at the Jiuzhou and Gongshan hydrologic stations, respectively.
Abrupt changes in the annual and winter streamflow at the Jiuzhou hydrologic station occurred in 1966 and 1967, respectively (Table 1). At the Gongshan hydrologic station, however, abrupt changes in the annual and autumn streamflow took place in 1998, whereas the change-point times of the spring, summer, and winter streamflow occurred inconsistently in 1995, 2000, and 1997, respectively.
2) Relation of the streamflow to the air temperature and precipitation
The relationships of streamflow to air temperature and precipitation are basin specific and season dependent (Table 2). Streamflow is fairly well correlated with air temperature and precipitation both for the annual and flood season, with values of percent variance explained ranging from 53% to 68%. At the upstream regions of both rivers, the relations of streamflow to air temperature and precipitation in the flood season are better than in the dry season. By contrast, the correlation of streamflow with air temperature and precipitation both for the annual and flood season are higher in the upstream region of the LMR than in that of the NSR. Compared among four seasons, a better correlation of streamflow with air temperature and precipitation is only found in summer (Table 2). Higher correlations between streamflow and both air temperature and precipitation are obtained in the upstream region of the LMR at all seasons exclusive of winter.
Statistics [percent variance explained (%)] on the correlation of streamflow to air temperature and precipitation at an annual scale in the upstream regions of the LMR and NSR by random forest regression. Negative percent variance represents a very poor fit.
To enhance the understanding of the relationships between changes in air temperature and precipitation and the related changes in streamflow, the relative contributions of the air temperature and precipitation to the streamflow both for the annual and flood season are illustrated in Fig. 8. Generally, precipitation has a higher relative importance than air temperature. In the upstream region of the LMR, changes in precipitation at the Deqin and Weixi stations make a small contribution to changes in streamflow, whereas changes in precipitation at the Changdu station have an important effect on streamflow (Fig. 8a). Changes in precipitation at the Suoxian and Basu stations are less important to streamflow from the upstream region of the NSR, whereas changes in precipitation at the Naqu and Dingqing stations have a relatively high importance on the streamflow. Air temperature in the flood season has less relative importance than annual air temperature in the two upstream regions (Fig. 8b).
Measure of the variable importance of the air temperature and precipitation at meteorological stations to streamflow at hydrologic stations for the (a) annual and (b) flood season. For each river upstream region, all meteorological stations were used to establish the random forest regression equation at an annual scale. IncNodePurity (increase in node purity) denotes the importance of the variable in decreasing node impurities from splitting on the variable.
Citation: Journal of Hydrometeorology 16, 5; 10.1175/JHM-D-14-0238.1
5. Discussion
a. Climate variability and related driving factors
As indicated by the preceding results, a persistent significant warming across the upstream regions of the LMR and NSR is confirmed by this study. The TFPW-MK test results show a rise in mean annual air temperature over the upstream regions of both rivers between the 1950s and 2010, especially for the period 1980–2010 with a median of up to 0.52°C decade−1 (Figs. 3, 4), which is higher than the previous estimation across the Qinghai–Tibet Plateau based on the data from the 1960s to mid-2000s (Guo and Wang 2012; Lan et al. 2010; Li et al. 2010; Wang et al. 2008; Zhang et al. 2011). This study observed an equivalent warming of the annual maximum air temperature and mean annual air temperature and a more rapid warming of the annual minimum air temperature. The median rise rate of the annual minimum air temperature reaches 0.83°C decade−1, approximately 1.6 times that of the mean annual air temperature. This substantial increase in annual minimum air temperature may be a vital attribute to the rise of the mean annual air temperature (Hu et al. 2012). This supports that the upstream regions of the rivers over the Qinghai–Tibet Plateau are some of the most sensitive areas to respond to global climate warming (Feng et al. 1998; Pan and Li 1996). Although the differences in the observed period among the stations have an important effect on the estimated warming rate, it is no doubt that significant air temperature warming has occurred extensively at the annual and seasonal scales since the 1980s or even earlier. It is worth noting that the temperature warming was exacerbating over the periods of study across the two upstream regions.
The climate warming trend persists throughout the annual cycle in the two upstream regions of both rivers. However, the warming trend rate is significantly different among the seasons. The strongest significant warming occurs in winter. Meanwhile, there is a prevalent warming in the other seasons. This result is similar to the previous findings over the Qinghai–Tibet Plateau (Fan and He 2012; Hu et al. 2012; Liu and Chen 2000; You et al. 2010; Zhu et al. 2001). The median warming rate of winter temperature between 1980 and 2010 for the upstream regions of the LMR and NSR is individually 0.68° and 0.76°C decade−1, which exceed greatly the previous estimation of the rise rate over the 1960s to early 2000s (Duan et al. 2006; You et al. 2010). These results reflect the fact that regional and seasonal differences in climate warming are very distinct in the study area and even across the whole Qinghai–Tibet Plateau. Although regional differences in the recent climatic warming over the Qinghai–Tibet Plateau have been attributed to elevation dependency in previous studies (Fan and He 2012; Liu and Chen 2000; Qin et al. 2009; Rangwala and Miller 2012; Zhu et al. 2001), the relevance of climate warming to elevation is not significant either at the annual or seasonal scale in this study (Figs. 1, 3, 6). It is plausible that other physical characteristics of the studied areas, except for elevation, also have strong impacts on climate warming (You et al. 2008, 2010). The contradictory results on elevation dependency of climate warming can be confounded by the effects of scales in spatial and temporal domains, especially in data-scarce alpine environments with sparse station coverage.
It is expected that climate warming will lead to an increase in precipitation and an intensification of the hydrological cycle (IPCC 2007; Mahlman 1997). With climate warming, prevailing positive trends in annual and seasonal precipitation occur in the upstream regions of both rivers (Figs. 3, 6). However, most of them are not statistically significant. This is supported by previous results (Cao et al. 2006; Ding and Zhang 2008; Cuo et al. 2012; Xu et al. 2008; Zhang et al. 2011), which also reported no significant upward trends in precipitation over the northern and southern Qinghai–Tibet Plateau. Furthermore, the direction and magnitude of the annual and seasonal precipitation trends varies greatly among stations and over different periods of study. Therefore, it is still unclear whether these trends are part of a long-term cycle of natural variability or part of a trend associated with climate warming due to the limited available historical climate records across the complex topography and wide range of elevations of the study area (Xu et al. 2008). Future work should differentiate a climate-driven trend from natural variability in the annual and seasonal precipitation.
Trends in meteorological and hydrologic time series can be regarded as a combination of gradual and abrupt changes. The results based on the BFAST method show that the abrupt changes in air temperature and precipitation occur at some stations throughout the 1960s to the 1990s and that there are few shared change-point times (Table 1). This reflects the fact that these abrupt changes mainly result from local physical conditions rather than large-scale atmospheric circulation (Cuo et al. 2012).
b. Climate–streamflow linkages
Changes in streamflow are often regarded as an important indicator for evaluating the impact of climate change on freshwater resources (Arnell and Gosling 2013; Novotny and Stefan 2007). With a warmer and wetter climate in the upstream regions of the LMR and NSR, an overall tendency toward increasing annual and seasonal streamflow is observed at their own outlets (Fig. 7), which is different from the upstream region of the Huanghe River, which exhibit a persistent decreasing streamflow (Cao et al. 2006; Hu et al. 2011; Lan et al. 2010) and an abrupt change in the mid-1990s (Zhang et al. 2011). This reflects significant regional differences in trends of streamflow from the rivers on the Qinghai–Tibet Plateau.
Abrupt changes in streamflow over the upstream regions of the LMR and NSR were detected to occur in the mid-1960s and the late 1990s, respectively, which are less consistent with the change-point times of air temperature and precipitation (Table 1). Because of relatively less anthropogenic disturbances in the upstream regions of both rivers, these changes could be primarily climate driven. The results from the random forest regression also show that changes in streamflow can be fairly well explained by changes in air temperature and precipitation both for the annual and flood season, especially for the upstream region of the LMR (Table 2). In the regression model, precipitation exhibits a higher relative importance, whereas air temperature is less important to the observed streamflow increase. It can be inferred that the streamflow is predominantly affected by precipitation both for the annual and flood season. However, the correlation of streamflow to air temperature and dry season precipitation is very poor in the two upstream regions. In addition, synchronous abrupt changes in meteorological variables, especially in precipitation, have not occurred (Table 1). This reflects complex relationships between streamflow and climate variables. Except for precipitation, snowpack and glacier melting are important sources of streams in the alpine environment. With climate warming, intensified glacier retreat and permafrost degradation are occurring in the Qinghai–Tibet Plateau and surrounding regions, which yields a net increase in streamflow (Qiu 2008; Yao et al. 2007). Meanwhile, evapotranspiration, an important component of the water and energy balance, significantly decreased over the Plateau (Xie and Zhu 2012; Zhang et al. 2007). Therefore, further investigation is urgently needed to obtain a better understanding of the linkages of climate warming to changes in snowpack, permafrost, glacier and evapotranspiration and the resulting impacts on streamflow in the upstream regions of large rivers on the Plateau.
6. Conclusions
Based on high-quality monthly air temperature and precipitation data from 16 meteorological stations in the upstream regions of the LMR and NSR between the 1950s to 2010, significant climate warming and its related prevalent positive precipitation trends are revealed both at the annual and seasonal scale. Increasing air temperature trends are widespread across the study area and are persistent or even accelerating over the periods of study. Among the three annual air temperature variables, the upward trend rate of the annual minimum air temperature from 1980 to 2010 reaches more than 0.83°C decade−1, approximately 1.6 times that of the other two. Climate warming exhibits distinct seasonal differences. The strongest temperature warming occurs in winter, and the second strongest occurs in autumn. Unlike air temperature, the widespread precipitation increase occurs in spring. The regional differences in climate variability are pronounced. There is a greater temperature increase in the upstream region of the LMR, especially in winter. Comparing the two upstream regions, a substantial increase in precipitation paralleling climate warming, especially in spring, was observed at the watershed scale. However, a consistent abrupt change in meteorological time series was not detected.
Prevailing but few significant increasing trends in streamflow are observed for the study area, which are mainly climate driven. Abrupt changes in streamflow occurred in the mid-1960s and the late 1990s at the outlets of the upstream regions of the LMR and NSR, respectively. Streamflow is fairly well related to precipitation both for the annual and flood season, whereas its linkage to air temperature is relatively minor. Because of the important impacts of climate warming on the other water balance components, including snowpack and glacier melting and evapotranspiration decline, the linkage of streamflow to air temperature is more complex. Further investigation into the mechanisms controlling the impacts of climate warming on streamflow is urgently needed.
The LMR and NSR are important transboundary waterways and ecological corridors across southwest China and Southeast Asia. They are also the unique arena for multilateral water cooperation and multicultural, transboundary water resources and ecosystem management because of the large-scale hydropower development plans proposed for their main stems. This study may be of practical and scientific importance for guiding cascade hydropower development and transboundary water resources management under different climate change scenarios.
Acknowledgments
This research work was financially supported by the National Science and Technology Support Program of China (2013BAB06B03), the National Natural Science Foundation of China (41461017, U1202232, 41061010), China Huaneng Group Science and Technology Program (HNKJ13-H17-03), Candidates of the Young and Middle Aged Academic Leaders of Yunnan Province (2014HB005), and Program for Excellent Young Talents of Yunnan University. We thank the editor and the three anonymous reviewers for their valuable comments on the manuscript.
REFERENCES
Adam, J. C., Haddeland I. , Su F. , and Lettenmaier D. P. , 2007: Simulation of reservoir influences on annual and seasonal streamflow changes for the Lena, Yenisei, and Ob’ rivers. J. Geophys. Res., 112, D24114, doi:10.1029/2007JD008525.
Arnell, N. W., and Gosling S. N. , 2013: The impacts of climate change on river flow regimes at the global scale. J. Hydrol., 486, 351–364, doi:10.1016/j.jhydrol.2013.02.010.
Barnett, T. P., Adam J. C. , and Lettenmaier D. P. , 2005: Potential impacts of a warming climate on water availability in snow-dominated regions. Nature, 438, 303–309, doi:10.1038/nature04141.
Breiman, L., 2001a: Random forests. Mach. Learn., 45, 5–32, doi:10.1023/A:1010933404324.
Breiman, L., 2001b: Statistical modeling: The two cultures (with comments and a rejoinder by the author). Stat. Sci., 16, 199–231, doi:10.1214/ss/1009213726.
Campbell, I., 2009: Introduction. The Mekong-Biophysical Environment of an International River Basin, I. C. Campbell, Ed., Academic Press, 1–11.
Cao, J. T., Qin D. H. , Kang E. S. , and Li Y. Y. , 2006: River discharge changes in the Qinghai-Tibet Plateau. Chin. Sci. Bull., 51, 594–600, doi:10.1007/s11434-006-0594-6.
Christensen, N. S., Wood A. W. , Voisin N. , Lettenmaier D. P. , and Palmer R. N. , 2004: The effects of climate change on the hydrology and water resources of the Colorado River basin. Climatic Change, 62, 337–363, doi:10.1023/B:CLIM.0000013684.13621.1f.
CNMIC, 2005: Monthly Surface Climate Dataset in China (version 3.0), continuing from July 2005 (updated monthly). Chinese National Meteorological Information Centre, accessed 20 October 2011. [Available online at http://data.cma.gov.cn/data/.]
Cuo, L., Zhang Y. , Wang Q. , Zhang L. , Zhou B. , Hao Z. , and Su F. , 2012: Climate change on the northern Tibetan Plateau during 1957–2009: Spatial patterns and possible mechanism. J. Climate, 26, 85–109, doi:10.1175/JCLI-D-11-00738.1.
Cuo, L., Zhang Y. , Zhu F. , and Liang L. , 2014: Characteristics and changes of streamflow on the Tibetan Plateau: A review. J. Hydrol.: Reg. Stud., 2, 49–68, doi:10.1016/j.ejrh.2014.08.004.
Ding, Y., and Zhang L. , 2008: Intercomparison of the time for climate abrupt change between the Tibetan Plateau and other regions in China (in Chinese). Chin. J. Atmos. Sci., 32, 794–805.
Duan, A., Wu G. , Zhang Q. , and Liu Y. , 2006: New proofs of the recent climate warming over the Tibetan Plateau as a result of the increasing greenhouse gases emissions. Chin. Sci. Bull., 51, 1396–1400, doi:10.1007/s11434-006-1396-6.
Elmore, K. L., Baldwin M. E. , and Schultz D. M. , 2006: Field significance revisited: Spatial bias errors in forecasts as applied to the Eta Model. Mon. Wea. Rev., 134, 519–531, doi:10.1175/MWR3077.1.
Fan, H., and He D. , 2012: Regional climate and its change in the Nujiang River basin (in Chinese). Acta Geogr. Sin., 67, 621–630.
Fan, H., He D. , and Wang H. , 2015: Environmental consequences of damming the mainstream Lancang–Mekong River: A review. Earth Sci. Rev., 146, 77–91, doi:10.1016/j.earscirev.2015.03.007.
Feng, S., Tang M. , and Wang D. , 1998: New evidence for the Qinghai-Xizang (Tibet) Plateau as a pilot region of climatic fluctuation in China. Chin. Sci. Bull., 43, 1745–1749, doi:10.1007/BF02883978.
Gautam, M. R., Timilsina G. R. , and Acharya K. , 2013: Climate change in the Himalayas: Current state of knowledge. Policy Research Working Paper 6516, World Bank, 64 pp. [Available online at http://documents.worldbank.org/curated/en/2013/06/17935389/climate-change-himalayas-current-state-knowledge.]
Groisman, P. Ya., Knight R. W. , Karl T. R. , Easterling D. R. , Sun B. , and Lawrimore J. H. , 2004: Contemporary changes of the hydrological cycle over the contiguous United States: Trends derived from in situ observations. J. Hydrometeor., 5, 64–85, doi:10.1175/1525-7541(2004)005<0064:CCOTHC>2.0.CO;2.
Guo, D., and Wang H. , 2012: The significant climate warming in the northern Tibetan Plateau and its possible causes. Int. J. Climatol., 32, 1775–1781, doi:10.1002/joc.2388.
Guo, J., 1985: Hydrography of Western Sichuan and Northern Yunnan (in Chinese). Science Press, 158 pp.
Hamed, K. H., 2009: Enhancing the effectiveness of prewhitening in trend analysis of hydrologic data. J. Hydrol., 368, 143–155, doi:10.1016/j.jhydrol.2009.01.040.
Hannah, D. M., Brown L. E. , Milner A. M. , Gurnell A. M. , McGregor G. R. , Petts G. E. , Smith B. P. G. , and Snook D. L. , 2007: Integrating climate–hydrology–ecology for alpine river systems. Aquatic Conserv: Mar. Freshwater Ecosyst., 17, 636–656, doi:10.1002/aqc.800.
Hu, Y., Maskey S. , Uhlenbrook S. , and Zhao H. , 2011: Streamflow trends and climate linkages in the source region of the Yellow River, China. Hydrol. Processes, 25, 3399–3411, doi:10.1002/hyp.8069.
Hu, Y., Maskey S. , and Uhlenbrook S. , 2012: Trends in temperature and rainfall extremes in the Yellow River source region, China. Climatic Change, 110, 403–429, doi:10.1007/s10584-011-0056-2.
Huntington, T. G., 2006: Evidence for intensification of the global water cycle: Review and synthesis. J. Hydrol., 319, 83–95, doi:10.1016/j.jhydrol.2005.07.003.
Immerzeel, W. W., van Beek L. P. H. , and Bierkens M. F. P. , 2010: Climate change will affect the Asian water towers. Science, 328, 1382–1385, doi:10.1126/science.1183188.
Immerzeel, W. W., van Beek L. P. H. , Konz M. , Shrestha A. , and Bierkens M. , 2012: Hydrological response to climate change in a glacierized catchment in the Himalayas. Climatic Change, 110, 721–736, doi:10.1007/s10584-011-0143-4.
IPCC, 2007: Climate Change 2007: The Physical Science Basis. Cambridge University Press, 996 pp.
Karl, T. R., and Riebsame W. E. , 1989: The impact of decadal fluctuations in mean precipitation and temperature on runoff: A sensitivity study over the United States. Climatic Change, 15, 423–447, doi:10.1007/BF00240466.
Labat, D., 2008: Wavelet analysis of the annual discharge records of the world’s largest rivers. Adv. Water Resour., 31, 109–117, doi:10.1016/j.advwatres.2007.07.004.
Labat, D., 2010: Cross wavelet analyses of annual continental freshwater discharge and selected climate indices. J. Hydrol., 385, 269–278, doi:10.1016/j.jhydrol.2010.02.029.
Labat, D., Goddéris Y. , Probst J. L. , and Guyot J. L. , 2004: Evidence for global runoff increase related to climate warming. Adv. Water Resour., 27, 631–642, doi:10.1016/j.advwatres.2004.02.020.
Labat, D., Goddéris Y. , Probst J. L. , and Guyot J. L. , 2005: Reply to comment of Legates et al. Adv. Water Resour., 28, 1316–1319, doi:10.1016/j.advwatres.2005.04.007.
Lai, Z., 1996: Impact of climate variation on the runoff of large rivers in the Tibetan Plateau (in Chinese). J. Glaciol. Geocryology, S1, 314–320.
Lan, Y., Zhao G. , Zhang Y. , Wen J. , Liu J. , and Hu X. , 2010: Response of runoff in the source region of the Yellow River to climate warming. Quat. Int., 226, 60–65, doi:10.1016/j.quaint.2010.03.006.
Li, L., Yang S. , Wang Z. , Zhu X. , and Tang H. , 2010: Evidence of warming and wetting climate over the Qinghai-Tibet Plateau. Arct. Antarct. Alp. Res., 42, 449–457, doi:10.1657/1938-4246-42.4.449.
Liaw, A., and Wiener M. , 2002: Classification and regression by randomForest. R News, Vol. 2/3, R Foundation, Vienna, Austria, 18–22. [Available online at https://cran.r-project.org/doc/Rnews/Rnews_2002-3.pdf.]
Liu, X., and Chen B. , 2000: Climatic warming in the Tibetan Plateau during recent decades. Int. J. Climatol., 20, 1729–1742, doi:10.1002/1097-0088(20001130)20:14<1729::AID-JOC556>3.0.CO;2-Y.
Livezey, R. E., and Chen W. Y. , 1983: Statistical field significance and its determination by Monte Carlo techniques. Mon. Wea. Rev., 111, 46–59, doi:10.1175/1520-0493(1983)111<0046:SFSAID>2.0.CO;2.
Mahlman, J. D., 1997: Uncertainties in projections of human-caused climate warming. Science, 278, 1416–1417, doi:10.1126/science.278.5342.1416.
Milliman, J. D., Farnsworth K. L. , Jones P. D. , Xu K. H. , and Smith L. C. , 2008: Climatic and anthropogenic factors affecting river discharge to the global ocean, 1951–2000. Global Planet. Change, 62, 187–194, doi:10.1016/j.gloplacha.2008.03.001.
Milly, P. C. D., Dunne K. A. , and Vecchia A. V. , 2005: Global pattern of trends in streamflow and water availability in a changing climate. Nature, 438, 347–350, doi:10.1038/nature04312.
Nijssen, B., O’Donnell G. , Hamlet A. , and Lettenmaier D. , 2001: Hydrologic sensitivity of global rivers to climate change. Climatic Change, 50, 143–175, doi:10.1023/A:1010616428763.
Nohara, D., Kitoh A. , Hosaka M. , and Oki T. , 2006: Impact of climate change on river discharge projected by multimodel ensemble. J. Hydrometeor., 7, 1076–1089, doi:10.1175/JHM531.1.
Novotny, E. V., and Stefan H. G. , 2007: Stream flow in Minnesota: Indicator of climate change. J. Hydrol., 334, 319–333, doi:10.1016/j.jhydrol.2006.10.011.
Oguntunde, P. G., Abiodun B. J. , and Lischeid G. , 2011: Rainfall trends in Nigeria, 1901–2000. J. Hydrol., 411, 207–218, doi:10.1016/j.jhydrol.2011.09.037.
Oki, T., and Kanae S. , 2006: Global hydrological cycles and world water resources. Science, 313, 1068–1072, doi:10.1126/science.1128845.
Pan, B., and Li J. , 1996: Qinghai-Tibetan Plateau: A driver and amplifier of the global climatic change—III. The effects of the uplift of Qinghai-Tibetan Plateau on climatic changes (in Chinese). J. Lanzhou Univ., 32, 108–115.
Prasad, A. M., Iverson L. R. , and Liaw A. , 2006: Newer classification and regression tree techniques: Bagging and random forests for ecological prediction. Ecosystems, 9, 181–199, doi:10.1007/s10021-005-0054-1.
Qin, J., Yang K. , Liang S. , and Guo X. , 2009: The altitudinal dependence of recent rapid warming over the Tibetan Plateau. Climatic Change, 97, 321–327, doi:10.1007/s10584-009-9733-9.
Qin, N., Chen X. , Fu G. , Zhai J. , and Xue X. , 2010: Precipitation and temperature trends for the Southwest China: 1960–2007. Hydrol. Processes, 24, 3733–3744, doi:10.1002/hyp.7792.
Qiu, J., 2008: China: The third pole. Nature, 454, 393–396, doi:10.1038/454393a.
Rangwala, I., and Miller J. , 2012: Climate change in mountains: A review of elevation-dependent warming and its possible causes. Climatic Change, 114, 527–547, doi:10.1007/s10584-012-0419-3.
Robinson, R. A. J., and Coauthors, 2007: The Irrawaddy River sediment flux to the Indian Ocean: The original nineteenth-century data revisited. J. Geol., 115, 629–640, doi:10.1086/521607.
Sen, P. K., 1968: Estimates of the regression coefficient based on Kendall’s Tau. J. Amer. Stat. Assoc., 63, 1379–1389, doi:10.1080/01621459.1968.10480934.
Theil, H., 1992: A rank-invariant method of linear and polynomial regression analysis. Henri Theil’s Contributions to Economics and Econometrics, B. Raj and J. Koerts, Eds., Springer, 345–381.
Verbesselt, J., Hyndman R. , Newnham G. , and Culvenor D. , 2010: Detecting trend and seasonal changes in satellite image time series. Remote Sens. Environ., 114, 106–115, doi:10.1016/j.rse.2009.08.014.
Wang, B., Bao Q. , Hoskins B. , Wu G. , and Liu Y. , 2008: Tibetan Plateau warming and precipitation changes in East Asia. Geophys. Res. Lett., 35, L14702, doi:10.1029/2008GL034330.
Xie, H., and Zhu X. , 2012: Reference evapotranspiration trends and their sensitivity to climatic change on the Tibetan Plateau (1970–2009). Hydrol. Processes, 27, 3685–3693, doi:10.1002/hyp.9487.
Xu, K., Milliman J. D. , and Xu H. , 2010: Temporal trend of precipitation and runoff in major Chinese Rivers since 1951. Global Planet. Change, 73, 219–232, doi:10.1016/j.gloplacha.2010.07.002.
Xu, Z. X., Gong T. L. , and Li J. Y. , 2008: Decadal trend of climate in the Tibetan Plateau—Regional temperature and precipitation. Hydrol. Processes, 22, 3056–3065, doi:10.1002/hyp.6892.
Yao, T., Pu J. , Lu A. , Wang Y. , and Yu W. , 2007: Recent glacial retreat and its impact on hydrological processes on the Tibetan Plateau, China, and surrounding regions. Arct. Antarct. Alp. Res., 39, 642–650, doi:10.1657/1523-0430(07-510)[YAO]2.0.CO;2.
Yao, Z., Duan R. , and Liu Z. , 2012: Changes in precipitation and air temperature and its impacts on runoff in the Nujiang River basin (in Chinese). Resour. Sci., 34, 202–210.
You, Q., Kang S. , Pepin N. , and Yan Y. , 2008: Relationship between trends in temperature extremes and elevation in the eastern and central Tibetan Plateau, 1961–2005. Geophys. Res. Lett., 35, L04704, doi:10.1029/2007GL032669.
You, Q., Kang S. , Pepin N. , Flügel W.-A. , Yan Y. , Behrawan H. , and Huang J. , 2010: Relationship between temperature trend magnitude, elevation and mean temperature in the Tibetan Plateau from homogenized surface stations and reanalysis data. Global Planet. Change, 71, 124–133, doi:10.1016/j.gloplacha.2010.01.020.
Yue, S., Pilon P. , Phinney B. , and Cavadias G. , 2002: The influence of autocorrelation on the ability to detect trend in hydrological series. Hydrol. Processes, 16, 1807–1829, doi:10.1002/hyp.1095.
Zhang, L., Su F. , Yang D. , Hao Z. , and Tong K. , 2013: Discharge regime and simulation for the upstream of major rivers over Tibetan Plateau. J. Geophys. Res. Atmos., 118, 8500–8518, doi:10.1002/jgrd.50665.
Zhang, S. F., Hua D. , Meng X. J. , and Zhang Y. Y. , 2011: Climate change and its driving effect on the runoff in the “Three-River Headwaters” region. J. Geogr. Sci., 21, 963–978, doi:10.1007/s11442-011-0893-y.
Zhang, Y.-Q., Liu C. , Tang Y. , and Yang Y. , 2007: Trends in pan evaporation and reference and actual evapotranspiration across the Tibetan Plateau. J. Geophys. Res., 112, D12110, doi:10.1029/2006JD008161.
Zhang, Y.-Y., Zhang S. F. , Zhai X. Y. , and Xia J. , 2012: Runoff variation and its response to climate change in the Three Rivers source region. J. Geogr. Sci., 22, 781–794, doi:10.1007/s11442-012-0963-9.
Zhang, Y.-Y., Zhang S. F. , Xia J. , and Hua D. , 2013: Temporal and spatial variation of the main water balance components in the three rivers source region, China from 1960 to 2000. Environ. Earth Sci., 68, 973–983, doi:10.1007/s12665-012-1800-2.
Zhu, W., Chen L. , and Zhou Z. , 2001: Several characteristics of contemporary climate change in the Tibetan Plateau. Sci. China, 44, 410–420, doi:10.1007/BF02912013.
