Abstract

Seasonal and interannual variations of evapotranspiration (ET) and energy exchange were estimated over degraded grassland and cropland land surfaces in a semiarid region of northeastern China using the eddy covariance technique from 2003 to 2008. The peak daily ET, which occurred in August, was 1.5–4.5 mm day−1 for the degraded grassland and 1.5–5.5 mm day−1 for the cropland land surface. Annual cumulative ET was roughly equal to annual precipitation at both sites. However, the annual cumulative ET at the cropland site was slightly larger (about 10–30 mm) than it was at the grassland at the end of each year. More water might come from irrigation at seedtime and from the soil. With the factor analysis technique, the results revealed that the atmospheric water demand was the most important factor in the ET process on a half-hour time scale in this semiarid area. On a seasonal time scale, ET was greatly constrained by surface conductance and precipitation; on an annual time scale, ET was greatly constrained by the total amount of precipitation at both sites. The accuracy of ET estimation using the Penman–Monteith formula in this semiarid area was also discussed.

1. Introduction

Evapotranspiration (ET) plays a critical role in ecological and hydrological processes. It also has significant impacts on the evolution of the planetary boundary layer (Barr and Betts 1997; Hill et al. 2008) and influences local weather and climate (Dirmeyer 1994; Pielke et al. 1998; Betts et al. 1999; Sun and Wu 2001). Semiarid and arid areas cover nearly 40% of the earth’s continental surface (Verhoef et al. 1999). The ET in these areas is different from that in humid areas on seasonal and interannual scales; for example, the annual ET in semiarid areas uses a large part of the precipitation falling on the land surface (Sala et al. 1992), which suggests ET is approximately equal to precipitation. However, because most ET studies have focused on humid areas, information about the seasonal and interannual variations of ET and its behavior in semiarid areas is still lacking, especially in China, where studies of ET have focused on humid or semihumid cropland (Liu et al. 2002; Kang et al. 2003), grassland (Li et al. 2005), forest (Shi et al. 2008), and alpine meadow (Gu et al. 2008) ecosystems.

Evapotranspiration is considered to be the most controversial item in hydrological budgets because of its uncertain response to global warming (Brutsaert 2006; Golubev et al. 2001; Wetherald and Manabe 2002; Liepert et al. 2004; Liu et al. 2004; Ramanathan et al. 2001; Roderick and Farquhar 2002; Wild et al. 2004). The uncertainties originate from the indirect nature of the measurement of ET. The traditional method of estimating ET is based on pan evaporimeters, using the complementary relationship between pan evaporation and actual evaporation (Rana and Katerji 1996; Roderick and Farquhar 2002). This indirect method lacks consideration of biophysical mechanisms, that is, the stomatal conductance and leaf area index (LAI). A direct means of measuring flux using an eddy covariance (EC) technique has been developed in recent decades and is now used in networks such as Americaflux, Euroflux, and Asiaflux (Baldocchi et al. 2001; Valentini et al. 2000; Aubinet et al. 2000; Running et al. 1999). However, because of the short history of EC measurement, most studies of ET are based on observations that focus on just a single growing season or a single year. There have been a few multiyear ET studies based on EC (some covering more than 5 years), for example, a study of a grassland in California (Ryu et al. 2008) and a study of a larch forest in eastern Siberia (Ohta et al. 2008).

The ET is influenced by many biological and environmental factors (Jarvis and McNaughton 1986; Blanken et al. 1997; Wilson and Baldocchi 2000). Assuming advection can be ignored, environmental variables, such as precipitation and soil moisture, greatly constrain the upper limit of ET, whereas net radiation, wind speed, and air humidity affect the variations of ET. High values of vapor pressure deficit (VPD) can produce strong humidity gradients between the leaf and the atmosphere that enhance transpiration and simultaneously stimulate stomatal closure adjustments of the plants to reduce water loss (Baldocchi and Xu 2007; Dang et al. 1997; Cornic 2000). The annual precipitation is recognized as being highly variable in semiarid areas (Sala and Lauenroth 1982). The seasonal and interannual variations of the factors controlling ET and energy in these areas still need to be investigated. On the other hand, the controlling factors are interactive within the coupled biosphere–atmosphere system, making it difficult to interpret the ET as a result of the controlling factors. Bulk canopy parameters, including surface conductance, the Priestley–Taylor coefficient, and the decoupling coefficient, can be used to estimate the effects of these controls on ET. Multiyear eddy covariance observations can establish the magnitude and the seasonal or interannual variations of these parameters, thus improving the estimation of the boundary conditions used in weather forecasting and in hydrological and climate models (Wilson and Baldocchi 2000). Evaluation of these parameters may also be useful in examine the complex linkages between carbon and water cycles and trace gas fluxes (Baldocchi and Meyers 1998).

In the present study, the seasonal and interannual variations of ET and energy exchange were investigated based on 6 years of eddy covariance measurements over two semiarid land ecosystems in northeastern China. The data include two wet and dry cycles of precipitation within the 6-yr period compared with the climate average (390 mm). The main objectives of this research are 1) to document the variations of evapotranspiration and energy distribution at seasonal and interannual time scales, 2) to estimate the magnitudes of bulk physiological parameters and their seasonal and interannual variations, and 3) to investigate the primary environmental and biological factors controlling the variations of evapotranspiration over seasonal and interannual cycles.

2. Materials and methods

a. Site description

The Tongyu observation station, which is one of the reference sites of the Coordinate Energy and Water Cycle Observations Project (CEOP, http://www.ceop.net), is located in a semiarid area of northeastern China (44°25′N, 122°52′E, 184 m above sea level). The station includes two surfaces for intercomparison, that is, degraded grassland and cropland site. The mean annual precipitation observed from 1961 to 2002 at the Tongyu weather station, about 30 km northeast of the observation station, was 390 mm, of which approximately 80% occurred during the growing season (May–September). Average air temperatures for January and July were −15.7° and 24°C, respectively. The soils in this area are classified as sandy loam, and the terrain is fairly open and flat, with slopes less than 1° in all directions, thus minimizing terrain-related influences on the EC measurements. The distance between the grassland and cropland site is about 5 km. The grass is usually less than 10 cm high, and the vegetation coverage is less than 70% in the growing season. The main crop in the cropland is corn mixed with sunflower (account for about 30%). The maximum height of the corn reaches about 2 m in the growing season. After harvest, the soil is bare. Soil water for plant growth mainly depends on natural precipitation, except for some irrigation during seeding time in the cropland. Based on the footprint analysis method developed by Horst and Weil (1994), the peak for the flux footprint is located at approximately 40 and 50 m upwind for the grassland site and the cropland site, respectively. The 85% flux fetch was at the distance of about 500 and 600 m upwind in the unstable condition for the grassland site and the cropland site, respectively. The fetch is adequate for eddy covariance measurements.

b. Observations

Long-term and continuous EC measurements have been conducted since October 2002. Here we use the 6 years of continuous data from 2003 to 2008. Meteorological measurements were made with instruments installed on a 20-m tower at each site. Profiles of air temperature (HMP45C; Vaisala, Inc.), relative humidity (HMP45C), and wind speed (034A; Metone, Inc.) were measured at 1.76, 4.36, 8.36, 12.86, and 17.46 m for the degraded grassland site and at 2.36, 4.36, 8.36, 12.36, and 17.06 m for the cropland site. Wind directions (014A; Metone) were measured at the highest of each of the five levels at both sites. Barometric pressure was measured with a barometer (CS105; Texas Electronics, Inc.). Net radiation components, including incoming and outgoing solar radiation (CM21; Kipp and Zonen, Inc.) and incoming and outgoing longwave radiation (CG4; Kipp and Zonen), were measured on a mast at 2 m above the surface at the degraded grassland site and at 3 m above the surface at the cropland site. We also measured soil temperature profiles at depths of 2, 5, 10, 20, 50, and 80 cm with temperature sensors (STP01_L50; Hukseflux, Inc., and 107_L; Campbell Scientific, Inc.); soil heat flux at depths of 2 and 10 cm with soil heat plates (HFP01SC_L50; Hukseflux); and soil moisture profiles at depths of 5, 10, 20, 40, and 80 cm with time domain reflectometers (TDRs) (CS616; Campbell Scientific). A tipping-bucket rain gauge (TE525MM; Texas Electronics) was used to measure precipitation near the tower. The frequency for fast (EC) measurement was 10 Hz; for slow (meteorological) measurements it was 3 s per sample. All of these meteorological variables were recorded at 30-min resolution with a Campbell Scientific CR-23X datalogger. More details of the instrumentation are given by Liu et al. (2008).

Water vapor, CO2, and sensible heat fluxes were measured continuously using an eddy covariance measurement system that includes a 3D ultrasonic anemometer–thermometer (model CSAT3; Campbell Scientific) and an open-path infrared gas analyzer (IRGA) (model LI-7500; Li-Cor, Inc.). The two sensors were oriented toward the prevailing wind, and the distance between them was approximately 15 cm for the grassland and 12 cm for the cropland land to minimize the underestimation of turbulent fluxes (Lee and Black 1994). The instruments were mounted at a height of 2 m above the surface of the degraded grassland site and 3.5 m above the surface of the cropland site. Periodic checks were made to ensure that there was no signal drift and no degradation due to accumulation of dust and other aerosols on the acoustic anemometer heads. The IRGA were also calibrated every half-year against standard gases for CO2 calibration and were referenced to a portable dewpoint generator (Li-610; Li-Cor) for water vapor calibration. Raw data were recorded at 10 Hz using Campbell Scientific CR-5000 dataloggers with 1-GByte PC cards and transferred to the laboratory every month for postprocessing using the EdiRe software (developed by J. Moncrieff, School of Geosciences, University of Edinburgh).

c. Data processing and gap filling

Before calculation of the turbulent fluxes, raw data were screened for out-of-range records, and all possible sensor or logger malfunctions were also detected and removed. Double coordinate rotations were carried out so that the mean vertical wind velocity for 30 min was forced to be 0 (McMillen 1988; Kaimal and Finnigan 1994). Corrections to the original fluxes were made for high-frequency losses due to separation of sensors, path averaging, and sensor frequency response (Massman and Lee 2002). Water vapor and carbon dioxide fluxes were corrected for the effect of density fluctuations (Webb et al. 1980). Data quality and assurance were assessed using the stationarity and stability criteria suggested by Foken et al. (2004). Raw data within 30 min that did not meet the criteria were eliminated in the analysis and replaced by the gap-filling strategy. Energy closure of the EC measurement was checked using the energy balance ratio (EBR):

 
formula

where LE is the latent heat flux (W m−2), Hs is the sensible heat flux (W m−2), Rn is the net radiation (W m−2), and G is the soil heat flux (W m−2).

Data gaps were caused by system maintenance, power failures, sensor calibration, spike removal criteria, quality tests, and the manual discards described above, leading to a total gap percentage in the flux measurement of about 25% over the 6-yr study period. The resulting gaps in sensible heat fluxes and latent heat fluxes were filled by a multi-imputation method proposed by Hui et al. (2004). Variables—including net radiation, incoming shortwave radiation, air temperature, VPD, volumetric soil moisture content at a depth of 5 cm, and wind speed—were subjected to the imputations. Gap filling for other variables was based on linear interpolation for small gaps, usually less than a few hours, whereas larger gaps were filled using the mean diurnal method (Falge et al. 2001).

d. Factor analysis

A multivariate statistical analysis technique, namely factor analysis, is applied in order to identify the relative importance of each variable and to derive a set of factors controlling ET. The theory and implementation of this technique has been well described in Anderson (2003). Briefly, it reduces the dimensionality of the data by detecting the underlying structure of relationships so that the given data can be rearranged into a small number of independent variables. The analysis was done including two steps, namely principal component analysis (PCA) and factor analysis (FA). Based on the half-hour time series, Monthly average values of six meteorological variables, namely, precipitation (PPT), wind speed U, net radiation Rn, VPD, soil water content at 5 cm below the surface Ws, and air temperature Ta were considered into the investigation. PCA was performed on the correlation matrix of the variables for each site separately. A varimax factor rotation method suggested by Kaiser (1958) was used to produce a set of orthogonal factors that are independent or uncorrelated.

e. Calculations of bulk parameters

To interpret the controlling factors for seasonal and interannual variations of ET, three bulk parameters were investigated: the surface conductance, the decoupling factor, and the Priestley–Taylor coefficient. The ET can be estimated using the Penman–Monteith equation (Monteith and Unsworth 1990):

 
formula

where is the latent heat of vaporization (J kg−1), Rn is the net radiation (W m−2), G is the soil heat flux (W m−2), is the air density (kg m−3), Cp is the specific heat capacity of air at constant pressure (J kg−1 K−1), Δ is the slope of the saturation vapor pressure–temperature curve (kPa K−1), is the psychrometric constant (kPa K−1), VPD is the water vapor pressure deficit (kPa), and Ga is the aerodynamic conductance (m s−1). Therefore, the surface conductance Gs (mm s−1) can be calculated by inverting Eq. (1):

 
formula

The quantity Ga is calculated by the following equation (Monteith and Unsworth 1990):

 
formula

where u* is the friction velocity (m s−1) measured by the ultrasonic anemometer, and U is the mean wind speed (m s−1) interpolated from the layer nearest to the ultrasonic anemometer in the wind profile.

The decoupling coefficient, which indicates the degree of canopy decoupling from the bulk air, is subsequently calculated by the following equation (Jarvis and McNaughton 1986):

 
formula

The Priestley–Taylor coefficient is defined as the ratio between the measured evaporation () and the equilibrium evaporation () (Priestley and Taylor 1972):

 
formula

It provides a method for comparing the measured evaporation with the climatologically expected evaporation, assuming a closed volume and constant net radiation over a wet ecosystem (McNaughton and Spriggs 1986). Here is calculated by the following equation (Priestley and Taylor 1972):

 
formula

The Jarvis-type surface conductance is calculated using the equation proposed by Jarvis (1976):

 
formula

where Gsm is maximum stomatal conductance, LAI is leaf area index, and the function-dependent variables PAR, T, D, ψ, and C are photosynthetically active radiation, air temperature, specific humidity deficit, leaf water potential, and canopy-space concentration of carbon dioxide, respectively.

To avoid the numerical instability when the denominator approaches 0, the daytime values of three bulk parameters (Gs, Ω and ) from 1000 to 1400 LT are used to calculate the monthly mean values in this study.

3. Results

a. Meteorological conditions

Weather conditions during the 6-yr period and the climatology at the Tongyu site are summarized in Tables 1 and 2. The mean annual precipitation was 296 mm for the 6 years, with a standard deviation of 85 mm. The year 2006 was considered to be a normal year in the study period. In the wet years (2003, 2005, and 2008), the average annual precipitation was 363 mm, and in the dry years (2004 and 2007), it was 202 mm. Abnormally high precipitation occurring in July (149 mm) and August (127 mm) of 2008 contributed to the very high overall precipitation in that year. The 6-yr average air temperature was 6.7° ± 0.6°C, which was slightly higher than the 42-yr average value (5.7° ± 0.9°C). The monthly average daily air temperature in the growing season was usually above 20°C and decreased to less than 0 from December to February (Fig. 1a). The annual average wind speed, integrated global solar radiation and vapor pressure deficit were shown in Table 2. No major differences in these measures were observed during the 6 years. Generally, the soil water content at 5 cm varied with the change of precipitation, increasing abruptly in May or June when the rainy season started and reaching its peak in the middle of the growing season (Fig. 1b). The monthly average volumetric soil water content ranged from 0.15 to 0.3 in the growing season.

Table 1.

Monthly precipitation and air temperature from the Tongyu weather station.

Monthly precipitation and air temperature from the Tongyu weather station.
Monthly precipitation and air temperature from the Tongyu weather station.
Table 2.

Summary of meteorology and bulk parameters at the grassland and cropland sites over 6 years.

Summary of meteorology and bulk parameters at the grassland and cropland sites over 6 years.
Summary of meteorology and bulk parameters at the grassland and cropland sites over 6 years.
Fig. 1.

(a) The monthly average daily maximum and minimum air temperature at the Tongyu weather station. (b) The monthly average volumetric soil water content at 5 cm below the ground surface.

Fig. 1.

(a) The monthly average daily maximum and minimum air temperature at the Tongyu weather station. (b) The monthly average volumetric soil water content at 5 cm below the ground surface.

b. Vegetation cover

The vegetation canopy of the two sites is described using the normalized difference vegetation index (NDVI) and the enhanced vegetation index (EVI). The NDVI and the EVI are alternative measurements of the vegetation canopy associated with the leaf area index and the percentage of vegetation cover (Hunsaker et al. 2003; Nagler et al. 2005). The data were obtained from the Moderate Resolution Imaging Spectrometer (MODIS) on the Earth Observing System (EOS)-1 Terra satellite, with a high resolution of 250 m, at 16-day intervals. Downscaling with atmospheric correction to the two sites has been made before this analysis (Hadjimitsis et al. 2010; Hwang et al. 2011). Seasonal variations of the NDVI and the EVI were similar; the major difference between the two vegetation indexes was that the EVIs were generally smaller than the NDVIs (Fig. 2). Values of the NDVI at the two sites ranged from 0.4 to 0.8 in the growing season (Fig. 2a). The NDVI at both sites rapidly increased from May to June after the germination of the plants. NDVI increases because of the buildup of canopy structures, that is, stem and leaf. Short periods of maximum NDVI occurred in the middle of July followed by a gradual decline as the soil moisture was exhausted (Fig. 1b). After the grasses wilted and the corn was harvested, the NDVI stayed around 0.2 in the winter. The interannual variation in the NDVI was associated with differences in soil water availability for the plants. For example, the maximum NDVI of 0.62 recorded in 2005 at the grassland site was almost twice the maximum recorded in 2004, which is consistent with the difference in soil water content. Seasonal or interannual changes in the EVI were the same as the changes in the NDVI except for the magnitude (Fig. 2b).

Fig. 2.

(a) Normalized difference vegetation index and (b) enhanced vegetation index at 500-m resolution and 16-day intervals derived from MODIS.

Fig. 2.

(a) Normalized difference vegetation index and (b) enhanced vegetation index at 500-m resolution and 16-day intervals derived from MODIS.

c. Surface energy closure

Lack of energy balance closure is common with eddy covariance measurement systems. The primary sources of energy imbalance could be sampling errors, instrument bias, neglected energy sinks, high- or low-frequency loss, and advection (Wilson et al. 2002). We evaluated the surface energy closure at our sites using Eq. (1) based on 30-min-average fluxes. The average value of the EBR was 0.81 for the two study sites (Table 3), which was near the average found at the FLUXNET sites (0.8; Wilson et al. 2002). A minimum EBR of approximately 0.75 was observed in the nongrowing season of 2003. The seasonal variations in monthly average diurnal cycles of the energy balance components are shown in Fig. 3. On the diurnal time scale, the four energy balance components gradually increased after sunrise and typically reached their peak values at midday at the grassland site. The maximum value of net radiation was about 400 W m−2 in the growing season, whereas it was about 250 W m−2 in the nongrowing season. The latent heat flux accounted for 18%–68% of the net radiation, whereas the sensible heat flux accounted for 16%–58% of the net radiation during the growing season on the daily scale. Similar patterns of diurnal cycles were observed at the cropland site, and no major differences in the magnitude of the four energy components existed between the two sites. Notably, there was a small lag of about 2 h between maximum net radiation and maximum soil heat flux at the cropland site, which is likely because the vegetation cover was denser at this site and heat flux would take more time to transfer into the ground. On the monthly time scale, sensible heat and latent heat flux changed dramatically during the growing season (Fig. 3). The maximum value of latent heat flux in the average diurnal courses in July or August was approximately 200 W m−2, accounting for more than half of the available energy consumption. In dry years, on the other hand, the midday peak values of Hs and LE were roughly equal. In the nongrowing season, Hs became the dominant component of the energy balance at midday in both dry and wet years.

Table 3.

The average EBR in the growing season and nongrowing season during the 6-yr period.

The average EBR in the growing season and nongrowing season during the 6-yr period.
The average EBR in the growing season and nongrowing season during the 6-yr period.
Fig. 3.

Seasonal variation of average diurnal cycles of the energy balance components: Rn, G, LH, and Hs. Each column represents average diurnal cycles for a given year in a different season and different site. Each row represents (a) grassland in the growing season, (b) grassland in the nongrowing season, (c) cropland in the growing season, and (d) cropland in the nongrowing season.

Fig. 3.

Seasonal variation of average diurnal cycles of the energy balance components: Rn, G, LH, and Hs. Each column represents average diurnal cycles for a given year in a different season and different site. Each row represents (a) grassland in the growing season, (b) grassland in the nongrowing season, (c) cropland in the growing season, and (d) cropland in the nongrowing season.

d. Seasonal and interannual variations of ET

Seasonal variations of ET were strongly correlated with the distribution of precipitation (R2 = 0.86), and the peaks of daily ET rates occurred in the middle of the growing season with the summer rainfall (Fig. 4). Generally, the peak daily ET in the middle of the growing season was 1.5–4.5 mm day−1 for the degraded grassland and 1.5–5.5 mm day−1 for the cropland during 2003–08. The 2008 growing season had the highest peak value of daily ET (4.7 mm day−1 for the grassland and 5.4 mm day−1 for the cropland site). The cropland site generally had higher peak daily ET rates than the grassland site. In contrast, daily ET rates in both sites were usually below 0.5 mm day−1 in the winter (Fig. 4). On the interannual time scale, ET varied considerably from year to year because of the large variations of the precipitation (Fig. 4). The estimation of annual total ET using the gap-filling criteria described in section 2c had uncertainties of 12–23 mm for the grassland site and 17–29 mm for the cropland site (Table 2).

Fig. 4.

Seasonal and interannual variation in daily ET in the (a) grassland and (b) cropland. (c) Monthly precipitation derived from the Tongyu weather station.

Fig. 4.

Seasonal and interannual variation in daily ET in the (a) grassland and (b) cropland. (c) Monthly precipitation derived from the Tongyu weather station.

e. Bulk parameters

The surface conductance calculated with the inverted Penman–Monteith equation represents the canopy integration of stomatal conductance, because it also encompasses the nonlinear effects of evaporation and canopy structure (Wilson and Baldocchi 2000). In general, the monthly average bulk surface conductance rapidly increased from June, reached its peak value in July or August, and declined in September (Fig. 5). The precipitation at the beginning of the growing season brings the rapid growth of vegetation, which abruptly increases the LAI and Gs, and subsequently increases ET (Fig. 4). A strong correlation (R2 = 0.7) between ET and NDVI or EVI in the growing season was observed in this area (Fig. 6). The distribution of seasonal precipitation greatly influenced the maximum monthly Gs value in different years (Fig. 5). The highest maximum monthly Gs of 11.2 mm s−1 for the cropland and 10.1 mm s−1 for the grassland occurred in June and July of 2005, which corresponds to heavy summer rainfalls (126.8 mm in June, 114.3 mm in July) that supply more water for plant growth. High values of Gs occurred again in the 2008 growing season, but they were not as high as they were in 2005 (Fig. 5). In contrast, the maximum surface conductance was significantly lower in the growing season of the dry years (2004 and 2007), which physically limited ET. Over the annual cycle, the annual average surface conductance for the grassland ranged between 2.99 and 4.34 mm s−1 (Table 2). Corresponding values for the cropland were 2.87 to 4.5 mm s−1. Generally, the variation of the annual average surface conductance followed the precipitation cycle during the 6-yr period. The daily Gs reached a maximum of 28.3 mm s−1 for the grassland and 35.8 mm s−1 for the cropland (not shown in the graph). The results were within the ranges reported from previous studies in other regions, such as an annual grassland in California (25 mm s−1; Ryu et al. 2008) and an irrigated maize-based agroecosystem (29 mm s−1 for maize and 41 mm s−1 for soybean; Suyker and Verma 2008).

Fig. 5.

The monthly averages of Gs for the grassland and cropland during the 6-yr period.

Fig. 5.

The monthly averages of Gs for the grassland and cropland during the 6-yr period.

Fig. 6.

The relationship between the monthly ET and the monthly average (a) NDVI or (b) EVI at the grassland and the cropland site in the growing and nongrowing seasons.

Fig. 6.

The relationship between the monthly ET and the monthly average (a) NDVI or (b) EVI at the grassland and the cropland site in the growing and nongrowing seasons.

The dimensionless decoupling coefficient (Ω) represents the aerodynamic degree of coupling between the vegetation and the atmosphere and ranges from 0 (completely coupling) to 1 (decoupling). Stomatal control of evapotranspiration becomes weaker when Ω approaches 1. The annual average decoupling coefficient (Ω) ranged from 0.34 to 0.4 at our sites, whereas the average value of Ω during the 6-yr period was 0.37 and 0.38 for the grassland and the cropland, respectively (Table 2). The average values of Ω in the growing season were 0.64 and 0.68 for the grassland and the cropland, respectively.

The Priestley–Taylor coefficient is defined as the value of actual ET normalized to the equilibrium evaporation. Previous studies, both field observations and model studies, demonstrated that normally exceeds 1 on well-watered croplands or wet ecosystems, with a widely accepted value of 1.26 (De Bruin and Keijman 1979; Stewart and Rouse 1976; Shuttleworth and Calder 1979). However, is often below 1 in arid–semiarid ecosystems (Wilson and Baldocchi 2000). The deviations from the wet ecosystem condition are ascribed to the physiological constraints of vegetation or advection processes (Wilson and Baldocchi 2000). The annual average value of during the 6-yr period was 0.43 for the grassland and 0.44 for the cropland (Table 2). The monthly average values of in the growing season are shown in Fig. 7. In the growing season of 2005, the value was 0.89 at the grassland site, which was higher than in other years, with a maximum value of 0.97 in August (Fig. 7a), which is comparable to the value for a wet ecosystem. In contrast, in the dry years of 2004 and 2007, the values during the growing season were much lower at both sites than they were in the other years (Fig. 7). The average values of in the growing season were 0.76 and 0.77 for the grassland and the cropland, respectively (Table 2).

Fig. 7.

The monthly average Priestly–Taylor coefficients in the growing season over the 6-yr period in the (a) grassland and (b) cropland.

Fig. 7.

The monthly average Priestly–Taylor coefficients in the growing season over the 6-yr period in the (a) grassland and (b) cropland.

4. Discussion

a. The relationship between ET and precipitation

The relationship between cumulative ET and cumulative precipitation, which was first proposed by Amiro and Wuschke (1987), indicates whether the water is being stored or released from the ground. In wet years, the cumulative precipitation curve was generally above the cumulative ET curve in the summer, indicating that large rainfall events supplied more water to recharge the soil. During the dry years, the cumulative ET curve stayed near the cumulative precipitation curve in the summer, indicating that soil water coming from precipitation was rapidly consumed by ET. The total cumulative ET was close to the total cumulative precipitation in this area over the 6-yr period (Fig. 8). Similar relationships between cumulative ET and precipitation were reported for semiarid grassland in California (Ryu et al. 2008) and a northern temperate grassland (Wever et al. 2002). The cumulative ET at the cropland site was slightly larger (about 10–30 mm) than it was at the grassland at the end of each year. More water might come from irrigation at seedtime and from the soil. From this standpoint, the amount of precipitation greatly constrained the soil water availability for plant growth and the magnitude of the annual total ET in this semiarid area.

Fig. 8.

Comparison of cumulative evapotranspiration and cumulative precipitation at the grassland and cropland sites over the 6-yr period 2003–08.

Fig. 8.

Comparison of cumulative evapotranspiration and cumulative precipitation at the grassland and cropland sites over the 6-yr period 2003–08.

b. The meteorological factors controlling ET

The quantity ET is considered to be a multivariate phenomenon as it is influenced by many meteorological variables. In general, the meteorological variables involved in ET are precipitation, wind velocity, solar radiation, humidity, air temperature, ground cover characteristics, etc., which are found to be mostly intercorrelated. Knowledge of the relative effect of these variables on ET is very important in understanding the hydrological processes in semiarid areas. A summary of the results of PCA is shown in Table 4. The proportion of the cumulative variance by the first three components explained more than 90% for both sites with eigenvalues larger than 0.5. Therefore, three components can be considered for the further analysis with only a loss of information less than 10%. The results of rotated FA producing the factor loadings and communalities are summarized in Table 5. It is noted that a few variables have very high factor loadings on each factor, which indicate they are relatively significant on the corresponding factor. The first factor is composed of VPD, Rn, and Ta variables. The VPD variable contributed the highest loadings on this dimension, and the other two variables (Rn and Ta) combined can increase VPD to accelerate ET process. Hence, the first factor can be attributed to the atmospheric water demand component. The Ws variable loaded the most heavily on the second factor, and precipitation can recharge the soil water storage. Therefore, the second factor can be identified as the humidity component. The third factor had only one heavy loading from the U variable, and thus the third factor was determined as the wind speed component. The results of the analysis also revealed the atmospheric water demand, which consist of vapor pressure deficit, net radiation, and air temperature variables, was the most important factor in the ET process on half-hour time scale in this semiarid area.

Table 4.

Total variance explained and principal components.

Total variance explained and principal components.
Total variance explained and principal components.
Table 5.

Factor loadings of variable and communality.

Factor loadings of variable and communality.
Factor loadings of variable and communality.

c. The biological factors controlling ET

The three bulk parameters (Gs, Ω, and ) mentioned in section 3e revealed strong biological factors controlling ET via the change of surface conductance. The decoupling coefficient Ω and Priestley–Taylor coefficient were employed to identify the degree of coupling between ET and surface conductance. The small Ω values in Table 2 indicated ET was strongly controlled by surface conductance and ambient humidity deficit, but as Ω increased in the raining seasons, ET became more controlled by the available energy. The small values in Table 2, which were lower than 1 at most time of the year, suggested ET was greatly limited by equilibrium evaporation in this semiarid area. Further, the relationship between monthly average bulk surface conductance and the Priestly–Taylor coefficient is shown in Fig. 9. In these two semiarid ecosystems, Gs was strongly correlated with ; was sensitive to Gs when Gs was low, suggesting a strong physiological control of ET. As Gs increases, becomes independent of surface conductance, which suggests an increasing degree of coupling between ET and available energy. Overall, the small results of Ω and indicated that ET was strongly controlled by surface conductance in this semiarid area.

Fig. 9.

The relationship between monthly average bulk surface conductances and Priestly–Taylor coefficient in the (a) nongrowing season and (b) growing season.

Fig. 9.

The relationship between monthly average bulk surface conductances and Priestly–Taylor coefficient in the (a) nongrowing season and (b) growing season.

d. Comparison between the Penman–Monteith equation and EC for ET calculation

A comparison between the LE calculated using the Penman–Monteith equation (LE-PM) and the LE measured using the eddy covariance technique (LE-EC) was carried out to evaluate the accuracy of the Penman–Monteith model in this semiarid area.

The LE-PM was calculated with Eq. (2), with the minor difference that Gs was estimated using a Jarvis-type stomatal conductance model [Eq. (8)]. Jarvis (1976) proposed that Gs can be described as the maximum stomatal conductance Gsm, which decreases in proportion to the LAI and other climatic variables that influence stomatal opening. In the present study, Gs was calculated by Eq. (8), with the LAI calculated using NDVI data following Fan et al. (2009). The ET values derived from EC measurement within the wet periods (with precipitation) were excluded in this investigation. The results show that ET estimated by the Penman–Monteith equation was generally lower than the ET measured by EC in the growing season at both the grassland and the cropland site (Figs. 10a,c). The deviation from the measured ET was even larger in the nongrowing season (Figs. 10b,d). Therefore, more work is needed to increase the accuracy of ET estimation using the Penman–Monteith formula in this semiarid area, such as measurements of LAI and more accurate estimation of Gs.

Fig. 10.

Comparison between LE calculated using the (a),(b) Penman–Monteith equation (LE-PM) and (c),(d) LE measured using the eddy covariance technique (LE-EC) in the (left) growing season and (right) nongrowing season.

Fig. 10.

Comparison between LE calculated using the (a),(b) Penman–Monteith equation (LE-PM) and (c),(d) LE measured using the eddy covariance technique (LE-EC) in the (left) growing season and (right) nongrowing season.

5. Conclusions

The seasonal and interannual variations of evapotranspiration and energy exchange over degraded grassland and cropland ecosystem were investigated in a semiarid area of northeastern China based on 6 years of eddy covariance measurements. The amount and distribution of precipitation were found to have great impact on the seasonal and interannual variation of ET via various environmental and biological factors. The total cumulative ET was observed to be close to the total cumulative precipitation during the study period. Three environmental factors (atmospheric water demand, humidity, and wind speed) were recognized to have the largest influence on ET on half-hour time scale. Results of bulk parameters such as decoupling coefficient and Priestley–Taylor coefficient revealed strong biological control of ET via surface conductance in the study area. Our data indicate the variation of precipitation and surface conductance exerted important control on ET on seasonal time scale, whereas the total amount of precipitation greatly constrained ET on the annual time scale in this semiarid area.

Acknowledgments

This research was supported by NSFC 41021004 and the National Basic Research Program of China (2010CB951801, 2006CB500401). Thanks are extended to Dr. Li Wanbiao (Peking University) for providing the MODIS vegetation dataset. Thanks are also given for three anonymous reviewers’ valuable comments and suggestions.

REFERENCES

REFERENCES
Amiro
,
B. D.
, and
E. E.
Wuschke
,
1987
:
Evapotranspiration from a boreal forest drainage basin using an energy balance/eddy correlation technique
.
Bound.-Layer Meteor.
,
38
,
125
139
.
Anderson
,
T. W.
,
2003
: An Introduction to Multivariate Statistical Analysis. 3rd ed. John Wiley and Sons, 752 pp.
Aubinet
,
M.
, and
Coauthors
,
2000
:
Estimates of the annual net carbon and water exchange of forests: The EUROFLUX methodology
.
Adv. Ecol. Res.
,
30
,
113
175
.
Baldocchi
,
D.
, and
T. P.
Meyers
,
1998
:
On using eco-physiological, micrometeorological and biogeochemical theory to evaluate carbon dioxide, water vapor and trace gas fluxes over vegetation: A perspective
.
Agric. For. Meteor.
,
90
,
1
25
.
Baldocchi
,
D.
, and
L. K.
Xu
,
2007
:
What limits evaporation from Mediterranean oak woodlands—The supply of moisture in the soil, physiological control by plants or the demand by the atmosphere?
Adv. Water Resour.
,
30
,
2113
2122
.
Baldocchi
,
D.
, and
Coauthors
,
2001
:
FLUXNET: A new tool to study the temporal and spatial variability of ecosystem-scale carbon dioxide, water vapor and energy flux densities
.
Bull. Amer. Meteor. Soc.
,
82
,
2415
2434
.
Barr
,
A. G.
, and
A. K.
Betts
,
1997
:
Radiosonde boundary layer budgets above a boreal forest
.
J. Geophys. Res.
,
102
(
D24
),
29 205
29 212
.
Betts
,
A. K.
,
J. H.
Ball
, and
P.
Viterbo
,
1999
:
Basin-scale surface water and energy budgets for the Mississippi from the ECMWF reanalysis
.
J. Geophys. Res.
,
104
(
D16
),
19 293
19 306
.
Blanken
,
P. D.
, and
Coauthors
,
1997
:
Energy balance and canopy conductance of a boreal aspen forest: Partitioning overstory and understory components
.
J. Geophys. Res.
,
102
(
D24
),
28 915
28 927
.
Brutsaert
,
W.
,
2006
:
Indications of increasing land surface evaporation during the second half of the 20th century
.
Geophys. Res. Lett.
,
33
,
L20403
,
doi:10.1029/2006GL027532
.
Cornic
,
G.
,
2000
:
Drought stress inhibits photosynthesis by decreasing stomatal aperture—Not by affecting ATP synthesis
.
Trends Plant Sci.
,
5
,
187
188
.
Dang
,
Q. L.
,
H. A.
Margolis
,
M. R.
Coyea
,
S. Y.
Mikailou
, and
G. J.
Collatz
,
1997
:
Regulation of branch level gas exchange of boreal trees: Roles of shoot potential and vapor pressure differences
.
Tree Physiol.
,
17
,
521
536
.
De Bruin
,
H. A. R.
, and
J. Q.
Keijman
,
1979
:
The Priestley-Taylor model applied to a large shallow lake in the Netherlands
.
J. Appl. Meteor.
,
18
,
898
903
.
Dirmeyer
,
P. A.
,
1994
:
Vegetation stress as a feedback mechanism in mid-latitude drought
.
J. Climate
,
7
,
1463
1483
.
Falge
,
E.
, and
Coauthors
,
2001
:
Gap filling strategies for long term energy flux data sets
.
Agric. For. Meteor.
,
107
,
71
77
.
Fan
,
L.
,
Y.
Gao
,
H.
Brück
, and
C.
Bernhofer
,
2009
:
Investigating the relationship between NDVI and LAI in semi-arid grassland in Inner Mongolia using in-situ measurements
.
Theor. Appl. Climatol.
,
95
,
151
156
.
Foken
,
T.
,
M.
Göockede
,
M.
Mauder
,
L.
Mahrt
,
B.
Amiro
, and
W.
Munger
,
2004
: Post-field data quality control. Handbook of Micrometeorology, X. Lee, W. Massman, and B. E. Law, Ed., Kluwer, 181–208.
Golubev
,
V. S.
,
J. H.
Lawrimore
,
P. Ya.
Groisman
,
N. A.
Speranskaya
,
S. A.
Zhuravin
,
M. J.
Menne
,
T. C.
Peterson
, and
R. W.
Malone
,
2001
:
Evaporation changes over the contiguous United States and the former USSR: A reassessment
.
Geophys. Res. Lett.
,
28
,
2665
2668
.
Gu
,
S.
, and
Coauthors
,
2008
:
Characterizing evapotranspiration over a meadow ecosystem on the Qinghai-Tibetan Plateau
.
J. Geophys. Res.
,
113
,
D08118
,
doi:10.1029/2007JD009173
.
Hadjimitsis
,
D. G.
, and
Coauthors
,
2010
:
Atmospheric correction for satellite remotely sensed data intended for agricultural applications: Impact on vegetation indices
.
Nat. Hazards Earth Syst. Sci.
,
10
,
89
95
.
Hill
,
T. C.
,
M.
Williams
, and
J. B.
Moncrieff
,
2008
:
Modeling feedbacks between a boreal forest and the planetary boundary layer
.
J. Geophys. Res.
,
113
,
D15122
,
doi:10.1029/2007JD009412
.
Horst
,
T. W.
, and
J. C.
Weil
,
1994
:
How far is far enough: The fetch requirements of micrometeorological measurement of surface fluxes
.
J. Atmos. Oceanic Technol.
,
11
,
1018
1025
.
Hui
,
D.
,
S.
Wan
,
B.
Su
,
G.
Katul
,
R.
Monson
, and
Y.
Luo
,
2004
:
Gapfilling missing data in eddy covariance measurements using multiple imputation (MI) for annual estimation
.
Agric. For. Meteor.
,
121
,
93
111
.
Hunsaker
,
D.
,
P.
Pinter
,
E.
Barnes
, and
B.
Kimball
,
2003
:
Estimating cotton evapotranspiration crop coefficients with a multispectral vegetation index
.
Irrig. Sci.
,
22
,
95
104
.
Hwang
,
T.
,
C. H.
Song
,
P. V.
Bolstad
, and
L. E.
Band
,
2011
:
Downscaling real-time vegetation dynamics by fusing multi-temporal MODIS and Landsat NDVI in topographically complex terrain
.
Remote Sens. Environ.
,
115
,
2499
2512
.
Jarvis
,
P. G.
,
1976
:
The interpretation of the variations in leaf water potential and stomatal conductance found in canopies in the field
.
Philos. Trans. Roy. Soc. London
,
B273
,
593
610
.
Jarvis
,
P. G.
, and
K. G.
McNaughton
,
1986
:
Stomal control of transpiration: Scaling up from leaf to region
.
Adv. Ecol. Res.
,
15
,
1
49
.
Kaimal
,
J. C.
, and
J. J.
Finnigan
,
1994
: Atmospheric Boundary Layer Flows: Their Structure and Management. Oxford University Press, 289 pp.
Kaiser
,
H. F.
,
1958
:
The varimax criterion for analytic rotation in factor analysis
.
Psychometrika
,
23
,
187
200
.
Kang
,
S. Z.
,
B. J.
Gu
,
T. S.
Du
, and
J. H.
Zhang
,
2003
:
Crop coefficient and ratio of transpiration to evapotranspiration of winter wheat and maize in a semi-humid region
.
Agric. Water Manage.
,
59
,
239
254
.
Lee
,
X.
, and
T. A.
Black
,
1994
:
Relating eddy correlation sensible heat flux to horizontal sensor separation in the unstable atmospheric surface layer
.
J. Geophys. Res.
,
99
(
D9
),
18 545
18 553
.
Li
,
S. G.
,
C. T.
Lai
,
G.
Lee
,
S.
Shimoda
,
T.
Yokoyama
,
A.
Higuchi
, and
T.
Oikawa
,
2005
:
Evapotranspiration from a wet temperate grassland and its sensitivity to microenvironmental variables
.
Hydrol. Processes
,
19
,
517
532
.
Liepert
,
B. G.
,
J.
Feichter
,
U.
Lohmann
, and
E.
Roeckner
,
2004
:
Can aerosols spin down the water cycle in a warmer and moister world?
Geophys. Res. Lett.
,
31
,
L06207
,
doi:10.1029/2003GL019060
.
Liu
,
B. H.
,
M.
Xu
,
M.
Henderson
, and
W. G.
Gong
,
2004
:
A spatial analysis of pan evaporation trends in China, 1955–2000
.
J. Geophys. Res.
,
109
,
D15102
,
doi:10.1029/2004JD004511
.
Liu
,
C. M.
,
X. Y.
Zhang
, and
Y. Q.
Zhang
,
2002
:
Determination of daily evaporation and evapotranspiration of winter wheat and maize by large-scale weighing lysimeter and micro-lysimeter
.
Agric. For. Meteor.
,
111
,
109
120
.
Liu
,
H. Z.
,
G.
Tu
, and
W. J.
Dong
,
2008
:
Three-year changes of surface albedo of degraded grassland and cropland surfaces in a semiarid area
.
Chin. Sci. Bull.
,
53
,
1246
1254
.
Massman
,
W. J.
, and
X.
Lee
,
2002
:
Eddy covariance flux corrections and uncertainties in long-term studies of carbon and energy exchanges
.
Agric. For. Meteor.
,
113
,
121
144
.
McMillen
,
R. T.
,
1988
:
An eddy correlation technique with extended applicability to non-simple terrain
.
Bound.-Layer Meteor.
,
43
,
231
245
.
McNaughton
,
K. G.
, and
T. W.
Spriggs
,
1986
:
A mixed-layer model for regional evaporation
.
Bound.-Layer Meteor.
,
34
,
243
262
.
Monteith
,
J. L.
, and
M. H.
Unsworth
,
1990
: Principles of Environmental Physics. 2nd ed. Chapman & Hall, 291 pp.
Nagler
,
P. L.
,
J.
Cleverly
,
E.
Glenn
,
D.
Lampkin
,
A.
Huete
, and
Z.
Wan
,
2005
:
Predicting riparian evapotranspiration from MODIS vegetation indices and meteorological data
.
Remote Sens. Environ.
,
94
,
17
30
.
Ohta
,
T.
, and
Coauthors
,
2008
:
Interannual variation of water balance and summer evapotranspiration in an eastern Siberian larch forest over a 7-year period (1998–2006)
.
Agric. For. Meteor.
,
148
,
1941
1953
.
Pielke
,
R. S.
,
R.
Avissar
,
M.
Raupach
,
A. J.
Dolman
,
X.
Zeng
, and
A. S.
Denning
,
1998
:
Interactions between the atmosphere and terrestrial ecosystems: Influence on weather and climate
.
Global Change Biol.
,
4
,
461
475
.
Priestley
,
C. H. B.
, and
R. J.
Taylor
,
1972
:
On the assessment of surface heat flux and evaporation using large-scale parameters
.
Mon. Wea. Rev.
,
100
,
81
92
.
Ramanathan
,
V.
,
P. J.
Crutzen
,
J. T.
Kiehl
, and
D.
Rosenfeld
,
2001
:
Aerosols, climate, and the hydrological cycle
.
Science
,
294
,
2119
2124
.
Rana
,
G.
, and
N.
Katerji
,
1996
:
Evapotranspiration measurement for tall plant canopies: The sweet sorghum case
.
Theor. Appl. Climatol.
,
54
,
187
200
.
Roderick
,
M. L.
, and
G. D.
Farquhar
,
2002
:
The cause of decreased pan evaporation over the past 50 years
.
Science
,
298
,
1410
1411
.
Running
,
S. W.
,
D. D.
Baldocchi
,
D. P.
Turner
,
S. T.
Gower
,
P. S.
Bakwin
, and
K. A.
Hibbard
,
1999
:
A global terrestrial monitoring network integrating tower fluxes, flask sampling, ecosystem modeling and EOS data
.
Remote Sens. Environ.
,
70
,
108
127
.
Ryu
,
Y.
,
D. D.
Baldocchi
,
S.
Ma
, and
T.
Hehn
,
2008
:
Interannual variability of evapotranspiration and energy exchange over an annual grassland in California
.
J. Geophys. Res.
,
113
,
D09104
,
doi:10.1029/2007JD009263
.
Sala
,
O.
, and
W. K.
Lauenroth
,
1982
:
Small rainfall events: An ecological role in semiarid regions
.
Oecologia
,
53
,
301
304
.
Sala
,
O.
,
W. K.
Lauenroth
, and
W. J.
Parton
,
1992
:
Long-term soil water dynamics in the shortgrass steppe
.
Ecology
,
73
,
1175
1181
.
Shi
,
T.-T.
,
D.-X.
Guan
,
J.-B.
Wu
,
A.-Z.
Wang
,
C.-J.
Jin
, and
S.-J.
Han
,
2008
:
Comparison of methods for estimating evapotranspiration rate of dry forest canopy: Eddy covariance, Bowen ratio energy balance, and Penman-Monteith equation
.
J. Geophys. Res.
,
113
,
D19116
,
doi:10.1029/2008JD010174
.
Shuttleworth
,
W. J.
, and
I. R.
Calder
,
1979
:
Has the Priestley-Taylor equation any relevance to forest evaporation?
J. Appl. Meteor.
,
18
,
639
646
.
Stewart
,
R. B.
, and
W. R.
Rouse
,
1976
:
A simple method for determining the evaporation from shallow lakes and ponds
.
Water Resour. Res.
,
12
,
623
628
.
Sun
,
L.
, and
G. X.
Wu
,
2001
:
Influence of land evapotranspiration on climate variations
.
Sci. China, Ser. D: Earth Sci.
,
44
,
838
846
.
Suyker
,
A. E.
, and
S. B.
Verma
,
2008
:
Interannual water vapor and energy exchange in an irrigated maize-based agroecosystem
.
Agric. For. Meteor.
,
148
,
417
427
.
Valentini
,
R.
, and
Coauthors
,
2000
:
Respiration as the main determinant of carbon balance in European forests
.
Nature
,
404
,
861
865
.
Verhoef
,
A.
,
S. J.
Allen
, and
C. R.
Lloyd
,
1999
:
Seasonal variation of surface energy balance over two Sahelian surfaces
.
Int. J. Climatol.
,
19
,
1267
1277
.
Webb
,
E. K.
,
G. I.
Pearman
, and
R.
Leuning
,
1980
:
Correction of flux measurements for density effects due to heat and water vapour transfer
.
Quart. J. Roy. Meteor. Soc.
,
106
,
85
100
.
Wetherald
,
R. T.
, and
S.
Manabe
,
2002
:
Simulation of hydrologic changes associated with global warming
.
J. Geophys. Res.
,
107
,
4379
,
doi:10.1029/2001JD001195
.
Wever
,
L. A.
,
L. B.
Flanagan
, and
P. J.
Carlson
,
2002
:
Seasonal and interannual variation in evapotranspiration, energy balance and surface conductance in a northern temperate grassland
.
Agric. For. Meteor.
,
112
,
31
49
.
Wild
,
M.
,
A.
Ohmura
,
H.
Gilgen
, and
D.
Rosenfeld
,
2004
:
On the consistency of trends in radiation and temperature records and implications for the global hydrological cycle
.
Geophys. Res. Lett.
,
31
,
L11201
,
doi:10.1029/2003GL019188
.
Wilson
,
K. B.
, and
D. D.
Baldocchi
,
2000
:
Seasonal and interannual variability of energy fluxes over a broadleaved temperate deciduous forest in North America
.
Agric. For. Meteor.
,
100
,
1
18
.
Wilson
,
K. B.
, and
Coauthors
,
2002
:
Energy balance closure at FLUXNET sites
.
Agric. For. Meteor.
,
113
,
223
243
.