Surface energy exchanges above two grassland ecosystems on the Qinghai-Tibetan Plateau

Surface energy exchanges above two grassland ecosystems on the Qinghai-Tibetan Plateau S. Liu, S.-G. Li, G.-R. Yu, X.-M. Sun, L.-M. Zhang, Z.-M. Hu, Y.-N. Li, and X.-Z. Zhang Key Laboratory of Ecosystem Network Observation and Modeling & Synthesis Research Center of Chinese Ecosystem Research Network, Institute of Geographic Sciences and Natural Resources Research, Chinese Academy of Sciences, Beijing 100101, China Graduate University of Chinese Academy of Sciences, Beijing 100049, China Northwest Institute of Plateau Biology, Chinese Academy of Sciences, Xining 810001, China Key Laboratory of Ecosystem Network Observation and Modeling, Institute of Geographic Sciences and Natural Resources Research, Chinese Academy of Sciences, Beijing 100101, China Received: 3 June 2009 – Accepted: 13 August 2009 – Published: 21 September 2009 Correspondence to: S.-G. Li (lisg@igsnrr.ac.cn) Published by Copernicus Publications on behalf of the European Geosciences Union.


Introduction
The energy exchange between land surface and the atmosphere is an important process affecting water vapor exchange and carbon sequestration of ecosystems (Ryu et al., 2008), and is critical in driving climate change from the local to even global scales (Raupach, 1998;Cai et al., 2003).The Qinghai-Tibetan Plateau is the highest plateau in the world with a mean altitude of greater than 4000 m above sea level.It is regarded as a sensitive region to global climate change because of their vastness in area and unique geophysical locations.For example, meteorological records for the Qinghai-Tibetan Plateau show that the annual mean temperature increased by Figures

Back Close Full Screen / Esc
Printer-friendly Version Interactive Discussion 0.16 • C per decade and winter temperature increased by 0.32 • C per decade from 1955 to 1996, which are higher than the increase in annual mean temperature of the northern hemisphere and at the same latitude due to global climate change (Zheng and Yao, 2006).Climate change has the potential to affect energy, water vapor and carbon dioxide exchanges of the alpine grassland ecosystem on the Qinghai-Tibetan Plateau.
Recently, a large number of studies on energy and water vapor exchanges have been undertaken in many grassland ecosystems.However, most of these studies have been done in savanna areas, temperate grassland in North America and the steppe on the Mongolian Plateau (Baldocchi et al., 2004;Dugas et al., 1999;Wever et al., 2002;Li et al., 2006a;Hao et al., 2007), and only very few studies have focused on the Qinghai-Tibetan Plateau (Gu et al., 2005;Yao et al., 2008), where alpine grassland occupied more than 60% of the area (Zheng, 2000).It is reported that soil water content is a major factor affecting the energy partitioning in the alpine meadow ecosystem based on 1-year measurement (Gu et al., 2005).However, plant functional-type, weather, and soil physical properties have a profound effect on water and energy fluxes (Baldocchi et al., 2004), and the shift in vegetation types induced by climate may feed back to regional climate (Eugster et al., 2000;Beringer et al., 2005).A model study has suggested that enhanced warming would cause northward shift of the alpine meadow and a reduction in shrub-dominated alpine steppe on the Qinghai-Tibetan Plateau (Song et al., 2005).Therefore, the study on a variety of vegetation communities with continuous flux measurements will advance our understanding of the energy balance on the Qinghai-Tibetan Plateau which has a complicated landscape.
In this paper, we used 4-5 years continuous flux data measured with the eddy covariance technique, from the ChinaFLUX program (Yu et al., 2006), to explore energy exchanges in an alpine shrub-meadow ecosystem and an alpine meadow-steppe on the Qinghai-Tibetan Plateau.There were big differences in vegetation type and dynamics of soil moisture between these two grassland ecosystems due to the differences in climate and soil properties, which affected the radiation balance and surface exchanges of energy and water vapor.The main objective of this paper was to study the mechanisms controlling the surface energy exchanges and answer the following two questions: 1. What results in the differences in radiation balance over these two ecosystems?
2. What are the main variables affecting surface energy partitioning over these two ecosystems?
2 Materials and methods

Site description
The research was conducted at two ecosystems representing the most dominant grassland types on the Qinghai-Tibetan Plateau including an alpine shrub-meadow ecosystem at Gancaitan (referred to as AM) and an alpine meadow-steppe ecosystem at Dangxiong (referred to as AS), respectively.The two sites are both located in the western part of China Grassland Transect (CGT) with a humid and semi-humid continental climate.At the AM, the soil type is silty clay loam and the bulk density is 0.98 g cm 3 for the surface horizon (top 30 cm).At AS, the soil type is sandy loam, and there are scattered stones in the surface soil profile and the vegetation growth is limited with nutrients.Table 1 presents more detailed descriptions about climate and vegetation at the sites.Further details can be found in Shi et al. (2006) and Hu et al. (2008).

Field measurements
The eddy covariance technique was used to measure the fluxes of carbon dioxide, water vapor, and energy at above-mentioned two grassland ecosystems (Yu et al., 2006).All 30 min fluxes were calculated online.We also measured meteorological variables such as air temperature (T a ), relative humidity, precipitation and volumetric soil water content (SWC).Four components of surface radiation balance (up and down long-and short-wave radiation) at each site were measured with the net radiometer (Model CNR-1, Kipp & Zonen, NY, USA).Soil heat fluxes at the depth of 5 cm were measured by two soil heat plates (HFP01, Hukseflux Inc., Delft, The Netherlands).For details see Yu et al. (2006).NDVI has been widely used to estimate green biomass and leaf area index (LAI).In this study, NDVI during the growing season (May-September) was used to examine the growth of vegetation (10-day maximum-value composite (MVC) with 1 km resolution, http://free.vgt.vito.be/index.php)due to the lack of LAI measurements in some years.

Data processing
Three-dimensional rotation was applied to align the coordinate system with the mean wind (Wilczak et al., 2001), and the correction for density effect due to heat and water vapor transfer was made according to Webb et al. (1980) during processing the raw 30-min flux data.When filling missing and spurious data, linear interpolation (for <2 h gap) and the mean diurnal variation (MDV) methods were used (Fu et al., 2006).The closure of energy balance is often calculated to assess the performance of the eddy covariance system.There were two studies about the closure of energy balance for sites of ChinaFLUX (Li et al., 2005;Yu et al., 2006), and the energy balance ratio (EBR) of grassland ecosystems ranged between 0.83 and 0.91 (Yu et al., 2006).The imbalance of energy budget may affect the evaluation of sensible heat (H) and latent heat (LE), but it was prevalent in FLUXNET (Wilson et al., 2002).Introduction

Conclusions References
Tables Figures

Back Close
Full

Calculations
Decoupling coefficient (Ω) and canopy surface conductance (g s ) were calculated to assess the effect of physiological and environmental factors on water vapor exchange.
The half-hourly g s was derived from the Penman-Monteith equation (Monteith and Unsworth, 1990): where ρ, C p and VPD are the air density (kg m −3 ), the specific heat of air (J kg and the atmospheric saturation vapor pressure deficit (kPa), respectively.And γ is the psychrometric constant (kPa K −1 ), ∆ is the slope of saturation vapor pressure as a function of air temperature (kPa K −1 ), and g a is the aerodynamic conductance (m s −1 ) which is computed with the method described by Lloyd (1995).R n and LE is net radiation and latent heat flux (W m −2 ), respectively, and G is soil heat flux (W m −2 ).
The dimensionless Ω was computed by: In calculation of g s and Ω, only midday values (09:00 to 16:00) were used and those obtained during the time periods with recorded precipitation, inadequate turbulence (u * <0.1 m s −1 ) and low PAR (PAR<25 µmol −2 s −1 ) were excluded (Wever et al., 2002).
In addition, the daily mean reflectivity of short-wave radiation at the surface (albedo) is the average when downward short-wave radiation >200 W m −2 to minimize the effect of low solar angles (Li et al., 2006a).Introduction

Conclusions References
Tables Figures

Back Close
Full 3 Results and discussion

Climate conditions and vegetation development
The seasonal and inter-annual variations in T a , VPD, wind speed at 2.2 m (u), SWC and precipitation are shown in Fig. 1.The seasonal variation of T a at two ecosystems showed similar pattern, while T a was usually higher at AS than at AM.Both VPD and u in the growing season were larger at AS than at AM.The mean annual precipitation during the study periods was 507±33 mm at AM and 418±130 mm at AS, respectively.SWC was lower at AS due to the lower soil water-holding capacity.Inter-annual comparison of T a and precipitation in the growing season at AM and AS are shown in Table 2.The precipitation in the growing season of 2006 (only 207 mm) was significantly lower than the average (480 mm) at AS. Seasonal and inter-annual variations in NDVI are shown in Fig. 2. The NDVI was generally lower and its inter-annual and intra-seasonal variations were higher at AS than at AM.The mean NDVI in the growing season was 0.27 in 2004 and 2005, higher than in 2006 and 2007.This pattern roughly agrees well with the variation of precipitation at AS (Table 2).The average NDVI in May of 2007 was higher than those of other four years at AM.This is likely due to higher temperature observed in May of 2007 that enhanced plant growth.The average temperature in May of 2007 was 5.33 • C, while it ranged from −0.22 • C to 3.05 • C during the previous four years.Yang and Piao (2006) suggested that the largest NDVI increase during the growing season occurred in the spring is due to the advanced growing season and accelerated vegetation activity for the grasslands on the Tibetan Plateau.

Seasonal and inter-annual variations in energy fluxes and albedo
For the homogenous grassland ecosystems, net radiation (R n ) at the canopy surface can be derived from the radiation balance equation: Seasonal variations in the daily albedo during the growing season are shown in Fig. 4. Daily albedo varied from 0.08 to 0.15 at AM, and relatively very high (very low) albedo was observed after heavy snowfall (after strong rainfall) (Fig. 4).The albedo at AM was close to that of European sub-alpine wet grassland ecosystems (0.11 in August and September) (Rosset et al., 2001).The daily albedo was usually higher at AS than at AM, ranging from 0.12 to 0.23, which was close to those of an European grassland ecosystems (0.17-0.19) at the same latitude (Rosset et al., 1997), and a central Mongolian steppe (0.13-0.26) (Li et al., 2006a).The variation of albedo was closely associated with the variation of SWC and the vegetation status (Li et al., 2006a).For example, a significant linear relationship at AS was found between albedo and SWC at the 20 cm depth (SWC in %) (albedo=−0.003SWC+(0.208±0.002),n=591, adjusted R 2 = 0.504, F = 602.27,P <0.0001).Gu et al. (2005) suggested that the high summer-Introduction

Conclusions References
Tables Figures

Back Close
Full Screen / Esc Printer-friendly Version

Interactive Discussion
Figure 5 shows the seasonal variations of sensible heat (H), LE and G.The highest H values were often observed from mid April to mid June at both sites.With the increase of rainfall and growth of vegetation, LE increased gradually while H decreased.The H increased again after the peak growth.H was generally higher at AM than at AS (Fig. 5).LE was higher in May at AM than at AS due to the higher SWC at AM than at AS.The seasonal pattern of G was similar at both sites.Gwas usually negative from early November to early February, indicating that the ground was energy source for the atmosphere.Gwas generally higher at AS than at AM from late February to late March, which was consistent with the seasonal variation of K d (Fig. 3).During the study periods, average value of G/R n in the growing season at both sites kept almost constant of about 0.04.

Comparison of inter-annual variations in the Bowen ratio between two ecosystems
The Bowen ratio (β=H/LE) is widely used for comparing the energy partitioning among different ecosystems (Wilson et al., 2002).The inter-annual variations in β were low at AM, ranging between 0.55 and 0.68.Similarly, T a , NDVI and canopy surface conductance (g s ) varied little inter-annually at AM (Table 2).The average β in the growing season of 2007 was lowest, which may stem from the abnormal high temperature in May as mentioned above (Sect.3.1).The AM ecosystem is characterized by low temperature (Table 1).More water vapor was evapotranspirated with the advanced vegetation growth under the conditions of warm spring in 2007 at AM.The amount of ET was 95 mm in May of 2007, higher than previous four years (varied from 54 to 60 mm).
By contrast, at AS, β varied from 0.37 (2004) to 0.66 ( 2006), being consistent with the wide variation in precipitation, indicating that water availability controls the energy partitioning at inter-annual scale.Wilson et al. (2002) suggested that the large variability of β (0.34-1.91) between sites and years at grassland can be explained by two reasons: at first, the grassland ecosystem is highly sensitive to human disturbance (e.g.grazing); and secondly it is 9169 Introduction

Conclusions References
Tables Figures

Back Close
Full Screen / Esc

Printer-friendly Version
Interactive Discussion more sensitive to soil moisture than shrubs and forests due to the difference in the ability of roots to access deep soil water.In this study, the difference of β at two sites can be explained by climate, soil type and vegetation.On the one hand, Indian monsoon has a greater influence on precipitation in the south of the Qinghai-Tibetan Plateau (Tian et al., 2001).During the study periods, the coefficient of variation of the annual precipitation was 31.1% at AS, higher than that at AM (6.5%).On the other hand, soil water-holding capacity was lower at AS than at AM.In addition, root mass was mainly distributed in the soil layer of 0-10 cm at AS, while it was in 0-40 cm depth at AM (Li et al., 2006b), resulting in further intensification of water limitation at AS.The vegetation distribution may change due to the climate change on the Qinghai-Tibetan Plateau (Song et al., 2005).However, more attention should be given to the effect of grazing.On the Qinghai-Tibetan Plateau, a large area of grassland has been degraded in the past few decades because of overgrazing (Cui et al., 2007), e.g. the AS site in this study.Bremer et al. (2001) suggested that grazing decreased ET, and thus the β increased.The increase in β indicates a higher fraction of available energy (R n −G) is partitioned into H which will cause a positive feedback to atmospheric temperature.Eugster et al. (2000) suggested that the vegetation change from tundra to shrub or shrub to forest in arctic would cause a negative feedback because forests had a lower ET rate, which reduced the water vapor flux to the atmosphere as a strong greenhouse gas.There was little information about effect of vegetation change on ET on the Qinghai-Tibetan Plateau.From our study, we learned that a conversion from AM to AS or vice versa would alter surface energy balance and the partitioning of the available energy though the change of surface albedo, giving rise to a change in local or even regional climate.However, there was still some uncertainty about climate change and the feedback from vegetation change on the Qinghai-Tibetan Plateau.This deserves further study in the future.Introduction

Conclusions References
Tables Figures

Back Close
Full Screen / Esc Printer-friendly Version Interactive Discussion

Comparison of seasonal variations in the Bowen ratio between two ecosystems
The seasonal variation of daily β during the growing season at AM and AS is shown in Fig. 6a, exhibiting a reverse trend with plant growth.However, the range of variations in daily β was larger and the values of β often increased markedly when the SWC was low at AS versus at AM, which was consistent with the seasonal variation of daily surface conductance (g s ) during the growing season (Fig. 6b).At both sites, the g s was relatively high when NDVI was high and soil was moist.The maximum daily average g s could reach about 20 mm s −1 at both sites, close to that of an annual grassland ecosystem in California (Ryu et al., 2008), and higher than that of temperate grassland in Canada (∼9 mm s −1 , lower in droughty year) (Wever et al., 2002).
Figure 7 illustrates the relationship between the monthly average β and NDVI, surface conductance (g s ) at AM and AS.With the increase of g s and NDVI, indicative of the growth status of vegetation, β decreased, indicating that more water vapor was evapotranspirated into the atmosphere.The seasonal pattern of β suggests that the phenology of the vegetation was the major factor affecting the energy partitioning at both sites at intra-seasonal scale.This is consistent with previous studies in that vegetation characteristics strongly influence the energy partitioning (e.g.Li et al., 2006a;Hammerle et al., 2008).
The growth of vegetation not only affects the energy partitioning, but also has great influence on partitioning of evapotranspiration into its components (plant transpiration and soil evaporation).Hu et al. (2009) suggested that the variations in LAI could cause stronger effects on the ratio of evaporation (E) to evapotranspiration (ET) in the sparsecanopy ecosystems than in the dense-canopy ecosystems.It was found that E/ET in the growing season was above 58% at AS, and only about 42% at AM, respectively (Hu et al., 2009).
Figure 8 shows the relationship between the daily average β and SWC at AM and AS.β decreased with the increase of SWC at AS.However, there was no significant Introduction

Conclusions References
Tables Figures

Back Close
Full Screen / Esc

Printer-friendly Version
Interactive Discussion correlation at AM because water availability was not the limiting factor of energy partitioning.
Previous studies have shown that soil moisture is an important factor affecting energy partitioning, particularly in arid and semi-arid areas.For example, β decrease with the increase of SWC after the precipitation events over the semi-arid grassland ecosystems on the Mongolian plateau (Hao et al., 2007;Li et al., 2006a).Gu et al. (2005) also suggested that soil moisture was an important factor affecting the energy partitioning in another alpine meadow ecosystem dominated by Kobresia humilis close to AM.However, soil water was not the limiting factor of evapotranspiration at AM.It can be explained by the big differences in soil type (Mat-Cryic Cambisol at Gu et al. (2005), Mol-Cryic Cambisols at AM) and vegetation type between these two ecosystems resulted from the different direction and gradient of slope.

Comparison of biophysical control on energy partitioning at half-hourly scale between two ecosystems
Energy partitioning is principally associated with the complicated interaction between biotic and abiotic variables such as green leaf area index, plant functional type, net radiation, soil moisture availability and vapor pressure deficit (VPD) (Gu et al., 2006;Li et al., 2006a).However, effects of these factors on β are different at different time scales, e.g.LAI and SWC varied little at a 30-min resolution while R n and VPD varied more at a 30-min than at a daily scale.To further understand the mechanisms controlling the surface energy exchanges at two ecosystems, the effects of R n and VPD on energy partitioning were analyzed at half-hourly time step.

Effect of net radiation
The Bowen ratio slightly increased with R n at both sites (Fig. 9), but drought increased the range of variation at AS (Fig. 9a) which was similar to that at a temperate forest ecosystem (Gu et al., 2006).It is well known that both H and LE increase with R n ;

Conclusions References
Tables Figures

Back Close
Full Screen / Esc Printer-friendly Version Interactive Discussion however, their increases are not in proportion because LE is limited by the amount of water available in the soil and stomatal openness (Gu et al., 2006).Therefore, β tends to increase with the increase ofR n at AS and AM, and the trend was stronger under the conditions of drought at AS (Fig. 9a).Gu et al. (2006) suggested that the difference in the Bowen ratio between drought and non-drought conditions increased with R n and VPD because more available energy was partitioned into LE with the increase of R n , and this trend was intensified at higher levels of VPD.However, in this study, there was no significant difference in effects of R n on β at two levels of VPD (<1 kPa and >1 kPa) no matter what soil moisture conditions at AS and AM (data not shown), which means that the effect of R n on β did not depend sensitively on VPD.This may be related to the low values of VPD at both sites (generally below 1.5 kPa).

Effect of vapor pressure deficit
Figure 10 shows the effect of VPD on energy partitioning at two levels of SWC (nondrought and drought) at AS.Under non-drought conditions, there was no distinct effect of VPD on β at AS (Figures 10a), but β increased with increased VPD under drought conditions (Fig. 10b).In addition, the effect of VPD on β with different levels of R n can be ignored under all levels of soil moisture (data not shown).The effects of VPD on energy partitioning at two levels of R n (<250 Wm −2 and >500 Wm −2 ) at AM was shown in Fig. 11.Overall, β decreased with the increased VPD at AM, but the decrease was sharper at high levels of R n (>500 Wm −2 ) (Fig. 11b).And the effect of R n between 250 Wm −2 and 500 Wm −2 was medium (data not shown) at AM.
The influence of VPD on energy partitioning was stronger at AM than at AS under non-drought conditions (Figs. 10a and 11), and it was nonlinear at AM (Fig. 11).What caused the difference in effect of VPD on β at AS and AM?Studies showed that the nonlinear effect of VPD on β stems from two simultaneous but opposing effects of VPD on LE: on the one hand, LE increased with the increased VPD driving more water vapor out of the stomatal pore; on the other hand, high VPD reduced the stomatal 9173 Introduction

Conclusions References
Tables Figures

Back Close
Full conductance limiting the LE (Ball et al., 1987).The positive effect of increased VPD outweighs its negative effect on LE when VPD is low (Gu et al., 2006), which explained why β decreased with the increased VPD at AM. Soil evaporation dominated the evapotranspiration at AS as mentioned above due to low LAI (Hu et al., 2009).Therefore, the effect of VPD on stomatal conductance was weaker at AS versus at AM. Further, there was a high correlation between high VPD and severe drought resulting in more available energy was partitioned into H, and thus increased β at AS (like Fig. 10b).Therefore, the difference in effect of VPD on β resulted from the difference in vegetation cover between two sites.The decoupling coefficient (Ω) is often used to separate the effect of VPD on LE from R n (Gu et al., 2005;Hao et al., 2007).Ω varies from 0 to 1.A low Ω indicates a relatively high influence of VPD on ET, and a high Ω suggests that R n exerts the dominant influence on ET (Jarvis and McNaughton, 1986).The values of Ω were often over 0.7 during the peak growth stage at both sites (not shown in this paper), implying that evapotranspiration process at two sites is coupled to the radiation forcing.
However, Ω was low during the dry period at AS (e.g. in August 2006 and May-June in 2007), which indicates that the control of VPD on LE was strengthened.Surface conductance (g s ) and aerodynamic conductance (g a ) are two parameters determining Ω directly.Here, the effects of g s and g a on Ω were discussed with the data of 2004 at half-hourly scale.The dependencies of Ω on g a and g s were in opposite directions (Fig. 12).As g a decreased and g s increased, thus Ω increased, suggesting that evapotranspiration became more coupled with radiation forcing.The magnitude of g a depends on the aerodynamic properties (e.g.wind speed and roughness) of the surface.The wind speed was larger, resulting in lower g a at AS compared with AM.Therefore, the value of Ω was higher in 2004 at AS than AM (Table 2) due to the lower

Summary
The seasonal and inter-annual variations in surface energy exchanges at two grassland ecosystems on the Qinghai-Tibetan Plateau were explored.Relative to the alpine meadow-steppe (AS), solar radiation (K d ) was lower, but net radiation (R n ) was higher at the alpine shrub-meadow ecosystem (AM).This is because the surface reflectivity for short-wave radiation (albedo) and the ratio of upward long-wave radiation to downward long-wave radiation (L u /L d ) were lower at AM as compared with AS.During the growing season, daily albedo varied from 0.08 to 0.15 at AM and from 0.12 and 0.23 at AS in response to variations of NDVI and soil moisture.It decreased significantly with the increase of SWC at AS.The mean L u /L d ratio in the growing season was 1.23 at AM and 1.33 at AS, respectively.In the annual scale, annual precipitation was the most important factor controlling energy partitioning at AS, while temperature of May controlled the inter-annual variations in the Bowen ratio (β) at AM which is characterized by low temperature.Vegetation growth played an important role in controlling energy partitioning at seasonal scale at both sites.However, abnormally high β often occurred during the growing season under the drought conditions due to the low soil water-holding capacity at AS as compared with AM.In addition, the influence of VPD on energy partitioning was stronger under non-drought conditions at AM than at AS at half-hourly scale due to high vegetation cover.Introduction

Conclusions References
Tables Figures

Back Close
Full

Fig. 1 .
Fig. 1.Variation of (a) monthly average of daily air temperature (T a ), (b) monthly average of daily VPD, (c) monthly average of daily wind speed, (d) monthly average of daily volumetric soil water content at 20-cm depth (SWC), and (e) monthly sum of precipitation (PPT) at the shrub-meadow ecosystem (AM) and the alpine meadow-steppe ecosystem (AS).
K and L represent the short-wave radiation and long-wave radiation, respectively.And subscript n, d and u represent net, downward, and upward directions, respectively.Figure3shows the seasonal and inter-annual courses of K d and R n .Their seasonal pattern was similar at both sites.K d was higher at AS than at AM and the total K d in the growing season (May-September) was 3331.9±90.5 MJ m −2 at AS, andThe ratio of R n /K d was about 0.4 at AS, but 0.6 at AM.The mean albedo (=K u /K d ) and L u /L d ratio at AM in the growing season were 0.12±0.01 and 1.23±0.02,respectively, both being lower than those of AS (0.17±0.01 for K u /K d and 1.33±0.04for L u /L d ).The ratio of net long-wave radiation to net short-wave radiation was 31.1% at AS and 46.6% at AM, respectively.The lower albedo and lower L u /L d were responsible for higher R n at AM relative to the AS. where