Agricultural and Forest Meteorology 316 (2022) 108842

Contents lists available at ScienceDirect

Agricultural and Forest Meteorology

journal homepage: www.elsevier.com/locate/agrformet

The role of climate change and vegetation greening on evapotranspiration

variation in the Yellow River Basin, China
Fubo Zhao a, Shuai Ma a, Yiping Wu a, *, Linjing Qiu a, Wenke Wang b, Yanqing Lian c, Ji Chen d,
Bellie Sivakumar e
a
Department of Earth & Environmental Science, School of Human Settlements and Civil Engineering, Xi’an Jiaotong University, Xi’an, Shaanxi Province, 710049, China
b
School of Water and Environment, Chang’an University, Xi’an 710054, China
c
Hohai University, Nanjing, Jiangsu Province, 210024, China
d
Department of Civil Engineering, The University of Hong Kong, Pokfulam, Hong Kong, China
e
Department of Civil Engineering, Indian Institute of Technology Bombay, Powai, Mumbai, 400076, India

A R T I C L E I N F O A B S T R A C T

Keywords: Actual evapotranspiration (ET) plays a key role in the water cycle and energy balance, and it is also sensitive to
Climate change climate and land use changes due to its explicit link with multiple land surface processes. Understanding
Evapotranspiration spatiotemporal changes of ET and its drivers is of vital importance for water resources management. In this study,
Penman-Monteith-Leuning model
we estimated spatiotemporal changes in ET and quantified its drivers in the Yellow River Basin (YRB)—the
Vegetation greening
Water resources
second largest river basin of China—during the period of 1982–2016 using the modified Penman-Monteith-
Leuning (PML) model by simply incorporating land use dynamics. The PML model was validated against on-
site flux observations, global evaporation modeling results, and water balance-based observations. The valida­
tion using annual and monthly data from above-mentioned data showed that the modified PML model performs
well for both the entire YRB and sub-watershed scale. During the 35-year study period, the YRB experienced
significant climate warming and vegetation greening, reflected by significant increases in air temperature and
leaf area index, causing a significant increase of the basin-averaged ET (1.36 mm/yr, P<0.01). We conducted the
factorial experiments to attribute ET variations to climate and vegetation greening, and the results show that
climate warming, vegetation greening, and increased vapor pressure deficit were the major contributors to the
positive changes in ET, accounting for 45.6% (0.62 mm/yr), 31.6% (0.43 mm/yr), and 19.9% (0.27 mm/yr).
Variations in solar radiation and wind speed played minor positive and negative roles in annual ET variations.
Moreover, vegetation greening significantly contributed (92.3%) to the increase in the ratio of transpiration to
ET due to the increased available energy absorbed by canopy. This study provides a comprehensive assessment of
the effects of climate and vegetation on ET, and can help in formulating appropriate policies for water resource
management in the context of climate warming and revegetation programs.

1. Introduction has changed at both regional and global scales during the past several
decades as a result of climate warming and human activities (Jung et al.,
Actual evapotranspiration (ET) is a key factor linking soil, plants, and 2011; Pascolini-Campbell et al., 2021). Variations in ET can exert pro­
the atmosphere and plays an important role in many physical processes found impacts on multiple factors associated with human life, such as
of the Earth, such as the hydrological cycle, energy balance, climate and food supply, freshwater supply, and economic development (Li et al.,
socioeconomic systems, and biogeochemical cycling (Jung et al., 2010; 2018; Zhao et al., 2020). Therefore, understanding variations in ET and
Pascolini-Campbell et al., 2021; Savita et al., 2019). As ET is involved in its drivers is of vital importance for water resources management,
multiple processes in terrestrial ecosystems, it is sensitive to climate and especially in the context of climate change.
land use changes (Montaldo and Oren, 2018; Zhang et al., 2020; Zhang ET is a key component of water balance, and approximately 60% of
et al., 2001). Many previous studies have reported that the pattern of ET precipitation is returned to the atmosphere through ET, i.e., through

* Corresponding author.
E-mail address: rocky.ypwu@gmail.com (Y. Wu).

https://doi.org/10.1016/j.agrformet.2022.108842
Received 23 August 2021; Received in revised form 15 January 2022; Accepted 20 January 2022
Available online 29 January 2022
0168-1923/© 2022 Elsevier B.V. All rights reserved.
F. Zhao et al. Agricultural and Forest Meteorology 316 (2022) 108842

biotic transpiration (interception loss) and abiotic evaporation (soil 2021). With the rapid intensification of climate change and human ac­
evaporation) (Lian et al., 2018; Zeng et al., 2012). Biotic transpiration tivities, water shortage has become one of the most serious issues
and abiotic evaporation occur through different processes and are threatening sustainable development in the YRB (Feng et al., 2016;
influenced by various factors (Jiang et al., 2020). For example, biotic McVicar et al., 2007). In particular, the Grain for Green project imple­
transpiration is related to plant growth and carbon assimilation and thus mented in the Loess Plateau region (the middle portion of the YRB) has
may be primarily dominated by vegetation components such as the leaf significantly changed the conditions of vegetation cover and the water
area index (LAI), whereas abiotic evaporation may be influenced by cycle (Chen et al., 2015). In this context, large spatial and temporal
both climatic variables and LAI (Zuo et al., 2012). During the past variations and an increasing trend of ET have been reported over the
several decades, many methods have been developed and used to esti­ entire YRB during the past several decades (Jiang et al., 2020; Xie et al.,
mate ET at the site, watershed, and regional scales (Lei and Yang, 2010; 2019). However, although the combined effects of climate change and
Leuning et al., 2008; Zhao et al., 2018a). In situ measurements, such as vegetation greening on ET have been acknowledged, their individual
the eddy covariance technique, sap flow measurements, and Bowen contributions remain poorly understood. In-depth understanding the
ratio systems, are regarded as the most reliable approaches for esti­ individual impacts of climate and vegetation on ET is of vital importance
mating ET at the site scale (Wang et al., 2013; Wang et al., 2011). for formulating appropriate measures for water resources management
However, on-site measurements cannot represent total ecosystems in the YRB, especially when facing the rapid climate change and intense
because they are limited to small scale studies and difficult to upscale human land management programs.
(Liang et al., 2020a; Zhao et al., 2018a). Many process-driven hydro­ This study aimed to individually analyze the effects of climate and
logical models have also been widely used to estimate ET at the water­ vegetation on spatiotemporal variations of ET over the YRB using a PM-
shed or regional scales. For example, Zhao et al. (2018a) used the Soil based method (Penman-Monteith-Leuning, PML). The specific objec­
and Water Assessment Tool (SWAT) model (Arnold et al., 1998) to es­ tives are as follows: (1) test the applicability of the PML model to the
timate spatiotemporal changes of ET in a semi-arid watershed. Hadi­ YRB; (2) investigate spatiotemporal variations of ET in the YRB during
wijaya et al. (2021) used the Canadian Land Surface Scheme (GLASS) to the recent 35 years (1982–2016) using the PML model; and (3) attribute
estimate ET at a boreal forest site. Yeste et al. (2020) used the Variable the changes in ET to climate change and vegetation greening.
Infiltration Capacity (VIC) model to estimate ET in the Duero River Basin
and found that the VIC model can simulate ET dynamics with high ac­ 2. Materials and methods
curacy. These hydrological models are valuable for understanding the
mechanism of the water cycle at different scales. However, these models 2.1. Study area
have the shortcoming of uncertainties in the simulations because sci­
entific hypotheses are simplified (Zhao et al., 2018b). Different sets of The Yellow River is the second largest river in China with a length of
equations, parameter values, and input driving variables result in 5464 km. It originates from the Qinghai-Tibetan plateau and flows
different estimates of ET, especially for areas with complex terrains and across nine provinces in northern China. It is also the most important
climatic conditions (Liang et al., 2020a; Yen et al., 2015; Zhao et al., water source in northwestern and northern China. The YRB is charac­
2018b). terized by arid and semi-arid continental monsoon climate with a mean
Land surface information can be captured at large scales using sat­ annual precipitation of 495 mm (Yin et al., 2021). The mean annual air
ellite remote sensing. Therefore, it has been widely used to estimate ET temperature ranges from approximately -4 to 14 ◦ C and varies with both
in recent years, and many remote sensing (RS)-driven ET models have latitude and elevation. The major land cover types in the YRB are
been developed (Fang et al., 2016; Xiao et al., 2012). For example, the grassland, cropland, and forest, which accounted for 47%, 26%, and 7%,
Priestley-Taylor Jet Propulsion Laboratory (PT-JPL) model (Priestley respectively, of the total land cover in 2010 (Fig. 1b). Grasslands are
and Taylor, 1972), which was constructed using satellite-based data, has
been widely used to estimate ET. The model has been proven to be an
efficient tool in estimating ET and identifying its drivers at large scales.
Shao et al. (2019) used the PT-JPL model to investigate ET variations in
the Loess Plateau and found that vegetation greening was the major
driver of ET variation. Yang et al. (2019b) used the PT-JPL model to
estimate ET over global semi-arid regions. The Penman-Monteith
(PM)-based methods are also important RS-driven ET methods
(Monteith, 1965; Penman, 1948), and they are commonly used for
estimating ET and quantifying its drivers. PM-based methods have clear
physical meanings and explicitly account for radiative and aero­
dynamics controls on ET and are suitable for various climatic conditions
(Li et al., 2020; Zhang et al., 2010). Based on the PM equation, Leuning
et al. (2008) further improved a PM model by introducing a biophysical
model for surface conductance to account for the physical features of
canopy and soil water loss (Zhang et al., 2020). The improved PM model
requires only meteorological observations and satellite-based LAI as
inputs, and it has been widely used to estimate ET and attribute its
drivers at both the regional and global scales (Leuning et al., 2008; Li
et al., 2020; Zhang et al., 2020). However, previous studies usually
assumed that the land use conditions kept constant in estimating ET by
using PML model, which may lead to over- or underestimations of ET
and its attributions (Zhang et al., 2020).
The Yellow River Basin (YRB), which is located in the transitional
Fig. 1. Location and land cover types (2010) in the Yellow River Basin. The
zone of monsoon and continental climates, is one of the most important sub-watersheds numbered 1-14 are illustrated in Table 1, which are used for
basins in China (Yin et al., 2021). The YRB directly supports more than calibration. The 15 sub-watershed is the Headwater Region of the YRB (HYRB).
100 million people and has been recognized as the Cradle of Chinese The stars in b are the flux tower sites that are used for model verification. Pie
civilization and ‘Mother River’ basin (Jiang et al., 2020; Tian et al., charts indicate the proportions of land cover types in the YRB.

2
F. Zhao et al. Agricultural and Forest Meteorology 316 (2022) 108842

mainly distributed in the upper and middle regions of the YRB, crop­ GLEAM v3 was used to produce global ET data set spanning the period
lands are primarily located in the middle and lower reaches, and forests from 1980 to 2016 at a spatial resolution of 0.25◦ × 0.25◦ . The GLEAM
are mainly distributed in the middle reaches. ET has been widely used to validate hydrologic model at regional and
global scales and showed good agreements with water balance-based ET
and tower flux measurements (Bai et al., 2020; Dembélé et al., 2020;
2.2. Data Konapala et al., 2020; Miralles et al., 2011; Shao et al., 2019; Yang et al.,
2017). Especially, for example, Yang et al. (2017) evaluated GLEAM ET
2.2.1. Meteorological data using eddy covariance observations and found that the GLEAM had a
Climate data were obtained from the China Meteorological Forcing reasonable accuracy for most land cover types across China. Therefore,
Dataset (CMFD), which was produced by the Institute of Tibetan Plateau we used GLEAM ET data set to validate the PML model’s performance in
Research, Chinese Academy of Sciences (http://www.itpcas.ac.cn/) this study. Third, the annual streamflow from the 14 major watersheds
(Yang et al., 2010). The dataset was generated by combining remote in the middle reaches of the YRB were collected for validation (Liang
sensing products, reanalysis dataset, and in-situ observation data from et al., 2015) (Table 1). The streamflow were further converted to ET
weather stations (Yang and He, 2019). It covers the period of 1979–2018 based on the water balance (ET=P-Q), assuming that the annual water
with a spatial resolution of 0.1◦ , and provides the following seven storage was negligible in the long term following previous studies (Liang
near-surface meteorological elements: 2-m air temperature, surface et al., 2020b; Shao et al., 2021; Shao et al., 2019; Wu et al., 2017)
pressure, specific humidity, 10-m wind speed, downward shortwave (Figure S1). For example, according to Shao et al. (2021), ignoring ΔS at
radiation, downward longwave radiation, and precipitation rate. In this 10-year or longer scales has the least influence on water budget closure
study, we used air temperature, surface pressure, specific humidity, in the Loess Plateau region (the middle portion of the YRB). We also
wind speed, and radiation data to drive the ET model. It should be noted collected in-situ observed Es and Ec data to validate ET components in
that the 10-m wind speed was converted to 2-m wind speed using the YRB (Table S2).
equation 1 (Allen et al., 1998):
4.87 2.3. ET model
U2 = U10 (1)
ln(67.8z − 5.42)
ET consists of three components: transpiration from plant canopy
where U2 and U10 are the 2-m and 10-m wind speeds, z = 10 m. (Ec: mm/day), evaporation from soil (Es: mm/day), and evaporation
from canopy interception (Ei: mm/day) (equation 1).
2.2.2. Remote sensing data
ET = Ec + Es + Ei (2)
The Global Inventory Modeling and Mapping Studies (GIMMS) LAI
3g data with a spatial resolution of 0.0833◦ and a 15-day temporal Here, we employed the PML model to estimate Ec and Es (Leuning
resolution from 1982 to 2016 were used in this study (Zhu et al., 2013). et al., 2008). Ec was calculated as follows:
The GIMMS LAI 3g data were generated based on the GIMMS normal­ ⎛ ( ) ⎞
ized difference vegetation index and a neural network algorithm. In this s ρcp
R nc + 86400 Da Ga
1000 ⎜ γ γ ⎟
study, we used the Global Land Surface Satellite (GLASS) albedo data to Ec = ⎜ ⎟ (3)
λρw ⎝ s
+ 1 + GGac ⎠
estimate the net radiation during 1982–2016. GLASS albedo has been γ

proven to be highly accurate over the YRB regions (Liang et al., 2020a).
where λ is the latent heat of vaporization (MJ kg− 1), ρw is the water
2.2.3. Land cover data density (kg m− 3), γ is the psychrometric constant (kPa ◦ C− 1), s is the
To reflect land cover dynamics in ET estimation, we incorporated slope of saturated vapor pressure curve (kPa ◦ C− 1), Rnc is the net radi­
land cover change information in the PML model. To this end, land cover ation absorbed by canopy (MJ m− 2 d− 1), ρ is the density of air (kg m− 3),
maps for 1980, 1995, 2000, 2005, and 2010 with a resolution of 1 km cp is the specific heat of air at constant pressure (MJ kg− 1 ◦ C− 1), Da is the
were used. The maps were collected from the Institute of Geographic vapor pressure deficit (VPD), Ga si the aerodynamic conductance (m
Sciences and Natural Resources Research, Chinese Academy of Sciences s− 1), and Gc is the canopy conductance (m s− 1). Rnc and Rns can be
(Liu et al., 2005). In the maps, the land cover types mainly include estimated as follows:
grassland, cropland, forest, shrub, urban area, water body, lake, bare
land, and desert. Moreover, the land cover information represent five Rn = Rnshort + Rnlong (4)
periods: 1982–1989 (1980), 1990–1999 (1995), 2000–2004 (2000),
2005–2009 (2005), and 2010–2016 (2010) (Liang et al., 2020a). We also Rnc = Rn (1 − exp(− KA LAI)) (5)
collected land use product from Song et al. (2018), who provided the
fraction combination of vegetation functional types, including tree
Table 1
cover, short vegetation, and bare ground, covering the study period of
Long-Term hydrometeorological characteristics (1982–2009) of the fourteen
1982–2015.
watersheds in the middle reaches of the YRB.
No Catchment Area (km2) P (mm) Q (mm) ET (mm)
2.2.4. ET validation data
To validate ET, several types of data were collected. First, we 1 Huangfu (HFC) 3230 395.1 24.7 370.4
collected flux tower-based observations from measurement sites in and 2 Gushan (GSH) 1260 392.3 32.9 359.4
3 Kuye (KYH) 8621 389.2 43.1 346.1
around the YRB (see Fig. 1b). These sites include Haibei (HB, shrub), 4 Tuwei (TWH) 3307 396.6 70.8 325.8
Changwu (CW, cropland), and Shapotou (SPT, desert), in which the HB 5 Jialu (JLH) 1138 423.5 29.1 394.4
station belong to the ChinaFLUX system (Yu et al., 2006). The data of CW 6 Wuding (WDH) 24682 363.8 29.8 334.0
and SPT stations were extracted from Liang et al., (2020a) using GetData 7 Dali (DLH) 3861 476.7 32.9 443.8
8 Qingjian (QJH) 3600 481.4 34.4 446.9
Graph Digitizer software (version 2). Data in CW station is only available
9 Yanhe (YH) 5857 490.9 30.9 460.0
for 2006 and 2007 (January to October), while data in SPT is only 10 Beiluo (BLH) 4186 485.6 33.2 452.4
available for 2005 and 2006 (May to October). Second, we used both 11 Jinghe (JH) 43106 509.8 31.2 478.6
monthly and annual ET data from the Global Land Evaporation 12 Weihe (WH) 30122 480.5 48.0 432.5
Amsterdam Model (GLEAM) (v3) (Miralles et al., 2010) to validate the 13 Fenhe (FH) 38728 490.4 13.3 477.1
14 Xinshui (XSH) 4186 488.7 23.1 465.6
ET for both the main sub-watersheds (Fig. 1a) and the entire YRB.

3
F. Zhao et al. Agricultural and Forest Meteorology 316 (2022) 108842

Rns = Rn − Rnc (6) GLEAM-ET and three on-site observations (Moriasi et al., 2007). The
optimized gsx is shown in Table 2.
where Rn is the net radiation and KA is the light extinction coefficient
(0.6), and LAI is the leaf area index. Ga and Gc in equation 1 can be 2.4. Attribution method
calculated as follows:

k2 um We first approximately quantified the impact of climate change and

Ga = ln[(zm − d)/zom ] (7) land use change effects on ET variation from 1982–2016 using a simple
ln[(zm − d)/zov ]
method following Huang et al. (2020). We assumed that changes in ET in
[ ][ ] areas with significant land use change (Theil-Sen slope, P < 0.05) were
gsx Qh + Q50 1
Gc = ln (8) attributable to land use change, whereas ET changes occurred without
kQ Qh exp(− kQ LAI) + Q50 1 + Da /D50
land use change were attributable to climate change. The net ET change
for one land cover change was calculated by averaging ET trends of this
where k is the von Kármán constant; um (m s− 1) is the wind speed at
specific land cover change across all pixels in the study area. The overall
height zm (m); d is the zero plane displacement height (m); zom and zov
impact of land cover change was defined as the proportion of the ET net
are the roughness for momentum and water vapor, estimated by zom =
change in areas with a significant land cover change. The residual was
0.123h and zov=0.1zom, where h is the canopy height (Zhou et al., 2006);
defined as the overall climate change impact. Based on these results, we
gsx is the maximum stomatal conductance, which is required for cali­
conducted factorial experiments to attribute ET variations to specific
bration; Qh is the PAR at the top of canopy height (MJ m− 2 d− 1); Q50 is
climatic variables and LAI (a proxy of vegetation greening). In total, six
the absorbed PAR; and D50 is the Da at which stomatal conductance
experiments were performed to quantify the contributions of climate
reaches gsx/2.
and vegetation to ET variations: control experiment (CT) with pre­
Es is estimated as follows:
dictions driven by all varied climate variables and dynamic LAI, S1
⎛ ⎞
s
experiment with ET prediction using the mean air temperature during
1000 ⎜f γRns ⎟ the 1980s (1982–1985), S2 using the mean vegetation condition during
Es = ⎝ ⎠ (9)
λρw γs + 1 the 1980s, S3 using the mean VPD during the 1980s, S4 using the mean
Rn during the 1980s, and S5 using the mean wind speed during the
where f is the ratio of soil evaporation. It varies between 0 and 1 and can 1980s (Liang et al., 2020a). We considered air temperature and VPD
be estimated as follows: simultaneously to gain a comprehensive understanding of driving fac­
( ) tors of ET variation. For example, the increased air temperature can
P elevate the soil temperature and thus the soil water availability and even
f = min ,1 (10)
Eseq soil drought, which can exert profound impacts on ET (Du et al., 2021).
Although VPD is tightly coupled with soil moisture (Liu et al., 2020), it
where P is the total precipitation and Eseq is the total equilibrium soil majorly reflects the atmospheric water demand and humidity effects on
evaporation over the previous 30 days (Zhang et al., 2020). Eseq can be ET. The details of the experiments are shown in Table 3. Differences
calculated as follows: between the CT and S1–S5 can be considered the effects of climate or
A γs vegetation on ET variations during 1982–2016. For example, the effects
Eseq = ( ) (11) of air temperature on ET can be calculated as follows:
λ 1 + γs
ETTa = f (CT) − f (S1) (15)

A = Rn − G (12) 3. Results
G = Rn [τc + (1 − M)(τs − τc )] (13)
3.1. Model evaluation
where A is the available energy absorbed by both vegetation and soil, G
is the soil heat flux (MJ m− 2 d− 1), τc is the parameter with better Fig. 2 shows comparisons between monthly ET estimated by PML
vegetation cover (0.05), and τs is the parameter of the bare soil area and observed ET from three flux observations. The results show that the
(0.325). M is the vegetation coverage, which was calculated based on time series of ET variations in these three sites were captured by the PML
the normalized difference vegetation index value in the YRB (Bas­ model. The highest agreement between the estimated and observed ET
tiaanssen et al., 1998; Shao et al., 2019). was for HB station, with Pearson’s correlation coefficient (R) of 0.95 and
We used the Priestley-Taylor Jet Propulsion Laboratory model to RMSE of 9.63 mm month− 1. The lowest agreement was found for CW
estimate Ei: station, with a relatively lower R of 0.91 and higher RMSE of 12.08 mm
month− 1.
s We also compared the monthly ET simulations with ET products
Ei = fwet α Rnc (14)
s+γ from GLEAM across the 14 watersheds in the middle and upper reaches
of the YRB and the entire YRB (Fig. 3). The results show that the
where fwet is the relative surface wetness and α is the PT coefficient
(1.26).
Table 2
As shown by the above equations, only the gsx parameter requires
Optimized gsx in different land cover types in the
calibration. We calibrated gsx by comparing PML-ET with FLUX-ET,
YRB.
GLEAM-ET, and water balance-ET. Previous studies on ET estimation
usually assumed a constant gsx. However, this assumption may lead to Land cover types gsx

over- or underestimations of ET, especially in areas with significantly Forest 0.0016

changing land cover conditions (Zhang et al., 2020). Therefore, we Shrubland 0.0025
Grassland 0.0029
applied a simple modification to the PML model by incorporating land
Cropland 0.0035
use change. In this manner, the parameter gsx can vary with changing Urban 0.0199
types of land cover, and the modified PML can thus reflect land cover Water body 0.0034
dynamics. We calibrated the gsx by comparing modeled ET and Desert and Barren 0.0034

4
F. Zhao et al. Agricultural and Forest Meteorology 316 (2022) 108842

Table 3 with water balance-ET, although relatively large uncertainties existed in

Details of the factorial experiment design. some watersheds (e.g., HFC, DLH, YH, and JH). We also compared the
Experiments Abbreviation Description modeled Es and Ec (the key ET components) with observations during
the study period (Figure S3). Results showed that the PML model could
Control CT Prediction of ET using all dynamic climatic
experiment variable and LAI simulate Es and Ec with reasonable accuracy. Specifically, the model
Scenario ETTa Air temperature using the mean conditions of simulated Es with a small bias of 1.13 mm month− 1 and a high R of 0.89.
settings 1982-1985, the others same as the control The model could simulate Ec with a bias of -6.81 mm month− 1 and a
experiment moderate R of 0.77. These results indicated that the PML model could
ETLAI LAI using the mean conditions of 1982-1985, the
others same as the control experiment
simulate ET components with acceptable accuracy. Overall, the com­
ETVPD VPD using the mean conditions of 1982-1985, parisons demonstrated that the PML model was robust and can be used
the others same as the control experiment for further analysis of ET variation in the YRB.
ETRn Net radiation using the mean conditions of 1982-
1985, the others same as the control experiment
ETWn Wind speed using the mean conditions of 1982- 3.2. Changes in climatic variables and LAI
1985, the others same as the control experiment

Fig. 4 shows the spatial patterns of mean annual climatic variables

variations of ET simulated by PML were consistent with GLEAM-ET and LAI during 1982–2016. Air temperature exhibited clear spatial
across the 14 watersheds and the Headwater region of the YRB heterogeneity across the YRB. High values of mean annual air temper­
(HYRB) and the entire YRB. Among them, the highest correlation be­ ature were found in the middle and lower reaches of the YRB, with
tween PML-ET and GLEAM-ET was observed for the headwater region of values ranging from 12.43 to 17.92 ◦ C. The lower air temperature was
the YRB (HYRB) and the entire YRB, with correlations of 0.97 and RMSE found in the headwater region of the YRB, with values ranging from
of 6.21 mm month− 1. The lowest correlation was observed for Jialuhe -9.53 to -3.83 ◦ C. During the study period, air temperature showed a
watershed, with the lowest R of 0.78 and highest RMSE of 15.3 month− 1. significant increase across almost the entire basin, indicating a warming
For the entire YRB, PML-ET correlated well with GLEAM-ET with an R of trend (warming rate of 0.06◦ C/yr and P < 0.01) (Fig. 5a). The highest
0.94 and RMSE of 7.45 mm month− 1. The consistency between PML-ET LAI was found in the mountain areas in the middle reaches of the YRB,
and GLEAM-ET for the 14 watersheds and the entire YRB indicate that while the relatively lower values were found in the northern areas.
PML could have satisfactory performance in simulating monthly ET in During the study period, LAI showed a significant increase at a rate of
the YRB (Moriasi et al., 2007). In addition, we compared annual PML-ET 0.003 m2/m2/yr (P < 0.01) (Fig. 5b). The middle reaches (the Loess
variations against both the water balance-ET and GLEAM-ET variations Plateau) of the YRB exhibited the most significant increase in LAI and
(Figure S1). Across the 14 watersheds, PML-ET was roughly consistent small portions in the headwater region exhibited a decrease. The spatial
patterns of VPD varied with altitude, with lower VPD in headwater

Fig. 2. Monthly comparisons between modeled ET using PML and observed ET in three flux tower sites. (a) Haibei (HB), (b) Changwu (CW), and (c) Shapotou (SPT).

5
F. Zhao et al. Agricultural and Forest Meteorology 316 (2022) 108842

Fig. 3. Monthly comparisons between estimated ET and GLEAM-ET across the 14 sub-watersheds (see Table 1) in the middle portions of the YRB, the HYRB, and the
entire YRB.

regions and higher VPD at lower elevation. The spatial trend of VPD 3.3. Spatiotemporal changes in ET and its components
significantly varied across the YRB, with a relatively higher increasing
trend in the middle reaches and lower increasing trend in the headwater The spatial distributions of annual ET from 1982–2016 in the YRB
regions. VPD also showed a significant increasing trend, with a rate of are shown in Fig. 6a. Annual ET exhibited regional differences, with
0.003 kpa/yr (P < 0.01) during the study period (Fig. 5c). Rn was found higher values mainly in the southern region and lower values in the
to be higher mainly in the headwater regions, whereas the northwestern northern region. Annual ET varied from 120 to 900 mm, with a mean
margins exhibited lower Rn. Most regions in the middle and lower rea­ value of 387 mm. Mean annual ET mostly ranged from 200 to 500 mm,
ches showed significant increases in Rn and the upper-to-middle tran­ and only small parts in the northern region showed values lower than
sitional regions showed a decreasing trend. Overall, Rn showed a 200 mm. The rate of ET variation ranged from -2 to 6 mm/yr. Areas with
significant increasing rate of 0.006 MJ/m2/d/yr (P < 0.01) during significantly increased ET were mainly located in the middle and
1982–2016 (Fig. 5d). Wn showed a wide spatial variation, and most headwater regions of the YRB. Only small parts in the middle region
areas experienced an average Wn ranging from 0.79 to 1.84 m/s. The showed decreasing trends of ET. Overall, ET exhibited a significantly
upper-to-middle transitional regions showed a significant increasing increasing trend (P < 0.001) from 1982 to 2016 with a rate of 1.36 mm/
trend of Wn with a slope ranging from 0.013 to 0.045 m/s/yr, whereas yr (Fig. 6g).
large areas in the headwater and middle regions showed a decreasing The spatial patterns of Ec were roughly consistent with the distri­
trend. In particular, Wn showed periodic fluctuations during the study butions of LAI: areas with higher Ec showed higher LAI. Areas with high
period. Before 2007, Wn showed a slight decreasing rate of -0.0007 m/s/ Ec were mainly distributed in the middle and southern regions of the
yr, but it significantly increased to a rate of 0.04 m/s/yr (P < 0.01) after YRB, whereas those with low Ec were mainly distributed in the upper
2007 (Fig. 5e). and upper-to-middle transitional regions. Annual Ec varied from 6 to
above 247 mm with a mean value of 108 mm. Most parts of the YRB
showed significant increasing trends of Ec, especially the middle regions

6
F. Zhao et al. Agricultural and Forest Meteorology 316 (2022) 108842

Fig. 4. Spatial patterns of key climate variables and LAI and their changing trends during 1982–2016. The dotted areas in the right panel denote the significant
changing areas with P < 0.05. Ta, VPD, Rn, and Wn are the air temperature, vapor pressure deficit, solar radiation, and wind speed, respectively. LAI is the leaf area
index. Dashed black lines in the inserted density plots are the average values.

(the Loess Plateau region). Overall, the spatially averaged Ec showed a stable in other areas. Short vegetation cover (SVC) showed significant
significant increasing trend (P < 0.001) with a rate of 0.85 mm/yr increasing trend in large areas of the middle portion (the Loess Plateau
during 1982–2016. region) (Figure S2a and Figure S2b). The overlapping ratios between
The spatial patterns of Es were roughly opposite to those of Ec. Areas significant changing TC/SVC and ET variation were 16%/30%
with higher Es values were mainly distributed in areas with lower LAI, (Figure S2c). Note that the overlapping ratios indicated the percentage
especially the northern regions of the YRB. The magnitude of Es ranged of areas where significant land cover changes and ET changes occurred.
from 91 to above 335 mm, with a mean value of 264 mm, which is much Based on this ratio and ET trends, we approximately quantified the
higher than that of Ec. Significant increases in Es were mainly observed contributions of changes in land use and climate to ET variations. In
in the headwater regions of the YRB. In contrast to Ec, Es in the middle total, climate change and land use change approximately contributed to
regions (the Loess Plateau region) showed a significant decreasing trend 53.0% and 47.0% variations in ET (Figure S2d), indicating that climate
(Fig. 6f). In total, the time series of Es showed a slightly increasing trend change may play a more important role in driving ET variation during
(P = 0.58 and slope = 0.16 mm/yr) during 1982–2016. 1982–2016. We further attributed ET variations to specific climatic
variations and vegetation greening through the factorial experiment.
The contributions of climatic variables and LAI to ET variations were
3.4. Attributing ET variation to climate variation and vegetation greening estimated by comparing ET between CT and various scenarios (Fig. 7). In
S1, annual ET exhibited a significant increasing trend (slope = 0.73 mm/
The impacts of climate change and land use change were shown in yr, P < 0.01) but the magnitude was lower than that in CT. The
Figure S2. The tree cover (TC) showed significant increasing trends in contribution of air temperature change to annual ET was 0.62 mm/yr
the southern and eastern marginal areas of the YRB and kept relatively

7
F. Zhao et al. Agricultural and Forest Meteorology 316 (2022) 108842

Fig. 5. Time series of anomalies of climate variables and LAI anomaly during 1982–2016. Ta, VPD, Rn, and Wn are the air temperature, vapor pressure deficit, solar
radiation, and wind speed, respectively. LAI is the leaf area index.

(45.6%) (Fig. 7b). Similarly, annual ET showed a significant increasing exchange processes (Jung et al., 2011). Changes in ET may exert pro­
trend (slope = 0.93 mm/yr, P < 0.01) in S2, and the contribution of found influences on water balance and freshwater supply in both natural
changes in vegetation greening to annual ET was 0.43 mm/yr (31.6%) and human systems. In this study, we estimated ET across the YRB using
(Fig. 7d). Without the effects of VPD, annual ET increased significantly the modified PML model. The average annual ET from 1982 to 2016 was
at a rate of 1.08 mm/yr (P < 0.01) during 1982–2016. The absolute found to be 387 mm. Ec, Es, and Ei accounted for ~28% (108 mm),
contribution of VPD to annual ET was 0.27 mm/yr (19.9%) (Fig. 7f). ~68% (264 mm), and ~4% (15 mm), respectively, of the total ET
During 1982–2016, Rn and Wn had relatively weak effects on annual ET. variation, suggesting that more water is lost through soil evaporation
In S4 and S5 without Rn and Wn, annual ET increased at rates of 1.34 than through transpiration because most areas in the YRB have sparse
mm/yr (P < 0.01) and 1.43 mm/yr (P < 0.01), respectively. The change vegetation (Liang et al., 2020a). We compared the averaged ET to ET
in Rn exerted a very weak positive effect on ET variation (0.02 mm/yr), values determined in previous studies (Table S1) and found the modeled
whereas the change in Wn exerted a weak negative effect on annual ET ET was reasonable (Jiang et al., 2020; Liu et al., 2006; Pei et al., 2017).
variation (-0.07 mm/yr) (Fig. 7j). The ratio of Ec/ET was also reasonable compared with results derived
The spatial patterns of the contributions of climate and vegetation from global terrestrial ecosystems (0.20–0.65) (Jasechko et al., 2013).
variables to changes in ET during 1982–2016 are shown in Fig. 8. As These consistent results indicate that the PML model can serve as a
shown in Fig. 8a, Ta contributed positively to ET variations across useful tool for estimating ET in the YRB. ET showed a significant
almost the entire YRB (see also the inserted density plot), with negative increasing trend during 1982–2016, with a rate of 1.36 mm/yr (P <
contributions in only small portions in the middle and lower portions of 0.01), which is consistent with the results of previous studies (Feng et al.,
the YRB. The contribution of LAI to ET variation ranged from -4 to 10 2016; Jiang et al., 2020). For example, Jiang et al. (2020) used the
mm/yr, with a general contribution of 0–2 mm/yr. In comparison with Shuttleworth-Wallace model to determine the spatiotemporal vari­
air temperature, LAI contributed negatively to the ET variation in rela­ ability of ET in the YRB and found ET increased significantly at a rate of
tively larger areas, which are mainly distributed in the middle portions 1.72 mm/yr during 1980–2010. The relatively lower rate of increase in
of the YRB (Fig. 8c). The contributions of VPD change to annual ET this study may be attributable to the difference in the study period and
showed similar spatial patterns with Ta. The general contribution was model implemented. Nevertheless, the estimated ET trend was much
0–2 mm/yr and the largest contribution reached up to 10 mm/yr. The larger than the ensemble of all observation-based reconstructions of
changes in Rn and Wn contributed negatively to annual ET in larger global terrestrial ET over similar periods (7.65 ± 1.26 mm/decade),
areas. Negative contributions of Rn and Wn mainly ranged from -2 to suggesting that more water was lost via ET during the past 35 years in
0 mm/yr, whereas positive contributions mainly ranged from 0 to 2 the YRB in comparison with the global predicted amount.
mm/yr.

4.2. Effects of climate and vegetation on ET variation

4. Discussion

The results of this study reveal that increased air temperature

4.1. ET magnitude and change
significantly contributed to increases in ET during 1982–2016 and
vegetation greening played a secondary role. This implies that climate
ET plays a key role in global water cycling and surface energy
change possibly played the most important role in controlling ET during

8
F. Zhao et al. Agricultural and Forest Meteorology 316 (2022) 108842

Fig. 6. Spatiotemporal changes of ET and its key components (Ec and Es) during 1982–2016. The dotted areas in the right panel denote the significant changing areas
with P < 0.05. Ec and Es indicate the transpiration by plant canopy and soil evaporation. Dashed black lines in the inserted density plots are the average values.

the past 35 years, followed by vegetation greening. To frame the find­ to a semi-arid region, where solar radiation is sufficient for ET (Yang
ings, we detected the fluctuation of ET and associated environmental et al., 2019b). Thus, ET variation may be not sensitive to changes in Rn.
factors using their detrended time series under different magnitudes (< The positive effects on ET of increase Wn during recent years (after
25th and > 75th percentiles) (Figure S4). The results showed that the 2007) may be offset by the decreased ones during the past period
fluctuation of ET under both low and high levels were more consistent (1982–2006), leading to an overall negative effect on ET variation.
with climatic variable (especially the air temperature), indicating the In this study, vegetation greening was found to be one of the main
key role of climate in determining ET fluctuation. This result is in line driving factors of annual ET variation in the YRB, which is consistent
with the findings of Li et al. (2021), who reported that climate change with the findings of previous studies (Feng et al., 2016; Li et al., 2020;
(including increased precipitation and air temperature) is the major Shao et al., 2019). Increased leaf area, which is mainly induced by
contributor to increases in ET over the Loess Plateau region. Similar revegetation programs (Chen et al., 2019; Chen et al., 2020), would
results have also been reported for the southern regions of China, which promote plant transpiration and evaporation from intercepted precipi­
receive abundant precipitation (Zhang et al., 2020). On the one hand, tation (Donohue et al., 2010), leading to higher ET. However, in
increased air temperature could enhance both Es and Ec, thus indirectly different climatic zones, vegetation greening may play different roles in
increasing ET. For example, increased air temperature could also elevate influencing ET. In the study area, although LAI is a major contributor to
soil temperature and thus increase soil evaporation directly. On the ET variation, the magnitude of contribution was below that of air tem­
other hand, the increased air temperature can directly increase ET. For perature. Negative contributions of LAI to ET were found in small areas
example, the increased air temperature can elevate VPD and thus in the upper and middle regions (Fig. 8b). In these areas, ET showed
improve the air water demand, which is proven in this study and others slight decreasing trends and LAI showed slight increasing trends. These
(Yukimoto et al., 2019; Zhou et al., 2018). The increased VPD may affect contrary tendencies lead to slight negative contributions of LAI to ET.
stomatal conductance and thus plant transpiration. In addition, the Nevertheless, vegetation greening (increased LAI) significantly
warming induced by increased air temperature may also affect LAI and increased the proportions of Rnc/Rn (proportion of available energy
thus the plant phenology, which will then change the energy and water absorbed by canopy to the total solar radiation) and Ec/ET (Fig. 9). It is
balance at both the ecosystem and canopy levels (Yuan et al., 2020). noteworthy that the increased LAI dominated the trend of Ec/ET during
Jeong et al. (2009) reported that warming-induced earlier spring the study period (Fig. 9b). Moreover, changes in annual Ec/ET induced
phenology could largely increase ET across eastern Asia. Among climate by LAI were positive and mostly larger than those induced by climate
variables, Rn plays a minor role in driving ET variation. The YRB belongs change (Fig. 9c), which was also consistent with previous reports (Niu

9
F. Zhao et al. Agricultural and Forest Meteorology 316 (2022) 108842

Fig. 7. Attribution of ET changes. The left panel indicates the inter-annual variations of ET under CT and different scenarios and the right panel indicates the net
effects of climatic variables and vegetation on ET. ETTa, ETLAI, ETVPD, ETRn, and ETWn are same as those in Table 3.

et al., 2019). In summary, vegetation greening played an important role 4.3. Implications for vegetation restoration
in driving ET variation over the YRB and significantly altered the pro­
portions of ET components during the study period. However, it should The significant increase in ET and its associated driving factors may
be noted that different attribution results may occur when using have implications for vegetation restoration in the YRB, especially in the
different estimation methods. For example, using another PM-type context of climate change. First, the increase in ET may further reduce
model, Liang et al. (2020a) found that the vegetation change played a water availability, which may exert negative effects on the growth of
more important role in driving ET changes than that of climate change vegetation (Murray-Tortarolo et al., 2016). More importantly, the
during 1982–2012 in the Loess Plateau region. Comparisons of the dif­ newly-planted vegetation under the ‘grain for green’ program would
ferences of different methods (such as process-based hydrological model require more water than the native ones (Feng et al., 2016), leading to
and machine learning) in estimating ET can be a good topic in future more water consumption via ET. Thus, wide revegetation may lead to
study (Sun et al., 2008). sharp increases in ET and decreased water availability, which may
further influence the survival rate of plants. Considering this possibility,
further caution should be taken when implementing such programs in

10
F. Zhao et al. Agricultural and Forest Meteorology 316 (2022) 108842

Fig. 8. Difference between ET trends estimated in CT and different scenarios. ‘→’ indicates the contributions of climate and vegetation to ET variation during the
study period. Ta, VPD, Rn, and Wn are the air temperature, vapor pressure deficit, solar radiation, and wind speed, respectively. LAI is the leaf area index. Dashed
black lines in the inserted density plots are the average values.

Fig. 9. (a) Changes in proportions of available energy for plant canopy to net radiation and transpiration to evapotranspiration, (b) attribution of Ec/ET changes, and
(c) time series of ΔEc/ET induced by climate and vegetation during 1982–2016.

the YRB. Second, the increase in ET may influence vegetation growth by to changes in land surface types (Jeong et al., 2009). Changes in surface
altering the energy dynamics. In general, higher ET implies more latent temperature would influence photosynthesis and thus plant growth
heat and less sensible heat in a given radiation, which may further lead (Yang et al., 2019a). Third, ET variations would also impact vegetation

11
F. Zhao et al. Agricultural and Forest Meteorology 316 (2022) 108842

water use efficiency (WUE = gross primary production (GPP)/ET). For a Declaration of competing interest
given GPP, a higher ET indicates a lower WUE. Further, a lower WUE
may indicate a relatively lower carbon assimilation rate (Cheng et al., The authors declare that they have no known competing financial
2017), which may impact the vegetation growth. interests or personal relationships that could have appeared to influence
the work reported in this paper.

4.4. Limitations and future scope Acknowledgments

It should be noted that this study is limited by some shortcomings. This study was funded by the National Science Foundation of China
First, for vegetation data, we used the popular GIMMS LAI to estimate ET (31961143011 and 42101029), the China Postdoctoral Science Foun­
in the YRB (Zhu et al., 2016). However, although the GIMMS LAI has a dation (2020M683451), the Strategic Priority Research Program of the
relatively longer temporal coverage, it has a relatively coarse resolution. Chinese Academy of Sciences (XDB40020205), the Shaanxi Major
Different sources of LAI, such as GLOBALMAP and GLASS, may lead to Theoretical and Practical Program (20ST-106), the Key Laboratory of
different ET trends and attributions. Second, in estimating ET, only Eco-Environment and Meteorology for the Qinling Mountains and Loess
climate and vegetation factors were considered in the PML model Plateau, Shaanxi Meteorological Bureau (2021K-6), the Technology
without explicitly accounting for anthropogenic effects, such as reser­ Innovation Center for Land Engineering and Human Settlements,
voir operation and human water withdrawals, which may lead to un­ Shaanxi Land Engineering Construction Group Co., Ltd and Xi’an Jiao­
certainties in the estimations (Bai et al., 2020). Third, in attributing ET tong University (201912131-B2), the ‘Light of the West’ talent program
variation, we assumed that the influencing factors were independent of of the Chinese Academy of Science, the Key Research and Development
each other. In fact, these factors may interact with each other. For Project in Shaanxi Province (S2020-YF-GHZD-0061), and the Innovation
example, increased air temperature may increase VPD and promote Team of Shaanxi Province (2021TD-52). We also thank the HPCC Plat­
vegetation growth (Piao et al., 2015), and vegetation growth may in­ form in Xi’an Jiaotong University for computing equipment and com­
fluence wind speed and precipitation. The attributions of VPD to ET puter maintenance. The GIMMS LAI 3g were from https://drive.google.
trend may thus include fingerprint of air temperature. Finally, other com/open?id=0BwL88nwumpqYaFJmR2poS0d1ZDQ, the CMFD
factors that were not considered in this study, such as increased carbon climate data were from http://westdc.westgis.ac.cn/zh-hans/data/7a3
dioxide concentration, variations in precipitation, soil moisture condi­ 5329c-c53f-4267-aa07-e0037d913a21/, and the GLEAM data were
tions, and water availability (Aguilos et al., 2021) might also have from https://www.gleam.eu/.
contributed to the ET variation during the study period. For example, in
some individual years (e.g., 2001-2003), where the air temperature and
Supplementary materials
LAI kept at a relative low levels, the ET was in high level due to other
influencing factors. Future studies should consider more influencing
Supplementary material associated with this article can be found, in
factors in the estimation of ET to gain a more comprehensive conclusion.
the online version, at doi:10.1016/j.ejps.2020.105216.
The flexible attribution framework used in this study has potential to be
used in other places, where climate and vegetation has significantly
changed during the study period. The method is constructed based on References
limited climate and vegetation variables and is easy to use. Once the
Aguilos, M, Sun, G, Noormets, A, Domec, J-C, McNulty, S, Gavazzi, M, 2021. Effects of
modeling results are verified based on multi-source data, people can use land-use change and drought on decadal evapotranspiration and water balance of
this attribution method to identify the roles of climate and vegetation on natural and managed forested wetlands along the southeastern US lower coastal
ET trend. Our modeling results may also alert that the widespread plain. Agricultural and Forest Meteorology 303, 108381.
Allen, RG, Pereira, LS, Raes, D, Smith, M, 1998. Crop evapotranspiration-guidelines for
greening in China during past decades may elevate water consumption computing crop water requirements-FAO irrigation and drainage paper 56, 300.
via ET (Bai et al., 2020; Chen et al., 2019). FAO, Rome, p. 6541.
Arnold, JG, Srinivasn, R, Muttiah, RS, Willians, JR, 1998. Large area hydrologic
modeling and assessment - part I: model development. Journal Of The American
5. Conclusions Water Resources Association 34 (1), 73–89.
Bai, P, Liu, X, Zhang, Y, Liu, C, 2020. Assessing the Impacts of Vegetation Greenness
Change on Evapotranspiration and Water Yield in China. Water Resources Research
In this study, the PML model was modified simply by incorporating
56 (10), e2019WR027019.
land cover change to improve ET modeling. The modified PML model Bastiaanssen, WGM, Menenti, M, Feddes, RA, Holtslag, AAM, 1998. A remote sensing
was then used to estimate the actual ET variation across the YRB during surface energy balance algorithm for land (SEBAL). 1. Formulation. Journal of
1982–2016. The driving factors of annual ET variation were quantified Hydrology 212-213, 198–212.
Chen, C, Park, T, Wang, X, Piao, S, Xu, B, Chaturvedi, RK, 2019. China and India lead in
through factorial experiments. Model tests including on-site verification, greening of the world through land-use management. Nat Sustain 2, 122–129.
comparison with remote-sensing data, and verification against water Chen, Y, Chen, L, Cheng, Y, Ju, W, Chen, HYH, Ruan, H, 2020. Afforestation promotes
balance-based ET indicated that the modified PML could simulate ET the enhancement of forest LAI and NPP in China. Forest Ecology and Management
462.
across the YRB reasonably well. The results showed that annual ET Chen, Y, Wang, K, Lin, Y, Shi, W, Song, Y, He, X, 2015. Balancing green and grain trade.
significantly increased at a rate of 1.36 mm/yr (P < 0.001) during the Nature Geoscience 8, 739–741.
past 35 years. Increased air temperature, vegetation greening, and Cheng, L, Zhang, L, Wang, YP, Canadell, JG, Chiew, FHS, Beringer, J, 2017. Recent
increases in terrestrial carbon uptake at little cost to the water cycle. Nat Commun 8
increased VPD were found to be the major contributors to the increased (1), 110.
ET, accounting for 45.6%, 31.6%, and 19.9%, respectively. Changes in Dembélé, M, Ceperley, N, Zwart, SJ, Salvadore, E, Mariethoz, G, Schaefli, B, 2020.
Rn and Wn played minor roles in contributing to ET variation. Impor­ Potential of satellite and reanalysis evaporation datasets for hydrological modelling
under various model calibration strategies. Advances in Water Resources 143,
tantly, vegetation greening (increased LAI) significantly changed ET 103667.
components, accounting for 92.3% of the increase in the proportion of Donohue, RJ, McVicar, TR, Roderick, ML, 2010. Assessing the ability of potential
transpiration to ET. This study provides a comprehensive assessment of evaporation formulations to capture the dynamics in evaporative demand within a
changing climate. Journal of Hydrology 386 (1-4), 186–197.
the effects of climate and vegetation on the annual ET variation in the
Du, M, Zhang, J, Wang, Y, Liu, H, Wang, Z, Liu, C, 2021. Evaluating the contribution of
YRB during the past 35 years. On the whole, climate and vegetation different environmental drivers to changes in evapotranspiration and soil moisture, a
greening contributed significantly to the annual ET variation. Therefore, case study of the Wudaogou Experimental Station. Journal of Contaminant
future revegetation programs should pay more attention to the effects of Hydrology 243, 103912.
Fang, Y, Sun, G, Caldwell, P, McNulty, SG, Noormets, A, Domec, J-C, 2016. Monthly land
climate change and vegetation greening on water resources to ensure cover-specific evapotranspiration models derived from global eddy flux
sustainable growth of ecosystems in the YRB. measurements and remote sensing data. Ecohydrology 9 (2), 248–266.

12
F. Zhao et al. Agricultural and Forest Meteorology 316 (2022) 108842

Feng, X, Fu, B, Piao, S, Wang, S, Ciais, P, Zeng, Z, 2016. Revegetation in China’s Loess Niu, Z, He, H, Zhu, G, Ren, X, Zhang, L, Zhang, K, 2019. An increasing trend in the ratio
Plateau is approaching sustainable water resource limits. Nature Climate Change 6 of transpiration to total terrestrial evapotranspiration in China from 1982 to 2015
(11), 1019–1022. caused by greening and warming. Agricultural and Forest Meteorology 279.
Hadiwijaya, B, Isabelle, P-E, Nadeau, DF, Pepin, S, 2021. Can a physically-based land Pascolini-Campbell, M, Reager, JT, Chandanpurkar, HA, Rodell, M, 2021. A 10 per cent
surface model accurately represent evapotranspiration partitioning? A case study in increase in global land evapotranspiration from 2003 to 2019. Nature 593 (7860),
a humid boreal forest. Agricultural and Forest Meteorology 304-305. 543–547.
Huang, N, Wang, L, Song, X-P, Black, TA, Jassal Rachhpal, S, Myneni Ranga, B, 2020. Pei, T, Wu, X, Li, X, Zhang, Y, Shi, F, Ma, Y, 2017. Seasonal divergence in the sensitivity
Spatial and temporal variations in global soil respiration and their relationships with of evapotranspiration to climate and vegetation growth in the Yellow River Basin,
climate and land cover. Science Advances 6 (41), eabb8508. China. Journal of Geophysical Research: Biogeosciences 122 (1), 103–118.
Jasechko, S, Sharp, ZD, Gibson, JJ, Birks, SJ, Yi, Y, Fawcett, PJ, 2013. Terrestrial water Penman, HL, 1948. Natural evaporation from open water, bare soil and grass.
fluxes dominated by transpiration. Nature 496 (7445), 347–350. Proceedings of the Royal Society of London Series A Mathematical and Physical
Jeong, S-J, Ho, C-H, Jeong, J-H, 2009. Increase in vegetation greenness and decrease in Sciences 193 (1032), 120.
springtime warming over east Asia. Geophysical Research Letters 36 (2) n/a-n/a. Piao, S, Yin, G, Tan, J, Cheng, L, Huang, M, Li, Y, 2015. Detection and attribution of
Jiang, ZY, Yang, ZG, Zhang, SY, Liao, CM, Hu, ZM, Cao, RC, 2020. Revealing the spatio- vegetation greening trend in China over the last 30 years. Global Change Biology 21
temporal variability of evapotranspiration and its components based on an improved (4), 1601–1609.
Shuttleworth-Wallace model in the Yellow River Basin. J Environ Manage 262, Priestley, C, Taylor, R, 1972. On the Assessment of Surface Heat Flux and Evaporation
110310. Using Large Scale Parameters. mwr 100, 81–92.
Jung, M, Reichstein, M, Ciais, P, Seneviratne, SI, Sheffield, J, Goulden, ML, 2010. Recent Savita A, Marsland S, Dix M, Bi D, Dobrohotoff P, Fiedler R, 2019. CSIRO-ARCCSS
decline in the global land evapotranspiration trend due to limited moisture supply. ACCESS-CM2 model output prepared for CMIP6 FAFMIP.
Nature 467 (7318), 951–954. Shao, R, Zhang, B, He, X, Su, T, Li, Y, Long, B, 2021. Historical Water Storage Changes
Jung, M, Reichstein, M, Margolis, HA, Cescatti, A, Richardson, AD, Arain, MA, 2011. Over China’s Loess Plateau. Water Resources Research 57 (3), e2020WR028661.
Global patterns of land-atmosphere fluxes of carbon dioxide, latent heat, and Shao, R, Zhang, B, Su, T, Long, B, Cheng, L, Xue, Y, 2019. Estimating the Increase in
sensible heat derived from eddy covariance, satellite, and meteorological Regional Evaporative Water Consumption as a Result of Vegetation Restoration Over
observations. Journal of Geophysical Research 116. the Loess Plateau, China. Journal of Geophysical Research: Atmospheres 124 (22),
Konapala, G, Mishra, AK, Wada, Y, Mann, ME, 2020. Climate change will affect global 11783–11802.
water availability through compounding changes in seasonal precipitation and Song, X-P, Hansen, M, Stehman, S, Potapov, P, Tyukavina, A, Vermote, E, 2018. Global
evaporation. Nat Commun 11 (1), 3044. land change from 1982 to 2016. Nature 560.
Lei, H, Yang, D, 2010. Interannual and seasonal variability in evapotranspiration and Sun, G, Noormets, A, Chen, J, McNulty, SG, 2008. Evapotranspiration estimates from
energy partitioning over an irrigated cropland in the North China Plain. Agricultural eddy covariance towers and hydrologic modeling in managed forests in Northern
and Forest Meteorology 150 (4), 581–589. Wisconsin, USA. Agricultural and Forest Meteorology 148 (2), 257–267.
Leuning, R, Zhang, YQ, Rajaud, A, Cleugh, H, Tu, K, 2008. A simple surface conductance Tian, F, Liu, L-Z, Yang, J-H, Wu, J-J, 2021. Vegetation greening in more than 94% of the
model to estimate regional evaporation using MODIS leaf area index and the Yellow River Basin (YRB) region in China during the 21st century caused jointly by
Penman-Monteith equation. Water Resources Research 44 (10). warming and anthropogenic activities. Ecological Indicators 125.
Li, C, Zhang, Y, Shen, Y, Yu, Q, 2020. Decadal water storage decrease driven by Wang, W, Liao, Y, Wen, X, Guo, Q, 2013. Dynamics of CO2 fluxes and environmental
vegetation changes in the Yellow River Basin. Science Bulletin 65 (22), 1859–1861. responses in the rain-fed winter wheat ecosystem of the Loess Plateau, China.
Li, T, Xia, J, Zhang, L, She, D, Wang, G, Cheng, L, 2021. An improved complementary Science of the Total Environment 461-462, 10–18.
relationship for estimating evapotranspiration attributed to climate change and Wang, W, Zhao, Z, Kang, W, Tian, D, Xiang, W, Yan, W, 2011. Characteristics of latent
revegetation in the Loess Plateau, China. Journal of Hydrology 592. heat flux over Cunninghamia lanceolata plantations in Huitong county. Journal of
Li, X, He, Y, Zeng, Z, Lian, X, Wang, X, Du, M, 2018. Spatiotemporal pattern of terrestrial Central and Southern University of Forrest Technology 31 (05), 192–197.
evapotranspiration in China during the past thirty years. Agricultural and Forest Wu, J, Miao, C, Wang, Y, Duan, Q, Zhang, X, 2017. Contribution analysis of the long-term
Meteorology 259, 131–140. changes in seasonal runoff on the Loess Plateau, China, using eight Budyko-based
Lian, X, Piao, S, Huntingford, C, Li, Y, Zeng, Z, Wang, X, 2018. Partitioning global land methods. Journal of Hydrology 545, 263–275.
evapotranspiration using CMIP5 models constrained by observations. Nature Xiao, J, Chen, J, Davis, KJ, Reichstein, M, 2012. Advances in upscaling of eddy
Climate Change 8 (7), 640–646. covariance measurements of carbon and water fluxes. Journal of Geophysical
Liang, W, Bai, D, Wang, F, Fu, B, Yan, J, Wang, S, 2015. Quantifying the impacts of Research: Biogeosciences 117 (G1).
climate change and ecological restoration on streamflow changes based on a Budyko Xie, J, Xu, Y-P, Wang, Y, Gu, H, Wang, F, Pan, S, 2019. Influences of climatic variability
hydrologicalmodel in China’s Loess Plateau. Water Resources Research 51, and human activities on terrestrial water storage variations across the Yellow River
6500–6519. basin in the recent decade. Journal of Hydrology 579.
Liang, W, Zhang, W, Jin, Z, Yan, J, Lü, Y, Li, S, 2020a. Rapid Urbanization and Yang K, He J, 2019. (ed National Tibetan Plateau Data C). National Tibetan Plateau Data
Agricultural Intensification Increase Regional Evaporative Water Consumption of the Center.
Loess Plateau. Journal of Geophysical Research: Atmospheres 125 (23). Yang, K, He, J, Tang, W, Qin, J, Cheng, CCK, 2010. On downward shortwave and
Liang, W, Zhang, W, Jin, Z, Yan, J, Lü, Y, Li, S, 2020b. Rapid Urbanization and longwave radiations over high altitude regions: Observation and modeling in the
Agricultural Intensification Increase Regional Evaporative Water Consumption of the Tibetan Plateau. Agricultural and Forest Meteorology 150 (1), 38–46.
Loess Plateau. Journal of Geophysical Research: Atmospheres 125 (23), Yang, X, Yong, B, Ren, L, Zhang, Y, Long, D, 2017. Multi-scale validation of GLEAM
e2020JD033380. evapotranspiration products over China via ChinaFLUX ET measurements.
Liu, J, Liu, M, Tian, H, Zhuang, D, Zhang, Z, Zhang, W, 2005. Spatial and temporal International Journal of Remote Sensing 38 (20), 5688–5709.
patterns of China’s cropland during 1990–2000: An analysis based on Landsat TM Yang, Y, Roderick, ML, Zhang, S, McVicar, TR, Donohue, RJ, 2019a. Hydrologic
data. Remote Sensing of Environment 98 (4), 442–456. implications of vegetation response to elevated CO2 in climate projections. Nature
Liu, L, Gudmundsson, L, Hauser, M, Qin, D, Li, S, Seneviratne, SI, 2020. Soil moisture Climate Change 9 (1), 44–48.
dominates dryness stress on ecosystem production globally. Nat Commun 11 (1), Yang, Z, Zhang, Q, Hao, X, Yue, P, 2019b. Changes in Evapotranspiration Over Global
4892. Semiarid Regions 1984–2013. Journal of Geophysical Research: Atmospheres 124
Liu, S, Sun, R, Sun, Z, Li, X, Liu, C, 2006. Evaluation of three complementary relationship (6), 2946–2963.
approaches for evapotranspiration over the Yellow River basin. Hydrological Yen, H, Su, Y-W, Wolfe, JE, Chen, S-T, Hsu, Y-C, Tseng, W-H, 2015. Assessment of input
Processes 20 (11), 2347–2361. uncertainty by seasonally categorized latent variables using SWAT. Journal of
McVicar, TR, Li, L, Van Niel, TG, Zhang, L, Li, R, Yang, Q, 2007. Developing a decision Hydrology 531, 685–695.
support tool for China’s re-vegetation program: Simulating regional impacts of Yeste, P, García-Valdecasas Ojeda, M, Gámiz-Fortis, SR, Castro-Díez, Y, Esteban-
afforestation on average annual streamflow in the Loess Plateau. Forest Ecology and Parra, MJ, 2020. Integrated sensitivity analysis of a macroscale hydrologic model in
Management 251 (1-2), 65–81. the north of the Iberian Peninsula. Journal of Hydrology 590.
Miralles, DG, De Jeu, RAM, Gash, JH, Holmes, TRH, Dolman, AJ, 2011. Magnitude and Yin, L, Feng, X, Fu, B, Wang, S, Wang, X, Chen, Y, 2021. A coupled human-natural system
variability of land evaporation and its components at the global scale. Hydrology analysis of water yield in the Yellow River basin, China. Sci Total Environ 762,
And Earth System Sciences 15 (3), 967–981. 143141.
Miralles, DG, Gash, JH, Holmes, TRH, de Jeu, RAM, Dolman, AJ, 2010. Global canopy Yu, G, Wen, X, Sun, X, Tanner, BD, Lee, X, Chen, J, 2006. Overview of ChinaFLUX and
interception from satellite observations. Journal of Geophysical Research: evaluation of its eddy covariance measurement. Agricultural and Forest Meteorology
Atmospheres 115 (D16). 137 (3-4), 125–137.
Montaldo, N, Oren, R, 2018. Changing Seasonal Rainfall Distribution With Climate Yuan, M, Zhao, L, Lin, A, Li, Q, She, D, Qu, S, 2020. How do climatic and non-climatic
Directs Contrasting Impacts at Evapotranspiration and Water Yield in the Western factors contribute to the dynamics of vegetation autumn phenology in the Yellow
Mediterranean Region. Earth’s Future. River Basin, China? Ecological Indicators 112, 106112.
Monteith, J, 1965. Evaporation and environment. Paper Presented at Symp Soc Exp Biol. Yukimoto S, Koshiro T, Kawai H, Oshima N, Yoshida K, Urakawa S, 2019. MRI MRI-
Moriasi, DN, Arnold, JG, Van Liew, MW, Bingner, RL, Harmel, RD, Veith, TL, 2007. ESM2.0 model output prepared for CMIP6 CMIP.
Model evaluation guidelines for systematic quantification of accuracy in watershed Zeng, Z, Piao, S, Lin, X, Yin, G, Peng, S, Ciais, P, 2012. Global evapotranspiration over the
simulations. Transactions of the ASABE 50 (3), 885–900. past three decades: estimation based on the water balance equation combined with
Murray-Tortarolo, G, Friedlingstein, P, Sitch, S, Seneviratne, SI, Fletcher, I, Mueller, B, empirical models. Environmental Research Letters 7 (1).
2016. The dry season intensity as a key driver of NPP trends. Geophysical Research Zhang, D, Liu, X, Zhang, L, Zhang, Q, Gan, R, Li, X, 2020. Attribution of
Letters 43 (6), 2632–2639. Evapotranspiration Changes in Humid Regions of China from 1982 to 2016. Journal
of Geophysical Research: Atmospheres 125 (13).

13
F. Zhao et al. Agricultural and Forest Meteorology 316 (2022) 108842

Zhang, L, Dawes, WR, Walker, GR, 2001. Response of mean annual evapotranspiration to NOAA-AVHRR NDVI data to feed a distributed hydrological model over the Mekong
vegetation changes at catchment scale. Water Resources Research 37 (3), 701–708. River basin. Journal of Hydrology 327 (1-2), 151–173.
Zhang, Y, Leuning, R, Hutley, LB, Beringer, J, McHugh, I, Walker, JP, 2010. Using long- Zhou, S, Yu, B, Zhang, Y, Huang, Y, Wang, G, 2018. Water use efficiency and
term water balances to parameterize surface conductances and calculate evaporation evapotranspiration partitioning for three typical ecosystems in the Heihe River
at 0.05◦ spatial resolution. Water Resources Research 46 (5). Basin, northwestern China. Agricultural and Forest Meteorology 253-254, 261–273.
Zhao, F, Wu, Y, Qiu, L, Bellie, S, Zhang, F, Sun, Y, 2018a. Spatiotemporal features of the Zhu, Z, Bi, J, Pan, Y, Ganguly, S, Anav, A, Xu, L, 2013. Global Data Sets of Vegetation
hydro-biogeochemical cycles in a typical loess gully watershed. Ecological Indicators Leaf Area Index (LAI)3g and Fraction of Photosynthetically Active Radiation (FPAR)
91, 542–554. 3g Derived from Global Inventory Modeling and Mapping Studies (GIMMS)
Zhao, F, Wu, Y, Qiu, L, Sun, Y, Sun, L, Li, Q, 2018b. Parameter uncertainty analysis of the Normalized Difference Vegetation Index (NDVI3g) for the Period 1981 to 2011.
SWAT model in a mountain-loess transitional watershed on the Chinese Loess Remote Sensing 5 (2), 927–948.
Plateau. Water 10 (6), 690. Zhu, Z, Piao, S, Myneni, RB, Huang, M, Zeng, Z, Canadell, JG, 2016. Greening of the
Zhao, F, Wu, Y, Yao, Y, Sun, K, Zhang, X, Winowiecki, L, 2020. Predicting the climate Earth and its drivers. Nature Climate Change 6 (8), 791–795.
change impacts on water-carbon coupling cycles for a loess hilly-gully watershed. Zuo, D, Xu, Z, Yang, H, Liu, X, 2012. Spatiotemporal variations and abrupt changes of
Journal of Hydrology 581, 124388. potential evapotranspiration and its sensitivity to key meteorological variables in the
Zhou, MC, Ishidaira, H, Hapuarachchi, HP, Magome, J, Kiem, AS, Takeuchi, K, 2006. Wei River basin, China. Hydrological Processes 26 (8), 1149–1160.
Estimating potential evapotranspiration using Shuttleworth–Wallace model and

14