Climate change in the Tibetan Plateau Three Rivers Source Region: 1960–2009

Authors

  • LiQiao Liang,

    Corresponding author
    1. Key Laboratory of Tibetan Environment Changes and Land Surface Processes, Institute of Tibetan Plateau Research, Beijing, China
    2. Institute of Geographic Sciences and Natural Resources Research, Beijing, China
    • Correspondence to: LiQiao Liang, Key Laboratory of Tibetan Environment Changes and Land Surface Processes, Institute of Tibetan Plateau Research, Beijing, China. E-mail: liangliqiao@itpcas.ac.cn

    Search for more papers by this author
  • LiJuan Li,

    1. Institute of Geographic Sciences and Natural Resources Research, Beijing, China
    Search for more papers by this author
  • ChangMing Liu,

    1. Institute of Geographic Sciences and Natural Resources Research, Beijing, China
    Search for more papers by this author
  • Lan Cuo

    1. Key Laboratory of Tibetan Environment Changes and Land Surface Processes, Institute of Tibetan Plateau Research, Beijing, China
    Search for more papers by this author

ABSTRACT

Changes observed in nine meteorological variables obtained from the Three Rivers Source Region (TRSR) between 1960 and 2009 were investigated using a fitted linear model, Mann–Kendall test, moving t-test, and Morlet wavelet. Analysis of the regionally scaled annual series from 1960 to 2009 showed that minimum (Tmin), maximum (Tmax), mean (Tmean) air temperature, precipitation (P), potential evaporation (Ep), and sunshine hours (SH) increased while relative humidity (RH) and wind speed (W) decreased. Trends were significant at a 99% confidence level for air temperature and at a 95% confidence level for Ep and W. With the exception of SH, regional scale changes observed in all variables in the past decade (2000–2009) when compared to climate norms (means of all climatic variables) from 1961 to 1990 were consistent with their corresponding linear trends from 1960 to 2009. Tmax, Tmean, and drought index (DI) exhibited one climate jump, Tmin and RH two, and SH, Ep, and W three at a significance level of α = 0.05. On a regional scale, the period from 1986 to 1997 experienced a warmer, drier climate due to higher than average air temperatures, lower P, and higher DI compared to means from 1960 to 2009. The majority of meteorological variables of the TRSR experienced significant (α = 0.05) short periodical cycling between 2 and 5 years. In terms of spatial distribution, seven out of 12 meteorological stations underwent warmer and wetter periods from 1960 to 2009, whereas the other five situated in the southeastern section of the TRSR underwent warmer, drier periods.

1. Introduction

Climate change is virtually irreversible, a manifestation of multiple factors (da Silva, 2004), such as rising temperatures (Ding et al., 2006; IPCC, 2007) as well as changes in precipitation (Liu et al., 2008, Liang et al., 2011), reference crop evapotranspiration (Liang et al., 2010), wind speed (McVicar et al., 2008), drought events (Potop et al., 2008), and solar radiation (Wild, 2009) to name a few examples. Observations have unearthed climate change impacts on environments and natural resources. Some examples are the expansion and increasing number of glacial lakes, a trend in decreasing snow coverage, shorter lake and river ice freezing periods, glacial ablation, and increasing surface instability of permafrost as well as its declining area (IPCC, 2007). Climate change has also caused ecosystem degradation (Lian and Shu, 2009), loss in biodiversity that only time can recover (Klein et al., 2004, Lenoir et al., 2008), and observable changes to plant species (Lenoir et al., 2008; Wang et al., 2011).

Climate systems have changed on both regional and global scales over the past century, and such changes are expected to continue into the near future and beyond (IPCC, 2007). Regionally, there have been numerous reports of climate change in various parts of the world, such as China (Ding et al., 2006), Philippines (Jose et al., 1996), Thailand (Limsakul and Goes 2008), Korea (Ko et al., 2010), Australia (Roderick and Farquhar, 2004), Nigeria (Akinbode et al., 2008), Jordan (Abu-Taleb et al., 2007), Brazil (da Silva, 2004), France (Chaouche et al., 2010), and North America (Hughes and Diaz, 2008).

Like these other regions, the Tibetan Plateau has experienced the effects of climate change. Studies have shown that the region is particularly sensitive to climate change impacts (Zhang, 2005). Climate change in the Tibetan Plateau can induce important feedbacks because the plateau, being higher than the surrounding land mass, affects atmospheric circulation and, thus, climate systems in Asia as well as in the Northern Hemisphere through mechanical and thermodynamic mechanisms (Yanai et al., 1992; Lian and Shu, 2009; Boos and Kuang, 2010). The Three Rivers Source Region (TRSR) is located in the plateau interior. Up to now, climate change studies in the region have mostly focused on air temperature and precipitation (Li et al., 2004; Li et al., 2006; Yang, 2006; Li et al., 2007; Gao and Liu, 2008; Liu et al., 2009), which hardly represents a comprehensive picture of climate change in the region.

Climate change has and will continue to have an effect on ecological balances and ecosystem functions and threatens the ecological security of the TRSR as well as ecosystems situated downstream from it. The fragile and unique alpine ecosystem regulates available water resources, benefitting local and downstream populations. In order to better understand local ecosystem response to global climate change and to mitigate and adapt to the local climate and ecosystem changes occurring within the TRSR, it is important to understand the full spectrum of climate changes taken place, such as trends, variability, and the occurrences of abrupt changes and the periodicity of changes within the region from an ecosystem perspective, which has yet to be investigated. The objective of this study was to reveal detailed factors related to TRSR climate change from an ecosystem perspective by examining the full spectrum of changes occurring. Seven observed and two calculated meteorological variables relevant to TRSR ecosystem functions were examined. These variables include the monthly maximum (Tmax), minimum (Tmin), and mean (Tmean) air temperatures, wind speed (W), relative humidity (RH), sunshine hours (SH), precipitation (P), calculated potential evaporation (Ep), and the drought index (DI). Records from 1960 to 2009 were selected for this study, which is the longest timeframe that has been applied thus far to the region.

2. Study area and data

2.1. Study area

Being the source region of the Yangtze, Yellow, and Mekong rivers, the TRSR (lat 31°39′ ∼ 36°16′ N, long 89°24′ ∼ 102°23′ E) is located on the interior of the Tibetan Plateau, covering an area of 3.61 × 105 km2 (Figure 1). The source areas of the Yangtze, Yellow, and Mekong rivers account for 44, 46, and 10% of the TRSR, respectively. Average elevation is approximately 5000 m AMSL (above mean sea level), ranging between 3335 and 6624 m AMSL. Although it is situated in a sub-frigid and semiarid region of the Tibetan Plateau (Lin and Wu, 1981), the TRSR has an overall low annual air temperature range, a high daytime air temperature range, and receives approximately 470 mm in annual P. Roughly 86% of total P falls during the rainy season (May–September).

Figure 1.

The location of the TRSR (black shading) in China (light shading). The blowup map shows elevation, river basins, and the locations and names of meteorological stations situated within the study area. YRSR, source region of Yellow River; YTRSR, source region of Yangtze River; MRSR, source region of Mekong River.

Although sparsely populated (1.8 people/km2, Qinghai Bureau of Statistics, 2009) due to the challenges of the harsh environment, the TRSR is of great importance in terms of fresh water resources and its unique ecosystems. It provides greater than one-third (34.5%) of the annual Yellow River streamflow, approximately 7.2% annual Mekong River streamflow, and 1.2% annual Yangtze River streamflow (calculated using streamflow data from 1956 to 2005). The TRSR also contains lakes (5.1 × 103) and wetlands (7.33 × 104 km2) (Xu, 2007). The fragile and unique ecosystem is highly sensitive and vulnerable to climate change (Lian and Shu, 2009), and studies have shown that ecosystem degradation has occurred on the plateau due to climate change (Xiang et al., 2009; Li et al., 2010; Xiao et al., 2010). To preserve TRSR ecosystems, the Qinghai government implemented reforestation programmes starting in the late 1990s and established a nature reserve in 2000. The nature reserve, accounting for 42% of the total area, is rich in ecotypes and wetlands (Zheng and Cai, 2005).

2.2. Data

Observations including Tmin and Tmax at 2 m above the surface, mean W at 10 m, RH, SH, and P were obtained from the China Meteorological Administration (CMA). P was obtained from the monthly accumulation of daily values while the remaining five variables were obtained from the monthly means of daily values. All the above including calculated Ep, DI, and Tmean taken from the 12 meteorological stations situated within the study area were used to investigate TRSR climate change (Table 1). The longest available records were analyzed at individual stations, whereas time series from 1960 to 2009 were used on a regional scale. Regional climate data were derived by simply averaging the data from all 12 meteorological stations (locations of each station are provided for in Figure 1). All annual and seasonal time series were tested using one-sample Kolmogorov–Smirnov [Statistical Product and Service Solutions (SPSS) version 13.0], applying the null hypothesis of no difference between sample probability distributions and normal distribution. Null hypotheses for all time series were not rejected (α > 0.05). As a result, some of the following analyses applying normal distribution were deemed feasible.

Table 1. TRSR meteorological stations and their records and attributes, including the land cover type of each station
NameNumberBasinLongitude (°E)Latitude (°N)Altitude (m)Period of recordLand cover types
  1. Land cover information is provided by the University of MaryLand (http://modis.umiacs.umd.edu/data/landcover/data.shtml).

Maduo56 033Yellow River98.2234.9242721953–2009Open shrubland, grassland
Dari56 046Yellow River99.6533.7539681956–2009Grassland
Jiuzhi56 067Yellow River101.4833.4336291958–2009Wooded grass
Xinghai52 943Yellow River99.9835.5833231960–2009Shrubland
Banma56 151Yangtze River100.7532.9335301960–2009Grassland, wooded grassland, mixed forest
Tuotuohe56 004Yangtze River92.4334.2245331956–2009Open shrubland
Wudaoliang52 908Yangtze River93.0835.2246121956–2009Open shrubland
Yushu56 029Yangtze River97.0233.0236811951–2009Grassland
Qumalai56 021Yangtze River95.7834.1341751956–2009Grassland
Qingshuihe56 034Yangtze River97.1333.8044151956–2009Grassland
Nangqian56 125Mekong River96.4832.2036441956–2009Grassland
Zaduo56 018Mekong River95.3032.9040661956–2009Grassland

3. Methods

3.1. Calculating potential evaporation

Many equations have been developed to estimate potential evaporation. These include the aerodynamic approach that incorporates air humidity and wind speed (Baldocchi et al., 1996); temperature based equations (Thornthwaite, 1948; Hargreaves and Samani, 1985); radiation based equations that incorporate solar radiation and temperature (Doorenboos and Pruitt, 1977); and an equation that combines aerodynamic and radiation approaches together (Penman, 1948; Bormann, 2011). The Penman equation (Shuttleworth, 1993) is reported to be able to better describe evaporation dynamics within a changing climate (Donohue et al. 2010). The equation is described as follows (Shuttleworth, 1993):

display math(1)

where Ep is potential evaporation, mm day−1; Rn is net radiation at the surface, mm d−1; Δ is the slope of the saturated vapour pressure curve, kPa °C−1; γ is the psychrometric constant, which varies from 0.038 to 0.045 (corresponding to altitude); u2 is wind speed measured at 2 m, m s−1; D is the vapour pressure deficit es-e, kPa (here es is the saturated vapour pressure, and e is the actual vapour pressure which is calculated by e = RH/100es); and λ is the latent heat of vapourization, 2.5 MJ kg−1.

W at a height of 2 m was converted from measurements at 10 m above the surface based on the logarithmic wind speed profile equation provided by the FAO Penman–Monteith equation (Allen et al., 1998):

display math(2)

where z is the height of the measurement above the surface, m; uz is the measured wind speed at z, m above the surface, m s−1.

Net shortwave radiation (Sn) can be estimated by the following equation (Shuttleworth, 1993):

display math(3)

where α is albedo (Table 2) obtained from the MODerate Resolution Imaging Spectroradiometer (MODIS) albedo product (MCD43B3) (2001–2010), mean monthly albedo was averaged from daily albedo at each month for the period 2001–2010; n is bright sunshine hours per day, h; N is the total day length dependent on latitude and month, h; Ra is the extraterrestrial radiation dependent on latitude and month, MJ · m−2 d−1; as is the extraterrestrial radiation fraction on cloudy days (n = 0); and as + bs is the extraterrestrial radiation fraction on clear days. Depending on atmospheric conditions (humidity and dust) and solar declination (latitude and month), values of as and bs will vary. When no data on observable solar radiation are available and as and bs are not calibrated, as = 0.25 and bs = 0.5 are selected as conditions of averaged climate. For this study, validated values of 0.22 and 0.55 from Chen et al. (2004) were used for as and bs, respectively. Owing to the strong link between energy and evaporation, Rn can be expressed as an equivalent depth of evaporated water (in millimeters) by dividing Rn by ρwλ. Here, ρw (103 kg m−3) is the density of water. The computation of Ep follows the recommendations in Chapter 4 of the Handbook of Hydrology by Shuttleworth (1993).

Table 2. Mean monthly albedo values of each meteorological station
StationJanuaryFebruaryMarchAprilMayJuneJulyAugustSeptemberOctoberNovemberDecember
Maduo0.250.260.240.230.220.210.210.210.200.210.260.23
Dari0.230.230.200.190.190.190.200.190.180.190.200.19
Jiuzhi0.210.210.180.180.170.180.180.180.170.160.180.18
Xinghai0.240.230.230.230.210.200.190.180.180.190.200.23
Banma0.160.180.160.160.160.160.160.160.150.140.160.16
Tuotuohe0.230.240.240.230.220.200.190.190.190.210.240.23
Wudaoliang0.230.260.250.240.230.200.190.190.200.290.260.23
Yushu0.190.190.190.190.180.180.170.180.170.170.180.19
Qumalai0.220.240.220.220.200.190.190.190.190.200.230.21
Qingshuihe0.290.270.280.220.200.180.190.190.190.250.330.31
Nangqian0.180.180.190.190.180.170.170.170.160.160.160.17
Zaduo0.200.200.200.190.190.180.180.180.170.170.180.18

3.2. Drought index

DI is a numerical indicator representing the degree of dryness of the climate at a given location. P during the growing period (Ren and Shi, 1995) is often used to assess drought, but this method does not take into account water consumption. Methods applying air temperature, accumulated air temperature, and P (de Martonne, 1926; Selianinov, 1928; Kira, 1945; Ped, 1975) identify the significance of temperatures and yields more reliably and provide better estimations compared to using P alone in colder regions. However, this method cannot reflect water consumption accurately. Moreover, methods applying net solar radiation, P, and latent heat produced during evaporation (Budyko, 1974) also cannot accurately reflect water consumption because they ignore changes in sensible heat. Methods using P, evapotranspiration, and soil moisture (Palmer, 1965) can reflect water consumption more accurately than the methods mentioned above, but the availability of accurate measurements of soil moisture is difficult to attain in the Tibetan Plateau. The ratio of reference evapotranspiration obtained from relative humidity to that of dew point temperature (Allen et al., 1998; da Silva, 2004) reflects dryness. However, this method emphasizes the difference between Tmin and the dew point temperature without considering water supply. The ratio of Ep to P (UNEP, 1992; Arora, 2002; Bannayan et al., 2010) can be used to represent drought by estimating how much evaporative demand can be met by P. This method is widely used and is therefore applied here to calculate DI:

display math(4)

Ep is estimated by the Penman method (eq.1). DI has the same spatial and temporal scale as Ep and P, i.e. seasonal/annual DI is derived from seasonal/annual Ep and P, and regional DI is calculated from regional Ep and P.

3.3. Time series analysis methods

A linear fitted model is used to test against the null hypothesis slope by means of a two-tailed t-test with a confidence level of 95 or 99%. It is a common method used for statistical diagnosis in modern climatic analysis studies (Donohue et al., 2010; Liang et al., 2010; Liu et al., 2010a).

In order to explore the periodicity of TRSR meteorological variables, the complex Morlet wavelet method was used owing to its reliability in detecting signal periodicity (Rigozo et al., 2002; Liang et al., 2010, 2011). The Morlet wavelet developed by Torrence and Compo (1998) was used due to its widespread usage (Rigozo et al., 2002; EL-Askary et al., 2004; Liang et al., 2010; for more details see Liang et al., 2011). The Matlab wavelet package developed by Torrence and Compo was used (available online at http://paos.colorado.edu/research/wavelets/).

Climate jumps occur when a change in climate condition exceeds a certain threshold, triggering a transition to a new state at a rate determined by the climate change occurrence. These new states will arise unexpectedly, and human or natural systems may have difficulty adapting to them. Changes in any climate or variability measurement associated with a particular climate jump can be abrupt (Committee on Abrupt Climate Change, National Research Council, 2002). For this study, Mann–Kendall (Mann, 1945; Kendall, 1948; Sneyers, 1963; Goossens and Berger, 1986; Liu et al., 2008; Liang et al., 2010) and moving t-tests (Wei, 1999; Xiang and Chen, 2006; Zhao and Xu, 2006; Lei et al., 2007; Li and Jiang, 2007; Liang et al., 2010, 2011) were used to detect climate jumps for annual time series on a regional scale (for more details see Liang et al., 2010, 2011). These two approaches complement each other and thus provide a comprehensive climate jump analysis. The magnitude of climate jumps were calculated by subtracting prior jump time series averages from post jump time series averages. The proportion of the difference relative to prior jump averages was also calculated to describe climate jump magnitudes.

4. Results

4.1. Meteorological variable linear trends

Means, standard deviations, and trends in annual Tmin, Tmax, Tmean, RH, SH, W, Ep, P, and DI are listed in Table 3. Seasonal statistics of meteorological variables are provided for in Table 4. Figure 2 illustrates trends in meteorological variables from individual meteorological stations.

Table 3. Statistical information of annual meteorological time series at individual stations as well as the entire TRSR from 1960 to 2009
StationTminTmaxTmean
μ (°C)σ (°C)Trend (°C a−1)μ (°C)σ (°C)Trend (°C a−1)μ (°C)σ (°C)Trend (°C a−1)
Maduo−9.71.000.037**3.80.880.014−3.70.840.027**
Dari−7.00.890.040**6.80.780.023**−0.90.720.029**
Jiuzhi−5.80.940.051**9.30.720.010*0.60.700.037**
Xinghai−5.80.850.046**9.70.670.020**1.40.650.033**
Banma−3.60.570.024**11.50.710.027**2.80.570.025**
Tuotuohe−11.11.040.030**4.40.880.019**−4.00.850.025**
Wudaoliang−11.40.890.041**2.30.730.023**−5.40.700.028**
Yushu−3.20.960.042**11.80.760.024**3.20.810.033**
Qumalai−8.80.800.024**5.70.870.035**−2.20.770.034**
Qingshuihe−11.40.830.026**3.90.860.024**−4.60.780.025**
Nangqian−2.20.780.037**12.50.760.029**4.20.670.031**
Zaduo−5.40.820.040**8.40.760.020**0.60.760.031**
TRSR−7.10.760.039**7.50.670.025**−0.60.670.032**
StationPE pDI
μ (mm a−1)σ (mm a−1)Trend (mm a−2)μ (mm a−1)σ (mm a−1)Trend (mm a−2)μσTrend (a−1)
Maduo318660.8501013491.802**3.30.70−0.004
Dari548710.6261015411.469**1.90.270.001
Jiuzhi74898−1.1651007431.846**1.40.190.005**
Xinghai362711.2941131490.1363.30.75−0.011
Banma65774−0.105128332−0.0612.00.260.001
Tuotuohe286690.8311112551.377**4.11.13−0.002
Wudaoliang284551.292**999541.623**3.70.89−0.008
Yushu481690.1321147451.575**2.40.430.004
Qumalai405651.0321069450.4112.70.49−0.006
Qingshuihe510630.081908370.855**1.80.250.002
Nangqian532841.0741243490.4022.40.44−0.004
Zaduo529731.0041104460.1842.10.34−0.004
TRSR473440.5341089350.918**2.30.240.000
StationRHSHW
μ (%)σ (%)Trend (% a−1)μ (h d−1)σ (h d−1)Trend (h a−2)μ (m s−1)σ (m s−1)Trend (m s−1 a−1)
  1. μ, mean; σ, standard deviation. Tmin, annual mean of monthly minimum temperature; Tmax, annual mean of monthly maximum temperature; Tmean, annual mean of monthly mean temperature; P, annual precipitation; Ep, annual potential evapotranspiraiton; DI, annual drought index; RH, annual mean of monthly relative humidity; SH, annual mean of monthly sun shine duration; W, annual mean of monthly wind speed.

  2. *Denotes trends statistically significant at α = 0.05.

  3. **Denotes trends statistically significant at α = 0.01.

Maduo583.8−0.059*27751525.628**3.10.360.002
Dari612.5−0.047*24381252.962**2.10.27−0.002
Jiuzhi662.1−0.057**23171422.4032.00.210.001
Xinghai503.30.0182721122−3.806**2.30.350.000
Banma602.40.00123421190.1221.40.43−0.021**
Tuotuohe533.80.01829071354.080**4.00.540.002
Wudaoliang573.4−0.063*27761523.498**4.30.580.004
Yushu553.6−0.096**24611041.783*1.10.20−0.005**
Qumalai542.80.01127051291.5752.60.55−0.014**
Qingshuihe673.4−0.03125771391.3082.70.31−0.002
Nangqian532.40.01225781441.0571.60.31−0.012**
Zaduo563.60.0172429138−1.2281.90.33−0.012**
TRSR582.1−0.0262594901.1252.40.29−0.007*
Table 4. Precipitation, potential evaporation, and drought index statistical information during dry and wet seasons at each meteorological station as well as the entire TRSR from 1960 to 2009
Dry season StationPEpDI
μ (mm season−1)σ (mm season−1)Trend (mm season−1 a−1)μ (mm season−1)σ (mm season−1)Trend (mm season−1 a−1)μσTrend (a−1)
Maduo50190.193385301.005**8.93.44−0.032
Dari89230.411*414240.709**5.01.49−0.017
Jiuzhi140380.590444240.876**3.51.13−0.008
Xinghai36180.179480300.34118.615.55−0.159
Banma119310.44457918−0.1395.21.61−0.025
Tuotuohe22160.104437350.795**31.331.89−0.098
Wudaoliang23100.094391320.967**21.612.03−0.042
Yushu63240.451490260.687**8.93.59−0.046
Qumalai50160.048427290.504*9.73.96−0.007
Qingshuihe83220.218338240.479*4.41.48−0.003
Nangqian62240.509*552270.20710.64.66−0.103*
Zaduo74270.28646331−0.0067.12.73−0.039
TRSR68160.367*451220.499*6.91.84−0.032
Wet season StationPE pDI
μ (mm season−1)σ (mm season−1)Trend (mm season−1 a−1)μ (mm season−1)σ (mm season−1)Trend (mm season−1 a−1)μσTrend (a−1)
  1. μ, mean; σ, standard deviation.

  2. *Denotes trends statistically significant at α = 0.05.

  3. **Denotes trends statistically significant at α = 0.01.

Maduo269590.663628300.794**2.50.55−0.003
Dari458680.228601240.764**1.30.210.001
Jiuzhi60997−1.758*563260.973**0.90.160.005**
Xinghai326671.11265131−0.2092.10.52−0.008
Banma53870−0.558704240.0741.30.210.002
Tuotuohe264630.717675330.587*2.70.77−0.002
Wudaoliang261521.199**608340.661*2.40.64−0.007
Yushu41866−0.322657280.884**1.60.310.004
Qumalai356640.98464328−0.0871.90.42−0.006
Qingshuihe42862−0.142570240.3741.40.230.002
Nangqian470870.577691360.1951.50.34−0.002
Zaduo455710.718641270.1911.40.27−0.002
TRSR406410.165637220.4181.60.190.001
Figure 2.

Spatial distributions of temporal trends for each meteorological variable. A pentagon or a circle is placed around the dot if the trend is statistically significant at α = 0.05 or α = 0.01, respectively. Ellipses mark the southeastern section of the TRSR.

As shown in Table 3, mean annual TRSR air temperature was −0.6 °C. Half the meteorological stations measured Tmean below 0 °C. Regional mean temperature standard deviation was high (0.67 °C), indicating large interannual Tmean variation within the TRSR. Major increasing trends (α = 0.01) were found for Tmin, Tmax, and Tmean at most meteorological stations except for Tmax at both Jiuzhi (α = 0.05) and Maduo (insignificant). Trends in Tmin, Tmax, and Tmean for the entire TRSR were 0.039 °C a−1, 0.025 °C a−1, and 0.032 °C a−1, respectively. Trends in Tmin were higher than those for Tmax for all meteorological stations with the exception of Banma and Qumalai, implying that, in general, increases in Tmin contributed more to increases in Tmean in the TRSR. In terms of spatial distribution, the rate of warming was high in the eastern and southern sections of the TRSR (Figure 2(a)–(c)).

P decreased from the southeast to the northwest due to the soaring mountains that blocked moisture entering from the Bay of Bengal and the western Pacific, making it difficult for moisture to move into the northwestern section of the TRSR (Dettman et al., 2003; Yu et al., 2009). Mean annual P varied at individual meteorological stations within the TRSR (Table 3), decreasing from the eastern section to the north and west. The proportion of seasonal P (rainy season) in annual P exceeded 80% for all meteorological stations as a result of the prevailing monsoonal climate of the region, and reached 90% at three meteorological stations situated in the northern and western sections of the TRSR where annual precipitation is low. As shown in Tables 3 and 4, annual dry and rainy season P showed increasing trends at most meteorological stations, with only a few meteorological stations showing significant trends (α = 0.05 or α = 0.01). Trends in annual and rainy season P decreased from the northwest to the southeast of the TRSR (Figure 2(d) and (f)). This trend was the reverse for dry season P (Figure 2(e)). Trends in dry season P influenced annual P to a lesser extent due to the low proportion of dry season P in annual P. Annual P increased insignificantly (0.534 mm a−2) across the TRSR, which showed an approximate 26.7 mm a−1 increase from 1960 to 2009.

Variations in annual, dry, and rainy season Ep among the meteorological stations were lower compared to P (Tables 3 and 4). At most meteorological stations, annual, dry, and rainy season Ep showed increasing trends, being statistically significant at a 99 or 95% confidence level (Tables 3 and 4). An insignificant trend of 0.918 mm a−2 was determined for the TRSR, corresponding to the increase in Ep (45.9 mm a−1) that took place between 1960 and 2009. Spatial patterns of trends for annual, dry, and rainy season Ep varied greatly (Figures 2(g)–(i)). The meteorological stations that showed significant positive trends were located in the source regions of the Yellow and Yangtze rivers.

Mean annual DI was 2.3 for the TRSR, ranging from 1.4 to 4.1 between meteorological stations (Table 3). According to the climate classification report provided by Meng et al. (2004), 8 out of 12 meteorological stations exhibited semiarid climate since mean annual DI falls in the range of 1.7–3.0 (mean annual DI varies from 0.5 to 42.8 in China). The highest DI detected was mainly distributed within the northern section of the TRSR. Although trends during the rainy season and for the annual series on a regional scale were low (Tables 3 and 4), individual meteorological stations did exhibit variable characteristics. Seven meteorological stations showed weak decreasing trends, while the remaining five showed small increasing trends for both the annual series and the rainy season, where Jiuzhi (annual series) exhibited a low but significant increasing trend (α = 0.01) (Figure 2(j) and (k)). During the dry season, all meteorological stations showed decreasing trends, varying in spatial distribution (Figure 2(l)). Taking the entire TRSR into account, dry season DI decreased considerably at −0.032 a−1.

Annual RH appeared to have a high spatial variability. Annual mean RH decreased at six meteorological stations, out of which five showed significant decreasing trends at α = 0.01 or α = 0.05 (Table 4). Decreasing trends mainly appeared in the central and eastern sections of the TRSR (Figure 2(m)).

SH increased at all meteorological stations with the exception of Xinghai and Zaduo, the trend varying from −3.806 to 5.628 h a−2 (Table 3). Significant changes (α = 0.01 or α = 0.05) were detected at six meteorological stations, where increasing trends were measured at five out of the six. Large trends, both increasing and decreasing, were detected in the source region of the Yellow River and the western section of the Yangtze River source region (Figure 2(n)).

Mean W varied at individual meteorological stations, largely due to location, so dose W trend (varying from −0.021 m s−1 a−1 to 0.004 m s−1 a−1; Table 3). Statistically significant trends (α = 0.01) at five meteorological stations located in southern TRSR were all negative (Figure 2(o)).

In general, Tmin, Tmax, Tmean, P, Ep, and SH increased while RH and W decreased from 1960 to 2009. Trends were regionally significant at α = 0.01 for air temperatures, and at α = 0.05 for Ep and W. Significant trends (α = 0.05 or α = 0.01) in all nine variables appeared at least once at each meteorological station. In terms of spatial distribution, no uniform pattern was found for trends in all nine variables. The majority of meteorological stations showed significant positive temperature trends, whereas few stations showed significant trends for P and DI. In southeastern TRSR outlined by ellipses in Figure 2, annual P decreased at two meteorological stations and increased slightly at the other three (Figure 2(d)). Annual air temperature and DI increased (Figure 2(c) and (j)), resulting in a drier and warmer climate in the region throughout the study period (1960–2009). On the other hand, the other seven stations experienced warmer and wetter conditions for the same period as a result of increasing air temperature, increasing P, and decreasing DI.

4.2. Regional interannual variation of meteorological variables

Interannual variation of each variable is provided in Figure 3. Tmin, Tmax, and Tmean increased from 1960 to 2009, with sharp increases occurring in the late 1990s (Figure 3(a)–(c)). W was extremely high in 1969 and decreased thereafter (Figure 3(d)). After the late 1990s, W started to rise once again. The sharp change in annual W that happened in or around 1969 is attributable to the replacement in anemometers countrywide to some extent, which took place between 1969 and 1970 (Jiang et al., 2010). Annual RH fluctuated up and down before 2001 and appeared to decrease after 2001 (Figure 3(e)). SH increased up to the mid 1980s and decreased slightly thereafter (Figure 3(f)). P showed no apparent trend over the 50 year study period (Figure 3(g)). Ep increased sizably around 1969 (Figure 3(h)), which may be related to the extreme high W measured in 1969. Like P, DI showed no apparent trend over the 50 year study period, although it underwent low averages and low fluctuations in the first decade (1960–1969) and high averages and high fluctuations thereafter (Figure 3(i)). The TRSR was warm and dry throughout 1986–1997. This was due to rising temperatures and lower P (with the exception of abnormally high P detected in 1989 and high P in 1993) compared to 50 year means.

Figure 3.

Temporal variation in meteorological variables in the TRSR from 1960 to 2009. Solid line: time series of climate elements; dotted line: mean over the whole period; dashed lines: means in subsequences obtained from climate jump analysis.

Using Morlet wavelet transformation, significant periodicity (α = 0.05) was detected for all variables with the exception of W throughout the study period (1960–2009) as shown in Table 5 and Figure 4. In general, climate conditions in the TRSR were characterized by a significant (α = 0.05) short periodic cycle of two to five years. For example, a significant 3 year cycle was detected for Tmin, Tmax, and Tmean throughout 1995–2000 (Figure 4(a)–(c)). In addition, a significant cycle of 3.5 years was detected for Ep and RH (Figure 4(e) and (h)). SH, P, and DI appeared to have undergone two significant cycles less than 5 years in duration (Figure 4(f), (g), and (i)). During 2001–2009, Ep, RH, DI, SH, and P underwent significant periodical cycling. On the other hand, the relatively long periodical cycles detected for SH (15 years), P (8 years), and DI (8 years) appeared to be statistically insignificant. Periodicity is determined to be unstable through time due to the complexity of climate systems in the TRSR. Short periodic cycles indicated that all climate variables were mainly influenced by strong quasiperiodic climate patterns (e.g. El Niño-Southern Oscillation (ENSO) and quasi-biennial oscillation) (Mendoza et al., 2006; Yadava and Ramesh 2007), with the exception of W which exhibited no periodicity. Besides climate systems, solar activity (i.e. sunspots) (Mendoza et al., 2006; Yadava and Ramesh, 2007) can influence SH to some extent, as indicated by the statistically insignificant 15 year periodic SH cycle. In terms of cycling, climate systems dominated greater than half of the study period for Tmax, Ep, P, RH, and DI.

Table 5. Regional periodicity determined by Morlet wavelet transformation for each meteorological variable from 1960 to 2009
VariablePeriodical cycle (years)Time that significant periodical cycle appeared
  1. *Denotes periodical cycle statistically significant at α = 0.05

Tmin (°C)3*1995–2000
Tmax (°C)3*1965–969, 1981–1989, and 1995–2000
Tmean (°C)3*1982–1986 and 1995–2000
W (m s−1)
Ep (mm a−1)3.5*1964–1968, 1981–1992, and 2004–2008
SH (h d−1)2.5*1987–1992 and 2003–2009
5*
15
P (mm a−1)2*1978–1994 and 2002–2009
5*
8
RH (%)3.5*1965–1970, 1980–1989, and 2004–2009
DI2*1966–1997 and 2001–2009
3.5*
8
Figure 4.

The Morlet wavelet power spectrum of anomaly for the nine meteorological variables in the TRSR between 1960 and 2009.

All meteorological variable averages for the past decade (2000–2009) were compared with World Meteorological Organization (WMO) climate norms (means of all climate variables) throughout the study period (1961–1990) (WMO, 1984) to identify the most recent climate changes taking place in the TRSR. The difference was the average of the past decade (2000–2009) minus climate norms, while relative change was taken as the percentage change of recent climate data to climate norms. Table 6 reveals that Tmin, Tmax, and Tmean have increased with differences of 1.0 °C for Tmin and 0.8 °C for Tmax and Tmean. Annual P and Ep increased by 3.3 and 1.7%, respectively, and DI decreased by 1.7%. W, RH, and SH decreased by 8.4, 2.9, and 1.2%, respectively. With the exception of SH, changes observed for all 2000–2009 variables (compared to 1961–1990) were consistent with regional linear trends from 1960 to 2009. Generally, the last decade (2000–2009) was warmer, less windy, wetter, and dimmer compared to 1961–1990 climate norms.

Table 6. Regional climate in the past decade (2000–2009) compared to WMO climate norms during 1961–1990
 TminTmaxTmeanWRHSHPEpDI
1961–1990−7.17.5−0.62.457.4260247310852.32
2000–2009−6.18.30.22.255.8257248911042.28
Difference1.00.80.8−0.2−1.7−301618.4−0.04
Relative change to 1961–1990 climate norms (%)   8.42.91.23.31.7−1.7

4.3. Regional meteorological variable climate jumps

Regional climate jumps detected by the modified MK and MTT methods for each meteorological variable are provided for in Table 7. Figure 3 illustrates the means before and after climate jumps occurred. With the exception of P, climate jumps occurred for all meteorological variables investigated. Air temperature jumps were all upwards, with two jumps occurring for Tmin in 1971 and 1995, one jump for Tmax in 2001, and one jump for Tmean in 1997. The approximate 1 °C jumps that occurred during 1995–2001 for Tmin, Tmax, and Tmean indicated that warming primarily occurred after the late 1990s, which confirms that the climate in the last decade (2000–2009) has been much warmer than the climate was throughout 1961–1990 as described in the previous section. DI and Ep experienced one upward climate jump in 1967. Upward and downward climate jumps were detected for RH in 1982 and 2002, respectively. Three climate jumps were detected for W and SH. For W, upward climate jumps occurred in 1969 (due to some extent to anemometer replacement) and 2002. A downward jump occurred in 1975. For SH, upward climate jumps occurred in 1965 and 1976, while a downward jump occurred in 1986. The magnitude of jumps was large for DI, RH (second jump), SH (first jump), and Ep, in the range of 5.5–6.9%. The highest magnitude in terms of jumps occurred for W in the first jump (33.5%, a result of anemometer replacement to some extent) and the second jump (16.2%). The upward jump detected in 1971 for Tmin, 1969 for W, and 1965 for SH led to the upward jump for Ep in 1967, which then caused an upward jump for DI in 1967.

Table 7. Regional climate jumps detected by the modified MK and MTT methods for each meteorological variable from 1960 to 2009
VariableNumber of jumpsTiming direction (method)MagnitudePercentage (%)
  1. ↑, upward change; ↓, downward change.

Tmin (°C)21971↑(MTT), 1995↑(MK, MTT)0.54, 0.93 
Tmax (°C)12001↑(MK, MTT)1.12 
Tmean (°C)11997↑(MK, MTT)1.13 
W (m s−1)31969↑(MTT), 1975↓(MK, MTT), 2002↑(MTT)0.73, −0.47, −0.1833.5, 16.2, −7.4
E p (mm a−1)31967↑(MK, MTT)615.9
SH (h d−1)31965↑(MK, MTT), 1976↑(MTT), 1986↓(MTT)136, 96, −975.5, 3.7, 3.6
P (mm a−1)0
RH (%)21982↑(MK), 2002↓(MK, MTT)0.64, −3.301.1, 5.6
DI11967↑(MK)−0.15−6.9

5. Discussion

5.1. Temporal characteristics of TRSR climate change

The TRSR calls attention to the fact that the Tibetan Plateau is an amplifier of global climate change. The rate of warming (0.032 °C a−1) was higher for the TRSR compared to 0.59 °C 50 a−1 for the Tibetan Plateau (Liu et al., 2010b), 0.22 °C 10 a−1 for China (Ding et al., 2006), and the average rate of warming for the Northern Hemisphere as well as for the globe during a similar period (Ding et al., 2006). Moreover, the P trend rate of 0.534 mm a−2 was higher compared to the rate of 0.125 mm a−2 for the Tibetan Plateau (Liu et al., 2010b), 0.14 mm a−2 for China (Liu et al., 2010b), and 0.11 mm a−2 for land area worldwide (IPCC, 2007) in a similar period. Like the Tibetan Plateau (Yang et al., 2011), the TRSR also experienced warmer and wetter trends during 1984–2006 with the exception that the change rates for TRSR were greater. As for changes in W, many parts of the world reported to have experienced a deceleration pattern in W such as the one that had occurred in the TRSR during the last 30 to 50 years. For example, please see You et al. (2010b), Ren et al. (2005a, 2005b), Xu et al. (2006), Ko et al. (2010), Hobbins (2004), Tuller (2004), and McVicar et al. (2008).

The TRSR, being located within a unique geographical unit (the Tibetan Plateau), exhibited unique characteristics. Not all variables exhibited the same direction in change as seen in other parts of the world. For example, increased Ep in the TRSR was opposite to the decrease in averaged Ep across China in 1957–2006 (Zeng et al., 2007; Shen and Sheng, 2008; Liu et al., 2010a). Moreover, the increasing trend in SH from 1961 to 1998 in the TRSR was opposite to trends measured at most meteorological stations in China for the same period (Chen et al., 2006). Many parts of the world experienced a steady decline in global radiation from the late 1950s to 1980s, a phenomenon known as global dimming (Gilgen et al., 1998; Liepert, 2002; Pinker et al., 2005; Stanhill and Cohen, 2005; Cutforth and Judiesch, 2007; Wild, 2009). However, an increasing trend in SH was exhibited in the TRSR during the same period. The RH trend (−0.026% a−1) in the TRSR was lower than that of China (−0.042% a−1), calculated from 623 meteorological stations throughout 1960–2009. Also, for 1992–2001, the RH trend in the TRSR (0.134% a−1) was lower than and opposite to the trend measured in southwestern Nigeria (−0.331% a−1) (Akinbode et al., 2008). Moreover, for 1960–2005, the RH trend in the TRSR (0.006% a−1) was much lower than the trend measured in Jordan (0.14% a−1) (Abu-Taleb et al., 2007).

As previously stated, previous climate change study in the TRSR, mostly focused on air temperature and P. Air temperature trends derived here were consistent with that obtained in the TRSR by previous studies from the 1960s to 2000s (Li et al., 2004, 2006; Yang, 2006; Gao and Liu, 2008). For example, the mean air temperature trend of 0.032 °C a−1 determined in this study falls within the scope obtained by previous studies (0.025–0.035 °C a−1). As for P, a slightly increasing trend of 0.534 mm a−2 was higher than the range of −0.281–0.150 mm a−2 reported by the four other studies cited above. Apparently, differences in temperature and P trends were mainly caused by data from different meteorological stations and the length of records used in prior analyses.

5.2. Possible mechanisms for climate change in the TRSR

Climate change taking place in the TRSR is primarily a local response to global climate change. On the one hand, the climate in the TRSR is controlled by large-scale weather systems, including the North Atlantic Oscillation (NAO), the Arctic Oscillation (AO), the East Asia westerly jet (WJ), the ENSO, and East and South Asian summer monsoons. For example, W in the TRSR were mainly controlled by WJ and ENSO as correlation coefficients indicated, and decreasing W trends were attributable to significant decreasing WJ in summer (α = 0.05). Decreasing P trends in TRST were attributable to decreasing WJ and East and South Asian summer monsoon indices, which showed high correlation coefficients with P in summer (Cuo et al., 2012). Global warming induced changes to large-scale weather systems will certainly affect TRSR climate conditions. You et al. (2010a) suggested that the change in atmospheric circulation (sea level pressure, geopotential height, and wind field) was one of the important factors for rising temperatures. Also, local scale weather systems play a role in addition to large-scale weather systems (e.g. wind shear line, and plateau vortices which are shallow cyclonic vortices generated in situ over the Tibetan Plateau during the rainy season) in affecting TRSR W and P patterns in summer, as the changes in precipitation and wind were only weakly related to large-scale indices in summer (Cuo et al., 2012). On the other hand, studies have found that land surface change caused local climate change feedbacks to occur. The snow-albedo feedback and land use changes are important factors that led to warming trends in the Tibetan Plateau (Kang et al., 2010). Surface albedo decreases as snow melts, so that more solar radiation is absorbed at the surface and surface warming is enhanced in response (Giorgi et al., 1997). Land cover changes in the TRSR were induced by permafrost and grassland degradation, urbanization, deforestation, and desertification (Cui and Graf, 2009). Changes such as these influence climate through biogeophysical effects (Brovkin et al., 2004) such as surface roughness transformation, evapotranspiration, and albedo as well as through biogeochemical feedbacks (Stich et al., 2005) such as CO2 atmospheric emission rates (Houghton and Hackler, 2003). Further research will have to be carried out to better understand local mechanisms of each meteorological variable.

Trends in Ep and DI (the two variables calculated) can be understood by quantifying the contribution of each independent variable. Significant increases in Tmin, Tmax, and Tmean accompanied by increasing SH and decreasing RH led to increasing Ep. Increasing Tmean and SH and decreasing RH contributed 0.915, 0.305, and 0.122 mm a−2 to Ep, respectively. On the other hand, the decrease in W offset 0.424 mm a−2 of this Ep increase. Regional averaged annual P and Ep increases relative to regional long-term mean annual P (ΔP/μP) and Ep (ΔEpEp) were 4.1 and 4.2%, respectively, indicating that annual Ep consumption just offset annual P increases, resulting in little regional change in annual DI, and so is for rainy season DI. On the other hand, the 5.5% increase in Ep was overwhelmed by a 27.0% increase in P during the dry season, resulting in a decrease in DI.

5.3. Response of water resources on the climate change

Significant increases in air temperature (α = 0.01) and Ep (α = 0.05) combined with slight increases in P and SH affected hydrological cycling in the TRSR as well as areas downstream from it. The TRSR contains permafrost and glacial areas of 2.31 × 105 km2 and 1.65 × 103 km2, respectively. Warming trends caused an increase in the number of thawing days and, consequently, the degradation of permafrost and the thickening of its active layer (Xue et al., 2009; Liu et al., 2010c). Permafrost thawing modifies hydrological processes by changing soil moisture as well as levels of ground water, rivers, and lakes (Xue et al., 2009; Liu et al., 2010c). Similarly, warming results in glacial melting (glacier terminus retreatment) and causes negative glacier mass balances, especially during summertime (Yao et al., 2004, 2007, 2012; Pu et al., 2008). Glacier ablation enlarges lake areas (Yao et al., 2007) and increases streamflow for a short period of time. In the long run, however, fresh water storage will be lost. In the TRSR, Yangtze River streamflow increased by 28.4 × 106 m3 a−2 from 1956 to 2005 and Mekong River streamflow increased by 8.8 × 106 m3 a−2 from 1956 to 2000 (Liang, 2011), partly due to increasing melt water (Yao et al., 2004 and 2007). For the Yellow River, streamflow decreased by 47.3 × 106 m3 a−2 from 1956 to 2005 due to the increase in evapotranspiration consumption as a result of rising temperatures (Liang, 2011), despite the increase in melt water (Yao et al., 2004 and 2007). It is clear that changes in available water resources in the TRSR will have profound impacts on local and downstream populations and ecosystems (Kehrwald et al., 2008; Immerzeel et al., 2010).

Climate change has affected and will continue to affect ecosystems in the TRSR. Xiang et al. (2009), Li et al. (2010), Xiao et al. (2010), and Zhang (2011) reported that the area, structure, function, and processes of alpine wetlands have changed. Zhang (2011) also stated that wetlands in the Damqu River Basin located in the source region of the Yangtze River have experienced complex intercategory transformations, although the total wetland area has remained essentially unchanged throughout 1998 and 2007. Through experimental warming, Klein et al. (2004) found that warming caused large and rapid species loss. Warming may change species geographic ranges as being reported in temperate regions (Iverson and Prasad, 1998; Lenoir et al., 2008 and Moritz et al., 2008) and threaten biodiversity within the TRSR.

6. Conclusions

Temporal variations in meteorological variables in the TRSR from 1960 to 2009 were analyzed by a fitted linear model as well as modified MK and MTT methods. Four points can be drawn from the analysis. They are listed as follows:

  1. Seven out of the 12 meteorological stations in the TRSR have reported warmer and wetter conditions between 1960 and 2009, with the exception of the five situated in the southeastern section of the region where conditions are warmer and drier. Rates related to warmer and wetter conditions are higher in the TRSR than in the Tibetan Plateau and China as a whole. Warming in the TRSR intensified in the late 1990s. Similar to most parts of China, the TRSR has experienced less windy conditions, although this started to pick up in the early 2000s.
  2. Changes observed for most variables in the past decade (2000–2009) were consistent (compared to 1961–1990) with regional linear trends from 1960 to 2009, with the exception of SH. Tmin, Tmax, Tmean, and W changed considerably between 2000 and 2009, with air temperatures 0.8 °C higher and W 8.4% lower, respectively. The TRSR became notably warmer and less windy, and slightly wetter and dimmer between 2000 and 2009 compared to the period from 1961 to 1990.
  3. Climate jumps (α = 0.05) occurred for all meteorological variables with the exception of regionally scaled P. Tmax, Tmean, and DI experienced one climate jump, Tmin and RH experienced two, and SH, Ep, and W experienced three.
  4. Major periodical changes within a five year period (α = 0.05) were detected for all meteorological variables, with the exception of W.

Acknowledgements

The Special Foundation of the Ministry of Science and Technology of China (no. 2008FY110300-01) and the National Natural Science Foundation of China (no. 51109196) supported this research. We would like to thank the “Hundred Talent” program granted to Lan Cuo by the Chinese Academy of Sciences. We would also like to thank Zhihao Liang for providing the permafrost and glacial area data for the TRSR.

Ancillary