Science of the Total Environment 806 (2022) 151310

Contents lists available at ScienceDirect

Science of the Total Environment

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

Spatiotemporal dynamics of wetlands and their driving factors based on

PLS-SEM: A case study in Wuhan
Chao Wang a, Le Ma a, Yan Zhang a, Nengcheng Chen a,b,⁎, Wei Wang a
a
State Key Laboratory of Information Engineering in Surveying, Mapping and Remote Sensing, Wuhan University, 129 Luoyu Road, Wuhan 430079, China
b
National Engineering Research Center of Geographic Information System, China University of Geosciences, Wuhan 430074, China

H I G H L I G H T S G R A P H I C A L A B S T R A C T

• A framework was proposed to explore

the interaction between natural envi-
ronment and human activities on wet-
land changes.
• Four potential indirect impact pathways
from latent variables to wetlands
changes was discover.
• Climate change threatens wetlands.
• The study provides theoretical support
to protect wetlands under different con-
ditions.

a r t i c l e i n f o a b s t r a c t

Article history: Globally, wetlands have been severely damaged due to natural environment and human activities. Understand-
Received 16 August 2021 ing the spatiotemporal dynamics of wetlands and their driving forces is essential for their effective protection.
Received in revised form 25 October 2021 This study proposes a research framework to explore the interaction between the natural environment and
Accepted 25 October 2021
human activities and its impact on wetland changes, by introducing Partial Least Squares Structural Equation
Available online 29 October 2021
Modeling (PLS-SEM) and Geographically Weighted Regression (GWR) model, then applying the methodology
Editor: Fernando A.L. Pacheco in Wuhan, a typical wetland city in China. The validity and reliability evaluation indicated that the PLS-SEM
model is reasonable. The results showed that the area of wetlands in Wuhan decreased by 10.98% in
1990–2018 and four obvious direct pathways of influence were found. Positive soil and terrain conditions are
Keywords: conducive to maintaining wetlands, while rapid urbanization drastically reduce the distribution of wetlands. It
Wetland damage is remarkable that the impact of climate on wetlands is gradually shifting from positive to negative. Furthermore,
PLS-SEM four potential indirect impact pathways affecting wetland distribution shown that urbanization and climate en-
Urbanization hance the negative impact of terrain on wetland distribution, while their impacts on soil weaken soil's direct pos-
Climate change
itive impact. This study provides a quantitative methodology for determining the causes of wetland loss; it can
GWR
also be applied to other cities or regions, which is essential for applying more effective measures to protect wet-
lands.
© 2021 The Authors. Published by Elsevier B.V. This is an open access article under the CC BY license (http://
creativecommons.org/licenses/by/4.0/).

1. Introduction

Wetland ecosystems are an important part of aquatic and terrestrial

ecosystems and play an important role in hydrological regulation, biodi-
⁎ Corresponding author at: State Key Laboratory of Information Engineering in
Surveying, Mapping and Remote Sensing, Wuhan University, 129 Luoyu Road, Wuhan
versity conservation and the maintenance of regional ecological balance
430079, China. (Kumari et al., 2020; Meng et al., 2017). Meanwhile, the wetland and
E-mail address: [email protected] (N. Chen). ecosystem has the highest ecosystem services values of all types of

https://doi.org/10.1016/j.scitotenv.2021.151310
0048-9697/© 2021 The Authors. Published by Elsevier B.V. This is an open access article under the CC BY license (http://creativecommons.org/licenses/by/4.0/).
C. Wang, L. Ma, Y. Zhang et al. Science of the Total Environment 806 (2022) 151310

ecosystems (Davidson et al., 2019). However, in the context of global of wetland degradation considering political, economic, and social fac-
climate change and rapid urbanization, wetlands are being severely tors (Guo and Zhang, 2019; Li et al., 2020a; Song et al., 2021). This eval-
damaged. It is estimated that, in the past 150 years, more than 50% of uation method focuses more on the main risks to wetlands, such as
the world's wetlands have changed, been degraded, or disappeared natural hazards and anthropogenic factors. However, the causes and
(Ramsar Convention on Wetlands, 2018). The existing wetlands are their interactions affecting wetlands have been still unknown. Wetland
still deteriorating, and the loss of urban wetlands is particularly serious degradation occurs under the long-term integrated influence of the
due to the long-term extraction of groundwater, industrial pollution, natural environment and human activities. It is necessary to make a sci-
population growth, aquaculture, and other urbanization processes entific assessment of wetland degradation from a comprehensive per-
(Davidson, 2014; Xu et al., 2019; Yu et al., 2017). Therefore, it is essen- spective, rather than just based on the decline of a specific aspect.
tial to understand the evolutionary trends of wetland and to further ex- Compared with the above methods, Structural Equation Modeling
plore the driving factors of wetland degradation, which can help policy (SEM) is a quantitative research method based on statistical analysis
makers to formulate effective protection policies and management techniques to deal with multifactor causal relationships (Hayes et al.,
plans. 2017). On the basis of integrated path analysis, factor analysis, regres-
The natural environment and human activities are generally consid- sion analysis, and analysis of variance, by establishing a link between
ered the main driving factors leading to structural changes in wetland empirical data and theoretical analysis, the relationships between mul-
ecosystems (Asselen et al., 2013). First, climate is the basis for the for- tiple causes and consequences can be determined. SEM is used to per-
mation and development of wetlands (Neubauer and Verhoeven, form the estimation of latent variables and create a complex variables
2019), in which precipitation has a direct impact on changes, and tem- prediction model (Cepeda-Carrion et al., 2019). This method provides
perature and evapotranspiration regulate the distribution of wetlands a better understanding of the direct and indirect interactions between
by affecting the hydrological cycle (Havril et al., 2018; Hirmas et al., the factors, in addition to the results of a traditional multivariate statis-
2018). In addition, snow and glacier melt can also affect wetland sys- tical analysis (Ren et al., 2021; Schweiger et al., 2016). Considering that
tems (Baltaci et al., 2020; Ray et al., 2019). Second, wetland ecosystems wetland degradation is the result of the interaction between the natural
often contain abundant shallow groundwater resources, and the fluctu- environment and human activities, SEM model was used to explore the
ation in surface water also affects the distribution of wetlands to some compound factors and interactions affecting wetland changes in this
extent (Li et al., 2019). The distribution of wetlands can also be influ- study. In addition, Geographically Weighted Regression (GWR) was in-
enced by elevation and slope (Liu et al., 2013). In addition, the structure troduced to determine the local spatial variation in the driving factors.
and elements of the soil affect the survival of the microbial community Wuhan, as one of the megacities with the richest inland wetland re-
and therefore impact on the wetlands, e.g., the abundance of organic sources in China, is also one of the leading cities in the world for inland
matter in the soil is beneficial to the distribution of wetlands (Hinson urban wetland resources. Wuhan is located in central China and has
et al., 2017). Rapid urbanization increases the risk of wetland degrada- abundant water resources and wetland resources; it is known as the
tion (Mao et al., 2018b), and brings about local climate change (Baltacı “City of Hundred Lakes” (Luo et al., 2018). Due to its special geographi-
et al., 2017). Urban expansion, suburban development, and the replace- cal location and hydrological conditions, it not only requires flood con-
ment of natural wetlands by agricultural and other landscapes increases trol, but is a gathering place for migratory birds and other endangered
wetland vulnerability (Liu et al., 2013). The factors that affect wetland species (Aythya baeri, Acipenser sinensis, Neophocaena asiaeorientalis,
distribution are often synergistic, and the degree of influence will etc.) (Mei et al., 2019; Wang et al., 2021a; Weerasinghe et al., 2021).
change in different periods. With the rapid urbanization and industrial development of recent
The continuous degradation of wetlands and the decline or loss in decades, the reclamation of lakes for farmland, the reduction or even
ecological environment service function has led to a greater awareness disappearance of lakes, frequent waterlogging and the decline of endan-
of the importance of wetland ecosystems (Xu et al., 2019). In general, gered animals have become extremely serious. Consequently, wetland
existing research on wetland changes is focused on several aspects. protection is particularly important. A study of the factors affecting the
The first is analyzing wetland degradation from a microscopic perspec- change of wetland in Wuhan can reflect the actual situation of urban
tive. At the mechanism level, changes in soil structure and elements, bi- wetlands and provide reference prices for other wetland cities. There-
ological and ecological mechanisms, and biogeochemical processes fore, this study analyzes wetland changes and explores the effects of
were analyzed. In this process, the changes in plant and animal popula- the natural environment, human activities, and their interaction on
tions caused by wetland degradation are analyzed (An et al., 2019; the long-term spatial distribution of wetlands. The three objectives of
Bansal et al., 2019; Ma et al., 2020). However, this method relies on this study were to (1) construct an analysis framework to identify the
the selection of samples, which involves the selection of experimental driving factors of wetland changes; (2) quantify the direct and indirect
regions, and there is uncertainty to some extent. The results of the effects and the interaction of natural environment and human activities
study can only have reference value and significance for the region at on wetland changes; and (3) explore the spatial heterogeneity in the in-
this time, and it is difficult to adapt to different study areas. The second fluential factors of wetland distribution.
is using statistical and mechanical models to evaluate wetland degrada-
tion, such as partial least squares regression, correlation analysis, gray 2. Materials and methods
correlation, and multiple linear regression (Jia et al., 2017; Li et al.,
2020b; Lin et al., 2018; Zheng et al., 2017), to quantify the relationships 2.1. The study area
between wetland destruction and a single independent variable, such as
climate, urbanization, or terrain. This method is relatively simple, and The city of Wuhan (113°41′–115° 05′E, 29°58′–31°22′N) in central
the evaluation results are often one-sided. It is gradually being replaced China is rich in wetland resources and is a typical lake wetland city in
by more sophisticated methods (Wang et al., 2021b; Ye and Qiu, 2021). the middle and lower reaches of the Yangtze River (Fig. 1). Wuhan has
The third is using remote sensing images to monitor land use/cover 1.627 × 105 hectares of wetlands, accounting for 18.9% of the city's total
changes (Abdullah et al., 2019; Adnan et al., 2020; Cong et al., 2019; land area. Wuhan consists of seven central districts (includes Jiang'an,
Jia et al., 2021; Zhang et al., 2021), and to study the spatiotemporal evo- Jianghan, Qiaokou, Hanyang, Wuchang, Qingshan, and Hongshan dis-
lution of wetlands in a large area or at a large scale. These studies are tricts) and six suburban districts (includes Huangpi, Dongxihu, Caidian,
mainly to explore dynamic changes in wetland area, and cannot further Hannan, Jiangxia, and Xinzhou districts). The city has a humid continental
explore the causes of wetland degradation. The last is to explore the subtropical monsoon climate, with an average annual temperature of
driving forces of wetland change using wetland ecological risk assess- 16 °C and annual precipitation of 1200 mm (Fang et al., 2019). It hardly
ment methods, such as quantitative analysis of the driving mechanisms ever snows in winter. In 2018, the permanent population grew to 13.64

2
C. Wang, L. Ma, Y. Zhang et al. Science of the Total Environment 806 (2022) 151310

Fig. 1. The location of Wuhan City.

million, and the GDP reached 1.48 trillion RMB (Wuhan Statistics Bureau, water, reservoirs and ponds, tidal flats of water bodies, and marshes
2020). Meanwhile, rapid urbanization and industrialization have become were considered as wetlands in this study. During this study, meteoro-
a serious threat to the wetlands. Balancing economic and social develop- logical data were obtained from the National Meteorological Informa-
ment with healthy ecological development of wetlands is a necessary and tion Centre (http://data.cma.cn/), including the annual average
comprehensive task. Understanding the main driving forces and effects of temperature, precipitation, and evaporation. The operations of reading,
wetland loss is a priority for achieving this goal (Gao et al., 2020; Hu et al., merging, checking, counting, and generation of spatial interpolation
2017). batch codes were performed, and 1-km spatial raster data were interpo-
lated by ANUSPLINE 4.4 software (Price et al., 2000). Soil data included
2.2. Data sources the proportion of clay, sand, silt, organic carbon (OC), and soil pH. Soil
property data were collected from the Resource and Environment Sci-
The study used land-use, socioeconomic, terrain, soil, and meteoro- ence and Data Centre (http://www.resdc.cn/), and soil pH and organic
logical datasets. The land-use datasets were obtained from the Institute carbon were obtained from the Harmonized World Soil Database
of Geographic Sciences and Natural Resources Research, China (http:// (http://www.fao.org/soils-portal/soil-survey/). Elevation was gener-
www.resdc.cn/), including data from 1990, 1995, 2000, 2005, 2010, ated using ASTER DEM (http://www.gscloud.cn). Socio-economic data
2015, and 2018, with a spatial resolution of 30 m. The accuracy of the included the annual average population density, Gross Domestic Prod-
classification was in excess of 94.3%. These data were interpreted uct (GDP), and built-up areas. Population density data were derived
based on the Landsat 5 MSS, Landsat 7 ETM/TM, and Landsat 8 OLI re- from LandScan Global Population Data (https://landscan.ornl.gov/
mote sensing satellite data, and the classification results were verified landscan-datasets) and the Resource and Environment Science and
by unmanned aerial vehicle images and field investigations. The first- Data Center (http://www.resdc.cn/). The GDP data are the product of
level classification of land-use data includes cropland, grassland, wood- the GDP statistical value of the county administrative region where
land, water bodies, built-up land, and unused land, and there are 24 spe- the grid unit is located and the GDP weight ratio of the standardized
cific types (Table 1). Among them, paddy fields of cropland, natural land use type and inhabitant density (Xiangdi et al., 2011).

Table 1 2.3. Methodology

The land use classification system and its description.

Categories Types Description The study proposed a research framework to explore the interac-
tion between natural environment and human activities on wetland
Cropland Paddy field, dry land Land for growing crops.
Woodland Closed forest, shrub, sparse Forestry land for growing trees, changes. Human activities have accelerated the degradation of wet-
wood, other woodland shrubs, bamboo, etc. lands, and the natural environment (including climate, soil, terrain,
Grassland High-coverage grassland, Land mainly grows various types etc.) is an important part of construction and maintenance of wet-
medium-coverage grassland, of herbaceous plants, with a lands. The spatiotemporal evolution of wetlands was analyzed mac-
low-coverage grassland coverage of more than 5%.
Waterbody Natural water, reservoirs and Land for natural inland waters and
roscopically by spatial statistical methods. The Partial Least Squares
ponds, permanent glacial snow, water conservancy facilities. Structural Equation Model (PLS-SEM) and GWR were introduced to
tidal flats, beaches explore the composite factors affecting wetland changes and their
Built-up Urban areas, rural residential Includes urban and rural interactions, and to understand the local spatial variation of driving
land areas, other built-up land residential areas, industrial and
factors. This study takes Wuhan as a case to analyze, as is shown in
mining land, traffic land, among
others. Fig. 2. Since 1990, China's industry and economy have developed
Unused Sand, Gobi, saline alkali land, Land has not been used. rapidly, and the damage to wetlands has been significant. However,
land marshes, the changes in wetland area and patterns driven by multiple factors
barren land, bare rock, other have been slow and gradual. After China's accession to the Ramsar
unused land
Convention in 1992, more attention has been paid to wetland

3
C. Wang, L. Ma, Y. Zhang et al. Science of the Total Environment 806 (2022) 151310

Fig. 2. Overall framework of the wetland changes assessment.

protection (Sun et al., 2015). The spatial distribution of wetlands effects in 1990, 2000, 2010, and 2018 were also discussed. Finally,
in Wuhan for 1990, 2000, 2010, and 2018 was investigated to ex- the GWR model was used to determine the sensitivity and spatial
plore the evolution of wetlands and the driving factors affecting heterogeneity of the wetland distribution’s response to various
these changes in the past 30 years. First, five wetland types of factors.
multitemporal mappings were performed, and the spatial and tem-
poral evolution of wetlands were determined using spatial statistical 2.3.1. Identifying the degree of changes in wetlands
methods. Furthermore, a land-use transition matrix was used to an- Various wetlands provide different levels of ecosystem services. Nat-
alyze changes in wetland landscape types over the past 30 years, ac- ural water, marshes, and beaches, as natural wetlands, can provide hab-
cording to the spatiotemporal transformation among different itats for many animals and plants. Reservoirs can conserve water
wetland types and between wetlands and non-wetlands. At the resources and withstand natural disasters such as floods and droughts.
grid scale, correlation analysis of manifest variables was used to ex- Paddy fields, as human-made wetlands, have lost many of the ecological
plore the correlation between variables in PLS-SEM preliminarily. functions of other wetlands because they are relatively independent of
Wetland distribution data, urbanization data, soil data, meteorologi- the external environment. Therefore, the transition from other wetlands
cal data, and terrain data were used as latent variables to construct to paddy fields, similar to the transition from paddy fields to non-
the PLS-SEM collectively. Subsequently, the indicator loadings for la- wetlands, should also be considered a type of damage (Mao et al.,
tent variables were established, and the hypothesized relationships 2018a; Peng et al., 2020). We categorized wetlands as either functional
between exogenous variables (urbanization, soil, climate, and ter- or non-functional to identify the degree of wetland change; the latter
rain) and endogenous variables (wetland distribution) were calcu- included paddy fields, whereas other wetlands were considered
lated as path coefficients. As a result, the main driving forces and functional (Table 2). The transitions among functional wetlands, non-
effects on wetland changes were evaluated using the path coefficient functional wetlands, and non-wetlands were defined to describe the de-
and indicator loadings. The changes in driving paths and driving gree of wetland change, as shown in Table 3.

4
C. Wang, L. Ma, Y. Zhang et al. Science of the Total Environment 806 (2022) 151310

Table 2 Y ¼ Λy η þ ε ð5Þ
The wetland classification system and its description.

Level 1 Level 2 Description Eqs. (4) and (5) are exogenous and endogenous indicators, respec-
Non-functional Paddy fields Land with irrigation facilities and for growing tively. Λ refers to the relationship between the manifest and latent var-
wetlands aquatic crops. iables; δ and ε refer to the measurement error.
Functional Natural Linear water bodies formed naturally or The structural model (or inner model) is a path diagram reflecting
wetlands water excavated manually, and natural impoundment the relationship between the effects of latent variables (Hoyle, 1995;
areas.
Reservoirs Artificial impoundment areas.
Kline, 2015):
and Ponds
Tidal flats The tidal zone between the water level in normal η ¼ Βη þ Γξ þ ζ ð6Þ
and flooding periods in a river or lake.
Marshes Land with long-term humid, seasonal, or where η is the endogenous latent variable, ξ is the exogenous latent
perennial water
accumulation and moist plants.
variable, and Β represents the effect of the exogenous latent variables on
the endogenous latent variables. Γ represents the effect of some endog-
enous latent variables on other endogenous latent variables. ζ is the re-
2.3.2. Land-use transition matrix gression residuals.
The variation in wetlands was analyzed using the transition matrix In this study, PLS-SEM was used to evaluate the essential factors af-
proposed by Markov (Zhang et al., 2011). A Markov chain can be de- fecting wetland distribution. Compared to Covariance Based Structural
scribed as a series of states, and the model is the dynamic evolution of Equation Modeling (CB-SEM) (Hair et al., 2011), PLS-SEM relaxes the
land-use types in different periods by generating a land-use area transi- assumption of multivariate normal distribution in the parameter esti-
tion matrix. The expression is: mation process (Shen et al., 2016). The wetland ecosystem examined
2 3 in this study, the CB-SEM model, cannot guarantee this assumption of
P 11 ⋯ P 1n normality. Furthermore, PLS-SEM is more suitable for exploratory re-
6 7
P ij ¼ 4 ⋯ ⋮ ⋯ 5 ð1Þ search on small samples and can effectively evaluate the interaction be-
P n1 ⋯ P nn tween variables (Fan et al., 2016; Tenenhaus et al., 2005).
Many factors influence wetland change, including climate, soil struc-
N
∑ j¼1 P ij ¼ 1, i, j−1, 2, . . . , n and 0 ≤ P ij < 1 ð2Þ ture, characteristics of terrain, economics, policy, urbanization, and
more. These factors vary over time and from region to region. This
In this study, Pij is the transition probability from the ith to the jth study takes Wuhan as a case; 14 evaluation indicators were compiled
state; n represents the number of land-use types. Four states were ana- and calculated using temperature, precipitation, evaporation, clay,
lyzed in the periods 1990–2000, 2000–2010, 2010–2018, and 1990–2018. sand, silt, OC, pH, elevation, surface water scarcity, GDP, population,
The Markov chain model is a random process in which the state of and built-up areas as manifest variables, and climate, soil, terrain, and
the land use type at time t + 1 depends only on the current time t, inde- urbanization as latent variables in PLS-SEM. An integrated raster dataset
pendent of any time before t. The state transition of the land use type St with a cell size of 1 km × 1 km was utilized in the PLS-SEM model
between the start time t and the end time t + 1(St+1) can be expressed (Table 4). The climate, soil, population, GDP, and elevation data were
as follows: rasterized into grid layers, and the average pixel value for each grid
was taken as a sample set. In addition, the proportions of built-up
Stþ1 ¼ P ij  St ð3Þ area, surface water, and wetlands in each cell of the grid were calcu-
lated. The surface water scarcity (SW_S) index was calculated by taking
2.3.3. The PLS-SEM model specification negative values for the total proportion of surface water. Table S1 pre-
PLS-SEM is used to estimate the causal network among latent vari- sents the datasets from 1990, 2000, 2010, and 2018 in more detail.
ables. The latent variables can be expressed by a set of manifest vari- The value of each variable was a logarithmic transformation to reduce
ables. The manifest variables are quantities that can be directly the difference by an order of magnitude.
observed or measured, and are also called observed variables. Latent Some studies have shown that changes in wetlands are influenced
variables are those that cannot be directly observed but can be con- by terrain, soil, urbanization, and climate change. Moreover, recent
structed with one or more manifest variables through a certain theory studies have suggested that urban development and climate change
or hypothesis. PLS-SEM usually includes measurement models and
structural models. The measurement model (or external model) is an Table 4
explanatory model consisting of latent variables and manifest variables Modeling indicators.
(Ullman and Bentler, 2012). The following formulas can express this:
Latent Variable Manifest Variable Year
Temperature (℃)
X ¼ Λx ξ þ δ ð4Þ
Climate Precipitation (mm)
Evaporation (mm)
Clay (%)
Sand (%)
Table 3
Soil Silt (%)
The degree of change and definition of wetland types.
OC (%)
Degree of change in wetlands Definition pH
Severely damaged Functional wetland → Non-wetland Elevation (m)
Terrain
Slightly damaged Functional wetland → Non-functional wetland Surface water scarcity (proportion)
Non-functional wetland → Non-wetland GDP (billion yuan)
Unchanged Functional wetland → Functional wetland
Urbanization Population (10,000 people)
Non-functional wetland → Non-functional wetland
Slight recovery Non-wetland → Non-functional wetland Built-up area (proportion)
Non-wetland → Functional wetland Wetland area Wetland area (proportion)
Significant recovery Non-wetland → Functional wetland Note: 1990 ; 2000 ; 2010 ; 2018

5
C. Wang, L. Ma, Y. Zhang et al. Science of the Total Environment 806 (2022) 151310

impact terrain and soil, and have comprehensive effects on wetland accounting for 60.3% of the total wetland area in 2018. They were
change (Moomaw et al., 2018). Thus, their interactions, such as the mostly distributed in the suburban districts of Wuhan. The proportion
coupled impacts of urbanization–soil factors, urbanization–terrain fac- of natural water was ranked second. Wuhan is located at the confluence
tors, climate–terrain factors, and climate–soil factors, should be consid- of the Yangtze River and Han River, and has 187 lakes and abundant nat-
ered when assessing ecological changes. PLS-SEM was initialized based ural water resources. In addition, tidal flats and marshes are primarily
on the above assumptions and established indicators. distributed around rivers and lakes. In general, the wetlands in Wuhan
are widely distributed. In 2018, wetlands accounted for 53.01% of the
2.3.4. Spatial heterogeneity of wetland dynamics to driving factors based on administrative area of Wuhan.
GWR
GWR is usually used to quantify spatial heterogeneity by establish- 3.2. Identification and analysis of wetland dynamic types in Wuhan City
ing relationships between dependent and independent variables with
spatial attributes. A prerequisite for applying the GWR model to analyze The spatial distribution of wetland destruction and restoration types
is the existence of spatial correlation between the spatial distribution of over the last 30 years is shown in Fig. 5. The main types of wetland
wetlands and the driving factors. In this study, each independent vari- changes were slight damage and slight recovery. Slightly damaged re-
able in GWR either promoted or resisted wetland changes. GWR can gions were mainly distributed in the suburban districts, while severely
be given by damaged areas were relatively small, mainly in the central districts.
The slight recovery areas were widely distributed, but the recovery
k
yi ¼ β0 ðui , vi Þ þ ∑ j¼1 βj ðui , vi Þxij þ εi ð7Þ areas were small and mostly distributed along rivers and lakes. From
1990 to 2018, wetland destruction extended from central districts to
where yi is a dependent variable representing the wetland proportion in suburban districts, and the degree of destruction became increasingly
the ith cell, xij is the jth independent variable (driving factor) at position severe.
(ui, vi), and βj(ui, vi) is the intercept at location (ui, vi), which is usually The degree of wetland changes in Wuhan is shown in Fig. 6, and the
used for local estimation by the weighted least square method. k is the transfer destination for the loss of wetlands is shown in Fig. 7. From
total number of factors, and εi denotes the random error term. 1990 to 2000, wetlands showed an increasing trend. The area of recov-
ered wetlands was 14.25 km2 more than the damaged area. From 2000
3. Results and analysis to 2010, the status of wetlands showed a degradation trend, and wet-
land loss was 94.07 km2 more than that recovered, primarily due to
3.1. The spatiotemporal evolution of wetland landscape paddy fields' conversion to built-up land, reservoirs, and ponds, with
198.94 km2 and 99.57 km2, respectively, and most natural water was
From 1990 to 2018, the distribution of wetlands changed signifi- converted to reservoirs and ponds, totaling 69.48 km2. Wetlands
cantly in Wuhan (Fig. 3). Wetlands showed an overall trend of degrada- changed dramatically from 2010 to 2018, up to 471.22 km2 of wetlands
tion, and the wetland area was smallest in 2018, 565.22 km2 less than in was severely damaged, and the slightly recovered area increased to
1990. Among human-made wetlands, the paddy field area decreased by 360.30 km2. During this period, the area of paddy fields decreased sig-
638.43 km2. The area of reservoirs and ponds areas initially increased nificantly, primarily being converted to built-up land and dry farmland
and then decreased, but the area in 2018 was 155.98 km2 larger com- (237.18 km2 and 193.14 km2, respectively). Natural water, reservoirs,
pared with 1990. In natural wetlands, the tidal flat area increased by and ponds were invaded by paddy fields, comprising 40.24 km2 and
93.01 km2, and natural water and marsh areas decreased by 57.4 km2, respectively. In general, wetlands in Wuhan changed greatly
156.42 km2 and 19.36 km2, respectively, over the same period. The re- from 1990 to 2018, and the damaged area was greater than the recov-
sults show that the degradation of wetlands mainly manifested as re- ered area (approximately 414.96 km2). While wetlands degraded, wet-
ductions in paddy fields and natural water. By 2018, human-made land recovery was enhanced, especially from 2010 to 2018, indicating
wetlands accounted for 71.13% of the total wetland area, while natural that policies protecting the wetland environment have an obvious effect
wetlands accounted for 28.87%. and have been gradually strengthened over the past 30 years in Wuhan.
There were also differences in spatial distribution among the various It is necessary to further strengthen protection measures and continue
wetland types (Fig. 4). Wetlands were dominated by paddy fields, wetland reserve construction in the future.

Fig. 3. Changes in wetland areas in Wuhan, 1990–2018.

6
C. Wang, L. Ma, Y. Zhang et al. Science of the Total Environment 806 (2022) 151310

Fig. 4. Spatial distribution of wetlands in Wuhan during 1990–2018.

Fig. 5. The spatial distribution maps of wetland change in 1990–2000 (A), 2000–2010 (B), 2010–2018 (C), and 1990–2018 (D).

7
C. Wang, L. Ma, Y. Zhang et al. Science of the Total Environment 806 (2022) 151310

Fig. 6. Area of each type of wetland change (1990–2018).

3.3. The Main driving forces and driving effects of wetland distribution the value of GDP gradually exacerbating wetland loss. From 1990 to
2018, the correlation coefficient of built-up area, GDP, and population to
3.3.1. Correlation analysis between driving factors and wetland distribution wetland distribution decreased from −0.23 to −0.48, from −0.1 to
In this section, the relationship between the wetland area and the −0.29, and from −0.11 to −0.3, respectively. (3) The low elevation and
other 13 variables is explored. As shown in Fig. 8, there were correlation wide surface water area are beneficial for the distribution of wetlands.
coefficients for the four periods from 1990 to 2018. The results confirmed
that wetland area was significantly and positively correlated with tem- 3.3.2. PLS-SEM models
perature (tem), evaporation (eva), silt, surface water (SW), and OC, but Reliability and validity were assessed to effectively evaluate the
negatively correlated with precipitation (pre), elevation (ele), population credibility of the PLS-SEM model (Chen et al., 2015; Hair et al., 2019).
(pop), GDP, and built-up area (build). In addition, we found: (1) At a con- The reliability of internal coherence was evaluated using Composite Re-
fidence level of p = 0.05, the proportion of clay, OC, and pH in the soil liability (CR) (Jöreskog, 1969), which is generally expected to be greater
were positively correlated with wetland distribution. (2) The confidence than 0.7 and no less than 0.6. The Average Variance Extracted (AVE) is
level of urbanization increased from p = 0.5 to p = 0.001 in the past used to verify convergent validity, which is required to be greater than
30 years, with the continuous increase in population, built-up area, and 0.5 (Nasution et al., 2020).

Fig. 7. A Sankey diagram of land-use type conversions from 1990 to 2018. The percentage values on each side indicate the land-use type of the total area.

8
C. Wang, L. Ma, Y. Zhang et al. Science of the Total Environment 806 (2022) 151310

Fig. 8. Variable correlations and interactions from 1990 to 2018 (*p < 0.05, **p < 0.01, and ***p < 0.001).

As the data in this study satisfied the above conditions well, all factor these factors for wetland changes. The results indicate that urbanization
loadings for all indicators were between 0.489 and 0.961 in Table 5. This and terrain always had negative effects on wetland distribution, with
indicates that all the factor loadings satisfy the requirement for struc- path coefficients of −0.197, −0.305, −0.349, −0.426, and −0.470,
tural validity (Moeinaddini et al., 2020). The Variance Inflation Factor −0.446, −0.674, and −0.615 from 1990, 2000, and 2018, respectively.
(VIF) was used to detect the multicollinearity of the 13 variables, and Soil had a positive effect from 1990 to 2018, with path coefficients of
the results showed that the VIF values of the main manifest variables 0.230, 0.177, 0.207, and 0.231, respectively. Notably, the path coefficient
were from 1 to 5 (Table S2), indicating that there was no severe covari- of climate has changed, and the impact of climate on wetland distribu-
ance among the elements (Thompson et al., 2017). The path coefficients tion was positive in 1990, 2000, and 2010 (β=0.165, 0.159, 0.073 re-
between the latent variables are given in Table S3. Both the model exter- spectively), and negative in 2018 (β = − 0.141).
nal and internal loadings are significant (P < 0.01), and a T-statistic of The results of PLS-SEM model showed in Fig. 9. In 1990, in addition
greater than 1.96 was similarly confirmed. The β values of urbanization, to the four direct pathways affecting wetland distribution, urbanization
climate, soil, and terrain factors and their interactions are also statisti- had an indirect and negative effect on wetland distribution through its
cally significant. In summary, all the above indicators essentially meet effect on soils, and climate had a positive and indirect impact on wet-
the ideal values, proving that the evaluation model is reasonable and re- land distribution through its effect on terrain (Fig. 9a). In 2000, an indi-
liable. rect and positive pathway through urbanization impacted the terrain
and altered wetland distribution (Fig. 9b). In 2010, four indirect path-
3.3.3. Effects of latent variables and manifest variables on wetland distribution ways were identified. The negative impact of urbanization and climate
The PLS-SEM models of the relationships between latent variables on soil indirectly weakened the positive impact of soils on wetlands. Cli-
and manifest variables on wetland distribution from 1990 to 2018 are mate and urbanization positively indirect affected wetlands through
shown in Fig. 9. The PLS-SEM models explained 27.2%, 31.7%, 40.3%, their effect on the terrain (Fig. 9c). In 2018, urbanization did not indi-
and 45.5% of the variance occurring in 1990, 2000, 2010, and 2018, re- rectly affect wetland distribution through its impact on the terrain
spectively. This finding reflects the increasing explanatory power of (Fig. 9d).

9
C. Wang, L. Ma, Y. Zhang et al. Science of the Total Environment 806 (2022) 151310

Table 5
Reliability and validity assessment of the PLS-SEM model.

gep 1990 2000 2010 2018

Construct Loading AVE CR Loading AVE CR Loading AVE CR Loading AVE CR

Climate 0.679 0.861 0.569 0.792 0.850 0.944 0.690 0.869

Temperature 0.953 0.911 0.961 0.927
Precipitation 0.864 0.540 0.959 0.811
Evaporation 0.620 0.766 0.841 0.742
Soil 0.477 0.781 0.475 0.778 0.490 0.700 0.439 0.633
Clay 0.575 0.642 Delete* Delete*
Sand Delete* Delete* Delete* Delete*
Silt 0.747 0.748 0.665 0.809
OC 0.821 0.806 0.787 0.653
pH 0.589 0.551 0.648 0.489
Terrain 0.669 0.798 0.670 0.798 0.610 0.756 0.611 0.754
Elevation 0.922 0.927 0.843 0.897
Surface water 0.698 0.703 0.714 0.646
Urbanization 0.744 0.895 0.629 0.835 0.880 0.711 0.819 0.931
GDP 0.939 0.771 0.784 0.897
Population 0.947 0.789 0.708 0.919
Built-up area 0.674 0.684 0.791 0.898
Wetland area 1.000 1.000 1.000 1.000 1.000 1.000

Note: CR > 0.65; AVE > 0.45; Outer loadings > 0.55. Values that do not satisfy the conditional hypothesis or are not significant are omitted.

3.4. Influential factors of wetlands distribution from spatial heterogeneity part of Huangpi District, the eastern and northern parts of Xinzhou
District, and the southern part of Jiangxia District. The regression coeffi-
The results depicted in Fig. 10 show that the regression coefficients cient of GDP (Fig. 10b) ranged from −0.155 to 0.064. The regression co-
of GWR in 2018 showed significant spatial differences in terms of the efficient of the population (Fig. 10c) was −0.464 to 0.040, with the
degree of influence of different factors on wetland distribution. The re- negative impact gradually weakening from the central to suburban dis-
gression coefficient of the built-up area (Fig. 10a) ranged from −0.300 tricts. The regression coefficient of the proportion of OC (Fig. 10d) in soil
to 0.294, with an overall negative influence, and was the strongest in was −0.703 to 0.279, negatively correlated in Huangpi districts and
the central districts. The positive influence was mainly in the northern positively correlated in other districts. In contrast, the distribution of

Fig. 9. The PLS-SEM model diagram shows the relationship between each variable and the distribution of wetlands in 1990 (a), 2000 (b), 2010 (c), and 2018 (d). The circles represent
latent variables and the rectangles represent manifest variables. The arrows represent the link between them. The dotted orange lines indicate negative correlations and solid dotted
green lines indicate positive correlations.

10
C. Wang, L. Ma, Y. Zhang et al. Science of the Total Environment 806 (2022) 151310

Fig. 10. Spatial distribution regression coefficients and corresponding factor scores from GWR. Input factor scores: (a) Built-up area; (b) GDP; (c) population; (d) surface water; (e) OC; (f)
pH; (g) Clay; (h) silt; (i) evaporation; (j) precipitation; (k) temperature; (l) elevation.

11
C. Wang, L. Ma, Y. Zhang et al. Science of the Total Environment 806 (2022) 151310

the pH regression coefficient (Fig. 10e) was contrary to the OC. The scale wetland changes. This study deepens our understanding of the
amount of clay in the soil (Fig. 10f) showed a positive correlation in driving forces and effects of wetland changes, and has promotion and
the northeast and a negative trend in the southwest. The amount of application value for wetland protection.
silt (Fig. 10j) had mainly positive effects, and the influence in the From 1990 to 2018, the negative impact of urbanization on wetlands
north of Huangpi District was negative. The regression coefficient of became increasingly apparent. Besides the direct effects, the negative
precipitation (Fig. 10h) ranged from −0.300 to 0.294, decreasing from impact of urbanization on soil diminished the direct positive impact of
northwest to southeast along the gradient, with high values reached soil on wetland distribution, attributed to land use related to human ac-
at the junction of Jiangxia and Huangpi districts and Xinzhou district. tivities. Some studies have explained that human-made stress, such as
The regression coefficient of temperature (Fig. 10i) ranged from large-scale soil disturbance activities (land reclamation or road con-
−0.582 to 0.870, and decreased from north to south. Evaporation struction, etc.), results in nutrient loss, changes in the composition of bi-
(Fig. 10g) in the central region of Wuhan was negatively correlated ological communities, and a reduction in biodiversity (Orimoloye et al.,
and positively correlated to the north and south. The regression coeffi- 2020; Zhang et al., 2016). Economic activities, such as aquaculture, in-
cient of the surface water area (Fig. 10k) was −0.087 to 0.622, increas- dustry, agriculture, and urban wastewater, change the water cycle and
ing along the direction from the northwest to the southeast. The soil structure by decreasing the water quality. As a result, the positive
elevation (Fig. 10l) indicated that the higher the elevation, the less suit- impact on wetlands is weakened, and the habitat quality decreases.
able for the distribution of wetlands, and was thus negatively correlated. Therefore, in the development of land use and economic activities, we
It is worth pointing out that surface water, OC in soil, and climate factors should not only build wetland protection areas, prohibit the occupation
were the dominant factors affecting wetland distribution. of wetlands, and put in place other protection measures, but also pay at-
tention to the damage to soil environment, changes in microbial com-
4. Discussion munity, changes in the content of trace elements that are beneficial to
the environment, and so on.
4.1. Direct impact of factors on wetlands distribution The models from 2000 and 2010 showed that the effects of urbaniza-
tion on terrain exacerbated the negative impact of the terrain on wet-
Our results demonstrated that the more rapid urbanization, as indi- lands (Fig. 9b, c). This finding may be related to the acceleration of
cated by the higher density of population, GDP, and built-up area, the economic construction and urbanization during this period. Human de-
more severe the damage to wetlands (Fig. 8). This trend can also be velopment and other activities have caused excessive consumption of
seen in the weaker impact of peripheral areas, such as suburban districts water resources, which exacerbates the process of wetland degradation.
(Fig. 10a–c). Therefore, human activities are the main factors driving The reclamation of lakes and the occupation of lake wetlands have dam-
wetland loss. The expansion of built-up land has led to the destruction aged the structural integrity of wetlands. This has inspired us to protect
of many wetlands. The development of agriculture and aquaculture the wetlands by maintaining the content and sustaining the quality of
has led to the expansion of villages and damaged wetland resources. surface water and groundwater. The fragmentation of wetland land-
It was also found that temperature and evaporation were positively scapes has increased, and the ecological health of wetlands has declined.
correlated with wetland distribution and area during 1990–2018, In 2018, the influence of urbanization on terrain disappeared from the
whereas precipitation showed a negative correlation (Fig. 8) due to model. It has been shown that wetland protection and ecological restora-
the destruction of the ecosystem by rainstorms and floods. This finding tion policies were promoted accordingly, and the interactions between
was consistent with the results of Lemke and Michael (Lemke et al., latent variables affecting wetland distribution gradually stabilized.
2017). It is remarkable that the positive impact of climate factors on The response of wetland distribution to climate change was
wetlands gradually weakened from 1990 to 2010 and turned into a neg- highlighted in this study. The results showed that climate has a favor-
ative impact in 2018 (Fig. 9). This shows that the current climate is be- able effect on wetland distribution by regulating surface water content.
coming increasingly unfriendly to wetland distribution. The conclusions This was because climate change alters hydrological cycles, especially
obtained in the relevant studies are generally consistent: global altering surface water supply, runoff, evapotranspiration, and precipita-
warming and extreme weather may pose a threat to wetlands (Xu tion interception (Havril et al., 2018). We thus conclude that the loss of
et al., 2019). wetlands can be alleviated by improving the regulation of climate on
According to the analysis, we found that abundant surface water re- surface water and enhancing water purification capacity. In 2010, cli-
sources were beneficial to the distribution of wetlands and played an mate change began to alter wetland distribution by affecting the soil
important role in breeding wetland ecology (Fig. 8). Adequate water re- (Fig. 9c), and the negative impact increased in 2018 (Fig. 9d). This find-
sources contributed to the number of wetland reserves in Wuhan dur- ing indicates that factors conducive to wetland distribution in soil are
ing rapid urbanization. In addition, the results showed that a higher weakened in the current climate. Coupled with the gradual deterioration
proportion of organic carbon and silt in the soil was beneficial to the dis- of the direct impact of climate on wetlands, the negative impact of cli-
tribution of wetlands (Fig. 8). In addition, the soil property, organic car- mate on wetlands has intensified. Therefore, this phenomenon must
bon, pH value, and water content affect the diversity of microbial be considered in future studies. We need to pay attention to the impact
communities and jointly affect the ecological regulation capacity of wet- of global warming on the environment, and look for breakthroughs in
lands (Ma et al., 2020). Based on the above analysis, the protection of the research on how climate conditions affect wetland changes, to fur-
water resources should be further strengthened, and biological invasion ther analyze the principles and mechanisms of the influence process.
should be prevented to maintain organic matter in the soil to achieve
sustainable development of wetland resources. 4.3. Limitations and implications

4.2. Indirect impact of latent variables on wetlands distribution Different from existing research on wetland changes, we innova-
tively proposed a research framework to explore the interaction be-
The complex interaction between the natural environment and tween the natural environment and human activities in terms of the
human activities on wetlands is difficult to understand. This study inno- impact on wetland changes, which makes up for the lack of quantitative
vatively found the influence of the interaction between terrain, urbani- analyses in the existing research. This study focused on the importance
zation, climate, and soil on wetlands by using PLS-SEM. The interaction of quantitative driving factors of wetland change, thus providing a new
impact pathways were particularly complex in 2010 and 2000, closely perspective on wetland change. Secondly, this study incorporated natu-
related to accelerated urbanization in this period (Fig. 9). Among ral environment and human activities into the comprehensive analysis
them, urbanization factors play a dominant role in the interdecadal framework, and established a framework for evaluating the complex

12
C. Wang, L. Ma, Y. Zhang et al. Science of the Total Environment 806 (2022) 151310

causal relationship between wetland distribution and driving factors. Acknowledgments

Some indirect pathways affecting changes in wetland distribution
were discovered in the model that affected the distribution of wetlands This work was supported by the National Key R&D Program of China
to some extent. By constructing a balanced system between economic (No. 2018YFB2100603).
and social development and wetland ecosystem health, the impact of The authors would like to thank the editors and anonymous re-
human activities on wetland ecosystems can be controlled in a reason- viewers for their valuable comments and insightful ideas.
able range to achieve sustainable development of the environment and
society. This study also provides theoretical support for wetland conser- Appendix A. Supplementary data
vation in the 17 specific Indicators of the SDGs, such as SDG 6.6 (protect
and restore water-related ecosystems), SDG 15.1 (ensure the conserva- Supplementary data to this article can be found online at https://doi.
tion, restoration, and sustainable use of terrestrial and inland freshwater org/10.1016/j.scitotenv.2021.151310.
ecosystems and their services), and so on.
The causes of wetland changes often vary in different regions. By intro- References
ducing manifest and latent variables appropriate to the study areas, the
Abdullah, A.Y.M., Masrur, A., Adnan, M.S.G., Baky, M.A.A., Hassan, Q.K., Dewan, A., 2019.
framework can be extended to different study areas to provide targeted Spatio-temporal Patterns of Land Use/Land Cover Change in the Heterogeneous
guidance on wetland conservation. Nevertheless, the factors affecting Coastal Region of Bangladesh Between 1990 and 2017. 11 , p. 790.. https://www.
wetland distribution should be considered more comprehensively, espe- mdpi.com/2072-4292/11/7/790.
Adnan, M.S.G., Abdullah, A.Y.M., Dewan, A., Hall, J.W., 2020. The effects of changing land
cially factors related to biodiversity and water level. Moreover, although use and flood hazard on poverty in coastal Bangladesh. Land Use Policy 99, 104868.
SEM has the capacity to analyze a finite number of samples, obtaining https://doi.org/10.1016/j.landusepol.2020.104868.
more effective samples can improve the model's accuracy. In addition, An, J., Liu, C., Wang, Q., Yao, M., Rui, J., Zhang, S., Li, X., 2019. Soil bacterial community
structure in Chinese wetlands. Geoderma 337, 290–299. https://doi.org/10.1016/j.
more multisource data should be considered for joint use in the future,
geoderma.2018.09.035.
to facilitate a more comprehensive assessment of wetland changes. Asselen, S.V., Verburg, P.H., Vermaat, J.E., Janse, J.H., 2013. Drivers of wetland conversion:
a global meta-analysis. PLOS ONE 8, e81292. https://doi.org/10.1371/journal.pone.
5. Conclusions 0081292.
Baltacı, H., Kındap, T., Ünal, A., Karaca, M., 2017. The influence of atmospheric circulation
types on regional patterns of precipitation in Marmara (NW Turkey). Theor. Appl.
This study innovatively constructs a research framework to explore Climatol. 127, 563–572. https://doi.org/10.1007/s00704-015-1653-1.
interactions between the natural environment and human activities Baltaci, H., Silva, M.C., Gomes, H., 2020. Climatological conditions of the Black Sea-effect
snowfall events in Istanbul, Turkey. Int. J. Climatol. 2020, 1–12. https://doi.org/10.
on the wetland changes by introducing PLS-SEM and GWR model. The 1002/joc.6944.
interactions of the driving factors and the magnitude of the forces are Bansal, S., Lishawa, S.C., Newman, S., Tangen, B.A., Wilcox, D., Albert, D., Anteau, M.J.,
systematically explored from the perspective of the spatial and tempo- Chimney, M.J., Cressey, R.L., DeKeyser, E., Elgersma, K.J., Finkelstein, S.A., Freeland, J.,
Grosshans, R., Klug, P.E., Larkin, D.J., Lawrence, B.A., Linz, G., Marburger, J., Noe, G.,
ral dynamics of wetlands. Taking Wuhan as a case study, the degree and
Otto, C., Reo, N., Richards, J., Richardson, C., Rodgers, L., Schrank, A.J., Svedarsky, D.,
compound driving factors of wetland change from 1990 to 2018 were Travis, S., Tuchman, N., Windham-Myers, L., 2019. Typha (Cattail) invasion in North
analyzed. The results showed that the area of wetland decreased from American wetlands: biology, regional problems, impacts, ecosystem services, and
1990 to 2018 by 565.22 km or 10.98%, while restored area increased, es- management. Wetlands 39, 645–684. https://doi.org/10.1007/s13157-019-01174-7.
Bureau, Wuhan Statistics, 2020. Wuhan Statistical Bulletin. Statistics Bureau of Wuhan
pecially during 2010–2018. Focusing on the driving factors of wetland City. http://tjj.wuhan.gov.cn/tjfw/tjgb/.
changes, we indicated that rapid urbanization and a lack of water re- Cepeda-Carrion, G., Cegarra-Navarro, J.-G., Cillo, V., 2019. Tips to use partial least squares
sources threatened wetland distribution, while a higher proportion of structural equation modelling (PLS-SEM) in knowledge management. J. Knowl.
Manag. 23, 67–89. https://doi.org/10.1108/JKM-05-2018-0322.
silt or organic carbon promoted wetland distribution. Meanwhile, the Chen, X., Liu, X., Hu, D., 2015. Assessment of sustainable development: a case study of
climate has become increasingly unfriendly to wetland distribution. Wuhan as a pilot city in China. Ecol. Indic. 50, 206–214. https://doi.org/10.1016/j.
More importantly, the PLS-SEM model detected some indirect path- ecolind.2014.11.002.
Cong, P., Chen, K., Qu, L., Han, J., 2019. Dynamic changes in the wetland landscape pattern
ways that affect the distribution of wetlands. Urbanization increased
of the Yellow River Delta from 1976 to 2016 based on satellite data. Chin. Geogr. Sci.
the negative impact on wetland distribution through its impact on the 29, 372–381. https://doi.org/10.1007/s11769-019-1039-x.
terrain, while the impact of urbanization on the soil weakened the di- Davidson, N.C., 2014. How much wetland has the world lost? Long-term and recent
rectly positive effect of soil on wetlands. The negative effects of terrain trends in global wetland area. Mar. Freshw. Res. 65, 934–941. https://doi.org/10.
1071/MF14173.
on wetlands are enhanced by climate. In addition, the negative impact Davidson, N.C., van Dam, A.A., Finlayson, C.M., McInnes, R.J., 2019. Worth of wetlands: re-
of climate on wetlands is also exacerbated by the impact of climate on soil. vised global monetary values of coastal and inland wetland ecosystem services %J.
Overall, this study proposed an innovative method to effectively en- Mar. Freshw. Res. 70, 1189–1194. https://doi.org/10.1071/MF18391.
Fan, Y., Chen, J., Shirkey, G., John, R., Wu, S.R., Park, H., Shao, C., 2016. Applications of struc-
hance our understanding of the driving forces and effects of wetland tural equation modeling (SEM) in ecological studies: an updated review. Ecol. Pro-
change. These findings have important implications for the application cess. 5, 19. https://doi.org/10.1186/s13717-016-0063-3.
of more effective measures to protect wetlands under different condi- Fang, J., Hu, J., Shi, X., Zhao, L., 2019. Assessing disaster impacts and response using social
media data in China: a case study of 2016 Wuhan rainstorm. Int. J. Disaster Risk
tions and can promote the achievement of the ecological and environ-
Reduct. 34, 275–282. https://doi.org/10.1016/j.ijdrr.2018.11.027.
mental indicators in the UN SDGs 2030. Gao, X., Zhang, A., Sun, Z., 2020. How regional economic integration influence on urban
land use efficiency? A case study of Wuhan metropolitan areaChina. 90, 104329.
CRediT authorship contribution statement https://doi.org/10.1016/j.landusepol.2019.104329.
Guo, Z., Zhang, M., 2019. The conservation efficacy of coastal wetlands in China based on
landscape development and stress. Ocean Coast. Manag. 175, 70–78. https://doi.org/
Chao Wang: Methodology, Writing – review & editing, Funding ac- 10.1016/j.ocecoaman.2019.03.018.
quisition. Le Ma: Methodology, Writing – original draft, Writing – Hair, J.F., Ringle, C.M., Sarstedt, M., 2011. PLS-SEM: indeed a silver bullet. J. Mark. Theory
Pract. 19, 139–152. https://doi.org/10.2753/MTP1069-6679190202.
review & editing, Visualization. Yan Zhang: Writing – review & editing, Hair, J.F., Risher, J.J., Sarstedt, M., Ringle, C.M., 2019. When to use and how to report the
Formal analysis. Nengcheng Chen: Supervision, Writing – review & results of PLS-SEM. Eur. Bus. Rev. 31, 2–24. https://doi.org/10.1108/EBR-11-2018-
editing. Wei Wang: Project administration, Funding acquisition. 0203.
Havril, T., Tóth, Á., Molson, J.W., Galsa, A., Mádl-Szőnyi, J., 2018. Impacts of predicted cli-
mate change on groundwater flow systems: can wetlands disappear due to recharge
Declaration of competing interest reduction? J. Hydrol. 563, 1169–1180. https://doi.org/10.1016/j.jhydrol.2017.09.020.
Hayes, A.F., Montoya, A.K., Rockwood, N.J., 2017. The analysis of mechanisms and their
The authors declare that they have no known competing financial contingencies: PROCESS versus structural equation modeling. Australas. Mark.
J. AMJ 25, 76–81. https://doi.org/10.1016/j.ausmj.2017.02.001.
interests or personal relationships that could have appeared to influ- Hinson, A.L., Feagin, R.A., Eriksson, M., Najjar, R.G., Herrmann, M., Bianchi, T.S., Kemp, M.,
ence the work reported in this paper. Hutchings, J.A., Crooks, S., Boutton, T., 2017. The spatial distribution of soil organic

13
C. Wang, L. Ma, Y. Zhang et al. Science of the Total Environment 806 (2022) 151310

carbon in tidal wetland soils of the continental United States. Glob. Chang. Biol. 23, South Africa. J. King Saud Univ. - Sci. 32, 90–96. https://doi.org/10.1016/j.jksus.2018.
5468–5480. https://doi.org/10.1111/gcb.13811. 03.004.
Hirmas, D.R., Giménez, D., Nemes, A., Kerry, R., Brunsell, N.A., Wilson, C.J., 2018. Climate- Peng, K., Jiang, W., Deng, Y., Liu, Y., Wu, Z., Chen, Z., 2020. Simulating wetland changes
induced changes in continental-scale soil macroporosity may intensify water cycle. under different scenarios based on integrating the random forest and CLUE-S models:
Nature 561, 100–103. https://doi.org/10.1038/s41586-018-0463-x. a case study of Wuhan urban agglomeration. Ecol. Indic. 117, 106671. https://doi.org/
Hoyle, R.H., 1995. Structural Equation Modeling: Concepts, Issues, and Applications. SAGE 10.1016/j.ecolind.2020.106671.
Publications. https://books.google.com.sg/books?id=CoM5DQAAQBAJ. Price, D.T., McKenney, D.W., Nalder, I.A., Hutchinson, M.F., Kesteven, J.L., 2000. A compar-
Hu, S., Niu, Z., Chen, Y., Li, L., Zhang, H., 2017. Global wetlands: potential distribution, wet- ison of two statistical methods for spatial interpolation of Canadian monthly mean
land loss, and status. Sci. Total Environ. 586, 319–327. https://doi.org/10.1016/j. climate data. Agric. For. Meteorol. 101, 81–94. https://doi.org/10.1016/S0168-1923
scitotenv.2017.02.001. (99)00169-0.
Jia, Q., Yu, W., Zhou, L., Liang, C., 2017. Atmospheric and surface-condition effects on CO2 Ramsar Convention on Wetlands, 2018. Global Wetland Outlook: State of the
exchange in the Liaohe Delta wetlandChina. 9. https://doi.org/10.3390/w9100806. World’s Wetlands and their Services to People. Ramsar Convention Secretar-
Jia, M., Wang, Z., Mao, D., Ren, C., Wang, C., Wang, Y., 2021. Rapid, robust, and automated iat, Gland, Switzerland. https://global-wetland-outlook.squarespace.com/s/
mapping of tidal flats in China using time series Sentinel-2 images and Google earth u1sr9dxmee4mw443pvhl6jp7ap9l1o.
engine. Remote Sens. Environ. 255, 112285. https://doi.org/10.1016/j.rse.2021. Ray, A.M., Sepulveda, A.J., Irvine, K.M., Wilmoth, S.K.C., Thoma, D.P., Patla, D.A., 2019. Wet-
112285. land drying linked to variations in snowmelt runoff across grand Teton and yellow-
Jöreskog, K.G., 1969. Statistical Analysis of Sets of Congeneric Tests. Educational Testing stone national parks. Sci. Total Environ. 666, 1188–1197. https://doi.org/10.1016/j.
Service. https://books.google.ch/books?id=xK58oAEACAAJ. scitotenv.2019.02.296.
Kline, R.B., 2015. Principles and Practice of Structural Equation Modeling. Fourth edition.
Ren, L., Li, J., Li, C., Dang, P., 2021. Can ecotourism contribute to ecosystem? Evidence from
Guilford Publications. https://books.google.com.sg/books?id=Q61ECgAAQBAJ.
local residents' ecological behaviors. Sci. Total Environ. 757, 143814. https://doi.org/
Kumari, R., Shukla, S.K., Parmar, K., Bordoloi, N., Kumar, A., Saikia, P., 2020. Wetlands con-
10.1016/j.scitotenv.2020.143814.
servation and restoration for ecosystem services and halt biodiversity loss: an Indian
Schweiger, E.W., Grace, J.B., Cooper, D., Bobowski, B., Britten, M., 2016. Using structural
perspective. In: Upadhyay, A.K., Singh, R., Singh, D.P. (Eds.), Restoration of Wetland
equation modeling to link human activities to wetland ecological integrity. Ecosphere
Ecosystem: A Trajectory Towards a Sustainable Environment. Springer, Singapore,
7, e01548. https://doi.org/10.1002/ecs2.1548.
Singapore, pp. 75–85 https://doi.org/10.1007/978-981-13-7665-8_6.
Lemke, M.J., Hagy, H.M., Dungey, K., Casper, A.F., Lemke, A.M., VanMiddlesworth, T.D., Shen, W., Xiao, W., Wang, X., 2016. Passenger satisfaction evaluation model for urban rail
Kent, A., 2017. Echoes of a flood pulse: short-term effects of record flooding of the Il- transit: a structural equation modeling based on partial least squares. Transp. Policy
linois River on floodplain lakes under ecological restoration. Hydrobiologia 804, 46, 20–31. https://doi.org/10.1016/j.tranpol.2015.10.006.
151–175. https://doi.org/10.1007/s10750-017-3220-5. Song, F., Su, F., Mi, C., Sun, D., 2021. Analysis of driving forces on wetland ecosystem ser-
Li, Y., Zhang, Q., Lu, J., Yao, J., Tan, Z., 2019. 35, 25–36. https://doi.org/10.1002/rra.3389. vices value change: a case in Northeast China. Sci. Total Environ. 751, 141778. https://
Li, Z., Jiang, W., Wang, W., Chen, Z., Ling, Z., Lv, J., 2020a. Ecological risk assessment of the doi.org/10.1016/j.scitotenv.2020.141778.
wetlands in Beijing-Tianjin-Hebei urban agglomeration. Ecol. Indic. 117, 106677. Sun, Z., Sun, W., Tong, C., Zeng, C., Yu, X., Mou, X., 2015. China's coastal wetlands: conser-
https://doi.org/10.1016/j.ecolind.2020.106677. vation history, implementation efforts, existing issues and strategies for future im-
Li, Z., Liu, M., Hu, Y., Xue, Z., Sui, J., 2020b. The spatiotemporal changes of marshland and provement. Environ. Int. 79, 25–41. https://doi.org/10.1016/j.envint.2015.02.017.
the driving forces in the Sanjiang plain, Northeast China from 1980 to 2016. Ecol. Pro- Tenenhaus, M., Vinzi, V.E., Chatelin, Y.-M., Lauro, C., 2005. PLS path modeling. Comput.
cess. 9, 24. https://doi.org/10.1186/s13717-020-00226-9. Stat. Data Anal. 48, 159–205. https://doi.org/10.1016/j.csda.2004.03.005.
Lin, W., Cen, J., Xu, D., Du, S., Gao, J., 2018. Wetland landscape pattern changes over a pe- Thompson, C.G., Kim, R.S., Aloe, A.M., Becker, B.J., 2017. Extracting the variance inflation
riod of rapid development (1985–2015) in the ZhouShan Islands of Zhejiang prov- factor and other multicollinearity diagnostics from typical regression results. Basic
ince, China. Estuar. Coast. Shelf Sci. 213, 148–159. https://doi.org/10.1016/j.ecss. Appl. Soc. Psychol. 39, 81–90. https://doi.org/10.1080/01973533.2016.1277529.
2018.08.024. Ullman, J.B., Bentler, P.M., 2012. Structural Equation Modeling. Handbook of Psychology,
Liu, W., Liu, G., Liu, H., Song, Y., Zhang, Q., 2013. Subtropical reservoir shorelines have re- Second edition https://doi.org/10.1002/9781118133880.hop202023.
duced plant species and functional richness compared with adjacent riparian wet- Wang, D., Li, M., Xiong, C., Yan, Y., Hu, J., Hao, M., Liang, B., Chen, J., Chen, G., Yang, G., Li, Y.,
lands. Environ. Res. Lett. 8, 044007. https://doi.org/10.1088/1748-9326/8/4/044007. Zhang, J., Gulyaeva, M., Shestopalov, A., Shi, W., Bi, Y., Liu, H., Wang, H., Liu, D., Chen, J.,
Luo, Q., Zhang, X., Li, Z., Yang, M., Lin, Y., 2018. The effects of China’s ecological control line 2021a. Ecology of avian influenza viruses in migratory birds wintering within the
policy on ecosystem services: the case of Wuhan City. Ecol. Indic. 93, 292–301. Yangtze River wetlands. Sci. Bull. 66, 2014–2024. https://doi.org/10.1016/j.scib.
https://doi.org/10.1016/j.ecolind.2018.05.009. 2021.03.023.
Ma, L., Jiang, X., Liu, G., Yao, L., Liu, W., Pan, Y., Zuo, Y., 2020. Environmental factors and Wang, X., Liu, G., Xiang, A., Qureshi, S., Li, T., Song, D., Zhang, C., 2021b. Quantifying the
microbial diversity and abundance jointly regulate soil nitrogen and carbon biogeo- human disturbance intensity of ecosystems and its natural and socioeconomic driv-
chemical processes in Tibetan wetlands. Environmental Science & Technology 54, ing factors in urban agglomeration in South China. Environ. Sci. Pollut. Res. https://
3267–3277. https://doi.org/10.1021/acs.est.9b06716. doi.org/10.1007/s11356-021-16349-1.
Mao, D., Luo, L., Wang, Z., Wilson, M.C., Zeng, Y., Wu, B., Wu, J., 2018a. Conversions be- Weerasinghe, R., Li, H., Fan, L., Zou, Z., Fang, C., 2021. Research on water restoration sys-
tween natural wetlands and farmland in China: a multiscale geospatial analysis. Sci. tem in Wuhan City based on Pocket Park concept. E3S Web of Conferences 237
Total Environ. 634, 550–560. https://doi.org/10.1016/j.scitotenv.2018.04.009. https://doi.org/10.1051/e3sconf/202123701012.
Mao, D., Wang, Z., Wu, J., Wu, B., Zeng, Y., Song, K., Yi, K., Luo, L., 2018b. China's Wetlands Xiangdi, H., Yi, Z., Shixin, W., Litao, W., Yanfang, H., 2011. Spatialization approach to 1km
Loss to Urban Expansion. 29, pp. 2644–2657. https://doi.org/10.1002/ldr.2939. grid GDP based on remote sensing. 2011 International Conference on Multimedia
Mei, Z., Han, Y., Dong, L., Turvey, S.T., Hao, Y., Wang, K., Wang, D., 2019. The impact of fish- Technology, pp. 739–742 https://doi.org/10.1109/ICMT.2011.6001941.
eries management practices on the survival of the Yangtze finless porpoise in China.
Xu, T., Weng, B., Yan, D., Wang, K., Li, X., Bi, W., Li, M., Cheng, X., Liu, Y., 2019a.
Aquat. Conserv. Mar. Freshwat. Ecosyst. 29, 639–646. https://doi.org/10.1002/aqc.
Wetlands of international importance: status. 16, 1818. https://www.mdpi.
3078.
com/1660-4601/16/10/1818.
Meng, W., He, M., Hu, B., Mo, X., Li, H., Liu, B., Wang, Z., 2017. Status of wetlands in China:
Ye, Y., Qiu, H., 2021. Environmental and social benefits, and their coupling coordination in
a review of extent, degradation, issues and recommendations for improvement.
urban wetland parks. Urban For. Urban Green. 60, 127043. https://doi.org/10.1016/j.
Ocean & Coastal Management 146, 50–59. https://doi.org/10.1016/j.ocecoaman.
ufug.2021.127043.
2017.06.003.
Moeinaddini, M., Asadi-Shekari, Z., Aghaabbasi, M., Saadi, I., Zaly Shah, M., Cools, M., 2020. Yu, H., Zhang, F., Kung, H.-T., Johnson, V.C., Bane, C.S., Wang, J., Ren, Y., Zhang, Y., 2017.
Proposing a new score to measure personal happiness by identifying the contributing Analysis of land cover and landscape change patterns in ebinur Lake wetland Na-
factors. Measurement 151, 107115. https://doi.org/10.1016/j.measurement.2019. tional Nature Reserve, China from 1972 to 2013. Wetl. Ecol. Manag. 25, 619–637.
107115. https://doi.org/10.1007/s11273-017-9541-3.
Moomaw, W.R., Chmura, G.L., Davies, G.T., Finlayson, C.M., Middleton, B.A., Natali, S.M., Zhang, R., Tang, C., Ma, S., Yuan, H., Gao, L., Fan, W., 2011. Using Markov chains to analyze
Perry, J.E., Roulet, N., Sutton-Grier, A.E., 2018. Wetlands in a changing climate: sci- changes in wetland trends in arid Yinchuan plain, China. Math. Comput. Model. 54,
ence, policy and management. Wetlands 38, 183–205. https://doi.org/10.1007/ 924–930. https://doi.org/10.1016/j.mcm.2010.11.017.
s13157-018-1023-8. Zhang, Y., Dong, S., Gao, Q., Liu, S., Zhou, H., Ganjurjav, H., Wang, X., 2016. Climate change
Nasution, M.I., Fahmi, M., Jufrizen, Muslih, Prayogi, M.A., 2020. The quality of small and and human activities altered the diversity and composition of soil microbial commu-
medium enterprises performance using the structural equation model-part Least nity in alpine grasslands of the Qinghai-tibetan plateau. Sci. Total Environ. 562,
Square (SEM-PLS). J. Phys. Conf. Ser. 1477, 052052. https://doi.org/10.1088/1742- 353–363. https://doi.org/10.1016/j.scitotenv.2016.03.221.
6596/1477/5/052052. Zhang, M., Lin, H., Long, X., Cai, Y., 2021. Analyzing the spatiotemporal pattern and driving
Neubauer, S.C., Verhoeven, J.T.A., 2019. Wetland effects on global climate: mechanisms, factors of wetland vegetation changes using 2000–2019 time-series landsat data. Sci.
impacts, and management recommendations. In: An, S., Verhoeven, J.T.A. (Eds.), Wet- Total Environ. 780, 146615. https://doi.org/10.1016/j.scitotenv.2021.146615.
lands: Ecosystem Services, Restoration and Wise Use. Springer International Publish- Zheng, X.J., Sun, P., Zhu, W.H., Xu, Z., Fu, J., Man, W.D., Li, H.L., Zhang, J., Qin, L., 2017. Land-
ing, Cham, pp. 39–62 https://doi.org/10.1007/978-3-030-14861-4_3. scape dynamics and driving forces of wetlands in the Tumen River basin of China
Orimoloye, I.R., Kalumba, A.M., Mazinyo, S.P., Nel, W., 2020. Geospatial analysis of wetland over the past 50 years. Landsc. Ecol. Eng. 13, 237–250. https://doi.org/10.1007/
dynamics: wetland depletion and biodiversity conservation of isimangaliso wetland, s11355-016-0304-8.

14