Comparison of dual crop coefficient method and Shuttleworth–Wallace model in evapotranspiration partitioning in a vineyard of northwest China

Agricultural Water Management 160 (2015) 41–56

Contents lists available at ScienceDirect

Agricultural Water Management journal homepage: www.elsevier.com/locate/agwat

Comparison of dual crop coefficient method and Shuttleworth–Wallace model in evapotranspiration partitioning in a vineyard of northwest China Peng Zhao a , Sien Li a , Fusheng Li b , Taisheng Du a , Ling Tong a , Shaozhong Kang a,∗ a b

Center for Agricultural Water Research in China, China Agricultural University, Beijing 100083, China College of Agriculture, Guangxi University, Nanning, Guangxi 530005, China

a r t i c l e

i n f o

Article history: Received 11 April 2015 Received in revised form 6 June 2015 Accepted 27 June 2015 Keywords: Dual crop coefficient Eddy covariance Evapotranspiration partitioning Sap flow Shuttleworth–Wallace model Soil evaporation

a b s t r a c t The objective of this study was to evaluate the potential use of dual crop coefficient method in FAO56 (FAO dual-Kc ) and Shuttleworth–Wallace (S–W) model in estimating evapotranspiration (ET) and its components (plant transpiration and soil evaporation) of a vineyard in arid region of northwest China. Continuous measurements of ET with eddy covariance, plant transpiration (T) with sap flow system and soil evaporation (E) with micro-lysimeter in 2013 and 2014, were used to validate the performance of two approaches. Results indicate that sap flow system and micro-lysimeter can provide accurate measurements of T and E at hourly and daily scales if compared to eddy covariance, respectively. The FAO dual-Kc method in partitioning ET was acceptable when using the site-specific basal crop coefficient obtained from sap flow, with the slope and intercept of linear regression of 0.96 and −0.13 mm d−1 (R2 = 0.81) for ET, 0.92 and −0.07 mm d−1 (R2 = 0.76) for E, 0.93 and 0.16 mm d−1 (R2 = 0.80) for T, respectively. The S–W model can better estimate ET, but overestimated T and underestimated E when using site-specific soil surface resistance, with the slope and intercept of linear regression of 0.98 and 0.28 mm d−1 (R2 = 0.79) for ET, 0.49 and 0.42 mm d−1 (R2 = 0.46) for E, 1.10 and 0.38 mm d−1 (R2 = 0.81) for T, respectively. Both approaches had obvious discrepancies of E after rainfall and irrigation, especially the S–W model, and overestimated T after a snowfall. Sensitivity analysis indicates that estimated ET and its components were sensitive to soil field capacity and wilting point in both approaches, and in the S–W model, predicted T was also sensitive to leaf area index (LAI) and minimum stomatal resistance and predicted E sensitive to soil surface resistance and LAI. Thus two approaches can estimate ET with good accuracy, but the FAO dual-Kc method had higher accuracy in estimating E and T. © 2015 Published by Elsevier B.V.

1. Introduction There are abundant light resources suited for grape production, but limited water resources in the arid region of northwest China. In recent years, large areas of vineyard have been established in this region (Liu et al., 2006). Irrigation is essential to ensure grape production in such areas, and appropriate amount of irrigation water at right time directly increases wine quality (Intrigliolo and Castel, 2008). However, most of vineyard in this region is furrowirrigated with an empirically determined irrigation quota (Zhang et al., 2011). To develop rational irrigation strategy and achieve higher water use efficiency, an accurate estimation of actual evapotranspiration is necessary (Kang et al., 2004).

∗ Corresponding author. Fax: +86 10 62737611. E-mail address: [email protected] (S. Kang). http://dx.doi.org/10.1016/j.agwat.2015.06.026 0378-3774/© 2015 Published by Elsevier B.V.

Evapotranspiration (ET) can be divided into soil evaporation (E) and plant transpiration (T). In the vineyard, due to large fraction of exposed soil, soil evaporation can account for 40% of ET using furrow irrigation (Zhang et al., 2010, 2011), 30% of ET using drip irrigation (Yunusa et al., 2004; Poblete-Echeverría et al., 2012) and 77% of ET using flood irrigation (Lascano et al., 1992). And the function of E and T is different, T is associated with plant productivity, while E does not directly contribute to plant productivity (Kool et al., 2014), so T is considered as the desirable component but E as undesirable component (Agam et al., 2012). A better understanding of ET components and how much water is used through plant transpiration can help investigate if irrigation can be improved and available water can be used more efficiently (Zhao et al., 2013; Kool et al., 2014). ET partitioning is possible using micro-meteorological measurements (Bowen ratio, eddy covariance system), eco-physiological techniques (sap flow, stable isotopes), and hydrological balance

42

P. Zhao et al. / Agricultural Water Management 160 (2015) 41–56

methods (lysimeter, soil water budget) (Kang et al., 2003a,b; Williams et al., 2004; Yepez et al., 2005; Er-Raki et al., 2010; Zhang et al., 2011). However, these techniques are often expensive and require specific equipments, thus they are limited in agricultural water management, so the estimating E and T respectively with models has been done by many researchers (Ritchie, 1972; Shuttleworth and Wallace, 1985; Lascano et al., 1987; Allen et al., 1998). Since the S–W model and FAO dual-Kc method are widely used, and have been validated to estimate ET components in different ecosystems (Hu et al., 2009, 2013; Er-Raki et al., 2010; Ding et al., 2013; Poblete-Echeverría and Ortega-Farias, 2013; Zhu et al., 2013, 2014), the two approaches were used to partition ET into soil evaporation and plant transpiration in a furrow-irrigated vineyard in this study. The S–W model is an approach to directly estimate ET components based on two Penman–Monteith equations, i.e., one for the plant and the other for the soil surface (Monteith, 1965), and the two components are weighted by a set of coefficients that represent the combination of soil and canopy resistances (Shuttleworth and Wallace, 1985). This model provides the possibility to partition ET into plant and soil components through the use of surface resistances to regulate the transfer of energy from plants (rsc ) and soil (rss ), and aerodynamic resistances (raa , rac , ras ) to regulate the transfer between these surfaces and the atmosphere (Farahani and Bausch, 1995). Compared to other direct models, the S–W model are more widely used for crops with a partial ground cover, such as orchards, and have achieved acceptable results (Ortega-Farias et al., 2007, 2010; Zhang et al., 2008, 2009; Hu et al., 2013). Like most models, S–W model has limitation due to the need of measured data above canopy and several hardly obtained input parameters in field condition. Apart from the direct models, the Food and Agricultural Organization of the United Nations (FAO) has developed an indirect method to estimate ET (Doorenbos and Pruitt, 1977) and further developed in the Paper FAO-56, includes the single and dual crop coefficient methods (Allen et al., 1998). Unlike the single crop coefficient method that integrated both effects of plant transpiration and soil evaporation into a crop coefficient (Kc ), the dual crop coefficient method allows to separate soil evaporation and plant transpiration and divide Kc into basal crop coefficient (Kcb ) and soil evaporation coefficient (Ke ). Due to practical simplicity and robustness, and also a fewer data requirements than the other direct models, the FAO-56 dual crop coefficient method has been adopted and evaluated over several sparse crops, such peach orchard (Goodwin et al., 2006), apple orchard (Dragoni et al., 2004), olive orchard (Er-Raki et al., 2010) and vineyard (PobleteEcheverría and Ortega-Farias, 2013). However, the straightforward adoption of generalized crop coefficients (Kcb or Kc ) recommended by FAO-56 can lead to the errors in the estimation of ET and its components, because the dividing of crop growth period and associated crop coefficients are closely related to local climate and crop condition (Katerji and Rana, 2006; Poblete-Echeverría and OrtegaFarias, 2013). So the adjustment of crop coefficient curve based on locally observed data is needed to achieve accurate estimate of actual water requirement. Several researchers have improved the model performance by adjusting the recommended crop coefficient curves in FAO-56 (Rana and Katerji, 2008; Er-Raki et al., 2008, 2010; Poblete-Echeverría and Ortega-Farias, 2013). Several studies have evaluated the models performance in partitioning ET into soil and plant components for sparse vegetation (Zhang et al., 2009; Er-Raki et al., 2010; Hu et al., 2013; PobleteEcheverría and Ortega-Farias, 2013), mainly about FAO dual-Kc method and S–W model. However, there are fewer studies about the model evaluation using long-term independent measurements of ET and its components and the comparison of different models to partition ET. Therefore, the objective of this study was to evaluate

the S–W model and FAO dual-Kc method in estimating evapotranspiration and its components in a furrow-irrigated Merlot vineyard under arid condition using the measurements of an eddy covariance system, micro-lysimeter and sap flow system during two growing seasons (2013 and 2014), respectively, so as to provide accurate estimation of evapotranspiration and its components in vineyard of northwest China. 2. Materials and methods 2.1. Study site The study was carried out in a furrow-irrigated Merlot (Vitis vinifera L.) vineyard (37◦ 52 N, 102◦ 50 E, 1585 m a.s.l.), located at the Experimental Station of Water-saving in Agriculture and Ecology of China Agriculture University in the Shiyang river basin of northwest China during the 2013 and 2014 seasons. The grapevines were planted in east–west rows with a distance between and within rows of 2.7 m and 1.0 m in 1999. The grapevines were manually separated into two trunks at the ground surface and one trunk was considered the half vine. The grapevine branches were fixed on the three wires, and the wire heights were 0.5, 1.0 and 1.5 m, respectively. The experimental site is located in a continental temperate zone, with a mean annual precipitation of 164.4 mm, annual temperature of 8 ◦ C and annual sunshine duration of over 3000 h. Water is in severe shortage in this region, the groundwater table is below 25 m, and mean annual pan evaporation is about 2000 mm. The vineyard soil is irrigated desert soil (Siltigic-Orthic Anthrosols) and soil texture is sandy loam, with an average bulk density of 1.49 g cm−3 to a depth of 1.0 m. Field capacity and wilting point were 0.31 and 0.27, and 0.11 and 0.10 m3 m−3 for the 0–1.0 m layer and 0–10 cm layer, respectively. The vineyard was furrow-irrigated with a trapeziform ditch on south side of each row, with the depth, bottom width and surface width of 30, 90 and 100 cm, respectively (Zhang et al., 2011). Grapevines were irrigated 6 times during whole growth stage, i.e., April 27, May 25, July 1, July 30, August 25 and October 13, 2013, and April 22, May 25, July 2, August 4, August 31 and October 18, 2014, and irrigation quota is 70 mm each time. 2.2. Measurements 2.2.1. Soil water content and leaf area index Gravimetric soil water content to a depth of 1.0 m was determined every 4–6 d by the oven-drying method at an interval of 0.1 m, with six sampling points inside the vineyard, and then volumetric soil water content was calculated from gravimetric soil water content and bulk density of each layer. Volumetric soil water content at the depth of 0.1 m was continuously monitored using six ECH2 O sensors (5TE, Decagon Devices, Inc., USA) distributed in the ditch and ridge, and collected every 10 min using a data-logger (Environmental logging system, Decagon Devices, Inc., USA). Leaf area index (LAI) was estimated as a function of shoot length to total leaf area (Ortega-Farias et al., 2007). In this study, LAI were calculated as:

i

LAI =

1

LAsh

Av

(1)

LAsh = −358.1 + 23.1Lsh (Lsh > 25cm), or − 42.8 + 12.8Lsh (Lsh < 25cm)

(2)

where LAsh is total leaf area per shoot (cm2 ), Av vine area (cm2 ), i total shoot number per vine and Lsh shoot length (cm). To develop the correlation between LAsh and Lsh , a total number of 73 branches were randomly selected from the vineyard at initial and middle

P. Zhao et al. / Agricultural Water Management 160 (2015) 41–56

43

Table 1 Summary of monthly meteorological variables during grapevine growing seasons. Season/month

Tmean (◦ C)

Wind speed (m s−1 )

Solar radiation (MJ m−2 d−1 )

D (kPa)

ET0 (mm d−1 )

Rainfalla (mm)

2013 April May June July August September October

17.3 18.0 20.5 21.4 21.0 16.1 11.9

1.19 0.89 0.56 0.33 0.31 0.39 0.32

20.7 20.7 18.4 16.7 17.6 15.2 14.1

0.99 1.05 0.94 0.72 0.62 0.47 0.48

4.25 4.13 3.79 3.42 3.39 2.56 1.84

0.0 9.0 15.2 10.6 26.8 20.0 2.2

2014 April May June July August September October

14.3 16.3 19.7 21.4 18.6 15.8 11.6

1.11 1.02 0.72 0.40 0.43 0.32 0.39

20.6 21.6 19.7 21.5 17.8 14.1 13.9

1.24 1.45 1.16 1.19 0.82 0.84 0.85

3.80 4.18 4.07 4.17 3.28 2.40 1.80

0.8 19.0 22.6 57.4 94.6 11.5 14.2

a

Monthly totals; Experimental period were from April 25 to October 21 in 2013, April 27 to October 23 in 2014.

growth period in 2013, and the total leaf area of a shoot was measured by an AM300 portable leaf area meter (ADC BioScientific Ltd., UK). The total shoot length per vine of eight vines was measured every 7–10 days during the experimental period. 2.2.2. Meteorological data and eddy covariance All climatic parameters (solar radiation, air temperature, relative humidity and wind speed) needed to calculate daily reference evapotranspiration (ET0 ) by the FAO–Penman–Monteith equation (Allen et al., 1998) and precipitation were measured with a standard automatic weather station near the experimental vineyard at a height of 2.0 m above the ground. Table 1 summarizes the main meteorological parameters for each month during the two growing seasons. Evapotranspiration was measured using an eddy covariance (EC) system, which was located in the central south of the vineyard. The fetch length was from 300 to 1000 m. The EC system consists of a CSAT3 three-dimensional sonic anemometer (Gill Instruments, UK), open-path H2 O & CO2 analyzer (Li-CorInc, USA, Model Li7500), NR-LITE net radiometer (Kipp & Zonen, Holland), HMP45C temperature and humidity sensor, CS616 and CR5000 data logger (Cambell Scientific, USA). The sensors were installed at a 4.0 m height above ground level. Wind speed, ultrasound virtual temperature, the densities of atmosphere and water vapor were measured by CSAT3 and Li-7500 every 0.1 s to obtain latent and sensible heat above the canopy. Soil heat flux (G) was measured at two points, in the ditch and ridge, by heat flux plates (Model HFP01) installed at 5.0 cm below the ground surface. All probes were connected with data logger, and computed the average value of 30 min. The mean daily air temperature measured in the vineyard was used to calculate cumulative growing degree days (GDD) using a 10 ◦ C base (Poblete-Echeverría and Ortega-Farias, 2013). The procedures for correcting the EC measurements are as follows: planar fit method for coordinate rotation (Paw et al., 2000; Finnigan et al., 2003) and density correction according to the method of Webb et al. (1980). Once the observations in rainy days or five or more of the 48 daily observations in a day were excluded, the observations in these days were not used to compute daily ET. However, if fewer than five observations in a day were excluded, data were linearly interpolated to complete daily cumulative fluxes (Ortega-Farias et al., 2007). The accuracy of the EC measurements was evaluated by the energy balance closure, using a linear regression between turbulent energy fluxes (H + ET) and available energy (Rn − G) (Twine et al., 2000; Wilson et al., 2002). After performing the linear regression using half-hourly EC values, the intercept, slope and coefficient of determination (R2 ) for the

two years were 24.64 W m−2 , 0.80 and 0.89, respectively. The values were within the common results found in previous studies (Wilson et al., 2002; Li et al., 2005). For daytime EC-based data, the measured energy budget components were forced to close using ‘Bowen-ratio closure’ method proposed by Twine et al. (2000), assuming that Bowen-ratio is correctly measured by the EC system. During nighttime periods, the ‘residual-␭ET closure’ method proposed by Twine et al. (2000) was adopted in this study, which assumed that the EC-based H is accurately measured and solved for ␭ET as the residual to the energy balance equation (ET = Rn − G − H) (Li et al., 2013). 2.2.3. Sap flow Plant transpiration (T) was determined using sap flow system (Flow32-1K system, Dynamax, Houston, TX, USA). Eight vine trees were randomly chosen within a 30 m diameter circle from the EC system, and sap flow was monitored during May 3–October 12 in 2013 and May 12–October 23 in 2014. The gauges were installed at a height of >0.2 m above the ground surface on one of two trunks per vine. Irrigation may result in flooding and damage the sensors, so the sensors were disconnected before irrigation and reinstalled within 2–3 days after irrigation (Zhang et al., 2011). Gauge output were measured every 60 s and recorded as 15-min averages with a CR1000 data logger. Sap flow (L d−1 ) was scaled to tree transpiration (mm d−1 ) using the average ground area of each vine, and the vineyard transpiration was averaged from the eight monitored trees. 2.2.4. Soil evaporation Daily soil evaporation was measured by micro-lysimeter made from PVC tubes with a diameter of 10 cm and height of 20 cm. Eighteen micro-lysimeters were installed within sap flow rows at six different positions: with a distance of 40 cm between two adjacent micro-lysimeters on a vertical line between two grapevine rows. There are three replications of each position. Daily soil evaporation at each micro-lysimeter was obtained as the difference between the weights measured by an electronic scale with the precision of 0.1 g at 19:00, and averaged as vineyard soil evaporation. Some studies needed regular reinstallation of micro-lysimeter to minimize the difference between soil moisture inside and outside the tubes (Yunusa et al., 2004; Zhu et al., 2014). To investigate such effect on soil evaporation measurement, one replication was reinstalled every 3–5 d from April 30 to July 6 in 2013 in this study, and compared with three unchanged replications. The difference between two methods was smaller (Ereinstalled = 1.02Eunchanged − 0.11, R2 = 0.89, 49 observations), thus the unchanged micro-lysimeter was

44

P. Zhao et al. / Agricultural Water Management 160 (2015) 41–56

used. And diurnal variation of soil evaporation was also measured in 18 sunny days during the experiment period, two of the three replications were weighted once an hour in the daytime, and the hourly soil evaporation was calculated. 2.3. Evapotranspiration model 2.3.1. Shuttleworth–Wallace model The S–W model combines a one-dimensional model of soil evaporation and a one-dimensional model of plant transpiration. Surface and canopy resistances regulate the mass and heat transfer on soil and plant surfaces, and aerodynamic resistances regulate those between the two surfaces and the atmospheric boundary layer (Shuttleworth and Wallace, 1985). The two terms are calculated by the following equations: ET = E + T = Cs PMs + Cc PMc

(3)

where  is latent heat of vaporization (J kg−1 ), ET, E and T are evapotranspiration, soil evaporation and canopy transpiration, respectively. PMS and PMC are terms similar to those in the Penman–Monteith model to describe soil evaporation and canopy transpiration. Values of PMs , PMc , Cs and Cc are calculated as follows: A +



a cP D−ras (A−As ) raa +ras

PMs =

+ A +



PMc = +



Cs = 1 +



Cc = 1 +



1+

rss raa +ras

a cP D−rac As raa +rac



1+

rsc raa +rac

Rs Ra Rc (Rs + Ra ) Rc Ra Rs (Rc + Ra )





(4)

 (5)

−1 (6)

−1 (7)

where  is the slope of saturation vapor pressure curve (kPa K−1 ); A and As are available energy input above canopy and above soil surface (W m−2 ), respectively; cP is specific heat of dry air at constant pressure (J kg−1 K−1 ); a is air density (kg m−3 ); D is water vapor deficit (kPa);  is psychrometric constant (kPa K−1 ); rsc and rac are bulk resistances of canopy stomatal and boundary layer (s m−1 ), respectively; ras and raa are aerodynamic resistances from soil to canopy and from canopy to reference height (s m−1 ), respectively; rss is soil surface resistance (s m−1 ). Values of Ra , Rs and Rc are estimated as follows: Ra = ( + ) raa

(8)

Rs =

( + ) ras

+ rss

(9)

Rc =

( + ) rac

+ rsc

(10)

The available energy at canopy and soil surface is computed as: (11)

As = Rns − G

(12) (W m−2 ),

(W m−2 ),

where Rn is net radiation G is soil heat flux Rns is radiation reaching soil surface (W m−2 ), which can be calculated using Beer’s law as follows: Rns = Rn exp (−CLAI)

rsmin F1 LAI F2 F3 F4

rsc =

(14)

where rsmin represents the minimal stomatal resistance of individual leaves under optimal conditions. Fi (i = 1,2,3,4) is the stress function of a specific environmental variable: F1 =

f =



1+f

f + rsmin /rsmax



(15)

0.55Rsi 2 Rsl LAI

(16)

F2 = 1 − 0.0025D



A = Rn − G

determined as 0.39, and then used the measured radiation above and under the canopy on a sunny day to validate the value of C and found that the extinction coefficient was reliable. Also, OrtegaFarias et al. (2010) used daily values of Rns and Rn in a vineyard with a distance between and within rows of 2.5 m and 1.5 m, and estimated an extinction coefficient of 0.38 ± 0.05, which was similar to our result. In this study, the same approach was used to calculate the three aerodynamic resistances, i.e., rac , ras and raa , as Shuttleworth and Wallace (1985). Canopy resistance (rsc ), which is generally expressed as a function of solar radiation, water vapor deficit, air temperature and soil water content, can be calculated using the Jarvis approach by the following equations (Noilhan and Planton, 1989):

(13)

where LAI is leaf area index (m2 m−2 ), C is the extinction coefficient of light attenuation. Using the MAESTRA model (Wang and Jarvis, 1990), Zhang (2012) simulated the canopy radiation transmission in the same vineyard and the extinction coefficient was

F3 = 1 − 0.0016(298 − Ta )

(17) 2

(18)

where rsmax is maximum stomatal resistance (2000 s m−1 ), Rsi the incoming solar (short-wave) radiation (W m−2 ), and Rsl threshold radiation value above the stomata openness (30 W m−2 ). The term 2/LAI represents the shading between leaves, and the factor 0.55 is the PAR portion of solar radiation flux density. D is water vapor deficit (kPa) and Ta is air temperature (K). In this study, diurnal photosynthesis variation of vine leaves was measured on a sunny day every month from July to October in 2013 using a LI-6400 portable photosynthesis system (Li-CorInc, Lincoln, NE, USA), and found a minimal stomatal resistance of 136 s m−1 , which was not much different from the published data. Zhang et al. (2008) found a rsmin of 146 s m−1 in a nearby vineyard and Ortega-Farias et al. (2010) determined rsmin value as 144 s m−1 in a Merlot vineyard, so the value 136 s m−1 was used as the rsmin in this study. The effect of soil moisture stress on rsc can be computed with the normalized soil water content (F4 ) (Calveta et al., 1998): F4 =

i − WP FC − WP

(19)

where i is volumetric soil moisture content in the root-zone; FC and WP are the corresponding soil field capacity and wilting point. Soil surface resistance (rss ) regulates water vapor movement from interior to soil surface, and is assumed to depend strongly on the soil water content of upper layer (s ). Previous studies have been conducted to investigate rss and commonly express it as a function of s , such as linear (Thompson et al., 1981; Zhang et al., 2008), exponential (Sellers et al., 1992; Zhu et al., 2014), or power functions (Hu et al., 2009; Ortega-Farias et al., 2010). But the function type and unknown parameters can not be accurately determined unless the independent soil evaporation is available, because soil texture and the depth chosen to measure soil water content were variable among different studies, which may result in variable relationships between rss and s . So in this study, the measured daily soil evaporation in 2014 was used to calculate rss by solving Eq. (4), and then plotted rss (excluding the computed negative values) against

P. Zhao et al. / Agricultural Water Management 160 (2015) 41–56

3000

rss (s m-1)

2.3.2. FAO dual-Kc method In FAO dual-Kc method, actual ET is defined as the product of crop coefficient (Kc ) and reference evapotranspiration (ET0 ), and Kc is divided into soil evaporation (Ke ) and plant transpiration (Kcb ) (Allen et al., 1998).

Observed (Present work) Zhang et al.(2008) Ortega-Farias et al.(2010) Zhu et al.(2014) Fitted (Present work)

4000

ET = Kc ET0 = (Ks Kcb + Ke ) ET0 2000

1000

0 2.0

2.5

3.0

3.5

4.0

4.5

θsat/θs Fig. 1. Comparison of soil surface resistance (rss , s m−1 ) response to soil surface moisture conditions (sat /s ).

measured s to determine the function type and parameters. Here three common functions were chosen as follows: rss = a(

sat )+b s

rss = a(

sat ) s



rss

45

= exp

(20)

b

(21)

s a−b sat

(22)

where a and b are unknown parameters, s volumetric soil water content at the 10 cm depth, sat saturated water content of surface layer (0.43 m3 m−3 ). After parameterizing the three functions, the performance of each one was validated in the model and the power function was chosen to determine rss by Eq. (23). The observed and fitted relationships (with R2 = 0.47) between rss and sat /s were shown in Fig. 1, and compared to several published researches and found that the relationship was quite variable among different studies. rss = 1.265(

sat ) s

5.418

(23)

Finally, the constant input parameters used in S–W model are shown in Table 2.

(24)

where Kcb , Ke and Ks are basal crop coefficient, soil evaporation and water stress coefficients, respectively. Following the proposed segment approach in the FAO-56, the vine growing seasons were divided into four growth seasons: initial (Lini ), development (Ldev ), mid (Lmid ) and late seasons (Llate ). The initial and mid seasons are characterized by horizontal line segments, while the development and late seasons are characterized by rising and falling line segments. The FAO-56 and observed length of growth seasons for the vineyard are given in Table 3. As shown in Table 3, a large difference existed between the recommended and observed values, even within the observed two years. The differences in the length of growing stages between the experimental seasons and those recommended by the FAO-56 may be attributed to climatic conditions at the studied site. Previous studies also found these differences, and reported that climate change was expected to have significant effect on the length of growing length because of temperature variation (Webb et al., 2007; Tomasi et al., 2011). Thus the determination of the length growth stages is important in applying FAO method because the data are not transferable to other agro-climatic locations. De Medeiros et al. (2001) presented a method that express Kcb curves in terms of cumulative degree days (GDD), which have proven to be useful in PobleteEcheverría and Ortega-Farias (2013). Hence GDD was used at the end of each stage and found a slight variance within two years though the growing lengths were quite different (Table 3), which was probably caused by the climatic variation: the total rainfall during the experiment period was 84 mm in 2013 and 239 mm in 2014, respectively. The observed GDD were similar to a previous study in a Merlot vineyard (Poblete-Echeverría and Ortega-Farias, 2013) (Table 3), so dividing growing period based on field observations is necessary. In the segment approach, three critical Kcb values are required to generate the entire Kcb curve during the vine growing season. The standard values of Kcb recommended by FAO-56 were commonly adopted. However, this could produce large errors. Several researchers used the observed Kcb values based on sap flow measurements and can improve the performance of the dual-Kc method (Er-Raki et al., 2010; Poblete-Echeverría and Ortega-Farias, 2013).

Table 2 List of constants used in FAO dual-Kc method and S–W model. Symbol  FC S  WP S Ze Zr p REW  FC R  WP R C d n zo zo  rb rsmin rsmax k  sat a

Name Surface soil field capacity Surface soil wilting point Depth of soil surface layer Effective root-zone depth Evapotranspiration depletion factor Readily evaporable water Root-zone soil field capacity Root-zone soil wilting point Extinction coefficient Zero plane displacement Eddy diffusivity decay constant Roughness length of canopy Roughness length of bare soil Mean boundary layer resistance Minimum stomatal resistance Maximum stomatal resistance Von Karman’s constant Surface soil saturated water content

The h value is canopy height of grapevine (m).

Value 0.27 0.10 0.15 1.0 0.45 10 0.31 0.11 0.39 2/3 ha 2.5 0.05 ha 0.01 25 136 2000 0.40 0.43

Unit 3

−3

m m m3 m−3 m m – mm m3 m−3 m3 m−3 – m – m m s m−1 s m−1 s m−1 – m3 m−3

Source

Model

Measured Measured Allen et al. (1998) Measured Allen et al. (1998) Allen et al. (1998) Measured Measured Zhang (2012) Sene (1994) Shuttleworth and Wallace (1985) Sene (1994) Sene (1994) Shuttleworth and Wallace (1985) Measured Ortega-Farias et al. (2010) Shuttleworth and Wallace (1985) Measured

Dual-Kc Dual-Kc Dual-Kc Dual-Kc Dual-Kc Dual-Kc Dual-Kc , S–W Dual-Kc , S–W S–W S–W S–W S–W S–W S–W S–W S–W S–W S–W

46

P. Zhao et al. / Agricultural Water Management 160 (2015) 41–56

Table 3 Days of different growth stages for grapevines suggested by FAO-56 and observed for the furrow-irrigated Merlot vineyard during 2013 and 2014 seasons, and a comparison with a previous study in a Merlot vineyard. Source

Days of different growth stages (d)

Total days (d)

Lini

Ldev

Lmid

Llate

Recommended by FAO-56

30

60

40

80

210

Present study Observed 2013 Observed 2014 Observed average

18 (119) 18 (53) 18 (86)

46 (545) 50 (519) 48 (532)

59 (1215) 90 (1269) 75 (1242)

57 (1481) 22 (1291) 40 (1386)

180 180 180

Poblete-Echeverría and Ortega-Farias (2013) Observed 2007/08 31 (132) Observed 2008/09 38 (174) Observed average 35 (153)

47 (478) 41 (555) 44 (517)

63 (1205) 62 (1241) 63 (1223)

50 (1641) 42 (1653) 46 (1647)

192 184 188

Cumulative growing degree days (GDD) are shown in brackets. Lini is initial season, Ldev is development season, Lmid is mid season and Llate is late season, shown in FAO-56 (Allen et al., 1998).

Basal crop coefficient (Kcb )

0.8 0.7

Kcb FAO

mean experimental Kcb

experimental Kcb

(a)

0.6 0.5 0.4 0.3 0.2 0.1 0.0

115

135

155

175

195

215

235

255

275

295

Day of year Basal crop coefficient (Kcb )

0.8 0.7

(b)

0.6 0.5 0.4 0.3 0.2 0.1 0.0

115

135

155

175

195

215

235

255

275

295

Day of year Fig. 2. Observed basal crop coefficient (Kcb ) and Kcb in FAO-56 during the experimental period of vineyard in 2013 (a) and 2014 (b).

In this study, the observed Kcb was obtained by the following equation: Kcb

sap

=

Tsap ETo

Kcb value is attribute to low fractional cover (maximum fc ≈ 30%) observed in this study, similar results were also found in a Merlot vineyard (Poblete-Echeverría and Ortega-Farias, 2013). Thus in this study the observed lengths of growth stages and basal crop coefficients were adopted instead of the FAO recommended values in the dual-Kc method. The details of the calculation functions and procedures in the FAO dual-Kc method are available from Allen et al. (1998). The soil parameters used in the method are presented in Table 2. The fraction of soil surface wetted by irrigation was measured as 0.5, and the effective depth of vine root-zone was 1.0 m. The maximum average fraction cover (fc ) of the vineyard was measured about 0.3 when the LAI reaches the peak value, and zero at the beginning of the growing season. In this study, the successive measurement of fc was not conducted, thus assuming that fc was linearly correlated to LAI and then determined seasonal variation of vineyard fraction cover. And then a sensitivity analysis of fc variation was done for the model outputs, and found only a small influence. 2.3.3. Sensitivity analysis of models To evaluate the sensitivity of S–W model and FAO dual-Kc method to the uncertainties in the values of input parameters (Table 2) or resistance formulations, the percentage variation in the estimated evapotranspiration, soil evaporation and plant transpiration were computed, respectively. In this study, values of input parameters and resistances were varied by ±30%. The set of parameter and resistance values which we used as the reference set was the same as for the two years calculations (Ortega-Farias et al., 2010). The ET, E and T predicted with it was on average 2.66, 0.84 and 1.82 mm d−1 in 2013, 2.21, 1.08 and 1.15 mm d−1 in 2014 for the S–W model, respectively. For the FAO dual-Kc method, the ET, E and T predicted with it was on average 2.07, 0.84 and 1.23 mm d−1 in 2013, 2.10, 1.04 and 1.06 mm d−1 in 2014, respectively.

(25)

where Kcb sap is basal crop coefficient calculated using sap flow measurements, Tsap is vine transpiration (mm d−1 ) and ET0 is reference evapotranspiration (mm d−1 ). Based on the observed growth stages, the observed Kcb curves were obtained by averaging the observed Kcb values. Fig. 2 compares the basal crop coefficient (Kcb ) calculated in 2013 and 2014 by Eq. (25) with the FAO Kcb , and the observed Kcb curves were also shown. The recommended and observed values of Kcb are presented in Table 4, and also compared to several previous studies on grapevine. Kcb in FAO and observed Kcb had the same trend during two years. Nevertheless, mean observed Kcb was systematically smaller than that suggested by Allen et al. (1998). The lower

2.3.4. Evaluation of model performance In this study, statistical comparison included liner regression analysis, root mean square error (RMSE), mean absolute error (MAE) and percent mean absolute relative error (PMARE) (Mayer and Butler, 1993; Ali and Abustan, 2014). These statistical parameters are described as follows:



N i=1 (Ei

RMSE =

− Oi )2

N

1 |Ei − Oi | N

(26)

N

MAE =

i=1

(27)

P. Zhao et al. / Agricultural Water Management 160 (2015) 41–56

47

Table 4 Kcb values for different varieties of grapevine. Source

Basal crop coefficient Kcb ini

Recommended by FAO-56 Teixeira et al. (2007)a Allen and Pereira (2009) ˜ et al. (2012) Fandino Poblete-Echeverría and Ortega-Farias (2013)

0.15 0.55–0.57 0.20–0.25 0.08–0.11 0.20–0.27

Present work Observed 2013 Observed 2014

Kcb sap ini 0.03 0.05

Kcb mid

Variety

Training system

Vine Petite Syrah Vine ˜ Albarino Merlot

– Wire vertical trellis – Wire vertical trellis Wire vertical trellis

Merlot

Wire vertical trellis

Kcb late

0.65 0.72–0.73 0.40–0.70 0.27–0.57 0.50–0.55 Kcb sap mid 0.56 0.46

0.40 0.60–0.63 0.30–0.55 0.07–0.25 0.43 Kcb sap late 0.05 0.10

a Approximate values, exact values were not reported. Sub-indices sap represents Kcb values obtained using Eq. (25). Sub-indices ini, mid and late represent initial, mid and late seasons, respectively (Allen et al., 1998).

100 |Ei − Oi | N Oi N

PMARE =

8

(a)

(28)

where N is number of observations, Ei and Oi are estimated and observed values, respectively. 3. Results and discussion

Ely+Tsap (mm d-1)

1

3.1. Comparison of evapotranspiration measured by eddy covariance and the sum of transpiration by sap flow system and soil evaporation by micro-lysimeter

6

4

2

Ely+Tsap = 0.94 ETec R² = 0.89 RMSE=0.47 mm d-1

0

E

and 1.15 mm d−1 in 2014, with ETlyec (ratio of E to ET) of 42% and 54%, respectively. Also the hourly observed Ely + Tsap using diurnal measurements of soil evaporation and plant transpiration were close to ETec , the regression equation was Ely + Tsap =





0.95ETec R2 = 0.83, RMSE = 0.066mm/h , and diurnal variation of Ely + Tsap and ETec in the 18 sunny days was close (Fig. 4). Thus plant transpiration measured by sap flow system and soil evaporation measured by micro-lysimeter can provide reliable measurements of hourly and daily ET components in the studied vineyard. Hence the measured soil evaporation, plant transpiration and evapotranspiration were used to evaluate the performance of the ET partitioning models. 3.2. The performance of two approaches in ET partitioning 3.2.1. Shuttleworth–Wallace model Having parameterized the S–W model as described above, the model was used to simulate the half-hourly E (Eq. (4)), T (Eq. (5)) and ET (Eq. (3)) values (W m−2 ), respectively. Daily estimated E, T and ET were obtained by summing up half-hourly values. Fig. 5 shows seasonal comparison of observed ET, E and T and those estimated by the S–W model in the two seasons. And the performance of the S–W model in estimating ET and its components is shown in Table 5. A good agreement was found between observed and simulated ET values in both years (Fig. 5a and b), but some discrepancies were found between observed and estimated values after irrigation or

0

2

4 ETec (mm d-1)

6

8

8

(b)

Ely+Tsap (mm d-1)

After forcing the energy balance to be closed, vineyard evapotranspiration measured by the eddy covariance (EC) system (ETec ) was compared to the sum of transpiration (Tsap ) measured by sap flow system and soil evaporation (Ely ) measured by microlysimeter. Fig. 3 indicates that Ely + Tsap was close to ETec in 2013 and 2014. The regression was not statistically different from line 1:1, with root mean square error (RMSE) of 0.47 mm d−1 in two years, indicating a good agreement between the measured ET by two approaches. Averaged values of ETec , Ely and Tsap were 2.31, 0.97 and 1.21 mm d−1 in 2013, and 2.35, 1.26

6

4

2

Ely+Tsap = 1.01 ETec R² = 0.87 RMSE=0.46 mm d-1

0 0

2

4 ETec (mm d-1)

6

8

Fig. 3. Comparison of daily EC observed evapotranspiration (ETec ) versus sum of components (Ely + Tsap ) by micro-lysimeter and sap flow system in 2013 (a) and 2014 (b). Dashed line represents 1:1 line.

rainfall events, which were possibly attributed to the estimation error of soil evaporation (Fig. 5f). Also the S–W model overestimated ET from DOY (days of year) 149 to 158 in 2013, which was caused by the overestimation of transpiration (Fig. 5c). Thus the S–W model provides an accurate estimation of evapotranspiration, with RMSE, mean absolute error (MAE) and percent mean absolute relative error (PMARE) of 0.68 mm d−1 , 0.52 mm d−1 and 25.7%, respectively (Table 5). Ortega-Farias et al. (2010) also reported that the S–W model can estimate daily ET in a vineyard, with slope, RMSE, MAE and ID of 1.01, 0.51 mm d−1 , 0.41 mm d−1 and 0.88, respectively. Zhu et al. (2014) used the S–W model to simulate daily ET at a maize field in northwest China, and found that the estimated and observed values falls closely among the 1:1 line (slope = 1.01, intercept = 0.01, RMSE = 0.05 mm d−1 ). The S–W model in simulating daily evapotranspiration was also validated by other studies (Hu et al., 2009, 2013; Gharsallah et al., 2013; Zhu et al., 2013).

48

P. Zhao et al. / Agricultural Water Management 160 (2015) 41–56

0.8

(a)

Ely+Tsap (mm h-1)

0.6

0.4

0.2 Ely+Tsap = 0.95 ETec R² = 0.83 RMSE=0.066 mm h-1 0.0 0.2

0.0

0.4

0.6

0.8

ET components (mm h-1)

ETec (mm h-1) Ely

0.4

Tsap

ETec

Ely+Tsap

(b) 0.3 0.2 0.1 0.0 6:00

9:00

12:00 15:00 Time (hh:mm)

18:00

Fig. 4. Comparison of hourly EC observed evapotranspiration (ETec ) versus sum of components (Ely + Tsap ) by micro-lysimeter and sap flow system: (a) scatter plot; (b) mean diurnal variation. Dashed line represents 1:1 line.

Table 5 Statistical analysis of measured and estimated daily evapotranspiration, soil evaporation and transpiration with two approaches for the furrow-irrigated vineyard. Season

a

b

R2

RMSE

MAE

PMARE

ETKc 2013 2014 Overall

0.05 −0.48 −0.13

0.87 1.13 0.96

0.85 0.80 0.81

0.62 0.66 0.64

0.48 0.54 0.51

22.3 24.5 23.4

ETSW 2013 2014 Overall

0.48 0.14 0.28

0.99 0.94 0.98

0.85 0.75 0.79

0.75 0.59 0.68

0.57 0.47 0.52

29.2 22.1 25.7

EKc 2013 2014 Overall

−0.07 −0.06 −0.07

0.93 0.92 0.92

0.60 0.84 0.76

0.50 0.39 0.44

0.37 0.31 0.34

39.9 31.6 35.5

ESW 2013 2014 Overall

0.16 0.64 0.42

0.66 0.38 0.49

0.50 0.44 0.46

0.49 0.68 0.60

0.40 0.47 0.44

48.1 42.0 44.9

TKc 2013 2014 Overall

0.19 0.13 0.16

0.93 0.93 0.93

0.84 0.74 0.80

0.37 0.35 0.36

0.28 0.26 0.27

58.0 52.8 55.4

TSW 2013 2014 Overall

0.73 0.12 0.38

1.12 1.02 1.10

0.74 0.66 0.63

1.04 0.48 0.81

0.88 0.36 0.62

159.1 51.9 104.8

a intercept (mm d−1 ); b slope (dimensionless); R2 determination coefficient; RMSE root mean square (mm d−1 ); MAE mean absolute error (mm d−1 ); PMARE percent mean absolute relative error (%); ETKc , EKc and TKc are the estimated evapotranspiration, soil evaporation and transpiration by the FAO dual-Kc method, respectively; ETSW , ESW and TSW are the corresponding values estimated by the S–W model.

The fluctuation of measured and estimated daily soil evaporation is illustrated in Fig. 5e and f. The S–W model underestimated E significantly in both years, especially after rainfall or irrigation. Both observed and simulated E increased with the increased wetting area of soil surface layer, but the increment is different. Actual E measured by micro-lysimeter increased significantly after rainfall or irrigation, e.g., 0.22 mm d−1 before rainfall (DOY129) and 3.15 mm d−1 after rainfall (DOY131), or 1.07 mm d−1 before irrigation (DOY214) and 2.28 mm d−1 after irrigation (DOY220) (Fig. 5f). However, the corresponding simulated values were 0.41 mm d−1 before rainfall and 2.51 mm d−1 after rainfall, 0.74 mm d−1 before irrigation and 1.15 mm d−1 after irrigation (Fig. 5f). After comparing three commonly used soil surface resistance equations, a power function was chosen to determine rss (Eq. (23)) with lower R2 between observed rss and sat /s . As shown in Fig. 1, when surface soil was wet (sat /s <3.0) after rainfall and irrigation, the optimized rss was significantly higher than most of the observed values, which may underestimate E. Gharsallah et al. (2013) also found that soil evaporation estimated by S–W model were less variable around wetting events, which may use the improper soil surface resistance values. Also the soil parameters (field capacity and wilting point) measured at several limited points may not adequately represent the whole field. And the sensitivity analysis of rss and soil parameters variation in this model were conducted, results confirmed such effects (Table 6). From the regression analysis of E for two years data, the slope, intercept, RMSE, MAE and PMARE were 0.49, 0.42, 0.60, 0.44 mm d−1 and 44.9%, respectively (Table 5). The simulated transpiration (T) were higher than observed values, especially in 2013 (Fig. 5c and d). Regression  equations are as follows: Tsimulated = 1.12Tobserved + 0.73 R2 = 0.74, RMSE = 1.04mm/d, MAE = 0.88mm/d in  2013 and Tsimulated = 1.02Tobserved + 0.12 R2 = 0.66, RMSE = 0.48mm/d, MAE = 0.36mm/d in 2014 (Table 5). As mentioned above, soil parameters measured at several limited points may produce the errors in estimating vineyard transpiration. And the minimal stomata resistance (rsmin ) determined in this study was based on only four-day measurements, which may differ from the true value and cause model errors (Table 6). Previous studies also showed that for long-term simulations of ET and its components, systematic overestimation or underestimation may occur when constant parameters were used, thus it was important to consider seasonal changes of physiology-related parameters (e.g., minimal and maximum stomatal resistance, rsmin and rsmax ) (Hu et al., 2009; Zhu et al., 2013). Although the S–W model overestimated T, it can also provide an acceptable estimation of T due to the complexity and heterogeneity of vineyard canopy structure. After evaluating the performance of the S–W model, the model had a good estimation of ET and an acceptable estimation of T, but heavily underestimation of E in a furrow-irrigated vineyard. For sparse vegetation types like the row planting grapevines in this study, evapotranspiration is a complex function of vapor and energy balance of both the canopy and soil surface. Due to the heterogeneity of canopy and large diurnal changes in the exposure of plants and soil to solar radiation, it is difficult to accurately estimate the vineyard ET and its components using the energy balance method (Heilman et al., 1994; Trambouze et al., 1998; Ortega-Farias et al., 2007). 3.2.2. FAO dual-Kc method Seasonal evapotranspiration, soil evaporation and plant transpiration were calculated using the FAO dual-Kc method over grapevines during two consecutive growing seasons. Seasonal basal crop coefficients of vineyard for the simulation were as follows: in 2013, Kcb ini = 0.03, Kcb mid = 0.56 and Kcb late = 0.05, while in 2014, Kcb ini = 0.05, Kcb mid = 0.46 and Kcb late = 0.10 (Table 4).

P. Zhao et al. / Agricultural Water Management 160 (2015) 41–56

These values were derived based on sap flow measurements in two years, expressed as the ratio of measured transpiration to reference evapotranspiration (ETo ) (Eq. (25)). A good agreement was found between the observed and simulated daily ET in two seasons (Fig. 6a and b). The root mean square error (RMSE) between measured and predicted ET was 0.62 and 0.66 mm d−1 for 2013 and 2014, respectively, with slight underestimation in the first season and slight overestimation in the second season (Table 5). Some discrepancies between these two values were observed around rainfall events (DOY131 in 2014), and dur-

49

ing stress period (DOY152 to 176 in 2014) when average soil water content in root-zone was 0.16 m3 m−3 . There were no significant differences between measured and estimated ET after irrigation in both seasons, which was similar to a pervious study in a Merlot vineyard in Chile using dual-Kc method (Poblete-Echeverría and Ortega-Farias, 2013). However, Er-Raki et al. (2010) found significant discrepancies between measured and simulated ET in an olive orchard after irrigation. The discrepancies after rainfall could be caused by canopy interception of precipitation, which has been found about 8% for a heavily rainfall in a sparse olive orchard with

7

80

(a) 2013 Evapotranspiration (mm d-1)

60

5

50

4 40

3 30

2

20

1

Irrigation and Rainfall (mm)

70

6

10

0

0

110

140

170

200 Day of year

230

260

290

7

80 70

6

60

5

50

4 40

3 30

2

20

1

Irrigation and Rainfall (mm)

Evapotranspiration (mm d-1)

(b) 2014

10

0

0

110

140

170

200 Day of year

230

260

290

5

80

(c) 2013

Transpiration (mm d-1)

60 50

3

40

2

30 20

1

Irrigation and Rainfall (mm)

70

4

10

0

0

110

140

170

200 Day of year

230

260

290

Fig. 5. Comparison of daily observed (dot) and simulated (line) values of evapotranspiration, soil evaporation and plant transpiration by S–W model in 2013 and 2014.

50

P. Zhao et al. / Agricultural Water Management 160 (2015) 41–56

4

80

(d) 2014

3

60 50

2

40 30

1

20

Irrigation and Rainfall (mm)

Transpiration (mm d-1)

70

10

0

0

110

140

170

200 Day of year

230

260

290

5

80

(e) 2013

60 50

3

40

2

30 20

1

Irrigation and Rainfall (mm)

Soil Evaporation (mm d-1)

70

4

10

0

0

110

140

170

200 Day of year

230

260

290

5

80

(f) 2014

60 50

3

40

2

30 20

1

Irrigation and Rainfall (mm)

Soil Evaporation (mm d-1)

70

4

10

0

0

110

140

170

200 Day of year

230

260

290

Fig. 5. (Continued).

average LAI of 1.1 (Gómez et al., 2001). And Liu and Luo (2010) also showed that the FAO dual-Kc method is appropriate in estimating total ET but not accurate in simulating the peak value after rainfall. For the period of hydric stress (DOY152 to 176) mentioned above in 2014, the estimated ET was significant lower than the observed values (Fig. 6b), which may attribute to the errors in simulating transpiration, and will be discussed below. The FAO dual-Kc method provided a reliable result in partitioning ET into E and T in both years though some errors

occurred (Fig. 6c–f). Table 5 shows a close correlation between the simulated and observed transpiration in both seasons: Tsimulated = 0.93Tobserved + 0.19, RMSE = 0.37mm/d in 2013, while in 2014 Tsimulated = 0.93Tobserved + 0.13, RMSE = 0.35mm/d. But underestimation of T were observed during the above mentioned stress period (DOY152 to 176) in 2014, which may result from the estimating errors of soil water stress coefficient related to the rooting depth (Zr ) (Er-Raki et al., 2008, 2010). Thus a sensitivity analysis of Zr variation was done and found a slight effect of Zr on

P. Zhao et al. / Agricultural Water Management 160 (2015) 41–56

the simulated T (Table 6), indicating that the rooting depth was not a contributing factor to produce the errors in simulating T in this study, and the errors may possibly result from the differences between the observed Kcb values and mean observed Kcb curve used in the method during the stress period (Fig. 2b), which will underestimate T. Also the soil parameters measured at several limited points may not adequately represent the whole field, thus can produce the errors in simulating T. Sensitivity analysis results also indicate that soil field capacity at the root-zone has larger effect on the estimated T (Table 6). Compared to the simulated T, the dual-Kc method also provided an accurate estimation of E (Fig. 6e and f), despite some discrepancies around wetting events and a slight seasonal underestimation: Esimulated = 0.93Eobserved − 0.07, RMSE = 0.50mm/d in 2013, and Esimulated = 0.92Eobserved − 0.06, RMSE = 0.39mm/d in 2014 (Table 5). The higher precision in simulating T than E using dual-Kc method was also found over other sparse vegetation. Er-Raki et al. (2010) used the experimental Kcb curves obtained from sap flow system to partition ET in an olive orchard, and also found a better estimation of T (RMSE = 0.59mm/d) and a poor estimation of E (RMSE = 0.73mm/d). But for dense planting crops, like maize and wheat, the dual-Kc method can estimate both components well. Zhao et al. (2013) applied this method to a winter wheat-summer maize system, and compared the simulated and observed soil evaporation during three growing seasons: the regression slope was close to 1 and RMSE were 0.37, 0.45 and 0.49 mm d−1 for three growing seasons, respectively. Ding et al. (2013) also confirmed the availability of FAO dual-Kc method in ET partitioning in a maize field. As expected, both measured and simulated E increased after irrigation or rainfall, and the increment was similar, and only a few days the estimated values were lower than observed values (Fig. 6e and f).

51

3.2.3. Inter-comparison of two approaches Fig. 7 shows the daily ET and its components estimated by two approaches versus the measured values. In 2014, two approaches provided a good estimation of ET, while in 2013, the S–W model overestimated ET and the FAO dual-Kc method underestimated ET (Fig. 7a and b, Table 5). From the regression analysis of ET for two years data, RMSE, MAE and PMARE were 0.68, 0.52 mm d−1 and 25.7% for S–W model, and 0.64, 0.51 mm d−1 and 23.4% for FAO dual-Kc method. Thus FAO dual-Kc method performed better in estimating ET than S-W model. The FAO dual-Kc method had a good estimation of T and was closely to the observed value (Fig. 7e and f), with RMSE, MAE and PMARE of 0.36, 0.27 mm d−1 and 55.4%, but 0.81, 0.62 mm d−1 and 104.8% for S–W model (Table 5). However, both approaches significantly overestimated T during the period after DOY284 in 2014, when the observed T suddenly decreased to very low level after a snowfall (12 mm, DOY284) (Figs. 5 d and 6 d). The S–W model significantly underestimated E, especially around wetting events (Fig. 7c and d), with RMSE, MAE and PMARE of 0.60, 0.44 mm d−1 and 44.9%. However, the dual-Kc method provided an acceptable simulation of E, with a higher R2 (0.76) and lower RMSE, MAE and PMARE (0.44, 0.34 mm d−1 and 35.5%) compared to the S–W model. Thus the FAO dual-Kc method had better performance in estimating soil evaporation than the S–W model. Possible reason is that in the FAO dual-Kc method, soil evaporation was a function of daily variation in soil water content in the evaporative layer, while in the S–W model, soil evaporation depended on soil surface resistance (rss ), which was not accurate enough, especially around wetting events. Gharsallah et al. (2013) also found the daily values of soil evaporation estimated by the FAO dual-Kc method are more variable in time than those simulated by the S–W model, especially after rainfall and irrigation.

Table 6 Sensitivity analysis of predicted vineyard evapotranspiration, soil evaporation and transpiration by two approaches to uncertainties in input parameters and resistances. Values outside the brackets are for 2013 and inside the brackets are for 2014. Model

Input variables

Percentage of variation −30% ET

S–W

Dual Kc

Extinction coefficient of the crop for net radiation, C Zero plane displacement, d (m) Leaf area index, LAI Eddy diffusivity decay constant, n Roughness length of the canopy, zo (m) Roughness length of the bare soil, zo  (m) Root-zone soil field capacity,  FC R Root-zone soil wilting point,  WP R Mean boundary layer resistance, rb (s m−1 ) Minimum stomatal resistance, rsmin (s m−1 ) Aerodynamic resistance between canopy and reference level, raa (s m−1 ) Bulk boundary layer resistance in the canopy, rac (s m−1 ) Aerodynamic resistance between substrate and canopy, ras (s m−1 ) Bulk stomatal resistance of the canopy, rsc (s m−1 ) Soil surface resistance, rss (s m−1 ) Vineyard fractional cover, fc Root-zone soil field capacity,  FC R Root-zone soil wilting point,  WP R Surface soil field capacity,  FC S Surface soil wilting point,  WP S Effective root-zone depth, Zr (m) Readily evaporable water, REW (mm) Evapotranspiration depletion factor, p

+30% E

T

ET

E

T

−0.6 (−2.2) −0.1 (−1.8) 8.6 (5.4) 0.0 (−1.0) 0.0 (0.1) 0.0 (0.0) −13.2 (−10.9) −15.0 (−24.3) −0.4 (0.3) −8.9 (−7.5) 0.0 (0.3)

−6.6 (−7.0) −3.3 (−6.2) −12.5 (−10.7) −3.2 (−4.1) 1.0 (0.7) 0.4 (0.4) 5.7 (3.5) 6.1 (6.9) 1.2 (0.5) 3.9 (2.4) −2.9 (−1.0)

2.0 (2.3) 1.4 (2.3) 18.2 (20.4) 1.3 (2.0) −0.4 (−0.5) −0.3 (−0.3) −21.7 (−24.4) −24.6 (−53.5) −1.2 (0.0) −14.6 (−16.7) 1.4 (1.4)

0.9 (2.7) 0.1 (1.1) −10.4 (−5.9) 0.1 (0.9) 0.0 (−0.1) 0.0 (0.0) 24.2 (22.0) 8.3 (12.5) 0.5 (−0.3) 13.0 (11.6) −0.1 (−0.3)

8.2 (8.6) 0.4 (3.1) 16.3 (13.6) 0.9 (3.5) −0.3 (−0.9) −0.1 (−0.5) −9.9 (−6.7) −3.2 (−3.5) −1.3 (−0.5) −5.5 (−3.6) 3.5 (1.2)

−2.5 (−2.8) −0.1 (−0.8) −22.5 (−24.1) −0.4 (−1.6) 0.1 (0.7) 0.1 (0.5) 39.7 (48.9) 13.5 (27.4) 1.3 (0.0) 21.3 (25.8) −1.7 (−1.7)

0.5 (−0.3)

−1.3 (−0.5)

1.3 (0.0)

0.1 (0.9)

2.6 (3.6)

−1.1 (−1.7)

0.1 (−0.5)

13.6 (12.0) 5.3 (8.0)

−5.6 (−3.7) 19.0 (17.7)

22.3 (26.7) −1.0 (−1.1)

1.5 (2.0) −1.5 (1.5) 0.0 (1.2) −8.4 (−8.2) 1.7 (1.0) −1.6 (0.4) −0.4 (−0.7) −0.9 (−2.6)

3.7 (4.3) 0.0 (0.0) 0.0 (0.0) −20.7 (−18.0) 4.1 (2.5) 0.0 (0.0) −0.9 (−1.5) 0.0 (0.0)

0.0 (−0.3) −2.6 (2.9) 0.0 (2.5) 0.0 (1.6) 0.0 (−0.4) −2.7 (0.8) 0.0 (0.2) −1.5 (−5.2)

−0.4 (0.3)

1.2 (0.5)

−1.2 (0.0)

−1.5 (−2.3)

0.8 (1.2)

−9.2 (−7.7) −3.6 (−5.6)

4.0 (2.5) −13.0 (−12.4)

−15.2 (−17.3) 0.7 (0.9)

−1.6 (−2.2) −3.2 (−5.1) −0.3 (−2.2) 7.8 (5.1) −1.7 (−1.2) −2.4 (−0.5) 0.4 (0.7) 0.0 (1.6)

−4.0 (−4.7) 0.0 (0.0) 0.0 (0.0) 19.6 (12.8) −4.3 (−2.8) 0.0 (0.0) 0.9 (1.5) 0.0 (0.0)

0.0 (0.3) −5.4 (−10.1) −0.5 (−4.4) −0.1 (−2.5) 0.0 (0.4) −4.1 (−0.9) 0.0 (−0.2) 0.0 (3.1)

52

P. Zhao et al. / Agricultural Water Management 160 (2015) 41–56

3.2.4. Sensitivity analysis The analysis of sensitivity to uncertainties in input parameters and resistances for the two approaches is shown in Table 6. For the S–W model, the sensitivity of predicted ET, E and T to uncertainties

in d, n, zo , zo , rb , raa , rac and ras are small. Predicted T is very sensitive to root-zone soil parameters (FC R , WP R ), LAI and rsmin , for these parameters, the deviation ranged from 13 to 40%. Predicted E was sensitive to rss and LAI, for the two parameters, the deviation ranged

7

80

(a) 2013 Evapotranspiration (mm d-1)

60

5

50

4 40

3 30

2

20

1

Irrigation and Rainfall (mm)

70

6

10

0

0

110

140

170

200 Day of year

230

260

290

7

80

(b) 2014 Evapotranspiration (mm d-1)

60

5

50

4 40

3 30

2

20

1

Irrigation and Rainfall (mm)

70

6

10

0

0

110

140

170

200 Day of year

230

260

290

4

80

(c) 2013

3

60 50

2

40 30

1

20

Irrigation and Rainfall (mm)

Transpiration (mm d-1)

70

10

0

0

110

140

170

200 Day of year

230

260

290

Fig. 6. Comparison of daily observed (dot) and simulated (line) values of evapotranspiration, soil evaporation and plant transpiration by FAO dual-Kc method in 2013 and 2014.

P. Zhao et al. / Agricultural Water Management 160 (2015) 41–56

53

4

80

(d) 2014

3

60 50

2

40 30

1

20

Irrigation and Rainfall (mm)

Transpiration (mm d-1)

70

10

0

0

110

140

170

200 Day of year

230

260

290

5

80

(e) 2013

60 50

3

40

2

30 20

1

Irrigation and Rainfall (mm)

Soil Evaporation (mm d-1)

70

4

10

0

0

110

140

170

200 Day of year

230

260

290

5

80

(f) 2014

60 50

3

40

2

30 20

1

Irrigation and Rainfall (mm)

Soil Evaporation (mm d-1)

70

4

10

0

0

110

140

170

200 Day of year

230

260

290

Fig. 6. (Continued).

from 16 to 19%. E was also sensitive to root-zone soil parameters and extinction coefficient (C), for the two parameters, the deviation ranged from 6 to 10%. It is worth noting that soil evaporation is very sensitive to rss while evapotranspiration is not, indicating that the S–W model is acceptable in estimating ET using a recommended or unadjusted soil resistance equation, but the simulated E was significantly different from the observed values. Ortega-Farias et al. (2007) also found that the variation of soil surface resistance have slight effect on ET estimation in the S–W model, and the variation of modeled ET was less than ± 4% when rss were varied by ±30%. And

Zhu et al. (2014) indicated that the optimized S–W model cannot properly partition total ET into its components when using only the EC-measured ET data. For the FAO dual-Kc method, soil evaporation was sensitive to surface soil field capacity (FC S ), and the deviation ranged about 20% when FC S were varied by ±30%, while transpiration was sensitive to root-zone soil field capacity (FC R ), and the deviation was less than 10% when FC R were varied by ±30%. In both approaches, soil parameters can produce a large deviation of predicted ET and its components, indicating that it is

54

P. Zhao et al. / Agricultural Water Management 160 (2015) 41–56

7

7

(a) 2013

(b) 2014 6

ET simulated (mm d-1)

ET simulated (mm d-1)

6 5 4 3 2 1

5 4 3 2 1

Dual Kc S-W

0

Dual Kc S-W

0 0

1

2 3 4 5 ET observed (mm d-1)

6

7

0

4

2 3 4 5 ET observed (mm d-1)

7

(d) 2014 E simulated (mm d-1)

3

2

1

3

2

1

Dual Kc S-W

Dual Kc S-W

0

0 1 2 3 E observed (mm d-1)

0

4

1

0

2 3 E observed (mm d-1)

4

4

5

(f) 2014

(e) 2013 4

T simulated (mm d-1)

T simulated (mm d-1)

6

4

(c) 2013 E simulated (mm d-1)

1

3

2

3

2

1

1

Dual Kc S-W

Dual Kc S-W 0

0 0

1

2 3 T observed (mm d-1)

4

5

0

1 2 3 T observed (mm d-1)

4

Fig. 7. Comparison of daily observed and simulated values of evapotranspiration, soil evaporation and plant transpiration by two approaches in 2013 (left panel) and 2014 (right panel). Dashed line represents 1:1 line.

important to consider the spatial heterogeneity of soil and to obtain accurate soil parameters for sparse furrow-irrigated vineyard. 4. Conclusions The S–W model and FAO dual-Kc method provided better estimation of evapotranspiration (ET) except for several days of wetting events (after irrigation and rainfall), with RMSE and PMARE of

0.68 mm d−1 and 25.7% for the S–W model, and 0.64 mm d−1 and 23.4% for the FAO dual-Kc method. Using the site-specific basal crop coefficient (Kcb ) curves obtained from sap flow system, the FAO dual-Kc method obtained an acceptable result in ET partitioning. Predicted soil evaporation (E) and plant transpiration (T) were slightly lower than the measured values, with RMSE of 0.44 and 0.36 mm d−1 , PMARE of 35.5% and 55.4%, respectively. Using sitespecific soil surface resistance, the S–W model overestimated T,

P. Zhao et al. / Agricultural Water Management 160 (2015) 41–56

with RMSE and PMARE of 0.81 mm d−1 and 104.8%, but underestimated E, with RMSE and PMARE of 0.60 mm d−1 and 44.9%. Both approaches had significant errors of E around wetting events, and overestimated T after a snowfall. The estimated ET and its components were sensitive to soil field capacity and wilting point in both approaches, and in the S–W model, predicted E was also sensitive to soil surface resistance and leaf area index (LAI), and predicted T sensitive to LAI and minimum stomatal resistance. Therefore, two approaches can estimate ET with good accuracy, but the FAO dual-Kc method had higher accuracy in estimating E and T. Acknowledgements The research was supported by the National Natural Science Foundation of China (51321001, 91425302), and the 111 Program of Introducing Talents of Discipline to Universities (B14002). References Agam, N., Evett, S.R., Tolk, J.A., Kustas, W.P., Colaizzi, P.D., Alfieri, J.G., Mckee, L.G., Copeland, K.S., Howell, T.A., Chavez, J.L., 2012. Evaporative loss from irrigated interrows in a highly advective semi-arid agricultural area. Adv. Water Resour. 50, 20–30. Ali, M.H., Abustan, I., 2014. A new novel index for evaluating model performance. J. Nat. Resour. Dev. 04, 1–9. Allen, R.G., Pereira, L.S., 2009. Estimating crop coefficients from fraction of ground cover and height. Irrig. Sci. 28, 17–34. Allen, R.G., Pereira, L.S., Raes, D., Smith, M., 1998. Crop evapotranspiration-guidelines for computing crop water requirements. In: FAO Irrigation and Drainage Paper 56. FAO, Rome, Italy. Calveta, J.C., Noilhana, J., Roujeana, J.L., Bessemoulina, P., Cabelguenneb, M., Oliosoc, A., Wigneronc, J.P., 1998. An interactive vegetation SVAT model tested against data from six contrasting sites. Agric. For. Meteorol. 92, 73–95. De Medeiros, G.A., Arruda, F.B., Sakai, E., Fujiwara, M., 2001. The influence of crop canopy on evapotranspiration and crop coefficient of beans (Phaseolus vulgaris L.). Agric. Water Manage. 49, 211–224. Ding, R.S., Kang, S.Z., Zhang, Y.Q., Hao, X.M., Tong, L., Du, T.S., 2013. Partitioning evapotranspiration into soil evaporation and transpiration using a modified dual crop coefficient model in irrigated maize field with ground-mulching. Agric. Water Manage. 127, 85–96. Doorenbos, J., Pruitt, W.O., 1977. Crop water requirements. In: Irrigation and Drainage Paper FAO 24. Food and Agriculture Organization of the United Nations, Rome, Italy. Dragoni, D., Lakso, A.N., Piccioni, R.M., 2004. Transpiration of an apple orchard in a cool humid climate: measurement and modeling. Acta Hortic. 664, 175–180. Er-Raki, S., Chehbouni, A., Hoedjes, J., Ezzahar, J., Duchemin, B., Jacob, F., 2008. Improvement of FAO-56 method for olive orchards through sequential assimilation of thermal infrared-based estimates of ET. Agric. Water Manage. 95, 309–321. Er-Raki, S., Chehbouni, A., Boulet, G., Williams, D.G., 2010. Using the dual approach of FAO-56 for partitioning ET into soil and plant components for olive orchards in a semi-arid region. Agric. Water Manage. 97, 1769–1778. ˜ M., Cancela, J.J., Rey, B.J., Martínez, E.M., Rosa, R.G., Pereira, L.S., 2012. Fandino, ˜ vineyards Using the dual-Kc approach to model evapotranspiration of albarino ˜ with consideration of active ground cover. Agric. (Vitis vinifera L. cv. Albarino) Water Manage. 112, 75–87. Farahani, H.J., Bausch, W.C., 1995. Performance of evapotranspiration models for maize-bare soil to closed canopy. Trans. ASAE 38, 1049–1059. Finnigan, J.J., Clement, R., Malhi, Y., Leuning, R., Cleugh, H.A., 2003. A re-evaluation of long-term flux measurement techniques part I: averaging and coordinate rotation. Boundary-Layer Meteorol. 107, 1–48. Gómez, J.A., Giráldez, J.V., Fereres, E., 2001. Rainfall interception by olive trees in relation to leaf area. Agric. Water Manage. 49, 65–76. Gharsallah, O., Facchi, A., Gandolfi, C., 2013. Comparison of six evapotranspiration models for a surface irrigated maize agro-ecosystem in Northern Italy. Agric. Water Manage. 130, 119–130. Goodwin, I., Whitfield, D.M., Connor, D.J., 2006. Effects of tree size on water use of peach (Prunus persica L. Batsch). Irrig. Sci. 24, 59–68. Heilman, J.L., McInnes, K.J., Savage, M.J., Gesch, R.W., Lascano, R.J., 1994. Soil and canopy energy balances in a west Texas vineyard. Agric. For. Meteorol. 71, 99–114. Hu, Z.M., Yu, G.R., Zhou, Y.L., Sun, X.M., Li, Y.N., Shi, P.L., Wang, Y.F., Song, X., Zheng, Z.M., Zhang, L., Li, S.G., 2009. Partitioning of evapotranspiration and its controls in four grassland ecosystems: application of a two-source model. Agric. For. Meteorol. 149, 1410–1420. Hu, Z.M., Li, S.G., Yu, G.R., Sun, X.M., Zhang, L.M., Han, S.J., Li, Y.N., 2013. Modeling evapotranspiration by combing a two-source model, a leaf stomatal model, and a light-use efficiency model. J. Hydrol. 501, 186–192. Intrigliolo, D.S., Castel, J.R., 2008. Effects of irrigation on the performance of grapevine cv. Tempranillo in Requena, Spain. Am. J. Enol. Vitic. 59, 30–38.

55

Kang, S.Z., Gu, B.J., Du, T.S., Zhang, J.H., 2003a. Crop coefficient and ratio of transpiration to evapotranspiration of winter wheat and maize in a semi-humid region. Agric. Water Manage. 59, 239–254. Kang, S.Z., Hu, X.T., Du, T.S., Zhang, J.H., Jerie, P., 2003b. Transpiration coefficient and ratio of transpiration to evapotranspiration of pear tree (Pyrus communis L.) under alternative partial root-zone drying conditions. Hydrol. Process. 17, 1165–1176. Kang, S.Z., Su, X.L., Tong, L., Shi, P.Z., Yang, X.Y., Abe, Y.K., Du, T.S., Shen, Q.L., Zhang, J.H., 2004. The impacts of human activities on the water-land environment of the Shiyang River basin, an arid region in northwest China. Hydrol. Sci. J. 49, 413–427. Katerji, N., Rana, G., 2006. Modelling evapotranspiration of six irrigated crops under Mediterranean climate conditions. Agric. For. Meteorol. 138, 142–155. Kool, D., Agam, N., Lazarovitch, N., Heitman, J.L., Sauer, T.J., Ben-Gal, A., 2014. A review of approaches for evapotranspiration partitioning. Agric. For. Meteorol. 184, 56–70. Lascano, R.J., Van Bavel, C.H.M., Hatfield, J.L., Upchurch, D.R., 1987. Energy and water balance of a sparse crop: simulated and measured soil and crop evaporation. Soil Sci. Soc. Am. J. 51, 1113–1121. Lascano, R.J., Baumhardt, R.L., Lipe, W.N., 1992. Measurement of water flow in young grapevines using the stem heat balance method. Am. J. Enol. Vitic. 43, 159–165. Li, Z.Q., Yu, G.R., Wen, X.F., Zhang, L.M., Ren, C.Y., Fu, Y.L., 2005. Energy balance closure at ChinaFLUX sites. Sci. China Ser. D 48, 51–62. Li, S.E., Kang, S.Z., Zhang, L., Li, F.S., Hao, X.M., Ortega-Farias, S., Guo, W.H., Ji, S.S., Wang, J.T., Jiang, X.L., 2013. Quantifying the combined effects of climatic, crop and soil factors on surface resistance in a maize field. J. Hydrol. 489, 124–134. Liu, Y.J., Luo, Y., 2010. A consolidated evaluation of the FAO-56 dual crop coefficient approach using the lysimeter data in the North China Plain. Agric. Water Manage. 97, 31–40. Liu, C.M., Zhang, F., Jiang, J.F., Wei, Y.G., 2006. Influence of climate resources on brewing grape along desert area in Hexi Corridor. Agric. Res. Arid Areas China 24, 143–148. Mayer, D., Butler, D., 1993. Statistical validation. Ecol. Model. 68, 21–32. Monteith, J.L., 1965. Evaporation and environment. Symp. Soc. Exp. Biol. 19, 205–234. Noilhan, J., Planton, S., 1989. A simple parameterization of land surface processes for meteorological models. Mon. Weather Rev. 117, 536–549. Ortega-Farias, S., Carrasco, M., Olioso, A., Acevedo, C., Poblete, C., 2007. Latent heat flux over a Cabernet Sauvignon vineyard using the Shuttleworth and Wallace model. Irrig. Sci. 25, 161–170. Ortega-Farias, S., Poblete-Echeverría, C., Brisson, N., 2010. Parameterization of a two-layer model for estimating vineyard evapotranspiration using meteorological measurements. Agric. For. Meteorol. 150, 276–286. Paw, U.K.T., Baldocchi, D.D., Meyers, T.P., Wilson, K.B., 2000. Correction of eddy covariance measurements incorporating both advective effects and density fluxes. Boundary-Layer Meteorol. 97, 487–511. Poblete-Echeverría, C.A., Ortega-Farias, S.O., 2013. Evaluation of single and dual crop coefficients over a drip-irrigated Merlot vineyard (Vitis vinifera L.) using combined measurements of sap flow system and an eddy covariance system. Aust. J. Grape Wine Res. 19, 249–260. ˜ Poblete-Echeverría, C., Ortega-Farias, S., Zuniga, M., Fuentes, S., 2012. Evaluation of compensated heat-pulse velocity method to determine vine transpiration using combined measurements of eddy covariance system and microlysimeters. Agric. Water Manage. 109, 11–19. Rana, G., Katerji, N., 2008. Direct and indirect methods to simulate the actual evapotranspiration of an irrigated overhead table grape vineyard under Mediterranean conditions. Hydrol. Process. 22, 181–188. Ritchie, J.T., 1972. Model for predicting evaporation from a row crop with incomplete cover. Water Resour. Res. 8, 1204–1213. Sellers, P.J., Heiser, M.D., Hall, F.G., 1992. Relations between surface conductance and spectral vegetation indices at intermediate (100m2 to 15km2 ) length scales. J. Geophys. Res. 97, 19033–19059. Sene, K.J., 1994. Parameterisations for energy transfers from a sparse vine crop. Agric. For. Meteorol. 71, 1–18. Shuttleworth, W.J., Wallace, J.S., 1985. Evaporation from sparse crops-an energy combination theory. Q. J. R. Meteorol. Soc. 111, 839–855. Teixeira, A.H., de, C., Bastiaanssen, W.G.M., Bassoi, L.H., 2007. Crop water parameters of irrigated wine and table grapes to support water productivity analysis in the São Francisco river basin, Brazil. Agric. Water Manage. 94, 31–42. Thompson, N., Barrie, I.A., Ayles, M. 1981. The meteorological office rainfall and evaporation calculation system: MORECS. Hydrological Memorandum No. 45. Tomasi, D., Jones, G.V., Giust, M., Lovat, L., Gaiotti, F., 2011. Grapevine phenology and climate change: relationships and trends in the Veneto region of Italy for 1964–2009. Am. J. Enol. Vitic. 62, 329–339. Trambouze, W., Bertuzzi, P., Voltz, M., 1998. Comparison of methods for estimating actual evapotranspiration in a row-cropped vineyard. Agric. For. Meteorol. 91, 193–208. Twine, T.E., Kustas, W.P., Norman, J.M., Cook, D.R., Houser, P.R., Meyers, T.P., Prueger, J.H., Starks, P.J., Wesely, M.L., 2000. Correcting eddy-covariance flux underestimates over a grassland. Agric. For. Meteorol. 103, 279–300. Wang, Y.P., Jarvis, P.G., 1990. Description and validation of an array model-MAESTRO. Agric. For. Meteorol. 51, 257–280.

56

P. Zhao et al. / Agricultural Water Management 160 (2015) 41–56

Webb, E.K., Pearman, G.I., Leuning, R., 1980. Correction of flux measurements for density effects due to heat and water vapour transfer. Q. J. Royal Meteorol. Soc. 106, 85–100. Webb, L.B., Whetton, P.H., Barlow, E.W.R., 2007. Modelled impact of future climate change on the phenology of winegrapes in Australia. Aust. J. Grape Wine Res. 13, 165–175. Williams, D.G., Cable, W., Hultine, K., Hoedjes, J.C.B., Yepez, E.A., Simonneaux, V., Er-Raki, S., Boulet, G., de Bruin, H.A.R., Chehbouni, A., Hartogensis, O.K., Timouk, F., 2004. Evapotranspiration components determined by stable isotope, sap flow and eddy covariance techniques. Agric. For. Meteorol. 125, 241–258. Wilson, K., Goldstein, A., Falge, E., Aubinet, M., Baldocchi, D., Berbigier, P., Bernhofer, C., Ceulemans, R., Dolman, H., Field, C., Grelle, A., Ibrom, A., Law, B.E., Kowalski, A., Meyers, T., Moncrieff, J., Monson, R., Oechel, W., Tenhunen, J., Valentini, R., Verma, S., 2002. Energy balance closure at FLUXNET sites. Agric. For. Meteorol. 113, 223–243. Yepez, E.A., Huxman, T.E., Ignace, D.D., English, N.B., Weltzin, J.F., Castellanos, A.E., Williams, D.G., 2005. Dynamics of transpiration and evaporation following a moisture pulse in semiarid grassland: a chamber-based isotope method for partitioning flux components. Agric. For. Meteorol. 132, 359–376. Yunusa, I.A.M., Walker, R.R., Lu, P., 2004. Evapotranspiration components from energy balance, sapflow and microlysimetry techniques for an irrigated vineyard in inland Australia. Agric. For. Meteorol. 127, 93–107. Zhang, B.Z., Kang, S.Z., Li, F.S., Zhang, L., 2008. Comparison of three evapotranspiration models to Bowen ratio-energy balance method for a vineyard in an arid desert region of northwest China. Agric. For. Meteorol. 148, 1629–1640.

Zhang, B.Z., Kang, S.Z., Zhang, L., Tong, L., Du, T.S., 2009. An evapotranspiration model for sparsely vegetated canopies under partial root-zone irrigation. Agric. For. Meteorol. 149, 2007–2011. Zhang, B.Z., Kang, S.Z., Li, F.S., Tong, L., Du, T.S., 2010. Variation in vineyard evapotranspiration in an arid region of northwest China. Agric. Water Manage. 97, 1898–1904. Zhang, Y.Q., Kang, S.Z., Ward, E.J., Ding, R.S., Zhang, X., Zheng, R., 2011. Evapotranspiration components determined by sap flow and microlysimetry techniques of a vineyard in northwest China: dynamics and influential factors. Agric. Water Manage. 98, 1207–1214. Zhang, Y.Q., 2012. Evapotranspiration and canopy water-carbon coupled model of a vineyard in arid regions of northwest China. In: PhD Thesis. China Agricultural University (in Chinese). Zhao, N.N., Liu, Y., Cai, J.B., Rosa, R.D., Paredes, P., Pereira, L.S., 2013. Dual crop coefficient modelling applied to the winter wheat-summer maize crop sequence in North China Plain: basal crop coefficients and soil evaporation component. Agric. Water Manage. 117, 93–105. Zhu, G.F., Su, Y.H., Li, X., Zhang, K., Li, C.B., 2013. Estimating actual evapotranspiration from an alpine grassland on Qinghai–Tibetan plateau using a two-source model and parameter uncertainty analysis by Bayesian approach. J. Hydrol. 476, 42–51. Zhu, G.F., Li, X., Su, Y.H., Zhang, K., Bai, Y., Ma, J.Z., Li, C.B., Hu, X.L., He, J.H., 2014. Simultaneously assimilating multivariate data sets into the two-source evapotranspiration model by Bayesian approach: application to spring maize in an arid region of northwestern China. Geosci. Model Dev. 7, 1467–1482.