Vous êtes sur la page 1sur 11

Environ Earth Sci (2015) 74:5443–5452

DOI 10.1007/s12665-015-4557-6

ORIGINAL ARTICLE

Variability and sensitivity analyses of spring wheat


evapotranspiration measurements in Northwest China
Chunmei Wang1 • Qingyan Meng1 • Zewei Miao2 • Xingfa Gu1 • Tao Yu1 •

Yulin Zhan1 • Miao Liu1 • Lijuan Zheng1 • Qiyue Liu1

Received: 25 October 2014 / Accepted: 19 May 2015 / Published online: 4 June 2015
 Springer-Verlag Berlin Heidelberg 2015

Abstract The variability and sensitivity of crop evapo- Keywords Evapotranspiration  Variability  Temporal
transpiration (ET) measurements at field scale are still stability  Sensitivity  Soil moisture
poorly understood in the irrigated farmland of arid region
in Northwest China. The spatial and temporal dynamics
and sensitivity of field ET are fundamental for the scaling Introduction
up and validation of ET estimates from remote sensing
data. In the study, we analysed the dynamics, impact fac- Crop evapotranspiration (ET) is a crucial component of the
tors and sensitivity of spring wheat ET during the growing hydrological cycle and in farming system management,
season in Northwest China. Results indicated that there was especially in arid agricultural areas (Mauser and Schädlich
a significant effect of first irrigation event on the spatial and 1998; Singh et al. 2008; Ding et al. 2013a). Variability and
temporary variability of ET. At the tillering-shooting stage sensitivity analysis of crop ET are important in hydrologic
(before the first irrigation event), spatial variability of ET modelling, optimizing tillage management practices,
was the lowest and gradually increased with crop growth. improving crop water use efficiency, and predicting
In some experimental plots, spring wheat ET had a sig- responses of crop growth to natural and human-induced
nificantly higher temporal stability than other plots except disturbances (Lei and Yang 2010; Anderson et al. 2012).
for the tillering-shooting stage. The sample sites with Crop ET measurements and estimates are often con-
higher temporal stability could be used for long-term ducted at three scales, i.e., field scale, catchment scale and
monitoring samples and for up scaling of ET measure- regional scale. Because of differences in mechanisms
ments. In comparison with LAI and ET0, surface soil driving and ecological processes between small and large
moisture change Dh0–20 cm was the most sensitive variable scale, it is difficult to convert ET measurements and pre-
of ET measurements, which could be used as the auxiliary dictions among various scales (Pietroniro et al. 1994;
variable to improve the ET accuracy. With the soil mois- Vinnikov et al. 1996; Yoo et al. 1998; Pietroniro and
ture measurements, the relative error of ET was 9.4 % with Prowse 2002; Cosh et al. 2004). It has been shown that
only half the number of ET sampling data. crop ET measurements and estimates at large spatial scale
have to be calibrated and verified with in-site ground truth
measurements at the field scale (Jacobs et al. 2004; De
& Chunmei Wang Lannoy et al. 2006; French et al. 2012). Therefore, crop ET
wangcm@radi.ac.cn variability and sensitivity at the field level are fundamental
& Qingyan Meng for the scaling up of ET estimates and are essential for
mengqy@radi.ac.cn validation of ET estimates by space- or air-borne remote
1 sensing data. Recently, multivariate spatial statistics,
Institute of Remote Sensing and Digital Earth,
Chinese Academy of Sciences, Beijing 100101, China including covariance, self-variance, semivariograms, and
2 spearman’s rank coefficient have been adopted to deter-
Department of Agricultural and Biological Engineering,
University of Illinois at Urbana-Champaign, 1304 West mine spatial and temporal variability of atmospheric
Pennsylvania Avenue, Urbana, IL 61801, USA evaporation or vegetation ET at the basin or regional levels

123
5444 Environ Earth Sci (2015) 74:5443–5452

(Tong et al. 2007; Miao et al. 2009, 2011; Yang et al. 2012; temperature of 8 C and 150 frost-free days annually. The
Raziei and Pereira 2013). However, little attention has been region is arid, with a mean annual precipitation of 164 mm,
paid to the variability of crop ET at the irrigation field and a mean annual pan evaporation of 2000 mm measured
scale, especially to the impact factor and sensitivity anal- with a cylinder Class A evaporation pan. The groundwater
ysis in the arid farmlands. table averages below 30 m.
Arid areas account for one-third of the earth’s land sur- Spring wheat is one of the conventional staple crops in
face area. In the arid area of Northwest China, over-ex- Northwest China and feeds dozens of millions of people. In
ploitation of water resources has led to serious this region, spring wheat is usually grown one season from
environmental problems, e.g., soil salinization and deserti- March to July per year with deep ploughing. And residue is
fication, degradation of grassland, groundwater depletion, removed and no cover remains after sowing. Because of
and decrease of vegetation cover (Kang et al. 2004). So the low precipitation and high evaporation, the water require-
water resources management is very important for the ment of spring wheat is provided mainly by irrigation from
development of sustainable agriculture in the Northwest groundwater.
China. The objectives of this study were (1) to measure and In this study, spring wheat was sown in the experimental
estimate spring wheat ET in an arid agricultural area of field on March 15, 2010 and March 13, 2011 and harvested on
China, (2) to investigate the spatial and temporal variability July 20, 2010 and July 17, 2011. The irrigation regime and
of spring wheat ET at the field level, and (3) to analyse crop management practices are listed in Table 1. To eliminate
sensitivity of crop ET measurements and driving variables. the influences of tillage practices on ET measurements, tillage
This research provides an insight into crop ET dynamic practices, including land preparation, sowing, harvest and
measurements and accurate quantification of ET variability fertilization, were the same for all experimental plots.
and sensitivity, which are essential to improve efficiency of The surface of the experimental field is fairly flat with a
water use by crops, to optimize irrigation regimes, to select maximum difference in elevation of approximately 0.3 m
variable-control irrigation technology throughout the world. (Fig. 2). In the soil profile there is a layer of impermeable
clay below approximately 100 cm, which impedes deep
drainage to groundwater. To ensure homogeneous irriga-
Study area and sampling tion within the field, the depth of water layer above the
ground is controlled during the irrigation process.
Study area
Experimental design and sampling
The experiments were conducted at the Shiyanghe Exper-
imental Station for Water-saving at the Agriculture and Evapotranspiration measurements were made in a spring
Ecology of China Agricultural University, located in Gansu wheat field (length 195 m and width 105 m) at the exper-
Province of Northwest China (N 37520 , E 102500 , altitude imental station in 2010. The experimental field was split
1581 m), during 2010 and 2011 (Fig. 1). The site has a into 91 plots, each 15 m 9 15 m. To obtain a robust esti-
mean annual sunshine duration of [3000 h, a mean annual mate of spatial variability, 21 samples, spaced at 7.5-m

Fig. 1 Location of study area

123
Environ Earth Sci (2015) 74:5443–5452 5445

Table 1 Crop management


Year Sowing date Harvest date Irrigation events
and irrigation events during the
2010–2011 spring wheat Irrigation datea Irrigation quota (mm)b
growing seasons
2010 15 March 20 July May 1 (DOS = 47) 97.5
May 20 (DOS = 66) 97.5
June 11 (DOS = 88) 97.5
July 4 (DOS = 111) 97.5
2011 13 March 17 July May 2 (DOS = 50) 97.5
May 20 (DOS = 68) 97.5
June 9 (DOS = 88) 97.5
July 8 (DOS = 117) 97.5
a
DOS day after sowing
b
Irrigation quota was measured with pump meter, which was calculated from irrigation water amount in
m3 divided by area of the experimental field

A portable hydrosense instrument (CS620, Campbell Sci-


entific Inc., USA) was used to measure surface soil moisture
at the 0–20 cm depth (h0–20 cm) at the centre of each plot.
During the soil moisture measurements, three sets of hydro-
sense instruments were used simultaneously to sample data
for 30 min to minimize the effects of measurement time.
A Bowen ratio instrument (Campbell Scientific Inc., USA)
was installed near the south side of central spring wheat in
2010, which was to validate the ET measurements. Leaf area
index (LAI) was measured every 5 days with a canopy
analysis instrument (SunScan, SS1, England). The canopy
instrument was set up in the centre of each plot within a
Fig. 2 Relative elevation (m) of the experimental field
(195 9 105 m2, with the reference point at the northwest corner of 60 9 60 cm area. Climate variables such as free water sur-
the field. Filled circle is for grid samples and arrow is sampling face evaporation capacity, precipitation, and wind speed were
direction) collected from a nearby weather station with a Weather Hawk
intervals, were collected in both east–west and south–north instrument (Campbell Scientific, USA).
directions. The measurements were sequenced from the In 2011, the experiments were conducted to verify ET
northwest corner plot, and they were named in accordance measurements. In 2011, the experimental field was 180 m long
with the S-curve sampling direction (labelled as plots 1, (east–west direction) and 150 m wide (south–north direction).
2,…, and 112) (Fig. 2). The measurements started on April It was split into 144 plots, each 15 9 15 m. Ground measure-
17 (DOS = 33, DOS = day after sowing) and ended on ments were started on May 12 (DOS = 60) and ended on June
July 2 (DOS = 109), 2010. 7 (DOS = 86), 2011. Because of the impacts of weather and
Soil moisture measurements were collected before and instrument availability, the 4 times measurements were con-
after each irrigation event, including 5 times for wheat ducted during two periods (DOS = 60–67 and
seeding-tillering stage (DOS = 33–43), 4 times for wheat DOS = 71–86). Results demonstrated that there were no sig-
tillering-shooting stage (DOS = 51–61), 4 times for nificant differences in soil texture for the top soil layer (the silt
shooting-heading stage (DOS = 71–81), and 3 times for loam according to the international system of soil classification
wheat heading-milk stage (DOS = 99–109). Because of standard), crop management or irrigation method between the
instrument problems, 97 soil moisture measurements were experimental fields in 2010 and 2011.
valid (data for plots 8, 15, 19, 20, 28, 36, 50, 51, 56, 63, 77,
88, 91 and 105 were not available). Volumetric soil
moisture in the root zone (0–100 cm) was measured Methods
through PVC tubes in the plots using a portable device
Diviner 2000 (Sentek Pty. Ltd., Australia). Every 3–5 days ET estimations with the water budget approach
measurements were obtained at 10-cm intervals to a depth
of 100 cm. Additional soil moisture data were collected Crop ETDt was estimated with the water budget approach
before and after irrigation events and after rainfall events. (Tong et al. 2007). And the 10-day cumulative ET (ET10)

123
5446 Environ Earth Sci (2015) 74:5443–5452

was as the variable in this study. The ET values measured During the spring wheat growing seasons in 2010 and
with the water budget approach are affected by individual 2011, there were four irrigation events (each *97.5 mm;
errors generated from the water budget components (Tong the irrigation quota was measured with a pump meter,
et al. 2007). So the estimated ET with the water budget which was calculated from the volume of irrigation water
approach must be verified through the measurements from amount in m3 divided by area of the experimental field).
other methods. Four precipitation events occurred with 4.5, 2, 4 and
In this study, the experimental site was fairly flat 18 mm of rainfall. Only the precipitation events of 2 and
(Fig. 2), and precipitation amount was low. An insulated 4 mm took place in the measurement period. Soil moisture
edge was created between the experimental field and the content in the 0–100 cm layer varied greatly and depended
neighbouring farmland, so runoff (RDt) from irrigation was on irrigation scheduling (irrigation quota and timing). Soil
negligible. Contribution from the water table (GDt) was moisture content had a peak value after irrigation and
ignored because the water table was more than 30 m deep decreased gradually thereafter.
at the study site. The drainage (DDt) could also be ignored
according to soil moisture in the soil profile after irrigation Spatial variability
(Fig. 3) and because a layer of impermeable clay occurred
below *100 cm. Moreover, deep drainages estimated with Spatial variability and dependence of ET were expressed by
Darcy Flux Law were 0.13, 1.53, 1.82, and 1.85 mm for the the empirical semivariances (Burgess and Webster 1980):
stages of seeding-tillering, tillering-shooting, shooting-
1 X Nh
heading, and heading-milk, respectively, and the ratio D10 cðhÞ ¼ ½Zðxi þ hÞ  Zðxi Þ2 ð1Þ
(a 10-day cumulative drainage) to ET10 was \6 % 2Nh i¼1
(Table 2). Therefore, the drainage (D10) was not consid- where xi is the ith sampling location, c(h) is the semivari-
ered in the paper. ance at lag distance h, Z is ET value, Nh is the number of
paired data at distance h (xi, xi ? h).
In the semivariogram model, nugget is the jumping
variance between short sample distances at the origin, led by
sample error and variance for short distances. Range is the
distance in which the difference of the variogram from the
sill becomes negligible. When the distance between sample
points is larger than the range value, the samples become
independent. Sill is the variogram value at the distance of
the range. The semivariance model has been categorized
into a spherical model, an exponential model, a Gaussian
model, a nugget model and a power model (Zhang 2005).

Temporal stability

As a spatial distribution index of water management,


temporal stability (also known as time stability, rank sta-
bility, or temporal persistence) reveals consistency of
spatial measurements of a variable over time (Vachaud
et al. 1985; Bertuzzi et al. 1994). Two techniques, Spear-
Fig. 3 Spatial variation of soil moisture at 0–100 cm soil profile
after irrigation (May 20, 2010) man’s rank correlation and temporal analysis of the

Table 2 Drainage estimated


Growing stages (DOS) D10 with the Darcy ET10 (mm) D10/ET10 (%)
with the Darcy Flux Law and
Flux Law (mm)
the calculation of D10/ET10
(assuming positive upward Seeding-tillering (33–43) -0.13 5.49 2.37
direction)
Tillering-shooting (51–61) -1.53 26.41 5.79
Shooting-heading (71–81) -1.82 36.83 4.94
Heading-milk (99–109) -1.85 37.35 4.95
DOS day after sowing, D10 10-day cumulative drainage, ET10 10-day cumulative evapotranspiration

123
Environ Earth Sci (2015) 74:5443–5452 5447

differences between individual and spatial average values, (called ‘‘sensitivity coefficient’’ hereinafter) has been
are often employed to determine temporal stability of ET widely used in ET studies (McCuen 1974; Qiu et al. 1998;
or soil moisture measurements. Temporal stability (cs) of Hupet and Vanclooster 2001), though the sensitivity coef-
ET measurements was calculated using Spearman’s Rank ficient did not account for the interactions among variables
coefficient cs: (Miao et al. 2011).
A positive or negative sensitivity coefficient of a vari-
P
m
6 ðETi;t  ETi;t0 Þ2 able indicates that ET will increase or decrease as the
i¼1 variable increases. The larger the sensitivity coefficient, the
cs ¼ 1  ð2Þ
mðm2  1Þ more sensitive to a specific variable the crop ET will be. In
where m is total sample size, ETi,t and ETi,t0 are ET values fact, the sensitivity coefficient is the slope of the tangent at
of the ith location at time t and t0 . The more stable the ET the origin of the sensitivity curve. The coefficient is
measurements between time t and t0 , the closer to 1 the accurate enough to represent the slope of the sensitivity
stability value (cs) will be. curve within a certain ‘‘linear range’’ around the origin.
The corresponding relative mean error di and variance The width of the range depends on the degree of non-
error r(d)2i were calculated as follows: linearity of the sensitivity curve. When a sensitivity curve
is linear, the sensitivity coefficient represents the change in
1X nt
ETi;t  ETt ET caused by any perturbation of the variable. If the sen-
di ¼ ð3Þ
nt t¼1 ETt sitivity curve is significantly non-linear, the predictive
power of the sensitivity coefficient will be limited to small
nt  2
1 X ETi;t  ETt perturbations only.
rðdÞ2i ¼  di ð4Þ
nt  1 t¼1 ETt

where di is the change in degree of soil moisture relative to Results and discussion
the mean value, and r(d)2i is the temporal stability of crop
ET measurements. To control the standard error and to Spatial variability analyses of crop ET
obtain the mean of ET estimates, the root mean standard
error was calculated as follows: Spring wheat ET10 estimated from the water budget
approach was compared with the measurements obtained
RMSEi ¼ ðdi þ rðdÞ2i Þ1=2
2
ð5Þ with Bowen ratio instrument. ET10 estimates were 5.49,
In the study, we used extreme differences rather than the 26.41, 36.83, and 37.35 mm for the stages of seeding-til-
standard deviation of relative differences (Vachaud et al. lering, tillering-shooting, shooting-heading, and heading-
1985; Van Pelt and Wierenga 2001), because the repeated milk, respectively. These estimates were comparable to the
number of ET measurements was low (only four) and measurements from the Bowen ratio instrument, which
probability distributions were not normal. were 5.25, 24.2, 33.58, and 34.98 mm for the same four
stages, respectively.
CV values of cumulative ET showed a decreasing trend
Sensitivity coefficient
with the increasing of ET observation interval. At the til-
lering-jointing and shooting-heading stages, when the
In ET and eco-hydrological studies, a ‘‘sensitivity curve
observation interval of ET was 4-, 6-, 11-, 14-day, CV
method’’ has been adopted to detect correlations between
values were 0.494, 0.487, 0.462, 0.445 and 0.482, 0.454,
dependent variables and independent variables (Saxton
0.339, 0.328, respectively. Likewise, at the heading-milk
1975; Singh and Xu 1997; Qiu et al. 1998; Beres and
stage, CV values were 0.495, 0.460 and 0.431 for 7-, 12-
Hawkins 2001; Hupet and Vanclooster 2001; Anderton
and 17-day cumulative ET. During the entire growing stage
et al. 2002; Goyal 2004; Miao et al. 2011). To eliminate the
in 2010, CV values of ET ranged from 0.33 to 0.49, which
effects of variable dimension on sensitivity of ET mea-
is highly spatially heterogeneous even for a small homo-
surements, a partial derivative of each variable was nor-
geneous field conforming to the previous report
malized as follows:
! (Raghuwanshi and Wallender 1997).
D ET=ET oET Vi Spatial variability of ET increased as the spring wheat
SVi ¼ lim ¼  ð6Þ
DVi !0 DV i= oVi ET grew. Cumulative ET10 during the growing seasons fol-
Vi
lowed a normal distribution and fit a spherical model sig-
where SVi is sensitivity coefficient of Vi, and Vi is the ith nificantly (Table 3). At the seeding-tillering stage, ET10
variable (Beven 1979). The normalization that calculated was only 5.49 mm with a low nugget and sill value. With
the ‘‘non-dimensional relative sensitivity coefficient’’ crop growth, the sill value of ET10 increased and the

123
5448 Environ Earth Sci (2015) 74:5443–5452

Table 3 Probability distribution and semivariogram of spring wheat ET estimates at various growing stages
Growing stages (DOS) Data distribution Model Nugget (mm)2 Sill (mm)2 Range (m) R2 ET10 (mm) r (mm)

Seeding-tillering (33–43) Normal Spherical 1.99 2.8 110 0.72 5.49 1.65
Tillering-shooting (51–61) Normal Spherical 82 160 75 0.75 26.41 12.25
Shooting-heading (71–81) Normal Spherical 70 230 55 0.83 36.83 14.87
Heading-milk (99–109) Normal Spherical 70 210 65 0.85 37.35 14.40

Table 4 Spearman rank coefficient of spring wheat ET estimates at various growing stages
Growing stage (DOS) Seeding-tillering (33–43) Tillering-shooting (51–61) Shooting-heading (71–81) Heading-milk (99–109)

Seeding-tillering (33–43) 1
Tillering-shooting (51–61) 0.345 1
Shooting-heading (71–81) 0.464 0.658 1
Heading-milk (99–109) 0.317 0.515 0.64 1

Fig. 4 Rank distribution of (a) (b)


relative errors of spring wheat 1.5 1.5

Mean relative difference


Mean relative difference

ET10: a at various growth 1 1


stages; b except the seeding-
tillering stage (after the first 0.5 0.5
irrigation)
0 0

-0.5 -0.5

-1 -1

-1.5 -1.5
0 20 40 60 80 100 0 20 40 60 80 100
Rank Rank

correlation distance decreased quickly. The large spatial measurements. Therefore, sampling locations 94 or 14
variability of ET has also been reported by Warrick (2003) could be used to estimate and calibrate the mean ET with a
and Brocca et al. (2006). relative error range of ±0.1. Some sampling sites were
obviously less stable (e.g., site 99) with a relative error of
Temporal stability analyses of crop ET 100 %.
Temporal stability of ET measurements varied with
Temporal stability is used to determine representative sites spring wheat growing stages. ET measurements were more
of ET measurements over time. The higher the temporal temporally stable after the seeding-tillering stage (after the
stability of ET measurements at a sampling site, the better first irrigation). The relative error ranges of 7 locations
the site is for long-term ET observations (Sokal and Rohlf were \0.1, including sampling sites 64, 98, 47, 55, 94, 30
1995; Jara et al. 1998; Hupet and Vanclooster 2004). In this and 111. Interestingly, sample site 94 was temporally
study, Spearman’s rank correlation coefficient cs was cal- stable for the entire growing seasons (Fig. 4b) and could be
culated to determine temporal variations of crop ET mea- used for long-term ET measurements.
surements at the sample sites. Seen from Table 4, all cs
were significant at the 99 % confidence level, which sug- Impact factor analysis of crop ET
gest that there existed temporal stability of ET measure-
ments, although cs between the seeding-tillering and other Crop ET increased with an increase of atmospheric evap-
growth stages were low (Table 4; Fig. 4a). oration (ET0) and LAI. The correlation coefficient (R) be-
Temporal stability of ET measurements varied with tween ET10 and ET0 was 0.82, whereas R was 0.88 between
sampling sites (Fig. 4a). Among all of the observations at ET10 and mean LAI. Spring wheat ET10 was linearly cor-
various growth stages, the average relative errors obtained related with mean LAI at a confidence interval of 95 %.
from 67 sampling plots were less than 30 % (Fig. 4a). The These results conform to previous literature that found crop
range of relative error for 80 sampling plots was [0.3. Of ET increased with the increasing ET0 and LAI
them, sites 94 and 14 represented well the mean ET \3.0 m2 m-2 (Al-Kaisi et al. 1989; Ding et al. 2013b).

123
Environ Earth Sci (2015) 74:5443–5452 5449

Fig. 5 Relationship between


crop ET10 and surface soil
moisture change: a seeding-
tillering stage; b tillering-
shooting stage; c shooting-
heading stage; and d heading-
milk stage

A linear relationship was found between crop ET10 and


changes in surface soil moisture Dh0–20 cm at different
growing stages (Fig. 5). The linear regressions were sig-
nificant at the 95 % confidence level. Linear proportional
coefficients between ET10 and Dh0–20 cm varied from 1.18
to 2.17. Previous literature had also reported that a strong
reciprocal interaction exists between ET and Dh0–20 cm
(Loague 1992; Hupet and Vanclooster 2004; McCabe and
Wood 2006; Gokmen et al. 2012). With the crop growth,
the correlations between Dh0–20 cm and ET10 decreased,
which may have been due to the aging of wheat leaves and
decreases in LAI (Hupet and Vanclooster 2004).
Also, measurements of ET10 in 2011 were used to verify
the relationship between Dh0–20 cm and ET10 in 2010.
Evapotranspiration estimates from the water budget
Fig. 6 Average daily sensitivity coefficients for Dh0–20 cm, LAI and
approach (ETw) match well with the ET estimated by the ET0
changes in soil moisture (ETr) with the measurements data.
The mean relative errors (RE) between ETw and ETr were
12 and 13 %, respectively. Therefore, the coordination from 0.05 to 0.23, with an average of 0.16. The average
method for estimating ET with changes in soil moisture sensitivity coefficient of ET10 to LAI ranged from 0.48 to
was reliable. 0.86, with an average of 0.57, and the average sensitivity
coefficient of ET10 to Dh0–20 cm ranged from 1.09 to 1.73,
ET improvement with the auxiliary variable with an average of 1.28 (Fig. 6). Thus, the variable
of Dh0–20 cm Dh0–20 cm was the most sensitive factor affecting ET10. The
high sensitivity coefficients of ET10 to Dh0–20 cm indicated
The variables Dh0–20 cm, LAI and ET0 all affected ET10. that the auxiliary variable of soil moisture measurement
The average sensitivity coefficient of ET10 to ET0 ranged would improve accuracy of ET10 estimates.

123
5450 Environ Earth Sci (2015) 74:5443–5452

Fig. 7 Effects of sample sizes


of spring wheat ET observations
on spatial distributions of the
relative error (RE) of ET
estimates with the coordination
method: a n = 24, b n = 48,
c n = 60 and d n = 80

Compared with in-field ET observations, surface soil especially after the tillering-shooting stage (after the first
moisture measurements are easier and more efficiently, so irrigation). In comparison with LAI and ET0, surface soil
Dh0–20 cm data could be used as an auxiliary variable to moisture change in the top 0–20 cm was the most sensitive
improve the accuracy of crop ET estimates. With surface variable of ET measurements during the spring wheat
soil moisture measurements, the relative error of ET esti- growing seasons. Compared with in-field ET observations,
mation was 9.4 % with only half the number of sampling surface soil moisture is easier and more efficiently mea-
data (n = 48) (Fig. 7). It is evident that the larger the sured, so the changes in soil moisture data could be an
sample size is, the lower the relative errors and variability auxiliary variable to improve the ET accuracy. This
of ET estimates will be. research provides an insight into crop ET dynamic mea-
surements and accurate quantification of ET variability and
sensitivity, which are essential to improve efficiency of
Conclusions water use by crops, to develop optimum irrigation regimes,
and to facilitate advanced variable-control irrigation tech-
The analyses of semivariance, Spearman’s rank coefficient,
nology and sustainable water management.
impact factors and sensitivity coefficient of crop ET were
conducted to investigate spatio-temporal variability attri- Acknowledgments The authors are grateful to the Experimental
butes and driving variables of ET measurements. We found Station of Water-Saving in Agriculture and Ecology in Gansu Pro-
that there was a large effect of the first irrigation event on vince for providing the data. This study was supported partly by the
National Natural Science Foundation (41471310&41371416), China;
the spatial and temporary variability of in-field ET mea-
the Youth Fund Project of Chinese Academy of Sciences
surements at different growing stages. During the spring (Y5SJ0600CX), China; China National Key project of High resolu-
wheat growing seasons, spatial variability of cumulative tion earth observation system ‘‘High resolution remote sensing
ET was the lowest at the tillering-shooting stage (before the monitoring and evaluation of water resources demonstration system’’
(08-Y30BO7-9001-13/15); China National Key project of High res-
first irrigation) and then gradually increased with crop
olution earth observation system ‘‘High resolution remote sensing
growth. In some experimental plots, spring wheat ET had a monitoring and evaluation of agricultural demonstration system’’ (09-
significantly higher temporal stability than other plots, Y30B03-9001-13/15).

123
Environ Earth Sci (2015) 74:5443–5452 5451

Appendix De Lannoy GJM, Verhoest NEC, Houser PR, Gish TJ, Van
Meirvenne M (2006) Spatial and temporal characteristics of soil
moisture in an intensively monitored agricultural field (OPE3).
Please see below the Appendix Table 5. J Hydrol 331:719–730
Ding R, Kang S, Li F, Tong L (2013a) Evapotranspiration measure-
Table 5 Definition of variables used in the study ment and estimation using modified Priestley–Taylor model in
an irrigated maize field with mulching. Agr For Meteorol
Variable Description
168:140–148
Ding R, Kang S, Vargas R, Zhang Y, Hao X (2013b) Multiscale
ET0 Atmospheric evaporation (mm)
spectral analysis of temporal variability in evapotranspiration
ET Crop evapotranspiration (mm) over irrigated cropland in an arid region. Agr Water Manag
ET4t Crop ET during Dt from time t1 to t2 (mm) 130:79–89
ET10 10-day cumulative crop evapotranspiration (mm) French AN, Hunsaker DJ, Clarke TR (2012) Forecasting spatially
distributed cotton evapotranspiration by assimilating remotely
ETi ET values of the ith location (mm)
sensed and ground-based observations. J Irrig Drain Eng
ETi,t ET values of the ith location at time t (mm) 138:984–992
ETw Observed crop evapotranspiration from water budget (mm) Gokmen M, Vekerdy Z, Verhoef A, Verhoef W, Batelaan O, van der
ETr Estimated crop evapotranspiration by the soil moisture Tol C (2012) Integration of soil moisture in SEBS for improving
changes (mm) evapotranspiration estimation under water stress conditions.
Remote Sens Environ 121:261–274
PDt Precipitation during Dt (mm) Goyal RK (2004) Sensitivity of evapotranspiration to global warming:
IDt Irrigation during Dt (mm) a case study of arid zone of Rajasthan India. Agr Water Manag
GDt Contribution from water table during Dt (mm) 69:1–11
Hupet F, Vanclooster M (2001) Effect of the sampling frequency of
RDt Water runoff during Dt (mm)
meteorological variables on the estimation of the reference
DDt Drainage during Dt (mm) evapotranspiration. J Hydrol 243:192–204
WDt Change of soil moisture content during Dt (mm) Hupet F, Vanclooster M (2004) Sampling strategies to estimate field
CV The coefficient of variation real evapotranspiration fluxes with a soil water balance
approach. J Hydrol 292:262–280
c(h) Semivariance at lag distance h Jacobs JM, Mohanty BP, Hsu E, Miller D (2004) SMEX02: field scale
rs Spearman’s Rank coefficient variability, time stability and similarity of soil moisture. Remote
di Relative mean error Sens Environ 92:436–446
Jara J, Stockle C, Kjelgaard J (1998) Measurement of evapotranspi-
r(d)2i Variance error ration and its components in a corn Zea Mays L field. Agr For
RMSEi Relative mean error di and variance error r(d)2i Meteorol 92:131–145
SVi Sensitivity coefficient of Vi Kang S, Su X, Tong L, Shi P, Yang X, Yukuo A, Du T, Shen Q,
Zhang J (2004) The impacts of human activities on the water–
land environment of Shiyang River Basin, an arid region in
Northwest China. Hydrol Sci J 49(3):413–427
References Lei H, Yang D (2010) Interannual and seasonal variability in
evapotranspiration and energy partitioning over an irrigated
Al-Kaisi M, Brun LJ, Enz JW (1989) Transpiration and evapotran- cropland in the North China Plain. Agr For Meteorol
spiration from maize as related to leaf area index. Agr For 150:581–589
Meteorol 48(1–2):111–116 Loague K (1992) Soil water content at R-5,1, spatial and temporal
Anderson RG, Jin Y, Goulden ML (2012) Assessing regional variability. J Hydrol 139:233–251
evapotranspiration and water balance across a Mediterranean Mauser W, Schädlich S (1998) Modelling the spatial distribution of
montane climate gradient. Agr For Meteorol 166–167:10–22 evapotranspiration on different scales using remote sensing data.
Anderton S, Latron J, Gallart F (2002) Sensitivity analysis and multi- J Hydrol 212–213:250–267
response, multi-criteria evaluation of a physically based dis- McCabe MF, Wood EF (2006) Scale influence on the remote
tributed model. Hydrol Process 16:333–353 estimation of evapotranspiration using multiple satellite sensors.
Beres DL, Hawkins DM (2001) Plackett-Burman technique for Remote Sens Environ 105(4):271–285
sensitivity analysis of many-parametered models. Ecol Model McCuen RH (1974) A sensitivity and error analysis of procedures
141:171–183 used for estimating evaporation. Water Resour Bull
Bertuzzi P, Bruckler L, Chanzy A (1994) Sampling strategies for soil 10(3):486–498
water content to estimate evapotranspiration. Irrig Sci 14:105–115 Miao Z, Xu M, Lathrop RG, Wang Y (2009) Comparison of the A-Cc
Beven K (1979) A sensitivity analysis of the Penman–Monteith actual curve fitting methods in determining maximum ribulose 1
evapotranspiration estimates. J Hydrol 44:169–190 5-bisphosphate carboxylase/oxygenase carboxylation rate,
Brocca L, Morbidelli R, Melone F, Moramarco T (2006) Soil potential light saturated electron transport rate and leaf dark
moisture spatial variability in experimental areas of central Italy. respiration. Plant Cell Environ 32:109–122
J Hydrol 333(2–4):356–373 Miao Z, Lathrop RG, Xu M, La Puma IP, Clark KL, Hom J,
Burgess TM, Webster R (1980) Optimal interpolation and isarithmic Skowronski N, Van Tuyl S (2011) Simulation and sensitivity
mapping of soil propertiesI the semi-variogram and punctual analysis of carbon storage and fluxes in the New Jersey
kriging. J Soil Sci 31:315–331 Pinelands. Environ Model Softw 26:1112–1122
Cosh MH, Stedinger JR, Brutsaert W (2004) Variability of surface Pietroniro A, Prowse TD (2002) Applications of remote sensing in
soil moisture at the watershed scale. Water Resour Res 40:12513 hydrology. Hydrol Process 16:1537–1541

123
5452 Environ Earth Sci (2015) 74:5443–5452

Pietroniro A, Soulis ED, Kowen N (1994) Deriving antecedent Tong L, Kang S, Zhang L (2007) Temporal and spatial variations of
moisture conditions from airborne SAR for input into a flood evapotranspiration for spring wheat in the Shiyang river basin in
forecasting model. International geoscience and remote sensing northwest China. Agr Water Manag 87:241–250
symposium. IEEE, pp 1435–1438 Vachaud G, Passerat de Silans A, Balabanis P, Vauclin M (1985)
Qiu G, Yanob T, Momiic K (1998) An improved methodology to Temporal stability of spatial measured soil water probability
measure evaporation from bare soil based on comparison of density function. Soil Sci Soc Am J 49:822–828
surface temperature with a dry soil surface. J Hydrol Van Pelt RS, Wierenga PJ (2001) Temporal stability of spatially
210:93–105 measured soil matric potential probability density function. Soil
Raghuwanshi NS, Wallender WW (1997) Field-measured evapotran- Sci Soc Am J 65:668–677
spiration as a stochastic process. Agr Water Manag 32:111–129 Vinnikov KY, Robock A, Speranskaya NA, Schlosser CA (1996)
Raziei T, Pereira L (2013) Spatial variability analysis of reference Scales of temporal and spatial variability of midlatitude soil
evapotranspration in Iran utilizing fine resolution gridded moisture. J Geophys Res 101:7163–7174
datasets. Agr Water Manag 126:104–118 Warrick AW (2003) Soil water dynamics. Oxford University Press
Saxton KE (1975) Sensitivity analysis of the combination evapotran- Inc, New York, pp 33–35
spiration equation. Agr Meteorol 15:343–353 Yang Y, Shang S, Jiang L (2012) Remote sensing temporal and
Singh VP, Xu C-Y (1997) Sensitivity of mass transfer-based spatial patterns of evapotranspiration and the responses to water
evaporation equations to errors in daily and monthly input data. management in a large irrigation district of North China. Agr For
Hydrol Process 11:1465–1473 Meteorol 164:112–122
Singh RK, Ayse I, Irmak S, Martin DL (2008) Application of SEBAL Yoo C, Valdés JB, North GR (1998) Evaluation of the impact of
model for mapping evapotranspiration and estimating surface rainfall on soil moisture variability. Adv Water Resour
energy fluxes in south-central Nebraska. J Irrig Drain Eng 21:375–384
134:273–285 Zhang R (2005) Applied geostatistics in environmental science.
Sokal RP, Rohlf FJ (1995) Biometry, 3rd edn. Freeman, New York, Science Press and Science Press USA Inc, Monmouth Junction,
pp 1–887 pp 91–92

123
Copyright of Environmental Earth Sciences is the property of Springer Science & Business
Media B.V. and its content may not be copied or emailed to multiple sites or posted to a
listserv without the copyright holder's express written permission. However, users may print,
download, or email articles for individual use.

Vous aimerez peut-être aussi