Journal of Hydrology 590 (2020) 125449

Contents lists available at ScienceDirect

Journal of Hydrology
journal homepage: www.elsevier.com/locate/jhydrol

Research papers

Trade-off between watershed water yield and ecosystem productivity along T

elevation gradients on a complex terrain in southwestern China
Ning Liua,b,c,d, Pengsen Sunb, , Peter V. Caldwellc, Richard Harperd, Shirong Liub, Ge Sune
⁎

a
Department of Forest Resources, University of Minnesota, Saint Paul, MN, USA
b
Research Institute of Forest Ecology, Environment and Protection, Chinese Academy of Forestry, Beijing, China
c
Coweeta Hydrologic Lab, Southern Research Station, U.S. Department of Agriculture Forest Service, Otto, NC, USA
d
College of Science, Health, Engineering and Education, Murdoch University, South Street, Murdoch, WA, Australia
e
Eastern Forest Environmental Threat Assessment Center, Southern Research Station, U.S. Department of Agriculture Forest Service, Research Triangle Park, NC, USA

ARTICLE INFO ABSTRACT

This manuscript was handled by Emmanouil Understanding the tradeoffs between water yield and ecosystem productivity is important for developing stra-
Anagnostou, Editor-in-Chief tegies for large scale ecological restoration worldwide. This study focused on a national forest protection project
in the Upper Yangtze River Basin where a logging ban was implemented in 1998. We used a hydrologic model
Keywords:
Elevation gradient and remote sensing data to study the interactions between water and carbon cycles along elevation gradients in
Water use efficiency the Minjiang watershed (MJ), where extensive deforestation and reforestation have occurred in the past seven
Water yield decades. Average annual evapotranspiration (ET), water yield, and gross primary productivity (GPP) from 2000
Carbon sequestration to 2015 were estimated as 429 mm yr−1, 555 mm yr−1, and 1002 g C m−2 yr−1, respectively. ET decreased
Trade-off sharply and consistently with increasing elevation, whereas GPP only decreased significantly in high elevation
areas (i.e., > 3,000 m), resulting in divergent trends of water use efficiency (WUE) with elevation. Evergreen
needleleaf forests (ENF) contributed 28% of water yield and 37% of GPP at the watershed scale, while grassland
(GRA) also contributed 28% of water yield, but only 20% of total watershed GPP. Moreover, runoff coefficients
showed strong negative correlations with GPP, suggesting a general trade-off relationship between water yield
and ecosystem productivity in MJ. Our results suggest that vegetation composition and elevation played a key
role in determining the relative ecological benefits for carbon and water in the study watershed with a complex
terrain.

1. Introduction 2019) and ‘Green for Grain’ project (Lü et al., 2012).
Thanks to global eddy flux measurements of carbon and water, the
There has been increasing interest in understanding interactions coupling of carbon (i.e., ecosystem productivity) and water has been
between water and vegetation at various scales in recent years (Jackson well established generally (Law et al., 2002; Sun et al., 2011), however
et al., 2005; Sun et al., 2011) under a background of dramatic global few studies have examined how water yield and ecosystem productivity
climate changes (Vasudevan et al., 2010). Carbon mitigation has been interact at the watershed scale in complex high elevation mountainous
proposed to moderate climate change by increasing the capacity of regions. Elevation is known to play a critical role in vegetation dis-
carbon sinks or reducing anthropogenic carbon emissions (IPCC, 2013). tribution and growth by affecting the water and solar energy supply
Afforestation and forest restoration are two common ways to achieve (Liu et al., 2017; Swetnam et al., 2017; Zhou et al., 2019). For example,
carbon mitigation and reduce soil erosion. However, trade-offs between leaf area and total live biomass of forests vary with increasing elevation
water yield and carbon sequestration have been recognized worldwide along the Qinghai-Tibet Plateau regions (Liu et al., 2017; Wang et al.,
(Jackson et al., 2005; Sun et al., 2017) with implications for human 2014). These differences in vegetation and climate result in variation in
water supply. For example, land cover change contributed to half of the watershed water balances along elevation gradients. However, patterns
observed change in streamflow, globally (Wei et al., 2018). Trade-offs of water and carbon balances and the factors that determined their
among water, carbon, soil erosion and crop production have been patterns along elevation gradients are not consistent. For example,
documented in China for several large-scale ecological restoration while Hu et al. (2018) reported decreasing evapotranspiration (ET)
projects, including the Three-North Shelterbelt Project (Deng et al., with increasing elevation at three forested sites in Mount Gongga, Ma

⁎
Corresponding author.
E-mail address: sunpsen@caf.ac.cn (P. Sun).

https://doi.org/10.1016/j.jhydrol.2020.125449
Received 10 April 2020; Received in revised form 16 August 2020; Accepted 20 August 2020
Available online 27 August 2020
0022-1694/ © 2020 Elsevier B.V. All rights reserved.
N. Liu, et al. Journal of Hydrology 590 (2020) 125449

et al. (2019) found that ET increased with increasing elevation at lower The major soil type in MJ is mountainous, brown, coniferous forest
elevations and then decreased with increasing elevation at higher ele- soil originating from limestone, phyllite, and basalt (Cui et al., 2012).
vations in the northeast of the Qinghai-Tibet Plateau. Temperature This soil is classified as silt-loam or loam with the United States De-
changes along elevation gradients was found to be the main contributor partment of Agriculture (USDA) soil texture triangle (Shirazi and
to these variations in ET and carbon sequestration in subalpine dark Boersma, 1984), which tends to have high runoff potential (Ross et al.,
coniferous forests in the southeastern Tibetan Plateau (Sun et al., 2018). The soil depth ranges from 60 to 100 cm at high elevation to
2020a, 2020b), while the dominant controls for ET switched from water 160–200 cm at low elevation (Dai et al., 2013).
to energy (shortwave radiation and air temperature) with increasing The main vegetation types of MJ are closed shrublands (CSH),
elevation in the northeastern Qinghai-Tibet Plateau (Ma et al., 2019). grassland (GRA) mainly composed of alpine meadow, broadleaf and
Clearly the patterns of water and carbon balances along elevation needleleaf mixed forest (MF), and evergreen needle leaf forest (ENF).
gradients and the factors determining their patterns for different ve- Today, 40.3% of the watershed is covered by forest, mostly ENF
getation types are complex and a comprehensive understanding of these (29.7%) and MF (6.9%). CSH and GRA account for 24.5% and 25.1% of
interactions will be essential for effective environmental conservation the watershed, respectively. Ten percent of the watershed is classified
projects. as barren/built-up lands (BSV) or permanent snow/ice (SNO). Most of
The 24 000 km2 Minjiang river watershed (MJ) located in the the vegetation in the watershed is distributed between 2500 and
subalpine zone of the southeast edge of Qinghai-Tibet Plateau plays a 4200 m elevation where water and energy availability are suitable for
strategic role in environmental protection as well as the economic and their growth.
social well-being for downstream communities. The MJ experienced
dramatic deforestation, with forest cover decreasing from 38.5% of the 2.2. Hydrologic balance modeling
watershed in the 1950s to 23% in the 1980s (Cui et al., 2012). A series
of reforestation and forest protection programs for soil conservation ET and water yield for each vegetation type were estimated by the
and water resource protection were launched in 1998, resulting in an monthly Water Supply and Stress Index (WaSSI) hydrologic model (Sun
increase in forest coverage to 34% by 2006 (Cui et al., 2012). Although et al., 2011). WaSSI has been broadly used across the conterminous
forest harvesting offset the decline in annual streamflow of the Minjiang United States for climate and water assessment (Caldwell et al., 2015; Li
River due to climate change before 2000 (Zhang et al., 2012), stream- et al., 2020). The core of the WaSSI model is an empirical monthly ET
flow has decreased significantly during the last two decades (Huang model derived from a data set of ecosystem‐level ET measurements
et al., 2014) due to the combined effects of climate change and forest based on eddy covariance or sapflow techniques, climate measure-
restoration. The large scale reforestation that replaced the native fir ments, and remotely sensed leaf area index (LAI) (Sun et al., 2011) (Eq.
(Abies faxoniana) species with spruce (Picea asperata) coupled with a (1)).
lack of management (e.g. thinning or pruning) resulted in a two-fold
ET = 0.174 P + 0.502 PET + 5.31 LAI + 0.0222 PET LAI
increase in ET as the spruce plantations are now in their fast-growing
stage (Zhang et al., 2020). Moreover, rising air temperatures have re- (1)
sulted in increased vegetation greening and ET (Sun et al., 2008). Un- R2 = 0.86, p < 0.0001, RMSE = 14.0 mm/month. Where PET is
derstanding these changes in water and carbon balances under land use potential evapotranspiration calculated by Hamon’s method (Hamon,
and climate change will be critical for local water resource and forest 1963); LAI is mean leaf area index, and P is precipitation.
management in the MJ watershed. The WaSSI model compares ET demand to soil water storage using
The complex terrain and diverse vegetation types in MJ provide an the Sacramento Soil Moisture Accounting model (SAC-SMA) (Burnash,
ideal site to study the interactions between water and carbon along 1995) and limits ET if soil water is insufficient to meet the demand. The
elevation gradients. We hypothesized that patterns of water and carbon basic simulation unit of the WaSSI model is a subcatchment, in which
balances and the factors that explained these patterns varied among each land cover type is simulated individually. Precipitation is parti-
vegetation types according to elevation gradients. Our objectives were tioned into effective rainfall and snow using an air temperature-based
to: (1) quantify the contribution of each vegetation type to water yield conceptual snow accumulation and melt model (McCabe and Wolock,
and ecosystem productivity in a watershed with complex terrain, (2) 1999). Then effective rainfall (precipitation + snow melt) is further
examine the sensitivity of different vegetation types to elevation with separated into actual ET and water yield (surface runoff and baseflow)
respect to their ecosystem water and carbon processes (including ET, based on the ET demand and soil water storage in the SAC-SMA model.
water yield, gross primary productivity (GPP) and water use efficiency The output of each land cover in each subcatchment is aggregated to
(WUE)), and (3) link vegetation composition to the relationship be- the subcatchment and watershed scale using an area-weighted aver-
tween water yield and carbon sequestration at the watershed scale. aging scheme.
Model inputs include monthly precipitation temperature, and LAI,
2. Study area and methodology land cover, and soil parameters. The land cover dataset used in this
study was developed using China’s HJ-1A/B and Landsat dataset. The
2.1. The Minjiang watershed point climate data were retrieved from the Chinese Meteorological Data
Sharing Service System (http://data.cma.cn/) and were interpolated to
The elevation of the Minjiang (MJ) watershed (30.68°N–32.83°N; 0.01° gridded datasets by Anusplin 4.1 using the two-dimensional thin
102.46°E–104.10°E) ranges from 726 m to 6118 m, with permanent plate smoothing splines method (Hutchinson, 1998). The monthly LAI
snow and glaciers in the peaks of some high mountains (Fig. 1). There was aggregated from the Moderate Resolution Imaging Spectro-
are two climate types in MJ, an alpine climate in high elevation radiometer (MODIS) MOD15A2H 8-day LAI product, which was re-
catchments and a subtropical climate in low elevation catchments and processed by Yuan et al. (2011).
downstream arid valleys. The climate of MJ was mainly affected by We used methods similar to Liu et al. (2013a) to calibrate and va-
westerly circulation, southwest and southeast Monsoons, with a mean lidate the performance of the WaSSI model in the MJ watershed.
annual temperature of 3.5 °C (−6.2 – 16.6 °C across the watershed) and Briefly, a priori soil parameters were generated from eight layer soil
mean annual precipitation of 990 mm yr−1 (750–1460 mm yr−1) be- particle-size distribution, soil depth, and other soil property data (Dai
tween 2000 and 2015. Precipitation varies spatially and temporally et al., 2013) using the methods developed in Anderson et al. (2006).
with sharp declines from the southwest to the northeast. The rainy Streamflow measurements at the outlet of the MJ watershed at the
season from May to October accounted for 75–90% of the total annual Zipingpu gauging station (Fig. 1) were divided into a calibration period
precipitation. (1990–2000) and a validation period (2001–2006) to evaluate the

2
N. Liu, et al. Journal of Hydrology 590 (2020) 125449

Fig. 1. The topography and spatial distribution of vegetation types of Minjiang watershed. BSV is barren or build-up area, CSH is closed shrubland, DBF is deciduous
broadleaf forest, EBF is evergreen broadleaf forest, ENF is evergreen needleleaf forest, GRA is grassland (mainly is alpine meadow), MF is broadleaf and needleleaf
mixed forest, and SNO is permanent snow/ice.

performance of the WaSSI model. The 11 soil parameters were adjusted assess model accuracy.
to achieve the best estimate of streamflow in the calibration period, and
were tested in the validation period. The watershed was divided into 2.3. Gross primary productivity (GPP) and water use efficiency (WUE)
537 subcatchments that were 44.7 km2 on average based on an 85 km2
optimal area threshold (Liu et al., 2013b) (Fig. 1). Four statistical in- GPP of each vegetation type in each subcatchment was derived from
dices, including the coefficient of determination (R2), percent bias (%), the MODIS MOD17A3 annual GPP product (Zhao et al., 2005). The
root mean square error (RMSE, mm) and Nash–Sutcliffe model effi- MOD17A3 GPP algorithm can be expressed as:
ciency coefficient (NSE) (Nash and Sutcliffe, 1970) were calculated to
GPP = × FPAR × PAR (2)

3
N. Liu, et al. Journal of Hydrology 590 (2020) 125449

Where GPP is gross primary productivity, ε is light use efficiency, PAR contributed the highest percentage of total watershed carbon seques-
is photosynthetically active radiation, and FPAR is the fraction of PAR tration (37%) and water yield (28%). Although GRA accounted for 25%
absorbed by the vegetation canopy. of the land area, this land cover type contributed 20% of the total
The original 500 m *500 m gridded MODIS annual GPP data was carbon sequestration while contributing the same proportion of water
downloaded from NASA’s Application for Extracting and Exploring yield as ENF (28%). CSH occupied 25% of the land area while con-
Analysis Ready Samples (AρρEEARS) (https://lpdaacsvc.cr.usgs.gov/ tributing 23% of the water yield and 26% of the carbon sequestration.
appeears/). The GPP of each vegetation type in each subcatchment was The remaining forest types (MF, DBF and EBF) contributed to wa-
aggregated from the mean annual MODIS GPP from 2000 to 2015. tershed water supply in proportion to their land coverage fraction
Annual Respiration (Re) data was derived from the difference between (Fig. 4d).
GPP and net primary productivity (NPP), which was derived from Forests generally had much higher ET than nonforest land cover
MODIS MOD17A2 NPP data. WUE of each vegetation type in each types (Fig. 5a). DBF ET was slightly greater than MF and EBF ET among
subcatchment was defined as the ratio of gross primary productivity forest types, the average of the top three forest types was about
(GPP) to WaSSI-predicted ET. 560 mm yr−1. The average ET of ENF, CSH, and GRA was 509, 487, and
429 mm yr−1, respectively (Fig. 5a). As the average annual precipita-
2.4. Trade-off between water yield and GPP at the watershed scale tion was quite similar among vegetation types (Fig. S1), water yield
among land cover types exhibited the opposite pattern as ET, with
The relationship between runoff coefficient and GPP was used to water yield greatest for GRA and lowest for DBF (Fig. 5b).
quantify the trade-off between water yield and carbon sequestration. Similar to ET, forest land GPP was greater than that of nonforest
The ratio of water yield to precipitation (runoff coefficient) was used to lands. EBF was the highest (1428 g C m−2 yr−1), followed by DBF and
quantify the water yield capacity for a subcatchment. The 537 sub- MF (~1360 g C m−2 yr−1) (Fig. 5c). ENF had the lowest GPP (1224 g C
catchments were divided into vegetation composition groups based on m−2 yr−1) among forest lands but higher than CSH (1076 g C m−2
the coverage of each vegetation type to evaluate the impact of vege- yr−1). GRA was the lowest among all vegetation types, with an average
tation composition on water and carbon relationships. The vegetation of 762 g C m−2 yr−1.
composition group of a subcatchment was defined as the dominant The difference between water yield and carbon sequestration among
vegetation type, if one existed. The dominant vegetation type in a vegetation types relates to differences in their water use efficiency. The
subcatchment was assumed to be the vegetation type that com- WUE differed among vegetation types (Fig. 5d) within a range of
prised > 60% of the total land area. If there was no dominant vegeta- 1.75–2.56 g C kg−1 H2O yr−1. The higher-productivity forests generally
tion type in the subcatchment, the subcatchment was assigned to the had higher WUE (~2.4–2.6 g C kg−1 H2O yr−1) than the lower-pro-
Mixed group. There were four vegetation composition groups in MJ, ductivity CSH (~2.2 g C kg−1 H2O yr−1) and GRA (~1.7 g C kg−1 H2O
including CSH dominant, ENF dominant, GRA dominant, and Mixed. yr−1). WUE of EBF was slightly higher than other forest types.

3. Results 3.3. Variation of water and carbon along elevation gradients

3.1. Model evaluation Spearman rank correlation was used to examine relationships be-
tween elevation and key structural, functional, and efficiency para-
The WaSSI model performed well over the calibration period (1990 meters involved in the water and carbon cycles. Water yield was posi-
to 2000) and the validation period (2001–2006) (Figs. 2 and 3). WaSSI tively related to elevation but negatively related to temperature, aridity
captured the general temporal dynamics of water yield at both the index (potential evapotranspiration/precipitation) and vapor pressure
monthly and annual scales, with monthly NSE of 0.78 and 0.72 for deficit (VPD) (Fig. 6). ET was positively related to temperature, aridity
calibration and validation periods, respectively, and annual NSE of 0.73 index and VPD but negatively related to elevation. GPP and Re showed
and 0.72 for calibration and validation periods, respectively (Fig. 3). strong positive correlation with LAI, but slight correlation with eleva-
The monthly and annual simulated water yield was highly correlated to tion and climate variables. WUE had a strong and consistent negative
the observed water yield, with R2 of 0.79 and 0.78, and RMSE of relationship with temperature and aridity index, but was only slightly
15.57 mm and 35.33 mm at the monthly and annual scale during the related to elevation (Fig. 6). Moreover, divergent correlation along
calibration period, respectively. Similarly, R2 was 0.75 and 0.72, and elevation gradients was found for LAI, GPP, Re and WUE (green box in
RMSE was 15.48 mm and 32.72 mm at the monthly and annual scale Fig. 6).
during the validation period, respectively (Fig. 3). Overall, the simu- Breakpoints in the relationship between environmental variables
lated water yield was very close to the observations, with a percent bias and elevation were tested using the “segmented” package in R 3.5 (R
of 0.6%. However, the model overestimated water yield for months Development Core Team, 2018). GPP and LAI had very similar re-
with very high precipitation in 1990, 1995, 2001 and 2003 (Figs. 2a lationships with elevation; these were stable for elevations below
and 3a), but underestimated water yield in 1992 and 1999, which could 3300 m and decreasing significantly (p < 0.001) for elevations above
be related to melting of glacial snow and ice that changed the annual 3300 m (Tables 1 and 2). In contrast, WUE had strong elevation-de-
water balance (Liu et al., 2008). While the model considers seasonal pendent relationships across all elevations, and 3500 m was the
snow melting processes, the permanent glacial snow and ice melting threshold where the relationships reversed from positive (below
was not considered. The averaged ET simulated by WaSSI for the entire 3500 m) to negative (above 3500 m) (Tables 1 and 2).
MJ from 2000 to 2006 was 426 mm yr−1, which was very close to the Although the breakpoints in water and carbon parameters varied
difference between annual precipitation and observed water yield among different vegetation types, the correlations between elevation
(415 mm yr−1). Despite the differences between simulated and ob- and those parameters were quite similar across vegetation types (Fig. 7
served water yield for some wet months, the WaSSI model reasonably and Table 2). ET was consistently and negatively correlated with ele-
estimated water yield for the MJ watershed (Fig. 2b). vation in all vegetation types (Fig. 7a). GPP slightly increased with
increasing elevation at low elevations for most vegetation types, but
3.2. Water and carbon for different vegetation types along elevation significantly decreased with increasing elevation for elevations >
gradients 3000 m (Fig. 7c). LAI of most vegetation types generally showed si-
milar negative relationships at high elevation but positive relationships
Forests accounted for approximately 51% of GPP (Fig. 4c) and 36% at low elevation, except in the case of DBF and EBF (Fig. S2). Across all
of water yield while covering 40% of the watershed (Fig. 4a and d). ENF vegetation types, the relationship between elevation and WUE varied

4
N. Liu, et al. Journal of Hydrology 590 (2020) 125449

Fig. 2. Time series of monthly (a) and annual (b) simulated and observed water yield (mm) of Zipingpu hydrologic station for the calibration period (1990–2000) and
the validation period (2001–2006).

according to their natural distribution pattern (Fig. 7d). For ENF, MF negative correlation with elevation when the elevation was higher than
and CSH that were distributed across both high and low elevation areas, 3600 m.
the increasing trend in WUE with increasing elevation began to de-
crease at around 3500 m (Fig. 7d, Tables 1 and 2). EBF and DBF were 3.4. Impact of vegetation composition on water yield and carbon
mostly distributed at low elevations (< 3500 m) and had a positive sequestration at subcatchment scale
correlation between WUE and elevation. On the contrary, GRA was
distributed mainly at high elevation (> 3000 m) and had a strong Water yield and GPP varied by dominant vegetation type and

Fig. 3. Monthly (a) and annual (b) scatterplots between the simulated and observed water yield of Zipingpu hydrologic station for the calibration period
(1990–2000) and the validation period (2001–2006).

5
N. Liu, et al. Journal of Hydrology 590 (2020) 125449

Fig. 4. Fraction of total watershed area (a),

precipitation (b), gross primary productivity
(c) and water yield (d) for different vegeta-
tion types in each 100-m elevation bin.
Labels on the bar are the percentage of each
vegetation of the total. CSH is closed
shrubland, DBF is deciduous broadleaf
forest, EBF is evergreen broadleaf forest,
ENF is evergreen needleleaf forest, GRA is
grassland, and MF is broadleaf and needle-
leaf mixed forest.

elevation over the 252 mixed vegetation subcatchments (Table 3). GRA type or subcatchment scale, were significantly affected by elevation. As
dominated subcatchments in high elevation areas had the lowest carbon we hypothesized, we found different breakpoints for different vegeta-
sequestration (GPP = 724 ± 140 g C m−2 yr−1) and highest water tion types along elevation gradients. However, similar water and
yield and runoff coefficient (0.61 ± 0.07). On the contrary, ENF carbon relations to elevation were observed among vegetation types.
dominated subcatchments had the highest carbon sequestration Water yield showed a strong positive correlation with elevation, while
(GPP = 1238 ± 100 g C m−2 yr−1) but lowest runoff coefficient GPP was divergently related to elevation between low elevation and
(0.49 ± 0.05). high elevation areas.
Standardized (Z-score) long term average runoff coefficient and GPP
exhibited a strong negative relationship when all subcatchments were
pooled (Fig. 8). Moreover, negative relationships were also seen in the 4.1. Effects of elevation on WUE
four vegetation composition groups independently, despite their dif-
ferences in either slopes or Z-score tendencies. Opposite trends in runoff We found that forest generally had much higher WUE than shrub-
coefficient and GPP suggested a trade-off relationship between water land and grassland along the elevation gradients. This result is con-
yield and carbon sequestration, and Z-score tendency can be used to sistent with results based on ChinaFlux (Li et al., 2018; Xiao et al.,
evaluate the relationship. Given the breakpoints of WUE for all vege- 2013) and global eddy covariance measurements (Sun et al., 2011;
tation types was 3500 m, we used this threshold to study the re- Zhang et al., 2016). For example, the simulated WUE of GRA
lationship between water and carbon in low and high elevation areas. (1.75 ± 0.4 g C kg−1 H2O yr−1) in this study is similar to the observed
There was no significant difference in Z-score tendencies among vege- WUE (1.72 g C kg−1 H2O yr−1) at the closest Haibei Alpine Tibet eddy
tation composition groups in the low elevation areas (elevation < covariance site, while the simulated WUE of DBF (2.42 ± 0.2 g C kg−1
3500 m) (Fig. 9a). However, in high elevation areas (elevation > H2O yr−1) in this study is close to the observed WUE (2.26 g C kg−1
3500 m), most ENF dominant subcatchments had relatively high H2O yr−1) at the nearest Hunan Yueyang eddy covariance site (Xiao
carbon sequestration and low runoff coefficient, while most GRA et al., 2013). Similarly, the simulated WUE of ENF (2.41 ± 0.2 g C
dominant subcatchments showed contrasting low carbon sequestration kg−1 H2O yr−1) is close to the range of observed WUE (2.5–3.1 g C
and high runoff coefficient (Fig. 9b). Mixed and CSH dominant sub- kg−1 H2O yr−1) along the elevation gradient from 2800 m to 3700 m in
catchments generally had either relatively high runoff coefficient or Mountain Gongga (Sun et al., 2020a, 2020b). However, ET observed at
carbon sequestration. Above 3500 m, ENF dominant subcatchments Mount Gongga (700–780 mm yr−1) (Hu et al., 2018), was higher than
shift rightward and downward (Fig. 9b), while no trends were detected the simulated ET (350–750 mm yr−1) in this study. This is because the
in CSH dominant and Mixed groups. mean annual precipitation in Mount Gongga is > 1600 mm yr−1, which
is much higher than MJ, and the observed sites are in a low elevation
area (3300 m).
4. Discussion Vegetation composition and climate conditions along elevation
gradients resulted in divergent trends in WUE between low elevation
The relationships between water and carbon, at either vegetation and high elevation. The breakpoint of WUE along the elevation gradient

6
N. Liu, et al. Journal of Hydrology 590 (2020) 125449

Fig. 5. Mean annual evapotranspiration (ET) (a), water yield (b), gross primary productivity (GPP) (c), and water use efficiency (WUE) (d) of each vegetation type
from 2000 to 2015 in the Minjiang watershed. Labels on the boxes are the mean. CSH is closed shrubland, DBF is deciduous broadleaf forest, EBF is evergreen
broadleaf forest, ENF is evergreen needleleaf forest, GRA is grassland, and MF is broadleaf and needleleaf mixed forest.

at approximately 3500 m was related to the vegetation distribution. In vegetation growth in high elevation areas (Xiao et al., 2013). In addi-
general, most of the forest was distributed at lower elevations, then tion, the high precipitation at high elevation leads to lower VPD which
transitioned from dark conifer forest to shrubland and grassland at further limits GPP (Fig. S2). Lastly, the steep slope and shallow soil
around 3500 m (Cui et al., 2012; Sun et al., 2020a, 2020b). This change associated with higher elevations also constrain the growth of vegeta-
in vegetation composition resulted in a decrease in WUE for higher tion (Zhang et al., 2013).
elevations. On the other hand, changes in climatic conditions with Among forest types, ENF showed the highest breakpoints along
elevation resulted in a gradual change in water and carbon parameters elevation (Table 1), while GRA was most sensitive to the temperature at
for all vegetation types. For elevations lower than the breakpoint of high elevation. The negative correlation between GRA GPP with ele-
WUE (3500 m), the increase in forest WUE was a result of the sig- vation (rho = −0.84) was much stronger than that of ENF
nificant decline in ET because GPP did not change with increasing (rho = −0.45). The break point of GRA GPP (~3600 m) was also
elevation (Table 2). Hu et al. (2018) reported a similar ET decline with higher than ENF (~3300 m). The different sensitivities and change
a rate of 9.72 mm per 100 m along an elevation gradient in Mountain points among vegetation types suggest potential dynamics in bound-
Gongga. In contrast to the northeast of Qinghai-Tibet where water aries between vegetation types under future climate change (Gao et al.,
availability limits ET in low elevation (Ma et al., 2019), we found ET 2019) with implications for the water balance in MJ.
was more strongly related to temperature than precipitation along
elevation gradients (Fig. 6), which is similar to the result of the Zhang 4.2. Trade-off between water yield and carbon sequestration
et al. (2020). In addition, ET for elevations between 2500 and 3500 m
was not water or energy limited and thus was suitable for forest growth Vegetation composition determined the water use efficiency at the
(Zhang et al., 2013). Therefore LAI, Re, and GPP of most of those ve- subcatchment scale and thus defined the relationship between water
getation types were not related to the elevation change in low and yield and carbon sequestration in MJ. The low GRA ET resulted in a
moderate elevation areas (Table 2 and Fig. S2). However, GPP declined high runoff coefficient but low GPP in GRA dominated subcatchments,
along the elevation gradient in high elevation areas (> 3500 m), while the high ENF ET led to the low runoff coefficient but high GPP in
leading to the dramatic decrease in WUE (Table 2). Although solar ENF dominated subcatchments. In general, the subcatchment benefit of
radiation was higher in high elevation area than lower elevation water yield or carbon sequestration depends on the proportion of forest
(Fig. 6), low temperature becomes the main constraining factor for coverage, i.e. higher forest coverage leads to higher carbon

7
N. Liu, et al. Journal of Hydrology 590 (2020) 125449

Fig. 6. Correlation matrix between eleva-

tion and climate variables (precipitation (P),
temperature (T), Solar radiation (Srd),
Aridity index (AI), vapor pressure deficit
(VPD)), leaf area index (LAI), water yield
(Q), and evapotranspiration (ET), gross pri-
mary productivity (GPP), respiration (Re)
and water use efficiency (WUE) for all ve-
getation types. In the upper-right panel, pies
reflect the Spearman’s correlation coeffi-
cient (rho); In the lower-left panel, ellipse
shows the distribution of the points with
fitted smooth lines. The green box highlights
some variables with two directions of the
relationship of them along elevation gra-
dients. (For interpretation of the references
to colour in this figure legend, the reader is
referred to the web version of this article.)

Table 1 Table 2
Breakpoint of elevation for gross primary productivity (GPP), respiration (Re), The Spearman rank correlation tests of elevation against evapotranspiration
water yield, and water use efficiency (WUE) across different vegetation types. (ET), gross primary productivity (GPP), respiration (Re), water yield, and water
use efficiency (WUE) for different elevation ranges (All – whole range, low –
Variables Breakpoint (m)
lower than breakpoint, high – higher than breakpoint in Table 1).
All MF ENF CSH GRA Variables Elevation No of rho
range points
LAI 3138 2932 3306 3312 3526 All MF ENF CSH GRA
GPP 3276 2993 3317 3324 3591
Re 3406 3153 3477 3407 3519 ET all 1941 −0.9 −0.85 −0.91 −0.9 −0.71
Water yield 3617 3028 3205 3267 3300 GPP all 1941 −0.79 −0.57 −0.61 −0.67 −0.85
WUE 3500 3197 3498 3372 3639 lower 889 – – – – –
higher 1052 −0.88 −0.67 −0.45 −0.69 −0.84
All is all vegetation types. CSH is closed shrubland, ENF is evergreen needleleaf Re all 1941 −0.52 −0.22 – −0.75 0.25
forest, GRA is grassland, and MF is broadleaf and needleleaf mixed forest. lower 1034 0.2 0.25 0.3 0.34 0.41
higher 907 −0.79 −0.49 −0.45 −0.77 −0.28
Water yield all 1941 0.54 0.23 0.33 0.48 0.69
sequestration but lower water yield. This also can be verified by paired
lower 1300 0.31 – 0.36 0.35 –
catchments experiments where it has been shown that deforestation higher 641 0.5 0.38 – 0.29 0.63
leads to an increase in water yield while reforestation or plantation WUE all 1941 −0.35 0.5 0.5 – −0.77
results in a decrease in water yield (Zhang et al., 2017). Interestingly, lower 1166 0.47 0.58 0.74 0.61 0.39
the diverse mixed vegetation type showed not only balanced ecological higher 775 −0.84 −0.43 −0.45 −0.5 −0.8

benefits for water and carbon (Fig. 8), but was also stable with respect
Note: rho is Spearman’s correlation coefficient and only showed variables with
to elevation from below 3500 m to above 3500 m (Fig. 9). This might be the significance (p) of the Spearman test is < 0.05. All is all vegetation types.
related to the varied water use efficiency of different vegetation types. CSH is closed shrubland, ENF is evergreen needleleaf forest, GRA is grassland,
In addition, water generally flows from low water consumption vege- and MF is Broadleaf and needleleaf mixed forest.
tation (e.g. GRA and CSH) at the ridge to high water use vegetation in
the valley (forest types) (Fan et al., 2019). Therefore, managing the
percentage of forest coverage and their distribution in a watershed

8
N. Liu, et al. Journal of Hydrology 590 (2020) 125449

Fig. 7. Relationships between elevation and

evapotranspiration (a), water yield (b),
gross primary productivity (c), and water
use efficiency (d) for different vegetation
types. Those smooth lines for each vegeta-
tion type are plotted using “loess” method in
R 3.5, with confidence interval as the shade.
CSH is closed shrubland, DBF is deciduous
broadleaf forest, EBF is evergreen broadleaf
forest, ENF is evergreen needleleaf forest,
GRA is grassland, and MF is broadleaf and
needleleaf mixed forest.

Table 3
Distribution of elevation, precipitation (P), temperature (T), leaf area index (LAI), water yield (Q), gross primary productivity (GPP), and runoff coefficient (RC) for
different vegetation dominant subcatchments.
Vegetation Composition group Number of subcatchments Elevation P T LAI Q RC GPP

(m) (mm yr−1) (°C) m2/m2 (mm yr−1) (g C m−2 yr−1)

CSH dominant 40 2743 ± 1100 922 ± 103 6.6 ± 3.6 1.2 ± 0.2 454 ± 83 0.49 ± 0.06 1089 ± 173
ENF dominant 151 3104 ± 404 1007 ± 99 5 ± 2.2 1.5 ± 0.2 501 ± 93 0.49 ± 0.05 1238 ± 100
GRA dominant 71 3952 ± 286 942 ± 93 0.8 ± 1.6 0.8 ± 0.2 582 ± 117 0.61 ± 0.07 724 ± 140
Mixed 252 3207 ± 723 979 ± 116 4.5 ± 3.6 1.2 ± 0.3 531 ± 149 0.54 ± 0.10 1039 ± 231

CSH is closed shrubland, ENF is evergreen needleleaf forest, GRA is grassland, and Mixed is no dominant vegetation type.

could be considered in future forest management planning to maximize 5. Conclusions

both water yield and carbon sequestration.
In addition to vegetation composition, ecosystem services are also In this study, the WaSSI model and MODIS products were used to
affected by local climate conditions. Some subcatchments had the same examine the relationship between elevation and key structural, func-
trends between runoff coefficient and GPP (Fig. 8). We found that tional, and efficiency parameters associated with water and carbon
higher temperature and lower precipitation led to both low water yield cycles in MJ. A significant trade-off relationship between water yield
and carbon sequestration of some subcatchments dominated by mixed and GPP was detected along elevation gradients in MJ, and vegetation
minor vegetation types or shrublands in the valley of MJ known as the composition appeared to play a key role in determining the relative
“Dry valley” (Pang et al., 2008). The low precipitation directly reduced ecological benefits for carbon and water. Along the elevation gradients,
the runoff coefficients, and high temperature increased ET which fur- ET was significantly related to elevation change, whereas there was a
ther decreased the runoff coefficient. Moreover, this increase in ET did divergent response in GPP between low elevation and high elevation.
not increase GPP because of the high aridity index in this valley area Apart from vegetation distribution, climate variables, especially tem-
(elevation < 2000 m) (Fig. S2c). On the contrary, some ENF domi- perature, significantly affected water yield and carbon sequestration
nated subcatchments had both high water yield and carbon sequestra- and their interaction along elevation gradients. Managing the percen-
tion due to relatively high precipitation (Fig. 9a and Fig. S2a). How- tage of forest coverage and their distribution in a watershed could be
ever, warming could increase ET of ENF at high elevation, and this considered in future forest management planning to maximize both
might decrease water yield in the future (Goulden and Bales, 2014). water yield and carbon sequestration. The Qinghai-Tibet Plateau has
been experiencing considerable warming, therefore future research
could further investigate the sensitivity of different vegetation types to
climate change and how vegetation change affects the water balance

9
N. Liu, et al. Journal of Hydrology 590 (2020) 125449

Fig. 8. The correlation between standardized (Z-

score) runoff coefficient and gross primary pro-
ductivity (GPP) in the Minjiang watershed. Each
point represents a subcatchment. CSH is closed
shrubland, ENF is evergreen needleleaf forest, GRA
is grassland, and Mixed is no dominant vegetation
type. The ellipse was drawn by “stat_ellipse” in
“ggplot” package using R 3.5.

especially in high elevation areas. Acknowledgement

CRediT authorship contribution statement This study was funded jointly by the science foundation of Chinese
Academy of Forestry (CAFYBB2017MA009) and the National Key R&D
Ning Liu: Software, Investigation, Writing - original draft. Pengsen Program of China (2017YFC0505006) and supported by the Key
Sun: Conceptualization, Formal analysis, Writing - review & editing. Laboratory of Forest Ecology and Environment of the National Forestry
Peter V. Caldwell: Visualization. Richard Harper: Validation. and Grassland Administration, China.
Shirong Liu: Supervision. Ge Sun: Methodology.

Declaration of Competing Interest Appendix A. Supplementary data

Supplementary data to this article can be found online at https://

The authors declare that they have no known competing financial
doi.org/10.1016/j.jhydrol.2020.125449.
interests or personal relationships that could have influenced the work
reported in this paper.

Fig. 9. The correlation between standardized (Z-score) runoff coefficient and gross primary productivity (GPP) in different elevation ranges in the Minjiang wa-
tershed. Each point represents a subcatchment. CSH is closed shrubland, ENF is evergreen needleleaf forest, GRA is grassland, and Mixed is no dominant vegetation
type. The ellipse was drawn by the “stat_ellipse” function in “ggplot” package using R 3.5.

10
N. Liu, et al. Journal of Hydrology 590 (2020) 125449

References Ma, Y.J., Li, X.Y., Liu, L., Yang, X.F., Wu, X.C., Wang, P., Lin, H., Zhang, G.H., Miao, C.Y.,
2019. Evapotranspiration and its dominant controls along an elevation gradient in
the Qinghai Lake watershed, northeast Qinghai-Tibet Plateau. J. Hydrol. https://doi.
Anderson, R.M., Koren, V.I., Reed, S.M., 2006. Using SSURGO data to improve org/10.1016/j.jhydrol.2019.05.019.
Sacramento Model a priori parameter estimates. J. Hydrol. 320, 103–116. https:// McCabe, G.J., Wolock, D.M., 1999. General-circulation-model simulations of future
doi.org/10.1016/j.jhydrol.2005.07.020. snowpack in the Western United States. J. Am. Water Resour. Assoc. 35, 12. https://
Burnash, R.J.C., 1995. The NWS river forecast system. Comput. Model. Watershed doi.org/10.1111/j.1752-1688.1999.tb04231.x.
Hydrol. 311–366, 311–366. Nash, J.E., Sutcliffe, J.V., 1970. River flow forecasting through conceptual models part I -
Caldwell, P.V., Kennen, J.G., Sun, G., Kiang, J.E., Butcher, J.B., Eddy, M.C., Hay, L.E., a discussion of principles. J. Hydrol. 10, 282–290. https://doi.org/10.1016/0022-
Lafontaine, J.H., Hain, E.F., Nelson, S.A.C., Mcnulty, S.G., 2015. A comparison of 1694(70)90255-6.
hydrologic models for ecological flows and water availability. Ecohydrology 8, Pang, X., Bao, W., Wu, J., 2008. Reasons of dry valley climate characteristic and its
1525–1546. https://doi.org/10.1002/eco.1602. formation reason in upstream of Minjiang River. Resour. Environ. Yangtze Basin 17,
Cui, X., Liu, S., Wei, X., 2012. Impacts of forest changes on hydrology: A case study of 46–53.
large watersheds in the upper reaches of Minjiang River watershed in China. Hydrol. R Development Core Team, 2018. A Language and Environment for Statistical
Earth Syst. Sci. 16, 4279–4290. https://doi.org/10.5194/hess-16-4279-2012. Computing. R Found. Stat Comput.
Dai, Y., Shangguan, W., Duan, Q., Liu, B., Fu, S., Niu, G., 2013. Development of a china Ross, C.W., Prihodko, L., Anchang, J., Kumar, S., Ji, W., Hanan, N.P., 2018.
dataset of soil hydraulic parameters using pedotransfer functions for land surface HYSOGs250m, global gridded hydrologic soil groups for curve-number-based runoff
modeling. J. Hydrometeorol. 14, 869–887. https://doi.org/10.1175/JHM-D-12- modeling. Sci. Data. https://doi.org/10.1038/sdata.2018.91.
0149.1. Shirazi, M.A., Boersma, L., 1984. A unifying quantitative analysis of soil texture. Soil Sci.
Deng, C., Zhang, B., Cheng, L., Hu, L., Chen, F., 2019. Vegetation dynamics and their Soc. Am. J. https://doi.org/10.2136/sssaj1984.03615995004800010026x.
effects on surface water-energy balance over the Three-North Region of China. Agric. Sun, G., Caldwell, P., Noormets, A., McNulty, S.G., Cohen, E., Moore Myers, J., Domec, J.-
For. Meteorol. 275, 79–90. https://doi.org/10.1016/j.agrformet.2019.05.012. C., Treasure, E., Mu, Q., Xiao, J., John, R., Chen, J., 2011. Upscaling key ecosystem
Fan, Y., Clark, M., Lawrence, D.M., Swenson, S., Band, L.E., Brantley, S.L., Brooks, P.D., functions across the conterminous United States by a water-centric ecosystem model.
Dietrich, W.E., Flores, A., Grant, G., Kirchner, J.W., Mackay, D.S., McDonnell, J.J., J. Geophys. Res. 116, 1–16. https://doi.org/10.1029/2010jg001573.
Milly, P.C.D., Sullivan, P.L., Tague, C., Ajami, H., Chaney, N., Hartmann, A., Sun, G., Hallema, D., Asbjornsen, H., 2017. Ecohydrological processes and ecosystem
Hazenberg, P., McNamara, J., Pelletier, J., Perket, J., Rouholahnejad-Freund, E., services in the Anthropocene: a review. Ecol. Process. https://doi.org/10.1186/
Wagener, T., Zeng, X., Beighley, E., Buzan, J., Huang, M., Livneh, B., Mohanty, B.P., s13717-017-0104-6.
Nijssen, B., Safeeq, M., Shen, C., van Verseveld, W., Volk, J., Yamazaki, D., 2019. Sun, J., Sun, X., Hu, Z., Wang, G., 2020a. Exploring the influence of environmental factors
Hillslope hydrology in global change research and earth system modeling. Water in partitioning evapotranspiration along an elevation gradient on Mount Gongga,
Resour. Res. 1737–1772. https://doi.org/10.1029/2018WR023903. eastern edge of the Qinghai-Tibet Platea, China. J. Mt. Sci. 17, 384–396. https://doi.
Gao, M., Piao, S., Chen, A., Yang, H., Liu, Q., Fu, Y.H., Janssens, I.A., 2019. Divergent org/10.1007/s11629-019-5687-1.
changes in the elevational gradient of vegetation activities over the last 30 years. Nat. Sun, P., Liu, S., Jiang, H., Lü, Y., Liu, J., Lin, Y., Liu, X., 2008. Hydrologic effects of NDVI
Commun. 10, 1–10. https://doi.org/10.1038/s41467-019-11035-w. time series in a context of climatic variability in an Upstream Catchment of the
Goulden, M.L., Bales, R.C., 2014. Mountain runoff vulnerability to increased evapo- Minjiang River. J. Am. Water Resour. Assoc. 44, 1132–1143. https://doi.org/10.
transpiration with vegetation expansion. Proc. Natl. Acad. Sci. USA 111, 1111/j.1752-1688.2008.00256.x.
14071–14075. https://doi.org/10.1073/pnas.1319316111. Sun, X., Wang, G., Huang, M., Chang, R., Hu, Z., Song, C., Sun, J., 2020b. The asyn-
Hamon, W.R., 1963. Computation of direct runoff amounts from storm rainfall. Int. Assoc. chronous response of carbon gain and water loss generate spatio-temporal pattern of
Sci. Hydrol. Publ. 63, 52–62. WUE along elevation gradient in southwest China. J. Hydrol. 581. https://doi.org/
Hu, Z., Wang, G., Sun, X., Zhu, M., Song, C., Huang, K., Chen, X., 2018. Spatial-temporal 10.1016/j.jhydrol.2019.124389.
patterns of evapotranspiration along an elevation gradient on Mount Gongga, Swetnam, T.L., Brooks, P.D., Barnard, H.R., Harpold, A.A., Gallo, E.L., 2017.
Southwest China. Water Resour. Res. 54, 4180–4192. https://doi.org/10.1029/ Topographically driven differences in energy and water constrain climatic control on
2018WR022645. forest carbon sequestration. Ecosphere. https://doi.org/10.1002/ecs2.1797.
Huang, X., Zhao, J., Li, W., Jiang, H., 2014. Impact of climatic change on streamflow in Vasudevan, D., Chua, E.Y.D., Davey, C.A., 2010. Crystal structures of nucleosome core
the upper reaches of the Minjiang River. China. Hydrol. Sci. J. 59, 154–164. https:// particles containing the “601” strong positioning sequence. J. Mol. Biol. 403, 1–10.
doi.org/10.1080/02626667.2013.853878. https://doi.org/10.1016/j.jmb.2010.08.039.
Hutchinson, M.F., 1998. Interpolation of rainfall data with thin plate smoothing splines. Wang, G., Ran, F., Chang, R., Yang, Y., Luo, J., Jianrong, F., 2014. Variations in the live
Part I: two dimensional smoothing of data with short range correlation. J. Geogr. Inf. biomass and carbon pools of Abies georgei along an elevation gradient on the Tibetan
Decis. Anal. 2, 139–151 https://doi.org/10.1.1.13.6255. Plateau, China. For. Ecol. Manage. https://doi.org/10.1016/j.foreco.2014.06.023.
IPCC, 2013. Climate Change 2013: The Physical Science Basis. Contribution of Working Wei, X., Li, Q., Zhang, M., Giles-Hansen, K., Liu, W., Fan, H., Wang, Y., Zhou, G., Piao, S.,
Group I to the Fifth Assessment Report of the Intergovernmental Panel on Climate Liu, S., 2018. Vegetation cover—another dominant factor in determining global
Change, Cambridge University Press, Cambridge, United Kingdom and New York, water resources in forested regions. Glob. Chang. Biol. 24, 786–795. https://doi.org/
NY, USA. https://doi.org/10.1017/CBO9781107415324. 10.1111/gcb.13983.
Jackson, R.B., Jobbágy, E.G., Avissar, R., Roy, S.B., Barrett, D.J., Cook, C.W., Farley, K.A., Xiao, J., Sun, G., Chen, J., Chen, H., Chen, S., Dong, G., Gao, S., Guo, H., Guo, J., Han, S.,
Le Maitre, D.C., McCarl, B.A., Murray, B.C., 2005. Atmospheric science: trading water Kato, T., Li, Y., Lin, G., Lu, W., Ma, M., McNulty, S., Shao, C., Wang, X., Xie, X.,
for carbon with biological carbon sequestration. Science (80-.) 310, 1944–1947. Zhang, X., Zhang, Z., Zhao, B., Zhou, G., Zhou, J., 2013. Carbon fluxes, evapo-
https://doi.org/10.1126/science.1119282. transpiration, and water use efficiency of terrestrial ecosystems in China. Agric. For.
Law, B.E., Falge, E., Gu, L., Baldocchi, D.D., Bakwin, P., Berbigier, P., Davis, K., Dolman, Meteorol. 182–183, 76–90. https://doi.org/10.1016/j.agrformet.2013.08.007.
A.J., Falk, M., Fuentes, J.D., Goldstein, A., Granier, A., Grelle, A., Hollinger, D., Yuan, H., Dai, Y., Xiao, Z., Ji, D., Shangguan, W., 2011. Reprocessing the MODIS Leaf
Janssens, I.A., Jarvis, P., Jensen, N.O., Katul, G., Mahli, Y., Matteucci, G., Meyers, T., Area Index products for land surface and climate modelling. Rem. Sens. Environ. 115,
Monson, R., Munger, W., Oechel, W., Olson, R., Pilegaard, K., Paw, U.K.T., 1171–1187. https://doi.org/10.1016/j.rse.2011.01.001.
Thorgeirsson, H., Valentini, R., Verma, S., Vesala, T., Wilson, K., Wofsy, S., 2002. Zhang, L., Sun, P., Liu, S., 2020. Growing-season transpiration of typical forests in dif-
Environmental controls over carbon dioxide and water vapor exchange of terrestrial ferent succession stages in subalpine region of Western Sichuan, China. Linye Kexue/
vegetation. Agric. For. Meteorol. 113, 97–120. https://doi.org/10.1016/S0168- Scientia Silvae Sin. 56, 1–9. https://doi.org/10.11707/j.1001-7488.20200101.
1923(02)00104-1. Zhang, M., Liu, N., Harper, R., Li, Q., Liu, K., Wei, X., Ning, D., Hou, Y., Liu, S., 2017. A
Li, C., Sun, G., Caldwell, P.V., Cohen, E., Fang, Y., Zhang, Y., Oudin, L., Sanchez, G.M., global review on hydrological responses to forest change across multiple spatial
Meentemeyer, R.K., 2020. Impacts of urbanization on watershed water balances scales: importance of scale, climate, forest type and hydrological regime. J. Hydrol.
across the conterminous United States. Water Resour. Res. 56. https://doi.org/10. https://doi.org/10.1016/j.jhydrol.2016.12.040.
1029/2019WR026574. Zhang, M., Wei, X., Sun, P., Liu, S., 2012. The effect of forest harvesting and climatic
Li, Y., Shi, H., Zhou, L., Eamus, D., Huete, A., Li, L., Cleverly, J., Hu, Z., Harahap, M., Yu, variability on runoff in a large watershed: the case study in the Upper Minjiang River
Q., He, L., Wang, S., 2018. Disentangling climate and LAI effects on seasonal varia- of Yangtze River basin. J. Hydrol. 464–465, 1–11. https://doi.org/10.1016/j.jhydrol.
bility in water use efficiency across terrestrial ecosystems in China. J. Geophys. Res. 2012.05.050.
Biogeosci. https://doi.org/10.1029/2018JG004482. Zhang, W., Ning, J., Song, K., Li, X., Wang, Y., 2013. The response of vegetation cover to
Liu, N., Sun, P., Liu, S., Sun, G., 2013a. Coupling simulation of water-carbon processes for the variation of heat and water conditions in Upper Minjiang Watershed, China. J.
catchment-calibration and validation of the WaSSI-C model. Chin. J. Plant Ecol. 37, Mt. Sci. 31, 280–286.
492–502. https://doi.org/10.3724/sp.j.1258.2013.00051. Zhang, Y., Song, C., Sun, G., Band, L.E., McNulty, S., Noormets, A., Zhang, Q., Zhang, Z.,
Liu, N., Sun, P., Liu, S., Sun, G., 2013b. Determination of spatial scale of response unit for 2016. Development of a coupled carbon and water model for estimating global gross
WaSSI-C eco–hydrological model—a case study on the upper Zagunao River wa- primary productivity and evapotranspiration based on eddy flux and remote sensing
tershed of China. J. Plant Ecol. 37 (2), 132–141. https://doi.org/10.3724/sp.j.1258. data. Agric. For. Meteorol. 223, 116–131. https://doi.org/10.1016/j.agrformet.2016.
2013.00014. 04.003.
Liu, P., Hao, L., Pan, C., Zhou, D., Liu, Y., Sun, G., 2017. Combined effects of climate and Zhao, M.S., Heinsch, F.A., Nemani, R.R., Running, S.W., 2005. Improvements of the
land management on watershed vegetation dynamics in an arid environment. Sci. MODIS terrestrial gross and net primary production global data set. Rem. Sens.
Total Environ. 589, 73–88. https://doi.org/10.1016/j.scitotenv.2017.02.210. Environ. 95, 164–176. https://doi.org/10.1016/j.rse.2004.12.011.
Liu, Y., Fan, N., An, S., Bai, X., Liu, F., Xu, Z., Wang, Z., Liu, S., 2008. Characteristics of Zhou, D., Hao, L., Kim, J.B., Liu, P., Pan, C., Liu, Y., Sun, G., 2019. Potential impacts of
water isotopes and hydrograph separation during the wet season in the Heishui River, climate change on vegetation dynamics and ecosystem function in a mountain wa-
China. J. Hydrol. 353, 314–321. https://doi.org/10.1016/j.jhydrol.2008.02.017. tershed on the Qinghai-Tibet Plateau. Clim. Change. https://doi.org/10.1007/
Lü, Y., Fu, B., Feng, X., Zeng, Y., Liu, Y., Chang, R., Sun, G., Wu, B., 2012. A policy-driven s10584-019-02524-4.
large scale ecological restoration: quantifying ecosystem services changes in the loess
plateau of China. PLoS One. https://doi.org/10.1371/journal.pone.0031782.

11