Catena 246 (2024) 108392

Contents lists available at ScienceDirect

Catena
journal homepage: www.elsevier.com/locate/catena

Quantifying the impact of natural and human activity factors on

desertification in the Qinghai-Tibetan Plateau
Ying Zhi a,b , Shulin Liu a,*, Tao Wang a , Hanchen Duan a , Wenping Kang a
a
Key Laboratory of Ecological Safety and Sustainable Development in Arid Lands, Northwest Institute of Eco-Environment and Resources, Chinese Academy of Sciences,
Lanzhou 730000, Gansu, China
b
University of Chinese Academy of Sciences, Beijing 100049, China

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

Keywords: Accurately determining of the impact of natural and human activity factors on desertification will facilitate the
Desertification development of effective strategies to combat desertification. Herein, the desertification dynamics of the
The Qinghai-Tibetan Plateau Qinghai-Tibetan Plateau (QTP) from 2000 to 2020 were inverted using the spatial combination of Albedo-MSAVI
Geodetector
features. Geodetector and structural equation modelling were used to quantify the independent and interactive
Structural equation model
effects of factors on desertification, revealing the total effect on desertification through direct and indirect effects.
Desertification was generally alleviated during the study period. The proportions of desertification undergoing
significant reversal and reversal were 1.4 % and 23.8 %, respectively. The vegetation type, precipitation, soil
type and temperature considerably affect the distribution of desertification compared to other driving factors.
Compared to a single factor, the mutual interactions between two factors have nonlinear enhancement effects on
the change in desertification. Climatic factors, natural environment and human activities directly influence
desertification, while natural environmental factors affect desertification indirectly by influencing human ac-
tivities and climate. Climatic factors make the greatest contribution to desertification of the QTP (62 %), fol-
lowed by human activities (− 36.8 %) and natural environment (8 %). Overall, the effects of human activities on
the desertification of the QTP cannot be ignored, and the warming and humidification of the climate have curbed
the process of desertification to a certain extent.

1. Introduction country, with 0.427 million km2 and 0.189 million km2 respectively,
accounting for 16.59 % and 7.36 % of the whole desertified land area in
Desertification is a serious threat to ecological social-economic se- China (Zan et al., 2023). The desertification on the Qinghai-Tibetan
curity (UNCED, 1992; Guo et al., 2017; Ontel et al., 2023). Desertifi- Plateau (QTP), often referred to as the “Third Pole of the Earth”,
cation aggravates the ecological condition in arid and semi-arid regions, significantly influences ecosystems both locally and in other regions by
seriously affecting ecosystem supply, regulation and other services, altering climate patterns that affect precipitation and temperature, as
which restricts sustainable socio-economic development and threatens well as potentially increasing aeolian activities such as dust storms that
the survival and development environment of human beings (Mirzabaev transport soil particles and nutrients over long distances, impacting air
et al., 2019). According to the United Nations Convention to Combat quality and soil fertility beyond the plateau (Li et al., 2001). The fragile
Desertification, desertification is defined as the land degradation in arid, and sensitive ecological base of the QTP and the frequent occurrence of
semi-arid and dry sub-humid areas resulting from various factors, extreme climatic events cannot be ignored in terms of their impact on
including climatic variations and human activities (UNCCD, 1994). By combating desertification (Li et al., 2022; Yang et al., 2023). Human
2019, the total desertified land area of China reached 2.574 million km2, activities driven by economic interests have exacerbated the process of
accounting for 77.59 % of the entire area of the desertification survey vegetation degradation, including reduced vegetation cover, lower
area and 26.81 % of the entire land area of China. The desertified land quality of vegetation, and decreased species diversity, leading to an
area of Tibet and Qinghai Province ranked among the top five in the aggravation of local desertification levels (Xu et al., 2015).

* Corresponding author at: Key Laboratory of Ecological Safety and Sustainable Development in Arid Lands, Northwest Institute of Eco-Environment and Re-
sources, Chinese Academy of Sciences, Lanzhou 730000, Gansu, China.
E-mail address: [email protected] (S. Liu).

https://doi.org/10.1016/j.catena.2024.108392
Received 1 April 2024; Received in revised form 31 August 2024; Accepted 8 September 2024
Available online 17 September 2024
0341-8162/© 2024 The Authors. Published by Elsevier B.V. This is an open access article under the CC BY-NC license (http://creativecommons.org/licenses/by-
nc/4.0/).
Y. Zhi et al. Catena 246 (2024) 108392

Clarifying the driving mechanisms of desertification evolution con- of desertification in the QTP from 2000 to 2020, (2) clarify the key
tributes to the development of measures to combat desertification in the factors affecting desertification change and the pathways of influence
QTP. The occurrence and development of desertification are mainly the and (3) recognise the direct and indirect impacts of the natural envi-
result of a combination of climate change and human activity (Olsson ronment, climatic factors and human activities on desertification
et al., 2005; Li et al., 2016). Many methods and models have been used change. The results of this study will provide practical information for
to detect the contribution of various impact factors to desertification. further understanding the mechanism driving desertification and pro-
Previous studies applied correlation analysis to obtain the relationship vide data support for the formulation of desertification control strategies
between the impact factors and desertification or principal component and environmental management measures on the QTP.
analysis to estimate the effects of various impact factors on desertifica-
tion (Zhang, 2000; Sun and Li, 2002; Li et al., 2007). Residual trend 2. Materials and methods
analysis models are also frequently used by many researchers to analyse
the influence of climatic factors and human activities on dryland 2.1. Study area
desertification (Burrell et al., 2017; Kundu et al., 2017; Liu et al., 2019).
However, these traditional statistical approaches often assume that the The QTP is situated in the area of 73◦ –105◦ E, 25◦ –40◦ N (Fig. 1a), and
relationship between the impacts of climatic and human activity factors is known as the “Roof of the World” and the “Third Pole of the Earth”
on land degradation and desertification is linear, which does not suffi- (Yao et al., 2012). The QTP is located in southwest China, including
ciently express the non-linearity and complexity of the response to the Tibet and Qinghai Province. The climate of the QTP is characterised by
driving factors and does not explain interactions and pathways between the following general features: strong radiation and sunshine, low air
the driving factors. The geographical detector model (GDM) is an temperature, low cumulative temperature, and temperature decreases
important method for spatial dissimilarity detection and driving force with height and latitude; thin air, low precipitation, and large seasonal
analysis (Wang et al., 2016). The GDM can explain the main drivers of variations in precipitation. The vegetation type of the QTP from south-
desertification robustly and directly without any linear assumptions and east to northwest, with the spatial distribution of water and thermal
reveal the effects of the interactions between driving factors on desert- conditions presents the vegetation distribution characteristics of forests,
ification, fully exploiting the correlation information between deserti- meadows, steppes and deserts, as well as the three-dimensional struc-
fication and drivers (Liang and Yang, 2016; Du et al., 2016; Hua and tural characteristics of vertical zones of broadleaf forests, needleleaf and
Hao, 2021). Studies using the geodetector method have found that the broadleaf mixed forests, needleleaf forests and scrubs. Human activities
interaction between two factors has greater explanatory power for are mainly in the river valleys, specifically in the Yarlung Zangbo River,
desertification than a single factor, and suggest that changes in desert- the Lhasa River, and the Nyang River areas of Tibet as well as the
ification result from the combined influence of three or more natural or grasslands in the northeast and the woodland areas in the southeast (Li
human factors, making the exploration of interactions among these et al., 2018; Sun et al., 2020).
factors an important area for future research (Guo et al., 2022; Wang
et al., 2023). Structural equation modelling (SEM) can decompose cor-
relation coefficients to obtain direct and indirect impacts of multiple 2.2. Data sources and pre-processing
influences on the target variable and the path of influence, effectively
quantifying the combined effects of multiple influences on the target 2.2.1. Satellite data
variable (Grace and Bollen, 2005). Feng et al. (2021) found through path The MODIS 16-day surface Albedo product (MCD43A3) was applied
analysis in structural equation modelling that human activities influence for Albedo inversion, and the MODIS eight-day reflectance product
aeolian desertification dynamics through direct impacts on the sandy (MOD09A1) was applied for the MSAVI inversion. Data were selected for
environment and indirect impacts on climate. The sensitivity and synthetic Albedo minima and the MSAVI means for each year during the
vulnerability of the QTP to climate change and anthropogenic distur- growing season (May–September), excluding the non-land features of
bances are more pronounced than in other low-latitude regions (Sun water, ice and snow. The download and processing of the above data
et al., 2021). Therefore, the combined use of GDM and SEM can provide were achieved through the GEE platform. The Albedo values (Liang,
a multifaceted and multidimensional view of the driving mechanisms 2001) and MSAVI values (Javzandulam et al., 2005) were calculated as
involved in the evolution of desertification in the QTP and formulate follows:
effective measures to combat desertification and protect and restore the
Albedo =0.16α1 + 0.291α2 + 0.243α3 + 0.116α4
regional ecosystems, which will help achieve the goal of sustainable (1)
+ 0.112α5 + 0.081α6 − 0.0015
development in the future.
The aims of this study are to (1) analyse the spatial pattern changes

Fig. 1. (a) Study area; (b) Location of verification points.

2
Y. Zhi et al. Catena 246 (2024) 108392

( √̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅ )
MSAVI = 2NIR + 1 − (2NIR + 1)2 − 8(NIR − RED) /2 (2) Albedo = α × MSAVI − b (3)

DDI = k × MSAVI − Albedo (4)

notes: α1 -α6 , White-sky Albedo band1-White-sky Albedo band6.
where k = − 1/α and α is the slope of the trend line fitted by Albedo and
2.2.2. Selection of driving factors MSAVI.
Desertification occurs and changes due to a combination of factors Considering the unique characteristics of the QTP soils, vegetation,
(Yang et al., 2022). Twelve representative, easily quantifiable and easily geomorphology, and climate, as well as the prevalence of primary bare
accessible factors were selected in terms of climate, vegetation, soil, deserts due to abundant Quaternary loose sediments as noted by Li et al.
topography and human activity. The vegetation type (VT), elevation (2018), we categorized the levels of desertification. Utilizing the natural
(Elev), soil type (ST), aspect (A) and slope (S) were selected as the base break (Jenks) method as described by Ma et al. (2011), we classified the
natural environment factors. Annual average precipitation (Pre), wind DDI values into six classes: bare desert, extremely severe desertification,
speed (WS), temperature (Tem) and potential evapotranspiration (PET) severe desertification, moderate desertification, low desertification and
are major climatic factors affecting desertification. Population (Pop) non-desertification. Bare desert and different levels of desertification
density, actual livestock carrying capacity (ALCC) and nature reserves interpretation markers were established by combining the above char-
(NRs) reflect human activities. Table 1 provides details of all driving acteristics (Appendix Table A.1, supplementary material, Yu et al., 2007;
factors. In ARCGIS 10.2, all driving factors were multi-valued. Using the Duan et al., 2019). The results of desertification classification were
Create Fishnet tool in ARCGIS 10.2, a regular grid of 10 km × 10 km was validated by selecting 303 field validation points and 1204 Google Earth
created. After excluding water, ice and snow areas, 25,858 sample validation points generated uniformly using ARCGIS 10.2 (Fig. 1b).
points were generated (Appendix Figure S1, supplementary material). Field surveys were conducted during the peak vegetation periods of
Subsequently, the Extract Multi Values to Points tool in ARCGIS 10.2 2019 to 2022, from June to September. The field validation points
was used to extract the values of the dependent and independent vari- recorded geographic coordinate information, surface landscape char-
ables based on the sample point locations. This facilitated driver analysis acteristics, desertification types, etc. The landscape photographs in
using GDM and allowed the use of these 12 driving factors as explana- Appendix Table A.1 were obtained from orthorectified photographs of
tory variables for desertification changes in SEM. the validation points taken by UAV.

2.3.2. Geodetector model

2.3. Methods
The Geodetector model (GDM) is an important method for spatial
dissimilarity detection and driver analysis (Wang et al., 2010). Factor
2.3.1. Feature space model
and interaction detection were used to analyse the influence of the
Studies have shown that multidimensional remote sensing informa-
driving factors on the spatial distribution pattern and desertification
tion from a combination of indicators can better reflect the surface cover
changes on the QTP.
status and desertification process than a single vegetation index (Feng
et al., 2018; Yang et al., 2022). Zeng et al. (2006) proposed a remote
(1) A factor detector is used to detect the extent to which a single
sensing model for desertification monitoring based on the Albedo-NDVI
driver influences the dependent variable. The factor is measured
feature space, the desertification difference index (DDI) model. MSAVI
using the q-statistics value, which was calculated using equation
was used to reflect the vegetation condition in this study owing to the
(5):
complexity and diversity of vegetation in the QTP. The Albedo-MSAVI
∑L
feature space, which represents the desertification trend, is divided h=1 Nh σ h
2

vertically, and different desertified lands can be effectively distin- q = 1− (5)

Nσ 2

guished and expressed by the DDI (Verstraete and Pinty, 1996).

ARCGIS10.2 was used to construct a 10 km × 10 km fishing network where h = 1, 2, 3, …; L denotes the hierarchy (i.e. category or grading) of
for the study area of the QTP, and 4000 sample points were randomly the dependent or independent variable and σ 2h and σ 2 are the variance of
selected to extract the corresponding MSAVI and Albedo values. The the hierarchy h and the entire region, respectively. Nh and N are the
linear regression equation between them was constructed by taking the number of cells in the hierarchy h and the entire region, respectively; the
Albedo-MSAVI as an example (Eq. (3). The DDI was based on the coef- value of q ranges from 0 to 1.
ficient α of the linear regression equation (Eq. (4).

Table 1
Driving factors applied for the desertification analyses.
Factor type Factor Spatial Year Data sources
resolution

Natural Elevation 1 km − Data Center for Resources and Environmental Sciences Chinese
environment Vegetation type 1:1 Million − Academy of Sciences (RESDC) (https://www.resdc.cn/)
Soil type 1:1 Million 1995
Slope 1 km −
Aspect 1 km −
Climate Precipitation 1 km 2000–2020
Temperature 1 km 2000–2020
Wind speed 1 km 2000–2020
Potential evapotranspiration 1 km 2000–2020 National Tibetan Plateau / Third Pole Environment Data Center (http://
data.tpdc.ac.cn)
Human activity Population 1 km 2000/2005/2010/2015/ Data Center for Resources and Environmental Sciences Chinese
2019 Academy of Sciences (RESDC) (https://www.resdc.cn/)
Nature reserves − 2018
Actual livestock carrying 1 km 2000–2019 National Tibetan Plateau / Third Pole Environment Data Center (http://
capacity data.tpdc.ac.cn)

Notes: 2019 data substituted due to lack of 2020 popolation data and actual livestock carrying capacity data.

3
Y. Zhi et al. Catena 246 (2024) 108392

(2) Interaction detection identifies the interactions between different 2.3.3. Structural equation model
factors and assesses whether the effect on the dependent variable
Y increases or decreases when factors X1 and X2 act together. 2.3.3.1. Structural equation model for desertification. SEM is a method
Appendix Table A.2 shows the basis of the assessment and the for hypothesizing and testing systematic relationships (Grace, 2006;
types of effects. Angelini et al., 2016). First, based on a literature review or a priori
knowledge, many assumptions on the relationship between variables
The GDM was implemented based on the ‘GD’ package in R 4.3.0. were made and a conceptual model was constructed. The SEM then
Pre, Tem, PET, WS, Elev, S, A, VT, ST, Pop, NRs and ALCC were selected substitutes the observed data to obtain the mathematical model. The
as independent variables in the model and DDI was used as a dependent model is continuously corrected according to the fitting effect, and the
variable. The effect or explanatory power of the independent factors on SEM model is constructed when the model is fitted satisfactorily (Yang
the dependent variable can be expressed using the value of the q-sta- et al., 2021).
tistics. The GDM requires discrete independent variables (Wang and Xu,
2017), and optimal discrete methods are used to reclassify continuous 2.3.3.2. Developing a graphical conceptual model. The following hy-
independent variables according to the principle of q-statistic max- potheses were made based on an understanding of the desertification
imisation. The optimal discrete method for Pre, Tem, PET, and Elev is driving mechanisms in the literature (Feng et al., 2021; Wang et al.,
standard deviation. The optimal discrete method for WS is equal inter- 2022). Taking the change in desertification in the QTP from 2000 to
val, for S it is geometric, and for Pop and ALCC it is quantile. The 2020 as the target variable, it was assumed that the continuously
vegetation types are classified into 12 categories. The soil types are changing climate and human activities directly affect desertification,
classified into 13 categories based on soil outline, and NRs are classified and the natural environment may directly influence desertification. The
into the core, buffer and experimental zones based on the zoning prin- natural environment causes changes in climate and human activities,
ciple. The spatial distribution of the driver factors is shown in Fig. 2. and human activities affect climate change, indirectly affecting desert-
ification. In the conceptual model of the SEM, the natural environment,
climatic factors, and human activities are potential variables that cannot

Fig. 2. Spatial distribution of the results of the optimal reclassification of the driving factors. (a) Precipitation, (b) Temperature, (c) Potential evapotranspiration, (d)
Wind speed, (e) Elevation, (f) Slope, (g) Aspect, (h) Soil type, (i) Vegetation type, (j) Population, (k) Nature reserves, (o) Actual livestock carrying capacity.

4
Y. Zhi et al. Catena 246 (2024) 108392

be observed directly. Therefore, the natural environment is reflected by moderate desertification land. The decrease in land with moderate
observational variables (i.e. Elev, S, A, VT and ST). Pre, Tem, PET and desertification was 28,604.07 km2, which was dominated by the shift to
WS from 2000 to 2020 are observational variables reflecting climate low desertification land. Severe desertification land was reduced by
change. The observational variables for human activities are the Pop, 32,156.33 km2, mostly transformed into low and moderate desertifica-
NRs and ALCC. tion land. Similarly, the area of extremely severe desertification land
and bare desert decreased. From 2000 to 2020, the level of desertifica-
2.3.3.3. Model validation. Based on the above hypotheses, SEM con- tion improved, with more areas converted from high to low levels than
struction was performed in AMOS 28.0 software. The model was in the reverse direction.
debugged by deleting or changing paths to achieve the best fit. The
goodness-of-fit of the SEM was measured using four parameters,
3.2. Independent and interactive influences on the drivers of
including the root squared mean error of approximation (RMSEA),
desertification change
goodness-of-fit index (GFI), adjusted goodness-of-fit index (AGFI) and
comparative fit index (CFI). A good model fit is indicated when the GFI,
3.2.1. Single factor analysis/Factor detection
CFI and AGFI values are greater than 0.9 and the RMSEA value is less
The ability of these factors to influence desertification on the QTP
than 0.08 (Grace and Keeley, 2006).
was VT>Pre > ST>Tem > PET>ALCC>WS>Elev > Pop > S>NRs > A
(Fig. 5); all of these driving factors were significant (p < 0.05). Fig. 5
3. Results
shows that VT has the largest q-statistics and is the main factor influ-
encing QTP desertification. At the same time, precipitation also has an
3.1. Spatiotemporal changes in desertification from 2000 to 2020
important influence on the process of desertification, second only to the
VT. Among the human activity factors, the q-statistics value of ALCC is
The classification results of the Albedo-MSAVI feature space model
greater than that of the NRs and Pop factors. The q-statistics for Elev, S
were evaluated using a confusion matrix based on a total of 1507 field
and A are relatively low, suggesting that topographic factors have weak
validation points and Google Earth validation points with an accuracy of
explanatory power for the distribution of desertification. Among the 12
88.07 % and a Kappa coefficient of 0.79. Fig. 3 shows the spatial dis-
influencing factors, the first three factors that strongly affect the distri-
tribution of desertification changes on the QTP from 2000 to 2020.
bution of desertification in the whole QTP are natural factors, probably
Throughout the study period, the spatial pattern of desertification on the
because most areas of the QTP are inaccessible.
QTP showed a reversal trend. Approximately 1.4 % of the desertified
area was significantly reversed and 23.8 % was reversed, mainly in the
3.2.2. Interactive detection
northwestern part of the QTP. The proportions of severely deteriorating
All the influencing factors shown in Fig. 6 passed the test of signifi-
and deteriorating desertification were 0.2 % and 2 %, respectively, with
cance, indicating that there are significant discrepancies between fac-
a sporadic distribution in the Qaidam Basin region and the southwestern
tors. Fig. 6 shows that the interaction of any two factors is augmented,
part of the QTP. A total of 72.6 % of the QTP region has a stable level of
indicating that both factors have a greater effect on the desertification
desertification.
process than a single factor. The factor combination with the strongest
Fig. 4 shows the state of mutual transfer between bare desert and
interaction was VT and ALCC, with a bivariable enhanced and q-statis-
desertification levels, and Table 2 lists the area transferred out and
tics value of 0.4486, indicating the joint influence of natural factors and
transferred in. Fig. 4 and Table 2 show that the non-desertification area
human activities. In addition, the composite of VT, Pre and other factors
in 2020 has increased by 250,900.30 km2 compared to 2000, of which
produced higher q-statistics values. VT is a bivariable that was enhanced
85.34 % was from the transfer of low desertification land. The overall
with all factors except A, which was nonlinearly enhanced. Pre was
decrease in low desertification land was 82,611.61 km2, mainly trans-
nonlinearly enhanced with A, WS and ALCC, which was a bivariable
formed into non-desertification land and partially transformed into
enhanced with other factors. The q-statistics for the interaction of A and

Fig. 3. Spatial distribution of desertification changes on the QTP from 2000 to 2020.

5
Y. Zhi et al. Catena 246 (2024) 108392

Fig. 4. Sankey map of bare desert and desertification transfer on the QTP from 2000 to 2020.

Table 2
Matrix of bare desert and desertification area transfer on the QTP from 2000 to 2020 (km2).
2020 Bare desert Desertification degree
2000
Extremely severe Severe Moderate Low Non

Bare desert 36458.36 33210.69 12872.28 4664.57 3232.06 1413.67

Extremely severe 7066.51 41041.19 65506.89 19950.72 8246.33 2922.52
Severe 1855.31 5320.57 45991.72 64509.30 40325.51 8164.49
Moderate 743.16 1180.70 5926.93 40906.48 94497.65 28834.85
Low 619.38 908.81 2474.46 10788.82 145249.70 231673.79
Non 234.46 418.36 1238.28 2665.81 17552.11 1513801.71

NRs was the lowest of all the interacting factors (q-statistics = 0.0252), and direct effects of the underlying variables on desertification were
and the interactions between S, A, NRs and the other factors produced combined to obtain the total effect of each latent variable (Fig. 8). The
low q-statistics values. Furthermore, nonlinear enhancement in- total impacts of changes in the natural environment, climatic factors and
teractions occurred between A and all other factors. human activity on desertification were 0.08, 0.62 and − 0.368,
respectively.

3.3. Direct and indirect influences on the drivers of desertification change 4. Discussion

Fig. 7 shows the results of the SEM. The mutual interactions among 4.1. Spatiotemporal variation of desertification on the Qinghai-Tibetan
factors are well supported in the model. The goodness-of-fit parameters Plateau
(i.e. GFI=0.966, CFI=0.964, AGFI=0.923, RMSEA=0.077) indicate a
good model fit. Overall, all path coefficients passed the test of signifi- Fig. 3 shows a reversal trend in the spatial pattern of desertification
cance, except for the coefficient of the path between the A and the on the QTP, with an overall reduction of desertified land in 2020
natural environment. Elev is the major factor influencing the potential compared with 2000. Zhang et al. (2018) found that the reversal of
variables of the natural environment with an impact coefficient of − aeolian desertification on the Qinghai-Tibetan Plateau became the
0.93, followed by ST with an impact coefficient of − 0.52. Among the dominant process after 2000, with a gradual decrease in the desertified
potential variables of climatic factors, Tem is the dominant factor with area, and attributed this to climate change favouring the recovery of
an impact coefficient of 0.98. In the case of human activities, the impact vegetation, the reduction of wind erosion and the implementation of
of NRs and Pop factors was greater. The natural environment, human ecological restoration projects. This is consistent with the findings of this
activities and climatic factors directly impact desertification, with in- study. In recent decades, the rate of warming on the QTP has been
fluence coefficients of − 0.16, − 0.42 and 0.62, respectively. The positive almost twice the global average rate (Kuang and Jiao, 2016; Yao, 2019).
effects of climate factors on the change in desertification are greater The vegetation on the QTP has improved as a result of the temperature
than the negative effects of human activities. increase, with an overall greening trend (Li et al., 2018; Jiao et al.,
The natural environment and human activities have an indirect 2021). Meanwhile, NDVI and LAI also reflect the plateau surface cover
impact on desertification change. The indirect influence of human ac- condition in a positive trend (Zhang et al., 2019; Wang et al., 2021).
tivity changes on desertification was 0.052 (Table 3), indirectly affecting
the desertification changes mainly through positive effects on climatic
factors. The indirect positive impact of the natural environment by
influencing human activities and climatic factors was 0.24. The indirect

6
Y. Zhi et al. Catena 246 (2024) 108392

Fig. 5. The q-statistics values of individual factors. Note: “**” represents p < 0.05.

Fig. 6. Interaction between pairwise factors. Note: “↑” and “↑↑” denote nonlinear enhanced and bivariable enhanced, respectively.

4.2. Impact factors on desertification change most of the steppe and alpine areas in the VT overlap with the areas
where desertification occurs in the study area. A significant correlation
Natural environmental variables have direct and indirect effects on was observed between ST and VT (Fig. 9). Different soil structures,
desertification on the QTP. VT had the highest q-statistics value in the moisture and nutrient contents lead to different vegetation distributions
single factor effect (Fig. 5), possibly because of the extent of the desert, (Piedallu et al., 2019). For example, arid soil, saline-alkali soil and

7
Y. Zhi et al. Catena 246 (2024) 108392

Fig. 7. Variable relationships of desertification changes and driving factors. Note: The arrows indicate the influence paths of the drivers, and the numbers are
standardized regression coefficients for the corresponding influence paths. Boxes indicate latent variables and ovals indicate observed variables. *, ** denote p <
0.05, p < 0.01, respectively.

Table 3
Direct, indirect and total effects of NE, CF, HA and DDI changes based on SEM paths.
Direct Indirect Total

Paths Coefficient Paths Coefficient Paths Coefficient

Effects on HA NE→HA 0.77 − − NE~HA 0.77

Effects on CF NE→CF 0.84 NE→HA→CF 0.064 NE~CF 0.904
HA→CF 0.08 − − HA~CF 0.08
Effects on DDI NE→DDI − 0.16 NE→CF→DDI 0.24 NE~DDI 0.08
NE→HA→DDI
NE→HA→CF→DDI
HA→DDI − 0.42 HA→CF→DDI 0.052 HA~DDI − 0.368
CF→DDI 0.62 − − CF~DDI 0.62

Note: natural environment (NE), climatic factors (CF), human activity (HA).

desert soil are not conducive to the survival and development of vege- regression coefficient of 0.98, Fig. 7). Thus, wet warming of the QTP
tation (Wang et al., 2022) and are more prone to desertification. promotes vegetation growth and alleviates desertification, while
Topographic factors (Elev, S and A) control the redistribution of water warmer temperatures and reduced precipitation exacerbate vegetation
and thermal conditions (Peng et al., 2019). Elev has a more direct in- degradation and desertification. The correlation between PET and Tem
fluence on vegetation growth and soil respiration than the S and A (Liu was as high as 0.93, again with strong explanatory power for desertifi-
et al., 2019; Zhang et al., 2020). In addition, Elev was significantly cation. WS has a negative effect on desertification (standardised
correlated with WS, Pre, Tem and PET (Fig. 9). Thus, although elevation regression coefficient of − 0.77, Fig. 7). The average elevation of the
has a low q-statistics value among the factors influencing desertification QTP is over 3500 m, and there is a clear positive correlation between
on the QTP, it has the highest path coefficient of influence (− 0.93) in the Elev and WS (coefficient = 0.689, Fig. 9). Increased WS exacerbates soil
natural environment. In particular, in the QTP, an Elev change can cause erosion and damages the vegetation cover, exacerbating desertification
a series of changes in vegetation, soil properties and climate, indirectly (Li et al., 2021; Zhai et al., 2021). The total contribution of climatic
affecting the spatial distribution of desertification. factors to desertification on the QTP was 62 % (Fig. 8), indicating that
Fragile and sensitive ecosystems make QTP regions susceptible to the humidification and warming of the climate have curbed the process
climate change (Sun et al., 2021). Precipitation is a major factor con- of desertification to a certain extent.
trolling vegetation changes in arid and semi-arid zones (Shi et al., 2021). Human activity, as an essential driver of desertification, with a
The study results show that precipitation has a relatively large impact contribution of − 36.8 % from direct plus indirect impacts, may have a
(q-statistics value = 0.298) on the desertification change among the greater impact on desertification on the QTP than predicted. In recent
selected climatic factors. The rate of warming on the QTP has been years, the population of the QTP has continued to grow (Li et al., 2017).
almost twice the global average rate over the last fifty years (Yao, 2019). Excessive cultivation, overgrazing and the cutting down of natural
Higher temperatures significantly lengthen the growing season of shrubs cause serious damage to vegetation, and excessive bare topsoil is
vegetation on the QTP, benefitting vegetation productivity (Wang et al., susceptible to erosion by strong winds, which makes desertification very
2017). The present study found that temperature had a very high posi- easy to occur (Lv and Yu, 2011). Li et al. (2018) reported the presence of
tive influence on desertification on the QTP in the SEM (standardised more severe human disturbances in the east and south of the QTP, and in

8
Y. Zhi et al. Catena 246 (2024) 108392

Fig. 8. Total contribution of driving factors to desertification.

Fig. 9. Correlation between different variables. Note: *, ** and *** denote p < 0.05, p < 0.01 and p < 0.001, respectively.

the central part of the Tibet Autonomous Region, where there is also a Mao et al., 2012; Zhang et al., 2015), to combat desertification. The
higher concentration of livestock (Fig. 2o). Human activities have actual number of livestock raised on the QTP has been controlled
increased desertification of approximately 35.9 % of the southern and (Fig. 10). The expansion of cities and infrastructure development in the
southwestern parts of the QTP, rising to 92.8 % in the northeastern part east–central, southeastern, and southwestern regions around Lhasa (Li
of the QTP (Li et al., 2016). Under the double pressure of changing et al., 2018) tends to trigger and aggravate desertification in the QTP.
climate and sustainable socio-economic development, the Chinese Therefore, appropriate measures should be taken in a targeted manner
government has adopted various policies, such as ‘converting pastures to to balance the links between urbanisation development, human activ-
grasslands’, “ecological migrants” and the setting up of NRs (Yeh, 2005; ities and the development of the regional ecosystem to prevent further

9
Y. Zhi et al. Catena 246 (2024) 108392

Fig. 10. Changes in the annual mean and maximum of ALCC on the QTP from 2000 to 2019.

desertification of the QTP region under human influence. significant advantages in exploring driving mechanisms. GDM quanti-
fied the independent variables and their interactions in the nonlinear
4.3. Interaction impacts of different factors on desertification change response to desertification (Zhu et al., 2020). The path coefficients of
SEM aim to capture the correlation of two factors by eliminating the
The occurrence, development and spatial distribution patterns of effects of other factors, and to estimate the total impact of the driving
desertification depend on the synergistic effect of many factors, not just factor by measuring the direct and indirect impacts (Yang et al., 2021).
the result of a single influencing factor (Li et al., 2016; Nie et al., 2021). This study had some limitations. First, the selection of human factors
This is confirmed by the results of GDM and SEM. Although the results was not exhaustive and lacked the relevant data, so it is challenging to
for the 12 influencing factors differ somewhat between GDM and SEM, quantify some human activity factors. Second, the sample data in the
this is primarily due to the different focuses of these methods. GDM model were not measured data but interpolated data. Moreover, because
emphasizes detecting and quantifying spatial heterogeneity and its numerical processing requires discretisation, considering accuracy and
impact on the dependent variable by categorizing data into discrete efficiency, although the optimal discretisation method was used, the
classes, which can influence results by highlighting certain patterns of results are affected by the varying classification criteria used in the
variation. SEM is a multivariate statistical technique that assesses the discretisation process.
structural relationships between observed and latent variables, Despite these limitations, this study used GDM and SEM methods to
capturing different aspects of these relationships. The unique contribu- deepen the understanding of the driving mechanisms behind desertifi-
tions of GDM and SEM to the analysis of desertification factors enable a cation changes and provided data support for ecological conservation on
more comprehensive and in-depth understanding of the causes of the QTP. In urban expansion, the loss of high-quality land resources,
desertification from different perspectives. This study shows that such as grasslands, farmland and forests, should be strictly limited. The
desertification change is affected by multiple effects of the climatic balance of human activities and the sustainable development of the
factors, natural environment and human activities, with a nonlinear ecological environment must be coordinated. Appropriate measures
response to the interactions among the 12 selected driving factors must be taken to prevent further degradation of the vegetation cover and
(Fig. 6) rather than a single superposition of impact coefficients. Natural exacerbation of desertification under human influence.
environmental factors as underlying drivers have relatively little direct
impact on desertification compared with human activities and climatic 5. Conclusions
factors, but the indirect pathways affecting desertification have the
largest effect (Table 4). In addition, human activities indirectly affect This study assessed the changes in desertification and its drivers on
desertification by influencing climatic factors, suggesting that human the QTP from 2000 to 2020, using GDM and SEM. Desertification
activities will accelerate the influence of climate change on the showed a general reversal trend, with significant reversal in 1.4 % and
biogeochemical cycling of the QTP, which is in line with the findings reversal in 23.8 % of the area, mainly in the northwest. Key drivers
reported by Chen et al. (2013). Overall, the results of GDM and SEM include VT, Pre, ST, and Tem. Interactions between drivers enhance
showed that various factors affect the distribution of desertification in desertification nonlinearly. Climatic factors contribute most to deserti-
the QTP through mutual promotion or mutual containment. fication changes (62 %), followed by human activities (− 36.8 %) and the
natural environment (8 %). Climatic factors, the natural environment,
4.4. Advantages and limitations of this study and human activities all directly impact desertification, with indirect
effects through interactions.
The interactions between the different factors driving desertification
on the QTP are complex. The combination of GDM and SEM has

10
Y. Zhi et al. Catena 246 (2024) 108392

Funding Jiao, K., Gao, J., Liu, Z., 2021. Precipitation drives the NDVI distribution on the tibetan
plateau while high warming rates may intensify its ecological droughts. Remote
Sens. 13, 1305.
This work was financially supported by the Second Tibetan Plateau Kuang, X., Jiao, J.J., 2016. Review on climate change on the Tibetan Plateau during the
Scientific Expedition and Research Program (Grant No. last half century. Geophys. Res. Atmos. 121, 3979–4007.
2019QZKK0305) . Kundu, A., Patel, N.R., Saha, S.K., Dutta, D., 2017. Desertification in western Rajasthan
(India): an assessment using remote sensing derived rain-use efficiency and residual
trend methods. Nat. Hazards. 86 (1), 297–313.
CRediT authorship contribution statement Li, S., Dong, Y., Dong, G., 2001. Sandy Desertification Problem and Sustainable
Development in Qinghai-Xizang (Tibet) Plateau. China Tibetan Research Press,
Beijing, China.
Ying Zhi: Writing – original draft, Validation, Methodology, Formal Li, S., Wang, Z., Zhang, Y., 2017. Crop cover reconstruction and its effects on sediment
analysis, Data curation, Conceptualization. Shulin Liu: Supervision, retention in the Tibetan Plateau for 1900–2000. J. Geog. Sci. 27, 786–800.
Funding acquisition, Conceptualization. Tao Wang: Conceptualization. Li, S., Wu, J., Gong, J., Li, S., 2018. Human footprint in Tibet: Assessing the spatial layout
and effectiveness of nature reserves. Sci. Total Environ. 621, 18–29.
Hanchen Duan: Validation. Wenping Kang: Validation. Li, J., Yao, Q., Zhou, N., Li, F., 2021. Modern aeolian desertification on the Tibetan
Plateau under climate change. Land Degrad. Dev. 32 (5), 1908–1916.
Li, Q., Zhang, C.L., Shen, Y.P., Jia, W.R., Li, J., 2016. Quantitative assessment of the
Declaration of competing interest relative roles of climate change and human activities in desertification processes on
the Qinghai-Tibet Plateau based on net primary productivity. Catena 147, 789–796.
The authors declare that they have no known competing financial Li, L., Zhang, Y., Liu, L., Wu, J., Wang, Z., Li, S., Zhang, H., Zu, J., Ding, M., Paudel, B.,
2018. Spatiotemporal patterns of vegetation greenness change and associated
interests or personal relationships that could have appeared to influence climatic and anthropogenic drivers on the Tibetan Plateau during 2000–2015.
the work reported in this paper. Remote Sens. 10, 1525.
Li, Q., Zhang, C.L., Zhou, N., Shen, Y.P., Wu, Y.Q., Zou, X.Y., Li, J., Jia, W.R., Wang, X.S.,
2018. Spatial distribution of aeolian desertification on the Qinghai-Tibet Plateau.
Data availability J. Desert Res. 38 (4), 690–700.
Li, S.C., Zhang, Y.L., Wang, Z.F., Li, L.H., 2018d. Mapping human influence intensity in
The authors do not have permission to share data. the Tibetan Plateau for conservation of ecological service functions. Ecosyst. Serv.
30, 276–286.
Li, R., Zhang, M., Konstantinov, P., Pei, W., Tregubov, O., Li, G., 2022. Permafrost
Acknowledgments degradation induced thaw settlement susceptibility research and potential risk
analysis in the Qinghai-Tibet Plateau. Catena 214, 106239.
Li, S., Zheng, Y., Luo, P., Wang, X., Li, H., Lin, P., 2007. Desertification in western Hainan
We would like to thank Kun Feng and Ruijie Lu for their assistance Island, China (1959 to 2003). Land Degrad. Dev. 18 (5), 473–485.
during fieldwork. Liang, S., 2001. Narrowband to broadband conversions of land surface albedo I:
algorithms. Rem. Sens. Environ. 76 (2), 213–238.
Liang, P., Yang, X., 2016. Landscape spatial patterns in the Maowusu (Mu Us) Sandy
Appendix A. Supplementary data Land, northern China and their impact factors. Catena 145, 321–333.
Liu, C.X., Melack, J., Tian, Y., Huang, H.B., Jiang, J.X., Fu, X., Zhang, Z.A., 2019.
Supplementary data to this article can be found online at https://doi. Detecting land degradation in Eastern China grasslands with time series
segmentation and residual trend analysis (TSS-RESTREND) and GIMMS NDVI3g
org/10.1016/j.catena.2024.108392. data. Remote. Sens. 11, 1014.
Liu, L., Wang, Y., Wang, Z., Li, D., Zhang, Y., Qin, D., Li, S., 2019. Elevation-dependent
References decline in vegetation greening rate driven by increasing dryness based on three
satellite NDVI datasets on the Tibetan Plateau. Ecol. Indic. 107, 105569.
Lv, C., Yu, B., 2011. The Control Technique and Mode of Land Degradation in Qinghai-
Angelini, M.E., Heuvelink, G.B.M., Kempen, B.M., Héctor, J.M., 2016. Mapping the soils
Tibet Plateau. Science Press, Beijing, China.
of an Argentine Pampas region using structural equation modelling. Geoderma 281,
Ma, Z., Xie, Y., Jiao, J., Wang, X., 2011. The construction and application of an Albedo-
102–118.
NDVI based desertification monitoring model. J. Procedia Environ. Sci. 10,
Burrell, A.L., Evans, J.P., Liu, Y., 2017. Detecting dryland degradation using time series
2029–2035.
segmentation and residual trend analysis (TSS-RESTREND). Rem. Sens. Environ.
Mao, X.F., Wei, X.Y., Xia, J.X., 2012. Evaluation of ‘ecological migrants’ adaptation to
197, 43–57.
their new living area in Three-River Headwater wetlands. China. Procedia Environ.
Chen, H., Zhu, Q., Peng, C., Wu, N., Wang, Y., Fang, X., Gao, Y., Zhu, D., Yang, G.,
Sci. 13, 1346–1353.
Tian, J., Kang, X., 2013. The impacts of climate change and human activities on
Mirzabaev, A.J., Wu, J., Evans, F., García-Oliva, I.A.G., Hussein, M.H., Iqbal, J., Kimutai,
biogeochemical cycles on the Qinghai-Tibetan Plateau. Global Change Biol. 19 (10),
T., Knowles, F., Meza, D., Nedjraoui, F., Tena, M., Türkes¸, R.J., V́ azquez, M., 2019.
2940–2955.
Weltz. Desertification. In: Climate Change and Land: An IPCC Special Report on
Du, Z., Xu, X., Zhang, H., Wu, Z., Liu, Y., 2016. Geographical detector-based
Climate Change, Desertification, Land Degradation, Sustainable Land Management,
identification of the impact of major determinants on aeolian desertification risk.
Food Security, and Greenhouse Gas Fluxes in Terrestrial Ecosystems.
PLoS One 11 (3), e0151331.
Nie, T., Dong, G., Jiang, X., Lei, Y., 2021. Spatio-temporal changes and driving forces of
Duan, H.C., Wang, T., Xue, X., Yan, C.Z., 2019. Dynamic monitoring of aeolian
vegetation coverage on the loess plateau of Northern Shaanxi. Remote Sens. 13 (4),
desertification based on multiple indicators in Horqin Sandy Land, China. Sci. Total
613.
Environ. 650, 2374–2388.
Olsson, L., Eklundh, L., Ardö, J., 2005. A recent greening of the Sahel—trends, patterns
Feng, J., Ding, J.L., Wei, W.Y., 2018. A study of soil salinization in Weigan and Kuqa
and potential causes. J. Arid Environ. 63, 556–566.
Rivers Oasis based on Albedo-MSAVI feature space. China Rural Water Hydropower.
Ontel, I., Cheval, S., Irimescu, A., Boldeanu, G., Amihaesei, V.A., Mihailescu, D.,
2, 147–152.
Nertan, A., Angearu, C.V., Craciunescu, V., 2023. Assessing the recent trends of land
Feng, K., Wang, T., Liu, S., Yan, C., Kang, W., Chen, X., Guo, Z., 2021. Path analysis
degradation and desertification in romania using remote sensing indicators. Remote.
model to identify and analyse the causes of aeolian desertification in Mu Us Sandy
Sens. 15 (19), 4842.
Land. China. Ecol. Indic. 124, 107386.
Peng, W., Kuang, T., Tao, S., 2019. Quantifying influences of natural factors on
Grace, J.B., 2006. Structural Equation Modeling and Natural Systems. Cambridge
vegetation NDVI changes based on geographical detector in Sichuan, western China.
University Press, Cambridge, UK.
J. Clean. Prod. 233, 353–367.
Grace, J.B., Bollen, K.A., 2005. Interpreting the results from multiple regression and
Piedallu, C., Cheret, V., Denux, J.P., Perez, V., Azcona, J.S., Seynave, I., Gégout, J.C.,
structural equation models. Bull. Ecol. Soc. Am. 86 (4), 283–295.
2019. Soil and climate differently impact NDVI patterns according to the season and
Grace, J.B., Keeley, J.E., 2006. A structural equation model analysis of postfire plant
the stand type. Sci. Total Environ. 651, 2874–2885.
diversity in California shrublands. Ecol. Appl. Publ. Ecol. Soc. Am. 16, 503–514.
Shi, S.Y., Yu, J.J., Wang, F., Wang, P., Zhang, Y.C., Jin, K., 2021. Quantitative
Guo, Q., Fu, B., Shi, P., Cudahy, T., Zhang, J., Xu, H., 2017. Satellite monitoring the
contributions of climate change and human activities to vegetation changes over
spatial-temporal dynamics of desertification in response to climate change and
multiple time scales on the Loess Plateau. Sci. Total Environ. 755, 142419.
human activities across the Ordos Plateau. China. Remote. Sens. 9 (6), 525.
Sun, W., Li, B., 2002. The relation between coupling among the principal components of
Guo, B., Wei, C., Yu, Y., Liu, Y., Li, J., Meng, C., Cai, Y., 2022. The dominant influencing
desertification factors and desertification in rear hills of Bashang since 1950.
factors of desertification changes in the source region of Yellow River: Climate
Geographical Res. 21, 391–398.
change or human activity? Sci. Total Environ. 813, 152512.
Sun, Y., Liu, S., Liu, Y., Dong, Y., Li, M., An, Y., Shi, F., Beazley, R., 2021. Effects of the
Hua, D., Hao, X., 2021. Spatiotemporal change and drivers analysis of desertification in
interaction among climate, terrain and human activities on biodiversity on the
the arid region of northwest China based on geographic detector. Environ.
Qinghai-Tibet Plateau. Sci. Total Environ. 794, 148497.
Challenges. 4, 100082.
Sun, Y.X., Liu, S.L., Shi, F.N., An, Y., Li, M.Q., Liu, Y.X., 2020. Spatio-temporal variations
Javzandulam, T., Tateishi, R., Sanjaa, T., 2005. Analysis of vegetation indices for
and coupling of human activity intensity and ecosystem services based on the
monitoring vegetation degradation in semi-arid and arid areas of Mongolia. Int. J.
fourquadrant model on the Qinghai-Tibet Plateau. Sci. Total Environ. 743, 140721.
Environ. Stud. 62 (2), 215–225.

11
Y. Zhi et al. Catena 246 (2024) 108392

UNCCD, 1994. United Nations Convention to Combat Desertification in those Countries Yao, T.D., 2019. Tackling on environmental changes in Tibetan Plateau with focus on
Experiencing Serious Drought and/or Desertification, Particularly in Africa; United water, ecosystem and adaptation. Sci. Bull. 64, 417.
Nations Convention to Combat Desertification. Yao, T.D., Thompson, L.G., Mosbrugger, V., Zhang, F., Ma, Y.M., Luo, T.X., Xu, B.Q.,
UNCED, 1992. Managing Fragile Ecosystems: Combating Desertification and Drought. Yang, X.X., Joswiak, D.R., Wang, W.C., Joswiak, M.E., Devkota, L.P., Tayal, S.,
United Nations Conference on Environment and Development. Jilani, R., Fayziev, R., 2012. Third pole environment (TPE). Environ. Dev. 3, 52–64.
Verstraete, M.M., Pinty, B., 1996. Designing optimal spectral indexes for remote sensing Yeh, E.T., 2005. Green governmentality and pastoralism in western China: ’converting
applications. IEEE Trans. Geosci. Remote Sens. 34 (5), 1254–1265. pastures to grasslands’. Nomadic Peoples. 9 (1–2), 9–30.
Wang, Z., Cui, G., Liu, X., Zheng, K., Lu, Z., Li, H., Wang, G., An, Z., 2021. Greening of the Yu, H.Y., Zhang, Z.D., Zhang, P.M., 2007. Evaluation and dynamic monitoring of land
Qinghai-Tibet Plateau and its response to climate variations along elevation desertification in Qinghai. Arid Zone Res. 24 (2), 153–158.
gradients. Remote Sens. 13, 3712. Zan, Guosheng, Wang, Cuiping, Li, Feng, Liu, Zheng, Sun, Tao, 2023. Key data results
Wang, J.F., Li, X.H., Christakos, G., Liao, Y.L., Zhang, T., Gu, X., Zheng, X.Y., 2010. and trend analysis of the sixth national survey on desertification and sandification.
Geographical detectors-based health risk assessment and its application in the neural For. Resour. Manag. 1–7.
tube defects study of the heshun region. China. Int. J. Geogr. Inf. Sci. 24, 107–127. Zeng, Y., Xiang, N., Feng, Z., Xu, H., 2006. Albedo-NDVI space and remote sensing
Wang, J., Xu, C., 2017. Geodetector: principle and prospective. Acta Geogr. Sin. 72 (1), synthesis index models for desertification monitoring. Sci. Geograph. Sin. 01, 75–81.
116–134. Zhai, X., Yan, C., Xing, X., Jia, H., Wei, X., Feng, K., 2021. Spatial-temporal changes and
Wang, Y., Xu, D., Li, L., 2023. Assessing the influences of natural and human factors upon driving forces of aeolian desertification of grassland in the Sanjiangyuan region from
desertification vulnerability in northern China during 1985–2010. Catena 233, 1975 to 2015 based on the analysis of Landsat images. Environ. Monit. Assess. 193,
107529. 1–16.
Wang, J., Zhang, T., Fu, B., 2016. A measure of spatial stratified heterogeneity. Ecol. Zhang, D., 2000. Quantitative analysis of influential factors on land desertification in
Indic. 67, 250–256. Qinghai Gonghe Basin. J. Desert Res. 20 (1), 59.
Wang, S., Zhang, B., Yang, Q., Chen, G., Yang, B., Lu, L., Shen, M., Peng, Y., 2017. Zhang, C.L., Li, Q., Shen, Y.P., Zhou, N., Wang, X.S., Li, J., Jia, W.R., 2018. Monitoring of
Responses of net primary productivity to phenological dynamics in the Tibetan aeolian desertification on the Qinghai-Tibet Plateau from the 1970s to 2015 using
Plateau. China. Agr. Forest Meteorol. 232, 235–246. Landsat images. Sci. Total Environ. 619, 1648–1659.
Wang, C., Zhao, W., Zhang, Y., 2022. The change in net ecosystem productivity and its Zhang, Y.L., Liu, L.S., Wang, Z.F., Bai, W., Ding, M., Wang, X., Zheng, D., 2019. Spatial
driving mechanism in a mountain ecosystem of arid regions. Northwest China. and temporal characteristics of land use and cover changes in the Tibetan Plateau.
Remote Sens. 14 (16), 4046. Chin. Sci. Bull 64, 2865–2875.
Xu, B., Li, J.Y., Jin, Y.X., Wang, Z.L., Qin, Z.H., Wu, M.Q., Wang, D.L., Zhang, J., Ma, H. Zhang, Y., Wu, X., Qi, W., Li, S., Bai, W., 2015. Characteristics and protection
L., Yang, X.C., 2015. Temporal and spatial variations of grassland desertification effectiveness of nature reserves on the Tibetan Plateau. China. Resour. Sci. 37 (7),
monitoring in Tibet of China. Int. J. Remote. Sens. 36 (19–20), 5150–5164. 1455–1464.
Yang, Z., Gao, X., Lei, J., Meng, X., Zhou, N., 2022. Analysis of spatiotemporal changes Zhang, Y., Zhao, W., Fu, L., Zhao, C., Jia, A., 2020. Land use conversion influences soil
and driving factors of desertification in the Africa Sahel. Catena 213, 106213. respiration across a desert-oasis ecoregion in Northwest China, with consideration of
Yang, L., Shen, F., Zhang, L., Cai, Y., Yi, F., Zhou, C., 2021. Quantifying influences of cold season CO2 efflux and its significance. Catena 188, 104460.
natural and anthropogenic factors on vegetation changes using structural equation Zhu, L., Meng, J., Zhu, L., 2020. Applying Geodetector to disentangle the contributions of
modeling: a case study in Jiangsu Province. China. J. Clean. Prod. 280, 124330. natural and anthropogenic factors to NDVI variations in the middle reaches of the
Yang, L., Zhao, G., Mu, X., Lan, Z., Jiao, J., An, S., Wu, Y., Miping, P., 2023. Integrated Heihe River Basin. Ecol. Indic. 117, 106545.
assessments of land degradation on the Qinghai-Tibet plateau. Ecol. Indic. 147,
109945.

12