Past and future spatiotemporal changes in evapotranspiration and effective moisture on the Tibetan Plateau


  • Yunhe Yin,

    Corresponding author
    1. Institute of Geographic Sciences and Natural Resources Research, Chinese Academy of Sciences, Beijing, China
    • Corresponding author: Y. H. Yin, Institute of Geographic Sciences and Natural Resources Research, Chinese Academy of Sciences, 11A Datun Road, Anwai, Beijing 100101, China. (

    Search for more papers by this author
  • Shaohong Wu,

    1. Institute of Geographic Sciences and Natural Resources Research, Chinese Academy of Sciences, Beijing, China
    Search for more papers by this author
  • Dongsheng Zhao

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


[1] Observed evaporative demand has decreased worldwide during the past several decades. This trend is also noted on the Tibetan Plateau, a region that is particularly sensitive to climate change. However, patterns and trends of evapotranspiration and their relationship to drought stress on the Tibetan Plateau are complex and poorly understood. Here, we analyze spatiotemporal changes in evapotranspiration and effective moisture (defined as the ratio of actual evapotranspiration (ETa) to reference crop evapotranspiration (ETo)) based on the modified Lund-Potsdam-Jena Dynamic Global Vegetation Model (LPJ). Climate data from 80 meteorological stations on the Tibetan Plateau were compiled for the period 1981–2010 and future climate projections were generated by a regional climate model through the 21st century. The results show regional trends towards decreasing ETo and statistically significant increases in ETa (p < 0.05) and effective moisture during the period 1981–2010 (p < 0.001). A transition from significant negative to positive ETo occurred in 1997. Additionally, a pronounced increase in effective moisture occurred during the period 1981–1997 because of significant decreased ETo before 1997. In the future, regional ETo and ETa are projected to increase, thus reducing drought stress, because of generally increased effective moisture. Future regional differences are most pronounced in terms of effective moisture, which shows notable increases in the northwestern plateau and decreases in the southeastern plateau. Moreover, the reduced magnitude of effective moisture is likely to intensify in the long term, due mainly to increased evaporative demand.

1 Introduction

[2] Mean global air temperature has increased approximately 0.74°C over the past 100 years (1906–2005) and is predicted to increase by 1.1°C to 6.4°C, between 1990 and 2100 [IPCC, 2007]. Climate change has strongly impacted both the water and carbon cycles, especially via processes involving evapotranspiration, which is a critical component of the water, energy, and carbon cycles [Jung et al., 2010]. At global scales, evapotranspiration levels are approximately 60%–65% of precipitation levels [Brutsaert, 2005]. In addition, evapotranspiration correlates strongly with net ecosystem production [Nemani et al., 2002].

[3] Numerous studies have recorded temporal variations in evapotranspiration at regional to continental scales, indicating significant decreasing trends of observed pan evaporation [Hobbins et al., 2004; Liu et al., 2004; Peterson et al., 1995; Roderick and Farquhar, 2002; van Heerwaarden et al., 2010; Yang and Yang, 2012] and estimated potential evapotranspiration during the past half-century [Chen et al., 2006; Yin et al., 2010]. However, large variations in estimates of land-surface evaporation obtained using different methodologies indicate that land evaporation contains large uncertainties [Mueller et al., 2011]. Fundamental questions regarding regional and global trends in evaporation (e.g., whether increasing or decreasing, and where and when such trends occur) are not fully resolved [Dolman and de Jeu, 2010]. For example, one study [Jung et al., 2010] found that actual evapotranspiration (ETa) in Asia had increased since 1998 whereas another study [Zeng et al., 2012] found a slight negative trend in this area during the same period.

[4] Future changes in regional and global climate will have notable potential impacts on the hydrologic cycle and will cause large uncertainties in future climate projections [Allen and Ingram, 2002; Arnell, 2011]. For example, Piao et al. [2010] predicted that future summertime warming rates will increase; such pronounced summer warming will inevitably enhance evapotranspiration, and increase the risk of agricultural water shortage in China. In addition, regional patterns of warming-induced changes in surface hydro-climatic conditions are complex and less certain than those for temperature [Milly et al., 2005].

[5] Although numerous studies have been conducted on spatiotemporal patterns of evapotranspiration, little is known about the magnitudes of such variations. In addition, few studies have considered the relationship between evapotranspiration and drought stress on the Tibetan Plateau. The Tibetan Plateau, which is the largest high-altitude land mass on earth, exhibits not only unique climatic features and physical environments, but is also extremely sensitive to global climate change [Zheng and Yao, 2006]. A rise in temperature of up to 0.3°C per decade has been recorded on the Tibetan Plateau over the past 50 years, and this increase is approximately three times the global rate of warming [Qiu, 2008]. The Tibetan plateau is also the “world water tower,” and the land-ocean-atmosphere interactions associated with the plateau generate a profound impact on the global climate and environment [Xu et al., 2008]. Thus, it is vital to evaluate the variability in evapotranspiration and effective moisture (expressed as the ratio of ETa to reference crop evapotranspiration (ETo)) on the Tibetan Plateau. High (low) ETa/ETo ratios indicate low (high) drought stress on plants [Shinker and Bartlein, 2010]. Understanding the changes in evapotranspiration and their effects on drought is critical to predicting climate change impacts on vulnerable ecosystems and the water cycle.

[6] At present, measurements of evapotranspiration are scarce [Seneviratne et al., 2010], particularly in the Tibetan Plateau region. The Lund-Potsdam-Jena (LPJ) model combines terrestrial vegetation dynamics, and land-atmosphere carbon and water exchanges in a modular framework [Sitch et al., 2003]. This model simulates the coupled terrestrial carbon and water cycles, and is therefore well suited to investigating biosphere-hydrosphere interactions [Gerten et al., 2004]. It comprehensively assesses ecosystem dynamics as well as interactions and feedbacks involving eco-hydrological processes. Thus, the model has been widely used to assess the impacts of climate change on global or regional water and carbon cycles [Doherty et al., 2010; Kaplan et al., 2012; Mahecha et al., 2010; Morales et al., 2005; Tao and Zhang, 2011]. These previous studies indicate that the LPJ model aptly describes the terrestrial water cycle. Furthermore, dynamic vegetation models, including vegetation dynamics and biogeochemical processes, have supplied us with an effective approach to project transient responses of the ecosystem to rapid climate change [Bachelet et al., 2003; Cramer et al., 2001]. Therefore, the LPJ dynamic model has an advantage over hydrological models in aspect of the regulating role of vegetation on evapotranspiration changes in the future.

[7] In this study, the LPJ model was used to estimate the impact of climate change on the water budget of the Tibetan Plateau. The primary objective was to investigate evapotranspiration changes during the period 1981–2010 using observed meteorological data, and to simulate spatiotemporal variability in evapotranspiration and effective moisture during the period 2011–2100 based on predictions of a regional climate model (RCM). The results have important implications for the promotion of ecosystem water-conservation function, ensuring the safety of water resources and ecosystems, and developing adaptation strategies to protect ecosystems that are vulnerable to the impact of climate change.

2 Materials and Methods

2.1 Study Area

[8] The Tibetan Plateau (27.07°N–39.50°N, 73.21°E–104.49°E) has an area of approximately 2.3 × 106 km2 and an average elevation of greater than 4000 m above sea level. The alpine ecosystems of the plateau have important ecological functions as they represent the headwaters of the Yangtze, Yellow, Yarlung Zangbo, and Lancang rivers. The climate of the Tibetan Plateau varies from warm and humid in the southeast to cold and dry in the northwest [Zheng, 1996]. Alpine grasslands, including meadow and steppe, are mostly distributed in the subcold regions of the central plateau. Relatively small areas of evergreen forest occur in the subtropical and temperate regions of the southeastern plateau. The boundaries of the Tibetan Plateau are determined according to the division of physico-geographical regions of China [Zheng, 1996] (see Figure 1).

Figure 1.

Location of the study area and meteorological observation stations on the Tibetan Plateau.

2.2 Evapotranspiration Simulations

[9] The LPJ model is a moderately complex dynamic global vegetation model, developed on the basis of the early equilibrium model BIOME3. The LPJ distinguishes 10 plant functional types (PFTs), each with different photosynthetic, phonological, and morphological characteristics [Sitch et al., 2003]. In our previous studies, the LPJ model has been carefully modified, calibrated, and successfully used to simulate terrestrial ecosystems on the Tibetan Plateau [Yin et al., 2013; Zhao et al., 2011a, 2011b]. To reflect the special characteristics of alpine ecosystems, two additional PFTs (representing shrubs and cold grasses) have been included in the model. In addition, the radiation-calibrated Penman-Monteith model is used to estimate ETo in the LPJ model [Yin et al., 2013]. The locations of radiation stations used for calibration are shown in Figure 1.

[10] In the hydrologic scheme of the LPJ, ETa (mm day−1) is estimated as the sum of interception losses and plant transpiration for all PFTs, including evaporation from bare soil. Usually, the total ETa does not exceed ETo. Specifically, interception loss (Ei, mm day−1) is determined from the canopy storage capacity (Si, mm day−1) and ETo (mm day−1), according to:

display math(1)
display math(2)
display math(3)

where Δ is the slope of the saturation vapor pressure versus air temperature curve (kPa °C−1), Rn is the net solar radiation at the plant surface (MJ m−2 day−1), G is the soil heat flux (MJ m−2 day−1), γ is the psychrometric constant (kPa °C−1), T is the mean air temperature at a height of 2 m (°C), U2 is the mean wind speed at a height of 2 m (m s−1), es is the saturation vapor pressure (kPa), ea is the actual vapor pressure (kPa), P is the precipitation (mm day−1), LAI is the leaf area index (%), fv denotes the fraction of the grid cell covered by vegetation (%), and i is the coefficient used for calculating interception and the dimensionless biome-dependent proxy for the rainfall regime.

[11] Soil evaporation (Es, mm day−1) is closely related to relative moisture levels in the upper 20 cm of the soil column (wr20, %), ETo, and fv, according to:

display math(4)

[12] Transpiration (Et, mm day−1), which is modeled as the minimum of an atmosphere-controlled demand function (D) and a plant-controlled supply function (S), is expressed by:

display math(5)

[13] The supply function is simulated by using the PFT-specific maximum transpiration rate and relative soil water content, and the demand function is simulated using ETo and the potential canopy conductance by coupling the photosynthesis and water balance modules of the LPJ. Transpiration is neglected if canopy storage evaporation equals ETo since there will be no day-time canopy-available energy remained for plant transpiration [Gerten et al., 2004]. Detailed descriptions of the LPJ model can be found in Gerten et al. [2004] and Sitch et al. [2003].

2.3 Data Sources and Modeling Protocol

2.3.1 Historical Meteorological and Hydrological Data

[14] We used high-quality monthly observations on maximum and minimum air temperatures, precipitation, mean relative humidity, sunshine duration, and mean wind speed obtained from the 80 meteorological stations during the period 1981–2010 (Figure 1). Data were provided by the China Meteorological Administration (CMA). Observation stations were deleted from the data set if: the station was built after 1981, the location of the station changed during the study period, the station was removed before 2010, or more than 5% of the data from the station was missing. Missing data were estimated by averaging the values obtained from the same station during other years. To meet the model input requirements, ground-based point meteorological data were interpolated on a 10 × 10 km grid using a thin plate spline method. Elevation is considered for temperature when interpolating.

[15] Annual runoff data of five hydrological stations, distributed at the Yellow River, the Yangtze River, the Langcang River, and the Yarlung Zangbo River on the Tibetan Plateau of China, were obtained for the period 1981–2000 from the Ministry of Water Resources. Since human water use and trans-basin water transfers are not significant, the runoff data of plateau catchments can be approximately regarded as under natural conditions. Locations of hydrological stations and catchments are shown in Figure 1.

2.3.2 Regional Climate Model Simulations

[16] Future climate data series were provided by the Institute of Environment and Sustainable Development in Agriculture, Chinese Academy of Agricultural Sciences. The research group predicted future climate data at a horizontal resolution of 50 × 50 km across China in the 21st century, using the Providing REgional Climates for Impacts Studies (PRECIS) system, a regional climate modeling system developed at the UK Met Office Hadley Centre for Climate Prediction and Research [Jones et al., 2004]. The PRECIS data set has been validated for its strong ability to simulate seasonal terrestrial climate variations [Xu et al., 2006; Zhang et al., 2006], and has been successfully applied to climate change impact assessments of ecosystem vulnerability [Wu et al., 2010], future crop production [Xiong et al., 2010], and heat waves [Yang et al., 2010]. For this assessment of evapotranspiration changes on the Tibetan Plateau, medium-high A2 and medium-low B2 scenarios from the Special Report on Emissions Scenarios (SRES) were used to give a range of different possible scenarios [Nakicenovic et al., 2000].

[17] Projected climate data for the A2 and B2 emission scenarios during the period 1981–2100 were used in this study, including maximum, minimum, and average temperatures, precipitation, mean relative humidity, mean wind speed, and net solar radiation. We calculated monthly mean anomalies based on differences between observed and PRECIS-generated data for the baseline period 1981–2010, and then added the anomalies to scenario data to provide model inputs for the period 2011–2100. Before applying the projections locally, the monthly PRECIS results were interpolated to high-resolution 10 × 10 km grid-scale data using a thin plate spline method. To assess future temporal variations, the period 1981–2010 (the 1990s) was taken as a baseline period, and compared with simulations for the periods 2011–2040 (the 2020s), 2041–2070 (the 2050s), and 2071–2100 (the 2080s).

[18] The PRECIS projections for the Tibetan Plateau show that annual mean temperature will increase by 4.01°C and 2.58°C under A2 and B2, respectively, and that annual precipitation will increase by 30.56% and 23.52% under A2 and B2, respectively, in the 2080s as compared with the baseline period.

2.3.3 Soil Data

[19] The soil texture data set for this study was derived from a digital map of soil texture types (1:14,000,000) [Zhang et al., 2004]. The data set contains information on the geographical distributions of different soil texture types and the proportions of mineral grains of different sizes in top soils. Soil textures were reclassified as clay, silt, sand, silty sand, sandy clay, silty clay, and clay with silt and sand based on the Food and Agriculture Organization of the United Nations classification standard [Ni et al., 2001]. The soil textures were transformed into grid-based formats and resampled on a 10 × 10 km grid using the ArcGIS system.

2.3.4 Simulation Protocol

[20] The LPJ was implemented for a 1000 year spin-up period to achieve equilibrium in stable vegetation structures and carbon pools. The model was then run in a dynamic mode to simulate the responses of ecosystems evapotranspiration to climate change over the Tibetan Plateau during the period 1981–2100 at a grid resolution of 10 km by 10 km. Monthly data were interpolated to provide a quasi-daily time series of climatic factors, such as temperature and precipitation; the data were disaggregated using a linear method or a stochastic weather generator in the LPJ model [Sitch et al., 2003].

2.4 Assessing Temporal Trends

[21] The slopes of linear regressions, obtained using the least squares method, were used to assess trends in the time series. Positive slopes indicate increasing trends, whereas negative slopes indicate decreasing trends. The significance levels of the trends were assessed using the nonparametric Mann-Kendall test, which is widely used to assess time series trends [Sneyers, 1990].

3 Results

3.1 Validation of ETa Simulations by the Modified LPJ Model

[22] The traditional water balance approach is ideal to estimate ETa on well-gauged basins at annual time scales, which can be used to validate the other direct ETa methods [Senay et al., 2011]. Figure 2a compares annual LPJ-modeled ETa against water balance ETa from five main catchments during the period 1980 to 2000. The water balance ETa is estimated by annual precipitation minus annual runoff, Q. The validation shows root-mean-square error (RMSE) of 58.07 mm/yr and mean bias error (MBE) of 29.16 mm/yr, indicating that the modified LPJ model can successfully estimate catchment scale ETa. Further analysis shows an overestimation of averaged LPJ-modeled ETa by 10.54% compared with the water balance estimates on a mean annual basis for the period 1981–2000.

Figure 2.

Validation of ETa simulations by the modified LPJ model. (a) Comparison with annual ETa from catchment water balance method, (b) comparison with the mean annual ETa for five catchments, and (c) model performance at two eddy covariance sites on the Tibetan Plateau.

[23] We further take a validation of the monthly ETa simulations with latent heat flux measured by eddy covariance technique at Haibei (lat. 37°36′N, long. 101°18′E, alt. 3250 m), an alpine meadow site on the northeastern temperate Tibetan Plateau [Gu et al., 2008], and Damxung (lat. 30°51′N, long. 91°05′E, alt. 4333 m), an alpine shrub-meadow site on the southern subcold Tibetan Plateau. As shown in Figure 2b, the modified LPJ model could satisfactorily estimate the seasonal variations of ETa for alpine ecosystems on the Tibetan Plateau.

3.2 Variations in Estimated Evapotranspiration During the Period 1981–2010

[24] Regional average evapotranspiration anomalies and linear trends in evapotranspiration on the Tibetan Plateau during the period 1981–2010 are shown in Figure 3. Variations in annual ETo are not statistically significant, although the linear regression shows a gently decreasing trend during this time period. Annual ETo on the Tibetan Plateau decreased significantly during the period 1981–1997 (p < 0.05) and increased significantly during the period 1998–2010 (p < 0.05).

Figure 3.

Regional average anomalies of (a) reference crop evapotranspiration, (b) actual evapotranspiration, (c) precipitation, and (d) effective moisture (in percent) on the Tibetan Plateau during the period 1981–2010. The smooth lines represent 5 year moving averages. The linear trends during the period 1981–2010 (dash-dotted line) are also shown. Slope values represent the results of linear regressions. * indicates trends significant at a level of 0.05. *** indicates trends significant at a level of 0.001.

[25] The LPJ-derived ETa on the Tibetan Plateau increased significantly during the period 1981–2010, at a rate of 0.58% per year (p < 0.05). This finding is consistent with that of Zhang et al. [2007], who found that the average ETa in 16 catchments on the eastern Tibetan Plateau significantly increased from 1966 to 2001; moreover, ETa in most plateau areas showed increasing trends. Figure 3b shows that prior to 1997, significant changes in the regional average ETa were absent, and ETa values were less than the mean in most years. Since 1997, ETa values have gradually increased, and positive anomalies have been present in most years; however the increasing trend in ETa is not significant. Figure 3c shows an increasing trend in annual precipitation on the Tibetan Plateau during the period 1981–2010, with an average increase of 0.19% per year.

[26] Figure 3d shows a significantly increasing trend in effective moisture on the Tibetan Plateau during the period 1981–2010, with an average increase of 0.41% per year (p < 0.001). The variations in effective moisture are more similar to those for ETa than for ETo. However, the dominant factors contributing to the increasing trend are different before and after 1997. During the period 1981–1997, decreased ETo was a more important influence on increased effective moisture, while during the period 1998–2010, positive changes in both ETo and ETa offset each other, resulting in a nearly even variation in effective moisture.

3.3 Projected Future Changes in ETo

[27] Figure 4 shows predicted changes in ETo on the Tibetan Plateau during three time periods, relative to the baseline period (1981–2010). Regional average ETo is predicted to increase on the Tibetan Plateau as a whole. Projections of ETo indicate: in the 2020s, an increase of 2.45% for the A2 and of 1.61% for the B2 emissions scenarios; in the 2050s, an increase of 6.36% for the A2 scenario and 3.51% for the B2 scenario; and in the 2080s, an increase of 11.72% for the A2 scenario and 5.31% for the B2 scenario. Larger differences between the predictions based on the two emissions scenarios are predicted by the end of the 21st century.

Figure 4.

Predicted changes in evapotranspiration and precipitation (in percent) on the Tibetan Plateau for three time periods relative to the baseline period 1981–2010: 2011–2040, 2041–2070, and 2071–2100. Results for the A2 (solid line) and B2 (dashed line) emissions scenarios are shown. The median values of the predicted changes are shown as solid lines; the box shows the 25th–75th percentile range, and whiskers extend to the 5th and 95th percentiles.

[28] Figure 5 shows the spatial distribution of projected annual ETo anomalies between historic and future periods for the A2 and B2 emissions scenarios. Values of ETo are likely to increase in most regions of the Tibetan Plateau, as compared with the baseline period. The greatest increases generally occur on the eastern and southern Tibetan Plateau, and in particular under the higher (A2) emissions scenario. However, ETo will decrease on the northwestern Tibetan Plateau under the B2 emissions scenario.

Figure 5.

Changes in annual reference crop evapotranspiration anomalies on the Tibetan Plateau between the baseline period (1981–2010) and three future time periods (2011–2040, 2041–2070, 2071–2100), modeled using climate change projections for the A2 and B2 emissions scenarios.

3.4 Projected Future Changes in ETa

[29] Figure 4 shows projected percentage changes in regionally averaged ETa on the Tibetan Plateau as simulated by the modified LPJ model. Future projections show a general increase in ETa, although substantial differences in the magnitudes of the increase occur in the two emissions scenarios. The projected increases in averaged ETa anomalies are: in the 2020s, 7.41% for the A2 scenario and 8.11% for the B2 scenario; in the 2050s, 19.93% for the A2 scenario and 17.96% for B2 scenario; and in the 2080s, 37.43% for the A2 scenario and 28.08% for the B2 scenario. The predictions of the high-emission A2 and low-emission B2 scenarios diverge considerably as warming progresses. Positive changes in precipitation are predicted, and the quantitative changes of P will be close to that of ETa in the future on the Tibetan Plateau.

[30] Figure 6 shows spatial variations in ETa anomalies simulated according to the modified LPJ model for the three future periods for the A2 and B2 emissions scenarios. In the 2020s, the ETa decreases in some areas of the southeastern plateau (a trend which is more obvious in the B2 scenario), while ETa anomalies increase over most of the remainder of the Tibetan Plateau. During the latter 2050s period, decreases in ETa in some areas are likely to change from negative to positive. During the 2080s, ETa values increase substantially in the northwestern part of the plateau for both emissions scenarios; however, ETa anomalies will still decrease in some small areas of the southeastern plateau for the B2 scenario.

Figure 6.

Same as Figure 5, but showing changes in actual evapotranspiration anomalies.

[31] Figure 7 shows the distributions of percent changes in precipitation anomalies for the future three time periods. In the 2020s, P will increase in most areas of the Tibetan Plateau, especially in northwestern plateau having significant increment. Negative changes are likely to occur in the headwater regions of the Yangtze River and the Langcang River and some areas of the southeastern plateau. As warming proceeds, future precipitation anomalies in arid northwestern areas with positive anomalies are projected to increase, while parts of humid southeastern areas with negative anomalies are likely to decrease. This will lead to a somewhat homogeneous spatial distribution in annual precipitation amount.

Figure 7.

Same as Figure 5, but showing changes in precipitation anomalies.

3.5 Projected Future Changes in Effective Moisture

[32] Effective moisture is predicted to increase in the future for the Tibetan Plateau as a whole (Figure 4). In the 2020s, the predicted mean effective moisture anomalies increase by 5.63% for the A2 and by 6.90% for the B2 emissions scenarios. In the 2050s, the predicted changes for the A2 scenario are weaker than those for the B2 scenario. In the 2080s, mean effective moisture anomalies are projected to increase by 25.25% for the A2 and by 23.02% for the B2 emissions scenarios.

[33] Figure 8 shows the distributions of percent changes in effective moisture anomalies between the baseline period and the future three time periods. In the 2020s, effective moisture will increase at approximately 62% of the grids on the Tibetan Plateau, with an average increase of 11.16%, for the A2 scenario, and at 66% of the grids, with an average increase of 12.86%, for the B2 scenario. The increases are mainly predicted to occur in the northwestern regions of the plateau where semi-arid to arid climates prevail, while decreases are predicted to occur in the southeastern plateau, where humid to subhumid climates prevail. The areal extent of regions with decreased effective moisture anomalies is likely to shrink in periods subsequent to the 2020s, especially for the B2 scenario, to less than 30% of the total area. The predicted decreases in the three future time periods are, for the 2020s, 2050s, and 2080s, −3.26%, −4.10%, and −6.64%, respectively, for the A2 scenario, and −4.49%, −5.39%, and −6.61%, respectively, for the B2 scenario. Net changes in central regions are positive but modest as compared with those in northerly regions.

Figure 8.

Same as Figure 5, but showing changes in effective moisture anomalies.

4 Discussion

[34] Our results showed a slightly decreasing trend in regionally averaged ETo during the period 1981–2010. Although ETo decreased significantly from 1981 to 1997 (p < 0.05), it has shown a significant increasing trend since 1998 on the Tibetan Plateau (p < 0.05). The transition from decreasing to increasing ETo has been reported for China in recent pan evaporation observational studies [Liu et al., 2011], which show a decrease of annual pan evaporation during the period 1960–1991, induced mainly by decreasing wind speed and decreasing solar radiation, and increases in ETo values since 1992, induced by rising temperatures.

[35] In a previous study, we found that decreasing wind speed was the primary cause of decreasing ETo on the Tibetan Plateau [Yin et al., 2010], which is in agreement with the results of other studies in the region [Chen et al., 2006; Zhang et al., 2009; Zhang et al., 2007]. The recent increase in ETo values may be attributed to the higher warming, and ETo increment is projected to be intensified for the next hundred years.

[36] Past increases in ETa estimated in this study are consistent with recent studies by Yang et al. [2011], who found increased evaporation at most of the stations on the Tibetan Plateau during the period 1984–2006, and by Gao et al. [2007], who found an increasing trend in west China during the period 1960–2002. However, ETa variations on the plateau differ from patterns of global ETa changes, which have been inferred to increase on average during the period 1982–1997, with no apparent subsequent increase in global evapotranspiration until 2008 [Jung et al., 2010]. This stresses the importance of regional perspectives for detecting and attributing changes in evapotranspiration.

[37] Causes of regionally observed evapotranspiration trends are debated. Solar radiation is the dominant source of energy at the land surface. Trends in radiation are expected to impact ETa in regions where ETa and radiation are correlated; however, in regions where the correlation is weak, trends in precipitation explain trends in ETa [Teuling et al., 2009]. Similarly, variations in incident solar radiation are shown to control the long-term variations of evapotranspiration in humid areas [Wang et al., 2010]. However, the Tibetan Plateau has experienced solar dimming for the past three decades due to the increase in water vapor amount and deep cloud cover [Yang et al., 2012]. There may be other mechanisms that contributed to the increased evapotranspiration over the Tibetan Plateau.

[38] A few studies have related portions of the increased ETa to changes in precipitation. For example, in the Mississippi River basin, an upward trend in ETa during the period 1949–1997 was driven primarily by increased precipitation and secondarily by human water consumption [Milly and Dunne, 2001]. Changes in precipitation play a key role in the changes of ETa estimated for most parts of China during the period 1960–2002, although in southeast China, changes in ETo appeared to be a major factor related to changes in ETa [Gao et al., 2007]. Soil water supply, closely related to precipitation, is the dominant factor in controlling long-term variations of evapotranspiration in arid areas [Wang et al., 2010]. Moreover, decreasing soil moisture supply is the main mechanism contributing to the cessation of the rising ETa trend in the Southern Hemisphere after 1998 [Jung et al., 2010]. On the Tibetan Plateau, increased precipitation during the last three decades may be a primary external driving mechanism for increased ETa values and reduced drought stress.

[39] Regional changes in precipitation character depend on the variability patterns of atmospheric circulation to a great extent [Trenberth, 2011]. Moreover, hydrologic cycle changes are generally attributed to changes in the strength of the monsoon in the Tibetan Plateau [e.g., Morrill, 2004]. Wind speed declination on the Tibetan Plateau may be due to the reduced strength of circulation as well [Zhang et al., 2009]. It has been widely discussed that wind stilling is the main driving force for the declining potential evapotranspiration on the Tibetan Plateau. Therefore, changes in atmospheric circulation are likely to affect evapotranspiration variations, together with the projected changes of regional P-ETa values as mentioned by Seager et al. [2007].

[40] Besides atmospheric and climatic mechanisms, vegetation also plays an important role in ETa changes. Many studies have discussed the trend toward longer growing season as a dynamic response of ecosystems to climate change. It was noted that the length of the growing seasons increased since the 1980s over the Northern Hemisphere indicated by satellite-measured normalized difference vegetation index [Jeong et al., 2011]. With changes in growing-season length, the prediction of a longer growing-season length would overall enhance evapotranspiration in humid regions of the Eastern US [White et al., 1999]. Moreover, vegetation structure dynamics would also exert potential influences on evapotranspiration. For example, a study by Piao et al. [2007] indicated that the net effect of increased vegetation leaf area index could provide a greater cumulative surface area for canopy water transpiration and interception, and hence enhance evapotranspiration. In some areas such as the headwater region of the Yangtze River, vegetation may be attributed to the ETa and effective moisture changes that are likely to increase while precipitation decreases. Further research will require attention to the complicated impacts of changes in climate and vegetation on evapotranspiration and hydrological cycle as well.

[41] Our results indicate that changes in effective moisture would have an overall northwest to southeast gradient on the Tibetan Plateau through the future hundred years for the A2 and B2 emissions scenarios. The changing pattern is corresponding to the northwest to southeast gradient in climate and ecosystems on the plateau. The climate gradient is mainly determined by topographic configuration and atmospheric circulation [Zheng, 1996]. In northwest regions with water-limited environment, positive changes in ETa and effective moisture are likely to occur and increasing precipitation could be the primary cause. While in southeast regions with energy-limited environment, effective moisture is projected to decrease; increased atmospheric water demand is likely to be attributed.

[42] It should be noted that our analysis is based on the simulation of dynamic interactions between natural terrestrial vegetation and water; the dynamic interactions are a critical prerequisite for realistic assessments of large-scale past and future changes in water supply and demand [Gerten et al., 2004]. Open water evaporation is absent in the LPJ model which will lead to a decrease in estimation of actual evapotranspiration in water bodies. Nevertheless, the present result of modeled ETa is larger than the water balance estimates. The uncertainties introduced by neglecting water bodies and other land cover types in the LPJ model need further investigation. Uncertainties in our analysis are also attributed to the quality of climate forcing data. The distribution of meteorological stations on the Tibetan Plateau is uneven, and stations are sparsely distributed on the western plateau, which may adversely affect the accuracy of regional averages [Liu and Chen, 2000], as well as the robustness of interpolation results. Although use of the RCM PRECIS improves the spatiotemporal details of climate projections inadequately resolved by general circulation models, it adds further uncertainties to projections of climate change at national scales [Xiong et al., 2007]. As stated by many other studies, uncertainties in LPJ-simulated changes also arise because human interventions, such as water withdrawal and land use change, have been ignored [Gerten et al., 2004]. However, the uncertainties are relatively small for the central and western regions of the Tibetan Plateau, where the influence of human activities is small. Further research is required to assess changes in evapotranspiration components, primary climatic driving forces, and future effects of vegetation dynamics on evapotranspiration changes.

5 Conclusions

[43] In this paper, a 30 year series of evapotranspiration on the Tibetan Plateau, analyzed using linear regression analysis and the nonparametric Mann-Kendall trend test, indicate a statistically significant decreasing trend in ETo during the period 1981–1997, and a transition to a significant increasing trend since the late 1990s. The LPJ-modeled ETa on the Tibetan Plateau increased significantly during the period 1981–2010 (p < 0.05), especially during the later 1990s and the 2000s. There is a slight and insignificant increasing trend in annual precipitation. In addition, effective moisture also showed a significant increasing trend during the period 1981–2010 (p < 0.001), a trend which was more obvious prior to 1997 because of decreasing ETo during this time.

[44] Future projections show that the significant increasing trend in ETa during the period 1981–2010 is likely to continue throughout the 21st century. In general, the predicted increments in both ETo and ETa are greater for the A2 scenario than for the B2 scenario, mainly reflecting the larger magnitude of temperature changes in the A2 than in the B2 scenarios. Effective moisture is also projected to increase, which indicates that drought stress will be lower for both the A2 and B2 scenarios. Spatially, effective moisture changes exhibit an overall northwest to southeast gradient, with increases in the northwest and decreases in southeast. In the 2050s and 2080s, areas with higher drought stress (decreases in effective moisture) in southeastern plateau regions are likely to shrink, mainly on account of enlargement of regions showing increased ETa. However, drought stress in southeastern plateau regions is likely to intensify, which is primarily related to increased evaporative demand. Our findings suggest that vulnerable alpine ecosystems are likely be negatively impacted by intensified drought stress in the southeastern region of the Tibetan Plateau under the influence of future climate change.


[45] This work was supported by the National Scientific Technical Supporting Programs of China (2012BAC19B02), the “Strategic Priority Research Program” of the Chinese Academy of Sciences, Climate Change: Carbon Budget and Relevant Issues (XDA05090304), and Commonweal Research Funding from Ministry of Environmental Protection of China (201009056). The authors thank Prof. Yinlong Xu from the Institute of Environment and Sustainable Development in Agriculture, Chinese Academy of Agriculture Sciences, for providing climate scenario data. The authors would like to thank Dr. Peili Shi from the Institute of Geographical Sciences and Natural Resources Research, Chinese Academy of Sciences, for providing latent heat flux data for the Damxung site of the ChinaFlux network.