Environmental Earth Sciences (2018) 77:358

https://doi.org/10.1007/s12665-018-7518-z

ORIGINAL ARTICLE

Characteristics of ecosystem water use efficiency in a desert riparian

forest
Xiaohong Ma1,2,3   · Qi Feng1,2 · Yonghong Su1,2 · Tengfei Yu1,2 · Ravinesh C. Deo1,2,4

Received: 13 September 2017 / Accepted: 26 April 2018

© Springer-Verlag GmbH Germany, part of Springer Nature 2018

Abstract
The Ejin oasis, located on the river sides of a hyper-arid region, represents some of the richest desert ecosystems, populated
with riparian forest. Water availability is the main limited factors for the distribution and growth of the forest. Currently, the
water–carbon process and mechanisms in the desert riparian forest remain uncertain. The study investigated the seasonal and
inter-annual variations in the gross primary productivity (GPP), evapotranspiration (ET) and ecosystem water use efficiency
(WUE), and their relationships with the event (e.g., irrigation) and climatic factors. The results showed that the desert riparian
forest produced new plant biomass with 0.79 g C kg−1 of water loss during the growing season. In the study period, monthly
GPP and ET was tightly coupled, and the mean cumulative GPP and ET from May–October were 676 g C m−2 and 864 mm,
respectively. Daily GPP and ET had similar relationships with global radiation (Rg), air temperature (Ta) and vapor pressure
deficit at monthly scale. Consequently, the seasonal variations in WUE were not obvious in the growing season, except in
September. In September, the value of WUE decreased significantly with an increase in Ta (for Ta > 15 °C) after irrigation,
which was attributable to the increased ET and a relatively steady GPP. It is possible that the time of irrigation is able to
affect the seasonal course of WUE. The study has implications that enable more rational water management to maintain the
the health and activities of the riparian forest in such a hyper-arid region.

Keywords  Ecosystem water use efficiency · The desert riparian forest · Irrigation effects

Introduction climate change (Feng and Fu 2013; Reed et al. 2012). Ripar-
ian corridors are the most diverse, productive and dynam-
Drylands encompass hyper-arid, arid, semiarid, and dry sub- ics system in hyper-arid regions, such as the lower reaches
humid areas and cover about 41% of the Earth’s land surface of the Heihe River Basin (HRB) in Northwestern China.
(Feng and Fu 2013). The ecosystems over drylands consti- Although riparian zones occupy a small proportion of the
tute a significant component of the global biogeochemical land area, there are lot of biological importance in the face
cycles (Reed et al. 2012), and are fragile and sensitive to of climate change and population growth. At the same time,
human activities often impact the vitality of the riparian
area by implementing water management strategies (Jackson
* Qi Feng et al. 2001). Understanding how the water–carbon process
qifeng@lzb.ac.cn and the mechanism factors that interact with the chang-
1
Northwest Institute of Eco-Environment and Resources, ing social demands and climate variability is important for
CAS, Lanzhou 730000, China improving the management of such vital ecosystems (Scott
2
Key Laboratory of Eco-hydrology of Inland River Basin, et al. 2004).
CAS, Lanzhou 730000, China As the second largest inland river basin in northwestern
3
University of Chinese Academy of Sciences, Beijing 100049, China, the Heihe River Basin (HRB) faced serious ecologi-
China cal and environment problem with increasing water con-
4
School of Agricultural, Computational and Environmental sumption in the 1990s (Cheng et al. 2014). The Ejin oasis,
Sciences, International Centre for Applied Climate Sciences located in the lower HRB, which is surrounded by a wide
(ICACS), Institute of Agriculture and Environment (IAg&E), desert, has experienced more serious ecological disasters.
University of Southern Queensland, Springfield, QLD 4300, Therefore, an ecological water diversion project (EWDP)
Australia

13
Vol.:(0123456789)
 358   Page 2 of 12 Environmental Earth Sciences (2018) 77:358

has been implemented by the Chinese government since the desert riparian forest and the response to environmen-
2002. Since then, the water has been imported into the lower tal factors has lot of scientific significance, especially for
reaches of HRB under strict artificial controls. In fact, the understanding the hyper-arid ecosystem’s productivity and
amount and the time of ecological water delivery can been water consumption. However, we known little about the
reflected by the variation of the groundwater depth (GWD). characteristics of seasonal and inter-annual variations of
Since then, the GWD has been shallower and the vegetation WUE and their response to changing human activities and
has experienced a favorable change in the lower reaches of environmental factors in the desert riparian forest. Fur-
HRB (Wang et al. 2013). The Ejin oasis located on the river thermore, how the climate change will affect the carbon
sides represents some of the richest desert ecosystems, popu- and water coupling process in the desert riparian forest
lated with riparian forest, which is internally controlled and remains unknown. Flanagan et al. (2017) found that the
interacts with surrounding deserts (Feng et al. 2012). After water-use rates of riparian cottonwood forests were high,
about 10 years later, we are still contemplating on the ques- even for the broad-leaf deciduous forest functional type.
tion: what is the state of water use efficiency (WUE) of the The high evapotranspiration (ET) was possible because of
vegetation? Since the implementation of the EWDP, part of the tree’s access to alluvial groundwater. So far, studies
the water imported into the lower reaches of the HRB was of water–carbon coupling process in riparian forests at
used to support the irrigation system for the riparian forest. the lower reaches of HRB have been mainly focused on
Desert riparian forest, as the primary body of the Ejin oases, the following aspects: First, the characterization of pho-
represents the majority of carbon–water coupling in hyper- tosynthesis and water use on leaf scale using traditional
arid spots. Although the area of the riparian forest in the Ejin site-experimentations of leaf gas exchange (Chang et al.
oasis is not very large, it does have a lot of biological impor- 2016) and second, the water sources and water use effi-
tance, since their growth and succession reflects the overall ciency of riparian forest using stable isotopes technique on
health and stability of the Ejin Oasis (Xi et al. 2010). It is leaf or plant scale (Cao et al. 2009; Si et al. 2014). These
thus essential to investigate the WUE of the desert riparian investigations have been characterized by small scale and
forest and its response to the environmental factors prevail- short time series. In addition, the characteristics of evapo-
ing in the Ejin oasis. transpiration and their responses to climatic factors in the
Recently, the knowledge of water and carbon cycling desert riparian forest at this region were analyzed by Yu
in dryland ecosystems has improved generally (Flanagan et al. (2017b). However, little is known about the charac-
et al. 2017; GrÜNzweig et al. 2003; Hinojo–Hinojo et al. teristics of the ecosystem-level WUE and response to bio-
2016; Liu et al. 2012a; Rodrigues et al. 2011; Scott et al. meteorological conditions in the riparian forests. Water
2006). Ecosystem water and carbon exchange are con- use efficiency at the ecosystem level is an important physi-
siderably affected by climate variability through a set of ological index to reflect the coupling relationship between
coupled physical and physiological processes (Mo et al. water and carbon cycle (Tong et al. 2014). When scaling
2017). The study of Rodrigues et al. (2011) reported that from the leaf to canopy or the ecosystem, the eddy covari-
the higher WUE and the steadier monthly pattern of WUE ance (EC) method was useful. The continuous and long-
were determined by the lesser monthly averaged vapor term observation of water and carbon exchange between
pressure deficit (VPD) and the reduction in water stress atmosphere and riparian forests is useful for studying the
in a eucalypt plantation. Liu et al. (2012b) showed that forest growth variations, and the process of carbon–water
the ecosystem’s WUE has increased in the wet year at a coupling including their response to climate changes at an
desert halophyte community, due to the adjustment of the ecosystem scale (Xie et al. 2016b).
community structure to the precipitation with more annual Studies focusing on the ecosystem’s WUE are crucial
or ephemeral plants in the wet year. Jia et al. (2016) found for assessing the appropriate amounts of ecological water
that the annual carbon and water fluxes appeared to be delivery and the adaptability of riparian forests to the cli-
suppressed in the years with low spring soil moisture, thus mate change accompanied by hydrological changes. In this
leading to a reduction in carbon sequestration capacity and study, we have measured and also computed the gross pri-
water use efficiency in a shrubland of Northern China. mary productivity (GPP), the evapotranspiration (ET) and
Quantitative information on the forest activities under dry the ecosystem water use efficiency (WUE) at an ecosystem
conditions may become relevant to the regions predicted scale in the desert riparian forest. Hence, the objectives of
to undergo increasing aridity (GrÜNzweig et al. 2003). In this research paper are as follows. (1) To determine the sea-
hyper-arid regions, water is the decisive controlling func- sonal and inter-annual variability of GPP, ET and WUE. (2)
tion of the biological system, which cause the vegetation To analyze the response of WUE to the events (i.e., irriga-
to reside on the water, and to form the desert riparian for- tion) and environmental factors (i.e., air temperature, global
est with unique regional characteristics (Huang 2016). radiation, vapor pressure deficit and leaf area index) to gain
The study of carbon and water coupling relationships in a better understanding and characteristics of the ecosystem

13
Environmental Earth Sciences (2018) 77:358 Page 3 of 12  358

water use efficiency in a desert riparian forest (Ejin oasis, a mean of 1.48 m. In addition, the salinity of the ground-
located on the river sides of a hyper-arid region in Northern water in Ejin oasis was about 2.9 g L−1, varying from 0.9 to
China). 10.4 g L−1. And the mean pH of the groundwater was 7.7.
The dominant anion species were S ­ O42−, ­Cl− and H
­ CO3−,
and the dominant dissolved ions within the groundwater sys-
Materials and methods tem were ­Mg2+ and ­Na+. The 60-year annual mean tempera-
ture is approximately 8.9 °C, with a mean monthly tempera-
Study site ture ranging from − 11.5 °C in January to 27.0 °C in July.
The study site is located in one of the best-grown riparian
The study has been conducted in a desert riparian forest at forests in the lower reaches of the HRB region. The domi-
the Qidaoqiao P. euphratica Forest National Natural Ref- nant plant species is Populus euphratica Oliv. (P. euphra-
uge (Fig. 1, 41°59′N, 101°10′E, 920.46 m elevation) (Yu tica) with a high density (320 stems ­ha−1): contributing to
et al. 2017b), located at the lower reaches of Heihe River approximately 75% of the total basal area in the study region
Basin (HRB). The terrain at this site is relatively regular (Yu et al. 2017b). The soil in the P. euphratica stand consists
with highly homogenous land surface cover, and the values of a sandy loam, of approximately 2 m deep and it has a
of the surface roughness length, zero-displacement height saturated volumetric water content of 0.35 m3 m−3 (Si et al.
and site’s slope are 1.38, 7.9 m and 0.51°, respectively. 2007).
The climate of the study site is hyper-arid (Si et al. 2007),
with a mean annual precipitation (P) of 37.8 mm and pan Flux and other measurements
evaporation (EP) of 2240.5 mm, respectively. The aridity
index is less than 0.05 in this study region. And most of the To investigate the characteristics of the ecosystem’s water
rainfall is depleted by evaporation before it actually infil- use efficiency in a desert riparian forest, water and carbon
trates into the deep soil layers (Yu et al. 2017a), the ground- fluxes were measured between the forest canopy and the
water as the only available water source, affect the growth atmosphere using the EC instruments placed at 20 m above
and the distribution of vegetation. At the study site, the the ground. The EC instruments included a three-dimen-
groundwater depth (GWD) fluctuated from 0.28 to 2.45 m sional sonic anemometer (CSAT3, Campbell Scientific,
during the growing season for the period 2013–2016, with Inc, Logan, UT, USA) used to measure the wind speed and

Fig. 1  The schematic diagram of the Heihe River Basin (HRB), selected Populus euphratica stand and Ejin meteorological station

13
 358   Page 4 of 12 Environmental Earth Sciences (2018) 77:358

virtual temperature, and an open-path ­CO2/H2O infrared gas ecosystem’s WUE has been defined as the ratio of GPP to
analyzer (LI-7500, LI-COR Inc. Lincoln, NE, USA) used ET, since the coupling of ET and GPP was generally stronger
to measure the concentrations of water vapor and carbon than those of ET and NEP (Kuglitsch et al. 2008; Xie et al.
dioxide fluctuations. Parallel to the fluxes measurements, 2016a). WUE is given by:
meteorological measurements made at this site included
GPP
the air temperature (Ta, °C) and the relative humidity (RH, WUE = , (1)
ET
%) measured by a relative humidity and temperature sensor
(HMP45C, Campbell, USA), whereas the solar radiation where GPP was obtained from the partitioned NEE (“Fluxes
components were measured by a 4-Component Net Radia- data processing and gap-filling ”), and the ET was calculated
tion Sensor (CNR4, Kipp & Zonen, Delft, NL). In addition, as:
one well was installed at about 30 m away from the EC
LE
tower; and the groundwater depth (GWD, m) was measured ET = , (2)
L𝜌w
automatically by the pressure transducers (HOBO- U20,
Onset Computer Corporation, Bourne, MA, USA) at 0.5 h where LE is the latent heat flux (MJ·m−2 h−1); L is the latent
intervals(Yu et al. 2017b). heat of vaporization of water (2.45 kJ g−1); and ρw is water
The leaf area index (LAI) data were obtained from the density (1 g cm−3).
Terra-MODIS products (MOD15A2H), which was an 8-day Next, the WUE at the ecosystem level was analyzed on
500-m resolution product (i.e., used at Validated Stage 2). daily to annual timescales. For a specific period, the WUE
can be calculated using the accumulations of GPP and ET,
Fluxes data processing and gap‑filling either on a daily scale using the daily sums of GPP and ET,
or on the growing season by integrating GPP and ET.
Raw ­CO2/H2O fluxes data were processed by the Eddypro
software (LI-COR Inc.) following standard quality con-
trol procedures, including the de-spiking, 2D coordinate
rotation, time lag removal, frequency response correction
Results
using model spectra and transfer functions (Moore 1986)
and air density corrections (Webb et al. 1980). After these
Microclimate
procedures, the half-hour net ecosystem exchange of C ­ O2
Annual mean Ta over the 4 years were all higher than the
(NEE, mg ­m−2 s−1), latent heat (LE, W m ­ −2) and sensible
mean from 1957 to 2016. There was no significant dif-
heat (H, W m−2) fluxes could be obtained. Next, the NEE
ference in temperature for the period 2013–2016, and the
was partitioned to acquire gross primary productivity (GPP)
annual mean Ta varied from 10.46 to 10.62 °C. During the
and ecosystem respiration (Re) (Reichstein et al. 2005). In
growing season (i.e., May–October), the lowest monthly
addition, a friction velocity (u*) threshold was used to filter
mean Ta was 9.07 °C in October 2016 and the highest was
the fluxes at night when the atmospheric turbulence was
28.87 °C in July 2016. The mean Ta were found to be 21.46,
not well developed. The threshold for u* was estimated by
21.29, 20.96 and 21.46 °C during the growing season of
the Moving Point Test as described in literature (Papale
the period 2013–2016, respectively. In addition, the annual
et al. 2006).
P were 34.2, 17.2, 70.1 mm and 52.1 mm for 2013–2016,
The EC and meteorological data gaps were filled with
respectively. And the mean of annual P for 1957–2016 was
methods considering the co-variation of fluxes with mete-
37.8 mm. The highest monthly mean of RH was 37.1% in
orological variables and the temporal auto-correlation of the
August of 2016 and Ep was 303.9 mm in August of 2015,
fluxes (Reichstein et al. 2005) according to the online proce-
while the mean Ep during the growing season in 2013–2016
dure (https​://www.bgcje​na.mpg.de/bgi/index​.php/Servi​ces/
was higher than the mean of 1957–2016 (Table  1). The
REddy​ProcW​ebMet​hod).
meteorological data were recorded by the Ejin weather sta-
tion and provided by China Meteorological Data Sharing
Definition and calculation of ecosystem water use
Services.
efficiency (WUE)

Water use efficiency (WUE) is an important indicator of Characteristics of GPP, ET and WUE

the characteristics in the water–carbon coupling process.
Although WUE can be calculated in very different ways During the growing season (i.e., May to October), the
depending on the temporal and spatial scales of interest, mean total ET (i.e., May to October) was 864 ± 137 mm
WUE is commonly considered as the ratio of carbon gains and GPP was 676 ± 87 g C m−2 for the period 2013–2016.
to water losses (Kuglitsch et al. 2008). In this paper, the The cumulative ET from May to October for 2013–2016

13
Environmental Earth Sciences (2018) 77:358 Page 5 of 12  358

Table 1  Monthly mean air May Jun Jul Aug Sep Oct May–Oct
temperature (Ta, oC), relatively
humidity (RH, %), and pan 2013 Ta 20.70 24.54 27.33 26.54 19.28 10.38 21.46
evaporation (Ep, mm)
2014 Ta 19.87 24.25 28.07 25.36 18.88 11.32 21.29
2015 Ta 20.18 24.40 27.17 25.85 18.01 10.17 20.96
2016 Ta 18.38 24.18 28.87 27.20 21.04 9.07 21.46
1957–2016 Ta 19.18 24.88 27.03 24.85 17.78 8.42 20.36
2013 RH 13.4 26.9 28.4 30.1 26.9 34.8 26.8
2014 RH 18.0 30.6 29.7 28.3 30.6 29.4 27.8
2015 RH 18.6 27.6 30.1 25.2 36.1 26.5 27.4
2016 RH 22.3 32.4 32.3 37.1 30.4 32.9 31.2
1957–2016 RH 21.4 25.7 31.8 33.5 32.6 34.3 29.9
2013 Ep 270.9 246.1 289.4 237.9 176.2 115.2 222.6
2014 Ep 277.3 245.4 293.2 266.1 207.1 151.6 240.1
2015 Ep 268.3 280.4 298.8 303.9 189.9 132.4 245.6
2016 Ep 243.5 272.3 285.4 285.3 217.4 129.5 238.9
1957–2016 Ep 301.5 333.6 338.8 299.1 211.5 130.0 269.1

250
Variations in daily GPP, ET and WUE

y = 0.68x + 15.43 The seasonal and inter-annual variations in GPP, ET

200 R² = 0.69 and WUE during the growing season from 2013 to
GPP (g.C.m-2.mon-1)

2016 were shown in Fig. 3. The maximum of daily GPP

150 and ET occurred in June and July, were averages of
7.52 g C m−2 day−1 and 10.04 mm day−1 for the period
100 2013–2016, respectively. The daily WUE did not reveal
obvious seasonal trends. The mean WUE was 0.79 g C kg−1
­H2O during the growing season for the period 2013–2016.
50
The groundwater, as the only available water source
affects the growth and distribution of riparian forest. The
0 amount and the time of irrigation can been reflected by the
0 50 100 150 200 250
variations in the groundwater depth (GWD). Chen et al.
ET (mm mon-1) (2003) showed that the GWD of 3.5 to 4.5 m is rational
for the growth of P. euphratica. In addition, other studies
Fig. 2  The couple relationship between monthly gross ecosystem pro- also showed that the apparent response of vegetation to the
duction (GPP) and evapotranspiration (ET) ground water fluctuations occured when the GWD had fluc-
tuated between 1.8 and 3.5 m in the lower reaches of HRB
(Wang et al. 2011). For this study, the GWD fluctuated from
were 660 ± 45, 914 ± 59, 939 ± 74 and 942 ± 44  mm, 0.28 to 2.45 m during the growing season for the period
respectively. And the cumulative GPP from May to Octo- 2013–2016. Figure 4 showed that there was no relationship
ber for 2013–2016 were 606 ± 46  g  C  m −2 , 636 ± 46, between GWD and WUE. Otherwise, the seasonal course of
662 ± 49, 802 ± 49  g  C  m −2 , respectively. The lowest WUE on a daily scale was affected by the time of irrigation.
values of GPP and ET occurred in 2013, otherwise the In fact, irrigation led to an increase in the available water
highest values of GPP and ET occurred in 2016. Monthly content of the desert riparian forest, so in this case, the level
GPP and ET were coupled linearly over the 4-year study of temperature and LAI acted to influence the variations
period (Fig. 2), and the slope of the relationship was 0.68, of GPP and ET, causing WUE to be changed. Firstly, it is
a value close to mean annual WUE. During the grow- observable that if the irrigation occurred on the later of the
ing season for the period 2013–2016, the mean values vibrant growing season (i.e., around August 27, 2016), the
of WUE ranged between 0.69  g  C  kg −1 ­H 2 O in 2014 WUE reduced very little or it was nearly a constant. This is
and 0.90 g C kg−1 ­H2O in 2013, with an overall mean of because increased available water and higher temperature
0.79 g C kg−1 ­H2O. led to a simultaneously increase in ET and GPP. Secondly, if
the irrigation had been implemented on the leaf senescence

13
 358   Page 6 of 12 Environmental Earth Sciences (2018) 77:358

2013 2014 2015 2016

40 Ta GWD
4

GWD (m)
Ta (°C)

20
2

0 0

400 Rg LAI 0.9

LAI
300
0.6
R g (W m )
-2

200
0.3
100

0 0.0
12

9
ET (mm d )
-1

8
GPP (g C m d )
-1
-2

0
3
WUE (g C kg H2O)

2
-1

0
121 166 212 258 304 166 212 258 304 166 212 258 304 166 212 258 304
DOY

Fig. 3  The seasonal and inter-annual variations of daily water use efficiency (WUE), gross primary productivity (GPP), evapotranspiration (ET),
global radiation (Rg), air temperature (Ta) and ground water depth (GWD)

period (i.e., September 16, 2014), the WUE was likely to the year 2016, with these values varying form 152.1 to
decrease with a decrease in LAI. Because the GPP decreased 196.3 g C m−2 month−1. In fact, the ET peaked in June for
with decreasing LAI, but the ET increased and retained a the year 2013 and 2016, in May for 2014, and in July for
higher value due to rise in the available water content. 2015. And the maximum of ET ranged between 153.8 and
225.8 mm month−1. In contrast to the GPP and ET, the WUE
Variation in monthly GPP, ET and WUE exhibited a peak in different months, and the values actually
fluctuated between 0.75 in July of 2015 and 1.19 g C kg−1
The variations in monthly GPP, ET and WUE over the ­H2O in September 2013. The variations in monthly WUE
study period were shown in Fig. 5. It is noticeable that the values were irregular, and this result concurred with the
GPP peaked in July for the period 2013–2015, in June for study of Tong et al. (2014) for a warm-temperate plantation.

13
Environmental Earth Sciences (2018) 77:358 Page 7 of 12  358

In accordance with this, the daily GPP, ET had a strong

2.5 positive linear relationship with Ta, Rg and VPD on the
growing season basis (Fig. 6). During the growing season,
2.0 both the GPP and the ET increased, presumably due to
the causative influence of Rg and Ta. At the monthly scale,
daily GPP and ET showed similar relations with Rg, Ta,
GWD (m)

1.5
and VPD, so that the variability in WUE was not obvious.
1.0 This indicates that the photosynthesis and evapotranspira-
tion were affected by the climatic factors, with each factor
making an approximately equal contribution. Moreover,
0.5
the relationship between GPP and Ta was slightly different
from that of ET and Ta, which appeared to be influenced by
0.0
irrigation. At the early phase of the growing season (i.e.,
0 1 2 3 4 5
-1 May), the ET increased with an increase in Ta, but the GPP
WUE (g C kg H2O)
did not vary as much with an increase in Ta. Consequently,
the WUE decreased slightly with an increase in Ta. During
Fig. 4  The relationship between daily water use efficiency (WUE) the vibrant growing season (i.e., June to August), both the
and groundwater depth (GWD)
GPP and the ET exhibited similar relationships with the
changes in Ta, and therefore, the WUE showed a modest
Climatic and biotic controls on the coupling variation (with an average of 0.91 g C kg−1 ­H2O). During
between GPP and ET the late growing season (i.e., September), the ET increased
significantly with an increase in Ta (for all Ta > 15 °C) after
Solar radiation, temperature and vapor pressure defi- the irrigation, however, the GPP did not changed with Ta.
cit are considered to be the main climatic factors driv- As a consequence, the WUE decreased significantly with
ing the photosynthetic and evapotranspiration processes. an increase in Ta.

300
2013 2014 2015 2016 300
(a) GPP ET
240 240

ET (mm month )
-1
GPP (g C m )

180 180
-2

120 120

60 60

0 0

1.2 (b)
WUE (g C kg H2O)

1.0
-1

0.8

0.6

0.4

Jun. Aug. Oct. Jun. Aug. Oct. Jun. Aug. Oct. Jun. Aug. Oct.
Month

Fig. 5  The variations of monthly gross primary productivity (GPP), evapotranspiration (ET) and water use efficiency (WUE) in 2013–2016

13
 358   Page 8 of 12 Environmental Earth Sciences (2018) 77:358

6
(a) May.
12
(b) May.
12
(c) May.
Jun. Jun. y = 0.36x - 1.72 Jun.
y = -0.02x+1.04 y = 0.14x+0.77
Jul. Jul. 2 Jul.
2
R = 0.07 Aug.
2
R = 0.17 Aug. R = 0.39 Aug.
5 10 10 y = 0.44x - 4.74
Sep. y = 0.06x +2.39 Sep. Sep.
y = -0.09x+2.78 Oct. Oct. 2 Oct.
WUE (g C kg H2O)

2 2
R = 0.01 R = 0.40
R = 0.58

GPP (g C m d )
4 8 8

-1

ET (mm d )
-1
-1

-2
3 6 6

2 4 4

1 2 2

0 0 0
0 5 10 15 20 25 30 35 0 5 10 15 20 25 30 35 40 0 5 10 15 20 25 30 35 40
Ta (°C ) Ta (°C) Ta (°C)
5
(d) May. (e) May. (f) May.
Jun. 12 Jun. 12 Jun.
Jul. Jul. Jul.
4 Aug. Aug. Aug.
Sep. Sep. Sep.
Oct. 9 9
WUE (g C kg H2O)

Oct. Oct.
GPP (g C m d )

3
-1

ET (mm d )
-1
-1

-2

6 6
2

1 3 3

0 0 0
0 100 200 300 400 0 100 200 300 400 0 100 200 300 400
-2
Rg (w m ) Rg (w m )
-2
Rg (w m )
-2

5
(g) May.
12 (h) May.
12 (i) May.
Jun. Jun. Jun.
Jul. Jul. Jul.
4 Aug. Aug.
10 10 Aug.
Sep. Sep. Sep.
Oct. Oct.
WUE (g C kg H2O)

Oct.
GPP (g C m d )

3 8 8
-1

ET (mm d )
-1
-1

-2

6 6
2
4 4
1
2 2

0 0 0
0 1 2 3 4 5 0 1 2 3 4 5 0 1 2 3 4 5
VPD (kPa) VPD (kPa) VPD (kPa)

Fig. 6  Relationships between daily gross primary productivity (GPP, radiation (Rg, W  m−2), and vapor pressure deficit (VPD, kPa) from
g  C  m−2  day−1), evapotranspiration (ET, mm  day−1), and water use May to October in 2013–2016
efficiency (WUE, g C k­ g−1 ­H2O) and air temperature (Ta, °C), global

Biotic controls on GPP, ET and WUE are shown in an increase in LAI for the months of May and September.
Fig. 7. Evidently, there was a strong relationship between By contrast, there were no obvious variation in GPP and
the GPP and LAI at the early and later phase of the grow- ET with an increase in LAI for the vibrant growing sea-
ing season (i.e., in May and September). The same effect son (i.e., the months of June–August). In this regard, the
was registered for the relationship between ET and LAI, values of WUE remained nearly constant from June to
but the relationship for this case was not as stronger com- August. For the desert riparian forest, the seasonal vari-
pared to that between the GPP and ET. In fact, the results ations in WUE were not so obvious, and these were not
showed that both GPP and ET increased significantly with controlled by LAI during the growing season.

13
Environmental Earth Sciences (2018) 77:358 Page 9 of 12  358

(a) GPPMay= 16.28x-1.13 this paper, we have performed extensive field experiments
9 May. 2
R = 0.47
to compute and compare the water use efficacy in the eco-
Jun.
Jul.
GPPJun= 14.62x-2.59 systems over drylands (Table 2).
2
During the growing season, the mean annual WUE was
GPP (g C m d )

Aug. R = 0.17
-1

6 Sep. GPPJul= 4.26x+2.18

0.79 ± 0.09 g C kg−1 ­H2O in the desert riparian forest over
-2

Oct. 2
R = 0.15 4 years. This value was lower than the values observed in an
GPPAug= 5.34x+0.67
3 2 Aleppo pine forest at the edge of Negev desert (3.3 g C kg−1
R = 0.15
GPPSep= 5.68x+0.45
­H2O, GrÜNzweig et al. 2003). Other regions showed rela-
2
R = 0.60
tively higher values as well, for example, in a natural ripar-
0 ian cottonwood forest (3.1 g C kg−1 ­H2O, Flanagan et al.
0.0 0.3 0.6 0.9 1.2 2017) and in a buffelgrass savanna at the edge of Sono-
(b) ETMay= 22.13x-1.14 ran Desert (1.7 g C kg−1 ­H2O, Hinojo-Hinojo et al. 2016)
12
May. 2
R = 0.21 (Table 2). However, the value of WUE recorded in this study
Jun.
Jul. ETJun= -6.5x-9.78 was close to the results reported for the other desert ecosys-
9
tems, i.e., 0.96 g C kg−1 ­H2O found by Scott et al. (2006) in
ET (mm d )

Aug. 2
-1

R = 0.03
Sep. ETJul= 6.03x+2.26 a Chihuahuan Desert shrubland; 0.76 g C kg−1 ­H2O found
6 Oct. 2
R = 0.01 by Liu et al. (2012a) for a saline desert dominated by the
ETAug= 9.83x-1.65
2
desert shrub Tamarix ramosissima Ledeb (Table 2). Lower
3 R = 0.13
ETSep= 8.57x-0.51 WUE in desert riparian forest (this study) was attributed to
2
R = 0.33
higher evapotranspiration and lower GPP in the hyper-arid
0 environment. The larger variability of WUE, as reported in
0.0 0.3 0.6 0.9 1.2 Table 2, is mainly attributable to the differences in climate
2.5
(c) conditions and vegetation types.
May.
2.0 Jun.
Jul. Climatic and biotic controls on the coupling
WUE (g C kg H2O)

Aug.
1.5 between GPP and ET
Sep.
Oct.
-1

1.0 The responses of GPP and ET to the climatic and biotic vari-
ables is known to determine the seasonal and inter-annual
0.5
variations in WUE. There was a strong linear correlation
between GPP and ET (Fig. 2), and the slope of this relation-
0.0
ship could be considered as an indicator of the ecosystem’s
0.0 0.3 0.6 0.9 1.2
LAI WUE (Law et al. 2002). Indeed, the study of Law et al.
(2002) found that the slope of the relationship was similar
Fig. 7  Relationships between daily gross primary productivity (GPP, between different biomes (i.e., slopes = 0.93 g C kg−1 ­H2O
g  C  m−2  day−1), evapotranspiration (ET, mm  day−1), and water use for grasslands, 0.87 for deciduous broadleaf forests, 0.85 for
efficiency (WUE, g C kg−1 ­H2O) and leaf area index (LAI) from May
to October in 2013–2016
crops, 0.65 for evergreen conifers) by comparing and ana-
lyzing the carbon and water exchange in forests, grasslands,
crops, across the temperate continental, oceanic, Mediter-
Discussion ranean, and the boreal climatic zones. This indicated that
physiological processes controlling water loss during carbon
Comparisons of WUE with other dryland ecosystems uptake reach a sustainable balance. The strong linear cor-
relation between GPP and ET also suggested that ET was a
Global drylands encompassing hyper-arid, arid, semiarid better proxy for water available to drive ­CO2 exchanges after
areas have expanded in the last 60 years and will continue hydrologic losses in a drier world (Biederman et al. 2016). In
to expand in twenty-first century, and the ecosystems over our research, the slope of the relationship between GPP and
drylands are fragile and sensitive to climate change (Feng ET was 0.68 in the desert riparian forest. The responses of
and Fu 2013). Recently, the eddy covariance measurements daily GPP and ET to climatic factors were similar at monthly
in drylands have provided valuable insights into the ecosys- scale (Fig. 6) and that both GPP and ET increased with an
tem carbon and water coupling relationships (Hinojo-Hinojo increase in Ta, Rg, and VPD (VPD < 3.5 kPa) on the growing
et al. 2016; Scott et al. 2006); however, the coupling rela- season basis.
tionship remains poorly identified and less understood for Previous studies have shown that the solar radiation, vapor
the desert riparian forest in the hyper-arid environment. In pressure deficit, air temperature, water availability, and the leaf

13
 358   Page 10 of 12 Environmental Earth Sciences (2018) 77:358

Table 2  Characteristics of mean annual air temperature(Ta), annual precipitation (P), leaf area index (LAI), and water use efficiency (WUE) in
dryland ecosystems
Ecosystem Location Ta (°C) P (mm) LAI WUE Calculation of WUE References
(g C kg−1
­H2O)

A natural riparian cottonwood 49.702°N, 112.863°W 5.9 380 1.4 3.1 GPP/ET Flanagan et al. (2017)
forest
The buffelgrass savanna 28°43′N, 110°33′W 22.7 344 1.74 GPP/ET Hinojo-Hinojo et al. (2016)
Aleppo pine forest at the edge 31°21′N, 35°03′E 270 3.3 NEE/ET GrÜNzweig et al. (2003)
of Negev desert
Eucalypt plantation 38°38 N 15.9 709 4.0 2.87 GPP/ET Rodrigues et al. (2011)
8°36 W
A saline desert shrub 44°170N, 87°560E 6.6 150 0.35 0.76 GPP/ET Liu et al. (2012a)
Chihuahuan Desert shrub 110°3′W, 31°44′N 17 322 0.96 GPP/ET Scott et al. (2006)
species
A desert riparian forest 41°59′N, 101°10′E 8.9 37.5 0.74 0.79 GPP/ET This study

area index are the primary factors affecting WUE (Li et al. and their responses to the autumn irrigation and envi-
2015; Tong et al. 2014; Xie et al. 2016a). A study of WUE for ronmental factors were analyzed over four years in the
a sparse vineyard in the arid northwest China showed that fac- desert riparian forest. Extensive field measurements and
tors as solar radiation, air temperature, water availability, and subsequent data analysis showed that the mean WUE
leaf area index all exerted a significant influences on vineyard’s ranged between 0.69 and 0.90 g C kg−1 ­H2O for the period
WUE dynamics (Li et al. 2015). And the WUE declined sig- 2013–2016, with a mean of 0.79 g C kg−1 ­H2O. The sea-
nificantly with the rising radiation, temperature and VPD. The sonal and inter-annual variations of WUE were not so
study of Tong et al. (2014) also showed that the WUE values obvious over the four year study period. The ecosystem
in a mixed planation dropped with increasing solar radiation WUE was not affected by climatic factor as R g, T a and
and VPD. These results were somewhat different compared VPD in the growing season in the desert riparian forest. In
to our study. In our study, there was no obvious relationship fact, the seasonal trend of WUE was affected by the time
identified between WUE and VPD, but the values of daily of irrigation. After the autumn irrigation, WUE declined
WUE kept higher value (ranging from 0.64 to 1.32 g C kg−1 significantly with an increase in air temperature. Moreo-
­H2O) when VPD was more than 3.8 kPa. For the case of the ver, the WUE was not also controlled by LAI at a daily
desert riparian forest, the WUE were nearly constant during scale according to the results of our study. In conclusion,
the growing season, mainly because the seasonal patterns of the WUE is not only highly vegetation dependent, but
daily GPP and ET were similar (Fig. 6). Therefore, there were also subject to the events (i.e., irrigation) and short-term
no obvious relations between WUE and radiation, air tempera- variations in meteorological conditions (Kuglitsch et al.
ture and VPD on the basis of the growing season. However, 2008). In accordance with the findings, it is necessary
it is important that after irrigation (where water was imported to investigate this problem over a long-term period to
under strict artificial controls since 2002 to relieve ecological ascertain the water–carbon coupling processes and the
problems), the hydrological conditions were therefore differ- underlying mechanisms, which are complex and challeng-
ent in this desert riparian forest. As a result, it is possible that ing for researchers, yet crucially important to unveil the
the irrigation activity had produced a significant on the the characteristics of ecosystem water use efficiency in the
seasonal course of daily WUE, which was affected by the time desert riparian forest deserts.
of irrigation (Fig. 3). Moreover, after autumn irrigation, Ta was
the main factor affecting the WUE (Fig. 6): the values of WUE Acknowledgements  This work was supported by National Key R&D
Program of China (2017YFC0404305). This work also was funded by
decreased significantly with increasing Ta. the National Natural Science Foundation of China (Nos. 31370467 and
41401033). The authors thank anonymous referees for their reading of
this manuscript and helpful reviews and comments.
Conclusions
Author contributions  Xiaohong Ma analyzed data and wrote the paper;
Qi Feng conceived the study; Yonghong Su and Ravinesh Deo revised
In this study, the seasonal and inter-annual variations of the paper; Tengfei Yu collected data and also revised the paper.
the gross primary productivity (GPP), evapotranspira-
tion (ET) and the ecosystem water use efficiency (WUE)

13
Environmental Earth Sciences (2018) 77:358 Page 11 of 12  358

Compliance with ethical standards  Li S et al (2015) Ecosystem water use efficiency for a sparse vineyard
in arid northwest China Agric Water Manag 148:24–33. https​://
doi.org/10.1016/j.agwat​.2014.08.011
Conflict of interest  The authors declare that they have no conflict of
Liu R, Li Y, Wang Q-X (2012a) Variations in water and C ­ O2 fluxes
interest.
over a saline desert in western. China Hydrol Process 26:513–522.
https​://doi.org/10.1002/hyp.8147
Liu R, Pan L-P, Jenerette GD, Wang Q-X, Cieraad E, Li Y (2012b)
High efficiency in water use and carbon gain in a wet year for a
References desert halophyte community. Agric For Meteorol 162:127–135.
https​://doi.org/10.1016/j.agrfo​rmet.2012.04.015
Biederman JA et al (2016) Terrestrial carbon balance in a drier world: Mo X, Liu S, Hu S, Xia J, Shen Y (2017) Sensitivity of terrestrial
the effects of water availability in southwestern North Amer- water and carbon fluxes to climate variability in semi-humid
ica. Glob Change Biol 22:1867–1879. https​://doi.org/10.1111/ basins of Haihe River. China Ecol Model 353:117–128. https​://
gcb.13222​ doi.org/10.1016/j.ecolm​odel.2016.09.003
Cao SK, Feng Q, Si JH, Su YH, Chang ZQ, Xi HY (2009) Relation- Moore CJ (1986) Frequency response corrections for eddy correlation
ships between foliar carbon isotope discrimination with potas- systems Bound Layer Meteorol 37:17–35 https:​ //doi.org/10.1007/
sium concentration and ash content of the riparian plants in the BF001​22754​
extreme arid region of China Photosynthetica 47:499–509. https​ Papale D et al (2006) Towards a standardized processing of net eco-
://doi.org/10.1007/s1109​9-009-0075-7 system exchange measured with eddy covariance technique: algo-
Chang Z, Ma Y, Zhou C (2016) Effects of elevated ­CO2 concentration rithms and uncertainty estimation. Biogeosciences 3:571–583
on water use efficiency of Tamarix ramosissima in an extremely Reed SC, Coe KK, Sparks JP, Housman DC, Zelikova TJ, Belnap J
arid region. Environ Earth Sci 75:1409. https​://doi.org/10.1007/ (2012) Changes to dryland rainfall result in rapid moss mortality
s1266​5-016-6221-1 and altered soil fertility. Nat Clim Change 2:752–755 http://www.
Chen YN, Chen YP, Li WH, Zhang HF (2003) Response of the accu- nature​ .com/nclima​ te/journa​ l/v2/n10/abs/nclima​ te159​ 6.html suppl​
mulation of proline in the bodies of Populus euphratica to the ement​ary-infor​matio​n
change of groundwater level at the lower reaches of Tarim River. Reichstein M et al (2005) On the separation of net ecosystem exchange
Chin Sci Bull 48:1995–1999. https:​ //doi.org/10.1007/bf0318​ 3993​ into assimilation and ecosystem respiration: review and improved
Cheng GD, Li X, Zhao WZ, Xu ZM, Feng Q, Xiao SC, Xiao HL (2014) algorithm. Glob Change Biol 11:1424–1439. https:​ //doi.org/10.1
Integrated study of the water–ecosystem–economy in the Heihe 111/j.1365-2486.2005.00100​2.x
River Basin. Natl Sci Rev 1:413–428. https:​ //doi.org/10.1093/nsr/ Rodrigues A et al (2011) Eight years of continuous carbon fluxes meas-
nwu01​7 urements in a Portuguese eucalypt stand under two main events:
Feng S, Fu Q (2013) Expansion of global drylands under a warm- drought and felling Agric For Meteorol 151:493–507. https​://doi.
ing climate. Atmos Chem Phys 13:10081–10094. https​://doi. org/10.1016/j.agrfo​rmet.2010.12.007
org/10.5194/acp-13-10081​-2013 Scott RL, Edwards EA, Shuttleworth WJ, Huxman TE, Watts C,
Feng Q, Peng JZ, Li JG, Xi HY, Si JH (2012) Using the concept of Goodrich DC (2004) Interannual and seasonal variation in fluxes
ecological groundwater level to evaluate shallow groundwater of water and carbon dioxide from a riparian woodland ecosystem
resources in hyperarid desert regions. J Arid Land 4:378–389 Agric For Meteorol 122:65–84 https​://doi.org/10.1016/j.agrfo​
Flanagan LB, Orchard TE, Logie GSJ, Coburn CA, Rood SB (2017) rmet.2003.09.001
Water use in a riparian cottonwood ecosystem: Eddy covariance Scott RL, Huxman TE, Cable WL, Emmerich WE (2006) Partitioning
measurements and scaling along a river corridor Agric For Mete- of evapotranspiration and its relation to carbon dioxide exchange
orol 232:332–348 https:​ //doi.org/10.1016/j.agrfor​ met.2016.08.024 in a Chihuahuan Desert shrubland Hydrol Process 20:3227–3243.
GrÜNzweig JM, Lin T, Rotenberg E, Schwartz A, Yakir D (2003) https​://doi.org/10.1002/hyp.6329
Carbon sequestration in arid-land forest. Glob Change Biol 9:791– Si JH, Feng Q, Zhang XY, Chang ZQ, Su YH, Xi HY (2007) Sap flow
799. https​://doi.org/10.1046/j.1365-2486.2003.00612​.x of Populus euphratica in a desert riparian forest in an extreme arid
Hinojo-Hinojo C, Castellanos AE, Rodriguez JC, Delgado-Balbuena J, region during the growing season. J Integr Plant Biol 49:425–436
Romo-León JR, Celaya-Michel H, Huxman TE (2016) Carbon and Si JH, Feng Q, Cao SK, Yu TF, Zhao CY (2014) Water use sources of
water fluxes in an exotic buffelgrass savanna Rangel Ecol Manag desert riparian Populus euphratica forests Environ Monit Assess
69:334–341 https​://doi.org/10.1016/j.rama.2016.04.002 186:5469–5477. https​://doi.org/10.1007/s1066​1-014-3796-4
Huang XR (2016) Progress on coupling relationship between carbon Tong X, Zhang J, Meng P, Li J, Zheng N (2014) Ecosystem water
and water of ecosystem in arid area. In: International conference use efficiency in a warm-temperate mixed plantation in the North
on geo-informatics in resource management and sustainable eco- China. J Hydrol 512:221–228. https​://doi.org/10.1016/j.jhydr​
systems. Springer, pp 456–465 ol.2014.02.042
Jackson RB, Car penter SR, Dahm CN, McKnight DM, Wang P, Zhang Y, Yu J, Fu G, Ao F (2011) Vegetation dynamics
Naiman RJ, Postel SL, Running SW (2001) Water in a induced by groundwater fluctuations in the lower Heihe River
changing. World Ecol Appl 11:1027–1045. https://doi. Basin, northwestern China. J Plant Ecol 4:77–90. https​://doi.
org/10.1890/1051-0761(2001)011[1027:WIACW]2.0.CO;2 org/10.1093/jpe/rtr00​2
Jia X et al (2016) Carbon and water exchange over a temperate semi- Wang Y, Feng Q, Chen LJ, Yu TF (2013) Significance and effect of
arid shrubland during three years of contrasting precipitation and ecological rehabilitation project in inland river basins in North-
soil moisture patterns. Agric For Meteorol 228:120–129. https​:// west China. Environ Manag 52:209–220. https​://doi.org/10.1007/
doi.org/10.1016/j.agrfo​rmet.2016.07.007 s0026​7-013-0077-x
Kuglitsch F et al (2008) Characterisation of ecosystem water-use effi- Webb EK, Pearman GI, Leuning R (1980) Correction of flux measure-
ciency of european forests from eddy covariance measurements. ments for density effects due to heat and water vapour transfer. Q
Biogeosci Discuss 5:4481–4519 J R Meteorol Soc 106:85–100. https​://doi.org/10.1002/qj.49710​
Law BE et al (2002) Environmental controls over carbon dioxide and 64470​7
water vapor exchange of terrestrial vegetation Agric For Meteorol Xi HY, Feng Q, Si JH, Chang ZQ, Cao SK (2010) Impacts of river
113:97–120. https​://doi.org/10.1016/S0168​-1923(02)00104​-1 recharge on groundwater level and hydrochemistry in the lower

13
 358   Page 12 of 12 Environmental Earth Sciences (2018) 77:358

reaches of Heihe River Watershed, northwestern China. Hydro- Yu TF, Feng Q, Si JH, Zhang XY, Zhao CY (2017a) Tamarix ramosis-
geol J 18:791–801. https​://doi.org/10.1007/s1004​0-009-0562-8 sima stand evapotranspiration and its association with hydrocli-
Xie J et al (2016a) Ten-year variability in ecosystem water use effi- matic factors in an arid region in northwest China J Arid Environ
ciency in an oak-dominated temperate forest under a warming cli- 138:18–26 https​://doi.org/10.1016/j.jarid​env.2016.11.006
mate Agric For Meteorol 218:209–217. https​://doi.org/10.1016/j. Yu TF, Feng Q, Si JH (2017b) Evapotranspiration of a Populus euphra-
agrfo​rmet.2015.12.059 tica Oliv. forest and its controlling factors in the lower Heihe River
Xie J et al (2016b) Seasonal variation in ecosystem water use effi- Basin. Northwest China Sci Cold Arid Reg 9:0175–0182
ciency in an urban-forest reserve affected by periodic drought.
Agric For Meteorol 221:142–151. https​://doi.org/10.1016/j.agrfo​
rmet.2016.02.013

13