Global Ecology and Conservation 59 (2025) e03585

Contents lists available at ScienceDirect

Global Ecology and Conservation

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

Original research article

Machine learning-based mapping wetland dynamics of the largest

freshwater lake in China
Fangyuan Bu a, Zhijun Dai a,b,* , Xuefei Mei a , Ao Chu c , Jinping Cheng d, Ling Lan e
a
State Key Laboratory of Estuarine and Coastal Research, East China Normal University, Shanghai 200062, China
b
Laboratory for Marine Geology, Qingdao Marine Science and Technology Center, Qingdao 266061, China
c
Institute of Water Science and Technology, Hohai University, Nanjing 210098, China
d
Department of Science and Environmental Studies, The Education University of Hong Kong, New Territories, Hong Kong, China
e
State Key Laboratory of Hydraulics and Mountain River Engineering, Sichuan University, Chengdu 610065, China

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

Keywords: Lake wetlands are critical components of freshwater ecosystems, providing crucial roles in water
Mudflat loss regulation and biodiversity conservation. However, these lake wetlands are experiencing perva­
Hydrological behavior sive and often irreparable losses due to anthropogenic activities and climate change. This study
Sediment discharge
utilizes remote sensing imagery from 1987 to 2023 based on a machine learning technique to
Vegetation expansion
Anthropogenic activities
analyze the dynamic changes in the wetlands of Poyang Lake, the largest freshwater lake in
China. The results revealed a substantial decline in Poyang’s wetland area, totaling 242.71 km2
over the study period. The mudflat areas within the wetlands demonstrated a marked reduction of
64 %, equating to a loss of 617.60 km², predominantly in the northern zone of the lake. In
contrast, vegetation coverage increased significantly by 36 %, rising from 1035.32 km² in 1987 to
1411.99 km² in 2023, characterized by a net gain of 361.05 km² primarily due to the
encroachment of mudflats. Increases in rainfall have expanded water bodies in dish-shaped lakes,
encroaching on the transitional mudflat areas, thus exacerbating wetland degradation. Addi­
tionally, intensified human activities, particularly the construction of the Three Gorges Dam, have
profoundly altered the river-lake topography gradient and enhanced Poyang Lake’s discharge
capacity into the Changjiang River. This alteration appears to be a primary driver of the observed
vegetation expansion. Concurrently, reservoir construction within the Poyang Lake basin has
trapped water and sediment, and sand extraction within the lake basin has water and sediment
entrapment, exacerbated by sand extraction activities that have directly diminished mudflat
areas. The present work highlights the ongoing degradation trends of lake wetlands and eluci­
dates the driving forces behind the evolution of Poyang Lake, providing valuable insights for
management and conservation strategies aimed at promoting the restoration and sustainable
development of lake wetlands.

1. Introduction

Lacustrine wetland, despite covering only 2 % of the earth’s land surface, plays an irreplaceable role in water provision and pu­
rification, regulating regional climate, supporting biodiversity, and providing habitat for endemic and endangered species (Kayranli

* Corresponding author at: State Key Laboratory of Estuarine and Coastal Research, East China Normal University, Shanghai 200062, China.
E-mail address: zjdai@sklec.ecnu.edu.cn (Z. Dai).

https://doi.org/10.1016/j.gecco.2025.e03585
Received 5 March 2025; Received in revised form 3 April 2025; Accepted 12 April 2025
Available online 14 April 2025
2351-9894/© 2025 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/).
F. Bu et al. Global Ecology and Conservation 59 (2025) e03585

et al., 2010; Reynaud and Lanzanova, 2017; Sterner et al., 2020; Bai et al., 2020; Mei et al., 2024). However, the global lake wetlands
are facing increasing threats from both climate change and human activities, resulting in unprecedented degradation that poses a
significant threat to local organisms and regional ecosystems (Weiming and Yao, 2011; Yuan et al., 2014; Lu et al., 2021; Zhong et al.,
2022). Consequently, a comprehensive assessment of lake wetlands’ response to climate changes and anthropogenic disturbances is
crucial for the protection and restoration of these vital ecosystems.
Climate change poses a significant threat to wetlands, primarily through alterations in rainfall patterns in terms of magnitude
reduction, frequency decrease, and intensity strengthening (White et al., 2022). For instance, Lake Chad has drastically shrunk from
25,000 km² in the 1960s to just 2000 km² in 2015, largely due to declining precipitation (Mahmood and Jia, 2019). Meanwhile, Lake
Urmia saw a significant reduction in the area starting in the mid-2000s, losing 88 % of its surface by 2015, primarily as a result of a
decrease in rainfall frequency (Radmanesh et al., 2022). In the Maumee River watershed, increased rainfall intensity has resulted in
higher runoff and discharge, exacerbating wetland degradation and threatening downstream ecosystems (Williams and King, 2020).
To make things worse, the future of lacustrine wetlands worldwide is even more pessimistic. Predicted by Xi et al., (2021) suggest a
6000 km2 lake wetlands are expected to disappear by 2100 due to climate change.
Recent scientific investigations have emphasized that extensive anthropogenic pressure contributed to the fragility of the lake
wetlands ecosystem (Ge et al., 2021; Wünnemann et al., 2024). Notable examples include the Nyando wetlands in Victoria Lake, which
lost 20.25 km2 of wetlands area between 1984 and 2010 due to cultivating the exposed land for agricultural production (Okotto-Okotto
et al., 2018). The Harike wetlands in Punjab of India experienced a 13 % reduction in their area from 1989 to 2010 due to agricultural
expansion within the catchment (Mabwoga and Thukral, 2014). Inle Lake wetlands area in Myanmar has decreased by 164 km2, which
is 4.2-fold higher in 2014 compared to 1989, primarily influenced by land use cover change (Karki et al., 2018). The persistent
recession of lake coastal wetlands due to anthropogenic destruction has been observed in the Great Lakes of the United States, with a
reduction of over 50 % in the past century (Gehring et al., 2020). Furthermore, the effects of dam operation on lake wetlands have been
well-documented, such as the Ithai barrage impacts of Loktak Lake and the TGD effects on Dongting Lake (Liu et al., 2023; Mahato
et al., 2023; Zeng et al., 2024).
Despite the gradual recognition of the impact of climate changes and human activities on lake wetlands, there has been insufficient
information on the dynamics of the wetlands. Traditional field surveys remain constrained by topography, climate, and accessibility,
fundamentally limiting the scope and comprehensiveness of monitoring efforts (You et al., 2023). Remote sensing has emerged as a
powerful tool for wetlands mapping (Zhao et al., 2023; Xiang et al., 2023). However, there remain the challenges in accurately
characterizing intricate wetland landscapes and improving classification accuracy. To address these challenges, diverse supervised
machine-learning methods such as support vector machines, random forests, as well as deep learning have been proposed (Hosseiny
et al., 2022; Zhao et al., 2022; Wang et al., 2023; Jahangeer et al., 2024). Previous studies have compared the typical machine learning
algorithms, including K-nearest neighbor, support vector machines, maximum likelihood, decision trees, and random forests, and
demonstrated that the random forests algorithm has superior performance in handling complex wetland landscapes and improving
classification accuracy (Amani et al., 2019; Long et al., 2021; Aslam et al., 2024).
Poyang Lake, located in the middle reaches of the Changjiang River, features extensive wetlands that provide a critical habitat for
millions of migrant birds during late autumn and winter and also supports a population of 44 million (Wu et al., 2014; Zhang et al.,
2015). However, due to climate changes and anthropogenic activities, the wetland of Poyang Lake presents ongoing degradation, with
the lake entering the dry season earlier and exposing it to drought longer (Mei et al., 2024). The advent of global climate change,
coupled with the increase in extreme weather events, has introduced unprecedented challenges to the Poyang Lake wetland (Wang
et al., 2019). Recent studies indicate that variability in precipitation and increased frequency of extreme weather events can cause fatal
damage to the Poyang Lake wetlands (Wang et al., 2019; Zheng et al., 2021). Moreover, intensive human activities will accelerate the
degradation of the wetlands (Huang et al., 2023). For instance, with the increased demand for industrial and agricultural water re­
sources, combined with sand excavation that enhanced Poyang Lake’s discharge ability into the Changjiang River, the wetlands
became drier (Lai et al., 2014; Lv et al., 2021). Meanwhile, previous studies have suggested that reservoirs in the basin can change the
sedimentation patterns (Mei et al., 2015). The TGD, commissioned in 2003 as the world’s largest hydroelectric dam, has trapped
massive sediment, leading to intensive riverbed erosion along the mid-lower Changjiang (Dai et al., 2018; Lou et al., 2022). As the
largest freshwater lake in China, Poyang Lake is freely connected with the Changjiang River, and its regulatory influence on the river is
undeniable. Most previous studies have primarily focused on the TGD’s influence on the hydrological aspects of Poyang Lake, such as
changes in the water area and water storage, or merely identified the downward migration of vegetation belts (Zheng et al., 2021;
Zhang et al., 2022; Zeng et al., 2024). However, there has been limited work on the spatial-temporal patterns and the underlying causes
of wetlands in Poyang Lake, let alone providing a comprehensive understanding of their changes across different zones.
Thereafter, this study aims to bridge this gap by 1) investigating the interannual variations in vegetation and mudflat coverage
within the wetlands of Poyang Lake; 2) examining the temporal and spatial transformations in vegetation and mudflat across different
zones of the lake, as well as transitions in land use; 3) identifying the possible drivers behind these dynamic changes. This knowledge
has a vital significance not only for the management and conservation of Poyang Lake wetlands but also for the management of similar
wetlands worldwide.

2. Materials and methods

2.1. Study area Poyang Lake

Located on the southern bank of the middle Changjiang River, Poyang Lake (28◦ 22’-29◦ 45’N, 115◦ 47’-116◦ 45’E) is one of two

2
F. Bu et al. Global Ecology and Conservation 59 (2025) e03585

remaining large-scale lakes that naturally connects to the Changjiang River (Fig. 1A). Dominated by a subtropical monsoon climate,
the lake catchment has the average annual temperature is 19.8 ℃. The average annual precipitation is 1638 mm (Zhang et al., 2024).
The lake receives catchment inflows from five major rivers, namely Ganjiang, Fuhe, Xinjiang, Raohe, and Xiushui, and discharge into
Changjiang River through a narrow channel at the northmost Hukou (Fig. 1A). Due to the interaction of river systems, Poyang Lake
distributed various semi-closed dish-shaped lakes. During flood season, a dish-shaped lake with Poyang Lake is an integrated lake;
during the dry season, hydraulic connections are interrupted, forming isolated sub-lakes (Zeng et al., 2024). The intra-annual lake
levels fluctuate between 8 and 22 m, creating an approximately 3000 km² wetlands system (Feng et al., 2013). The Poyang Lake
Ecological Wetland is one of six wetlands in the world. These extensive wetlands provide a habitat for millions of migratory birds,
making Poyang Lake one of the most renowned wetlands in the world (Liang et al., 2023).
The terrain of Poyang Lake inclined from south to north, shows obvious geological heterogeneity (Fig. 1B). The main lake’s
geological heterogeneity limits this division from accurately reflecting the spatial characteristics of Poyang Lake. Consequently, to
effectively analyze the spatial characteristics of mudflat and vegetation dynamic changes in Poyang Lake, our study divides the lake
into five zones based on its hydrological and topographical features (Fig. 1B). The northern lake zone is a region leading into
Changjiang River, located in the northern Songmen Mountain. The central lake zone is a lake-type water body, located in the southern
Songmen Mountain. The western lake zone is the estuarine delta, formed by the Ganjiang and the Xiushui. The southern lake zone is a
delta formed by the Ganjiang and Fuhe. The eastern lake zone is a deep-water area, that receives water from Raohe.

2.2. Materials

According to the Worldwide Reference System, one Landsat image (Path 121, Row 40) completely covers Poyang Lake. Given the
significant seasonal fluctuations in the water level of the lake, remote sensing images were captured during the dry season, with the
water level ranging between 11 and 12 m. A total of 14 Landsat (30 m spatial resolution) single-scene images, spanning from 1987 to
2018, were selected for analysis. Since 2019, Sentinel-2 (10 m spatial resolution) imagery has met the selection criteria for Poyang
Lake. The new imagery from Sentinel-2 can partly reduce the mixed pixel problem as compared to Landsat. Due to the similarities in
spectral and temporal resolution between Landsat and Sentinel imagery, as well as their comparable accuracy in wetland assessments
(Ballut-Dajud et al., 2025), the resolution of the images was deemed negligible in its impact on the results (Jing et al., 2019).
Consequently, Sentinel images were used for analysis post-2019 (Table 1). All images were preprocessed in Google Earth Engine (GEE)
(https://earthengine.google.com). Additionally, the monthly precipitation data for the Poyang Lake catchment (from 1956 to 2020)
were made available by the National Climate Centre of the Chinese Meteorological Administration (https://www.cma.gov.cn). Daily
water level at Duchang Station (from 1956 to 2023), as well as monthly water discharge and suspended sediment discharge at
Waizhou, Meigang, Hushan, Lijiadu, Wanjiabu, and Hukou (from 1956 to 2020) were collected from the Changjiang Water Resources
Commission (CWRC, https://www.cjw.gov.cn). Moreover, geomorphological information from 2000 to 2020 for Hukou and Datong
was also obtained from CWRC. The total capacity and number of reservoirs within the Poyang Lake Basin were taken from the Hy­
drological Bureau of the Jiangxi Province. All data underwent rigorous verification and uncertainty analysis according to government

Fig. 1. Study area of Poyang Lake. (A) Poyang Lake catchment and water system; (B) Poyang Lake basin morphology and divisions.

3
F. Bu et al. Global Ecology and Conservation 59 (2025) e03585

Table 1
Mudflat and vegetation area in Poyang Lake corresponding to water level at Duchang station, overall accuracy and Kappa coefficient.
Image time Water level Mudflat area Vegetation area Overall accuracy Kappa coefficient
(m) (km2) (km2)

1987–12–17 11.54 1002.04 1037.10 0.98 0.97

1990–01–23 11.62 887.22 1134.85 0.90 0.86
1990–12–19 11.34 1145.74 1064.90 0.93 0.91
1993–12–17 11.72 737.31 1214.76 0.97 0.95
1997–02–11 11.58 838.96 1392.40 0.96 0.95
2004–04–03 12.09 652.89 1339.34 0.98 0.96
2005–02–01 11.82 544.38 1607.60 0.97 0.95
2014–03–24 11.83 554.65 1342.19 0.94 0.91
2014–10–24 11.31 348.98 1278.00 0.95 0.92
2016–02–16 11.44 300.96 1509.22 0.97 0.96
2019–01–23 11.45 721.78 1177.60 0.90 0.87
2019–09–29 11.11 691.24 1214.22 0.93 0.90
2022–08–09 11.01 999.31 1160.74 0.99 0.98
2023–09–18 11.76 384.45 1411.99 0.97 0.96

protocols, ensuring the accuracy rate surpassed 95 %.

2.3. Methods

2.3.1. Creation of training samples

To investigate the evolution of the lacustrine wetlands, we categorized the Poyang Lake into three types: water, mudflats, and
vegetation. High-resolution and cloud-free images were chosen to generate training samples through visual interpretation. First, three
layers were constructed, each assigned specific category labels for identification. Then, dozens of training polygons were manually
chosen, with 1000 random points generated using the ‘random point’ command in GEE. The training dataset comprised 70 % of the
samples, with the remaining 30 % set as validation samples.

2.3.2. Computation of spectral indices

Various landforms have different spectral characteristics (Pike and Rozema, 1975; Jiang et al., 2015). To enhance the identification
of various landforms and increase the reliability of extraction result, in this study, five surface reflectances of blue (B2), green (B3), red
(B4), near-infrared (B5) and shortwave infrared (B6 and B7) were chosen, and four water and vegetation indices were applied for each
image on the GEE platform, such as normalized difference vegetation index (NDVI) (Tucker, 1979), enhanced vegetation index (EVI)
(Huete et al., 2002), modified normalized difference water index (mNDWI) (Xu, 2006) and land surface water index (LSWI) (Xiao et al.,
2004; Chandrasekar et al., 2010). NDVI and EVI are effective in distinguishing vegetation from mudflats and water, offering significant
advantages for vegetation extraction and feature recognition. mNDWI excels at separating water bodies, while LSWI is sensitive to
moisture content, enabling it to differentiate vegetation from mudflats and water. The formulas for calculating spectral indices are as
follows:
ρNIR − ρRed
NDVI = # (1)
ρNIR +ρRed

ρNIR − ρRed
EVI = 2.5 × # (2)
1.0 + ρNIR + 6.0ρRed + 7.5ρBlue

ρGreen − ρSWIR
mNDWI = # (3)
ρGreen + ρSWIR

ρNIR − ρSWIR
LSWI = # (4)
ρNIR + ρSWIR

Where ρGreen, ρRed, ρBlue, ρSWIR, ρNIR correspond to the wavelength ranges of the Landsat TM and OLI sensors: green (0.52–0.60 μm), red
(0.63–0.69 μm), blue (0.45–0.52 μm), shortwave infrared (1.55–1.75 μm), near-infrared (0.77–0.90 μm).

2.3.3. Random forest

Machine learning, a subset of artificial intelligence, has revolutionized the detection and classification of objects in images and
videos (Adhinata et al., 2024). Random forest, a machine learning technique, combines decision trees and the Bagging method with the
advantages of flexibility, robustness, utility, and efficiency. This ensemble classifier has been widely employed in classification tasks
involving multi-featured data. Consequently, random forest demonstrates outstanding performance in wetlands classification tasks
compared to other algorithms (Peng et al., 2022). Here, a random forest algorithm implemented within the GEE API utilizes four
spectral indices (NDVI, mNDWI, EVI, LSWI) along with reflectance values from different image bands as input datasets. The trained
random forest classifier was used to classify the entire image based on the spectral index values. Following the determination of the

4
F. Bu et al. Global Ecology and Conservation 59 (2025) e03585

vegetation and mudflat distribution and visual adjustments to ensure classification accuracy, maps depicting the changes in vegetation
and mudflat in Poyang Lake from 1987 to 2023 were obtained.

2.3.4. Landcover changes monitoring

Poyang Lake’s bottom becomes exposed during the dry season, creating a large wetland with diverse cover types. Here, according
to the random forest classification results, we investigate landcover transitions from 1987 to 2023. Furthermore, based on key points in
climate changes and human activities, we divided the time into three periods: 1987–2004, 2004–2013, and 2013–2023, analyzing
landcover and comparing landcover transitions across these periods. The land cover transitions are depicted using different colors to
illustrate spatial changes and highlight significant transformations.

2.3.5. Accuracy assessment

Ensuring the accuracy of remote sensing resolution and image processing necessitates thorough error analysis and verification
(Heale and Twycross, 2015). For Landsat image classification, the confusion matrix was employed to evaluate the performance of the
random forest classification algorithm (Huang, 2019). Accuracy evaluation typically includes user accuracy, producer accuracy,
overall accuracy, and Kappa coefficient. Table 1 presents the annual accuracy assessment of the random forest classification. Both the
overall accuracy and Kappa coefficient of each year surpass 0.90 and 0.86, respectively, exceeding the minimum standard of 85 %
(Kumar et al., 2021). This indicates that the result obtained from an image can be considered reliable. Additionally, to account for the
bias of the linear interpolation method, a 95 % confidence interval was constructed to enhance the reliability of the result. The for­
mulas for overall accuracy and Kappa coefficient are as follows:
TP + TN
Accuracy = # (5)
TP + TN + FP + FN
∑ ∑k
n ki=1 nii − i=1 (ni + n+i )
Kappa = ∑ k
# (6)
n2 − i=1 (ni + n+i )

Accuracy reflects the test accuracy of the classification results. TP (True Positives) and TN (True Negatives) are the samples whose
predicted values are consistent with the true values. FP (False Positives) and FN (False Negatives) are the samples whose predicted
values are opposite to the true values. The Kappa coefficient provides an overall measure of the classification accuracy. Where n is the
total number of validation pixels, nii is the number of correctly classified pixels for the i-th class; n+i is the total number of reference
pixels for the i-th class; and k is the number of classes.

Fig. 2. Temporal variation of wetlands area in Poyang Lake. A: Time variation of the total wetlands area; B: Represents the temporal variation of
wetlands area in northern lake zone.

5
F. Bu et al. Global Ecology and Conservation 59 (2025) e03585

3. Results

From 1987–2023, Poyang Lake’s wetlands area, consisting of mudflat and vegetation, exhibited fluctuating changes around an
average area of 1978.20 km², without significant trend change (Fig. 2A). Losses of wetlands continued in the northern lake zone, with
an average decrease rate of 4.08 km²/yr, resulting in a substantial 44.74 % reduction, from 352.65 km² in 1987–194.87 km² in 2023
(Fig. 2B). In central lake zone, western lake zone, eastern lake zone, and southern lake zone, the wetlands area remains unchanged.
However, these stability were caused by the opposite trends in vegetation and mudflat changes within wetlands as shown below.

3.1. Significant mudflat loss in Poyang Lake wetlands

From 1987–2023, the area covered by mudflats in the Poyang Lake experienced a consistent annual decline, and the overall mudflat
coverage decreased from 998.60 km² in 1987–384.45 km² in 2023 (Fig. 3A). Over the past 37 years, Poyang Lake experienced a
significant loss of 614.16 km², which accounts for approximately 62 % of the initial measurement mudflat area in 1987. The rate of
decrease in the lake’s mudflat area was calculated to be 5.86 % per year. However, this trend was not consistent throughout the entire
study period. The change in mudflat cover area in Poyang Lake can be divided into two distinct stages (Fig. 3A). Specifically, between
1987 and 2015, there was a rapid reduction in mudflat cover, with an average decrease rate of 26.34 km²/yr. This resulted in a loss of
approximately 70 % of the mudflat cover compared to the initial measurement in 1987. Between 2015 and 2023, a fluctuating trend in
mudflat cover emerged, with the area increasing from 299.88 km² to 384.45 km², showing a slight but not significant increase.
Notably, within this period, the increase in the area of the eastern lake zone’s mudflats contributed 39 % of the overall increase
(Fig. 3A).
Furthermore, the reduction in mudflat area exhibited significant region variations, occurring primarily and most notably in the
northern lake zone, central lake zone, and western lake zone (Fig. 3B-D). The most significant losses occurred in the northern lake zone,
where the mudflat area decreased at a rate of 6.62 km²/yr, resulting in a substantial 64.46 % reduction, from 328.23 km² in
1987–116.67 km² in 2023 (Fig. 3B). The central lake zone also experienced a considerable reduction in mudflat area, decreasing at a
rate of 1.90 km²/yr, leading to a 67.15 % reduction, from 154.17 km² in 1987–50.65 km² in 2023 (Fig. 3C). Similarly, the western lake
zone experienced a loss of mudflats at a rate of 1.48 km²/yr, leading to a 58.35 % reduction, from 129.01 km² in 1987–53.74 km² in
2023 (Fig. 3D).

3.2. Vegetation expansion in Poyang Lake wetlands

Between 1987 and 2023, the vegetation area in Poyang Lake increased by 36 %, from 1035.32 km² in 1987 to 1411.99 km² in

Fig. 3. Temporal variation of mudflat area in Poyang Lake. A: Time variation of the total mudflat area; B-D: Represents the temporal variation of
mudflat area in northern lake zone, central lake zone, and western lake zone, respectively.

6
F. Bu et al. Global Ecology and Conservation 59 (2025) e03585

2023, but it was worth noting that the area markedly increased by 54.87 % at a rate of 30.87 km²/yr during 1987–2003, followed by a
slight decrease from 1603.40 km² in 2003 to 1411.99 km² in 2023 (Fig. 4A). Although a slight decline after 2003, the average
vegetation area from 2003 to 2023 was still 140.47 km² larger than the average area from 1987 to 2003.
The increase in vegetation cover area exhibited significant regional variations, with the most notable changes occurring in the
northern lake zone and central lake zone (Fig. 4B-C). The vegetation area in the northern lake zone increased by 53.79 km² in 37 years
with an average rate of 2.54 km²/yr (Fig. 4B). Similarly, an apparent increase in the central lake zone vegetation area can be found
during the study period. The vegetation area expanded at a rate of approximately 3.35 km²/yr, increasing by a factor of 11 from
8.31 km² in 1987–107.80 km² in 2023 (Fig. 4C). In southern lake zone and eastern lake zone, although vegetation area did not present
a distinct trend during the period from 1987 to 2023, it increased by approximately 23 % and 33 % compared to 1987 (Fig. 4A). In
contrast, the vegetation cover area in western lake zone, far from Changjiang mainstream, has maintained a high level of stability over
the study period.

3.3. Spatial variations in mudflat and vegetation areas in Poyang Lake wetlands

Between 1987 and 2023, five zones exhibited distinct spatial patterns of gain and loss in wetlands area within Poyang Lake
(Fig. 6D). During this period, wetlands decreased mainly in northern lake zone (outlined in green in Fig. 5A), vegetation flourished in
the center areas of Poyang Lake because of early water retreat (outlined in green in Fig. 5B), while mudflat degradation in the lake,
especially in northern lake zone (outlined in green in Fig. 5C). These spatial changes in the lake displayed significant differences across
three distinct periods (Fig. 6A-C).
Specifically, from 1987 to 2004, the changes in mudflat and vegetation primarily occurred in the transitional regions in the western
lake zone, central lake zone, southern lake zone, and eastern lake zone. 334.70 km² of mudflat cover was converted to vegetation and
44.95 km² in reverse, coupled with landcover transitions involving water covers, this ultimately led to a net loss of 447.71 km² in
mudflat area and an increase of 305.07 km² in vegetation cover (Fig. 6A). From 2004–2013, there was widespread destruction of
mudflat in northern lake zone, resulting in a 45 % reduction. Concurrently, a remarkable vegetation expansion in the delta front of
southern lake zone, which increased by 12.09 km², representing a growth of 37.61 % relative to 2004 (Fig. 6B). Subsequently, from
2013 to 2023, the loss of mudflat in northern lake zone intensified, further decreasing by 31.58 % from the area in 2013. Meanwhile,
while the rate of vegetation expansion at the delta front in the southern lake zone slowed down, vegetation growth occurred on former
mudflats coverage regions of the central lake zone and the Dish-shaped regions of the western lake zone and southern lake zone
(Fig. 6C). This ultimately resulted in a net loss of 35.21 km² of mudflat area, while vegetation expanded by 135.37 km² in Poyang Lake
during this period.

Fig. 4. Temporal variation of vegetation area in Poyang Lake. A: Time variation of the total vegetation area; B-D: Represents the temporal variation
of vegetation area in northern lake zone and central lake zone, respectively.

7
F. Bu et al. Global Ecology and Conservation 59 (2025) e03585

Fig. 5. Gains and losses of wetlands area in Poyang Lake between 1987 and 2023. A: Wetlands, B: Vegetation and C: Mudflat. The green circles
highlight the core areas experiencing the most significant changes for each category.

4. Discussion

4.1. Precipitation change

Precipitation change has consistently been recognized as a pivotal factor in modulating the inundation conditions in lakes, with
precipitation changes identified as the primary determinant of its impact (Mei et al., 2015; Woolway et al., 2020; Tao et al., 2020).
While previous studies have extensively examined the relationship between precipitation and water area changes in Poyang Lake
(Feng et al., 2012; Chen et al., 2023; Ren et al., 2024), there remains a gap in understanding the influence of precipitation on the
dynamics of lake wetlands, especially sub-lakes changes.
From 1987–2020, the Poyang Lake Basin experienced relatively little variation in annual mean precipitation during the dry season
(Fig. 7A). However, monthly precipitation records over the Poyang Lake Basin indicated a slight but not significant increase in
November 2003–2020 compared to 1987–2002, resulting in a 25.6 mm increase in cumulative precipitation (Fig. 7B). Although this
change had a limited effect on the expansive main lake area, it likely contributed to the expansion of isolated dish-shaped lakes. One
prominent example is the eastern lake zone’s dish-shaped lakes. Prior to 2003, these lakes were characterized as transition zones
between water and vegetation. However, as precipitation increased slightly, these transitional zones were gradually converted into
open water bodies in 2014, a condition that has largely persisted to the present day (Fig. 8A).

4.2. Human activities within the Poyang Lake catchment

Human activities within the basin have a significant impact on the lake environment (Zhang et al., 2018; Liu et al., 2023; Wün­
nemann et al., 2024). In the Poyang Lake watershed, sand excavation and reservoir regulation are the primary anthropogenic ac­
tivities. Following a total ban on sand excavation of Changjiang River mainstream in 2000, intensive sand mining moved to Poyang
Lake, with an annual extraction rate of approximately 236 × 106 m3 (De Leeuw et al., 2010; Mei et al., 2015; Wang et al., 2020). Most
of the sand mining occurred along the outflow channel of the lake, leading to large areas of mudflat being directly converted into open
water (Fig. 8B). This observation aligns with our observation of the rapid reduction in mudflat cover area in the northern lake zone
(Figs. 3B and 5B).
In comparison to the local disturbance caused by sand extraction in the northern lake zone, the operation of the reservoir has an
impact on the entire lake and sediment inputs. Since 1975, more than 9600 reservoirs have been constructed within the Poyang Lake
Basin (Mei et al., 2018). We further present information regarding the capacity of large dams during 1987–2019, which constitute the
significant factor affecting sediment transport from the five sub-basins to the Poyang Lake (Fig. 9A). During this period, the capacity of
reservoirs within the Poyang Lake increased from 22.70 billion m3 in 1987–32.11 billion m3 in 2019. A noteworthy finding is a
significant correlation (R = 0.59, P < 0.05) observed between reservoir capacity and sediment discharge of the lake (Fig. 9B),
highlighting the prominent role of reservoir construction in intensifying sediment retention.
We conducted correlation analyses between sediment discharge from the five sub-basins and the respective mudflat and vegetation
areas in each sub-region of Poyang Lake from 1987 to 2019. The result showed that only the correlation coefficient between sediment
discharge and mudflat (or vegetation) area in specific sub-zones exceeded 0.45, indicating a potential influence of sediment discharge
on mudflat (or vegetation) dynamics in those regions. However, overall coefficients did not reach statistical significance in each zone.

8
F. Bu et al. Global Ecology and Conservation 59 (2025) e03585

Fig. 6. Land use change chord diagram and spatial distribution from 1987 to 2023.

Based on these findings, we can conclude that the decrease in sediment input caused by reservoirs has made a minimal impact on the
observed vegetation growth and mudflat reduction in these regions.

4.3. Dam operations along the Changjiang mainstream

Classical river-connected drain lakes are susceptible to human interference from the river environment (Chen et al., 2019; Mei
et al., 2024). Previous research has reported that the influence of the TGD on hydrological processes and wetlands ecosystems in the
Changjiang River Basin is immediate and fundamental (Li et al., 2015; Mei et al., 2015; Liu et al., 2023). The TGD affects the evolution
of the Poyang Lake wetlands through two primary mechanisms: sediment trapping and regulation of Changjiang River discharge.
The construction of the TGD has led to annually trapped 1.23 × 108 t of sediment, resulting in intensive riverbed erosion along the
middle and lower Changjiang River (Dai et al., 2018). Consequently, the topographic gradient between Hukou (the outlet of Poyang
Lake) and Datong (the Changjiang riverbed) has increased significantly, rising from 10.38 m in 2007–17.01 m in 2020 (Fig. 10). This
enhanced sediment discharge from Poyang Lake to the Changjiang River, combined with inadequate sediment input from sources,
converted the lake from a depositional system to an erosional system in 2003 (Fig. 11B). This factor likely contributed to mudflat
erosion in northern lake zone, central lake zone, and western lake zone, closely connected to the Changjiang River (Fig. 3B-D).
Moreover, the reduction of sediment deposition also facilitates vegetation extension towards the lake center, which has been
confirmed by previous studies (Wang et al., 2019; Zeng et al., 2024).
Similar to changes in sediment patterns, the TGD impounded water during September and October, significantly increasing the
lake-river water level gradient. This compelled the lake to discharge a greater volume of water to the Changjiang River, resulting in a
decrease in the lake’s water level by 1.38 m during the dry season (Mei et al., 2024). These factors had two significant effects on

9
F. Bu et al. Global Ecology and Conservation 59 (2025) e03585

Fig. 7. A) Average monthly precipitation for Poyang Lake Basin during dry season; B) Average precipitation for Poyang Lake Basin in November.

Fig. 8. Temporal and spatial changes of wetland in the Poyang Lake. Green is vegetation, grey is mudflat, blue is water.

vegetation growth within Poyang Lake. On the one hand, the lake entered its dry season 36 days earlier, prolonging the dry season by
approximately 50 days (Fig. 12). On the other hand, the decrease in water levels created suitable conditions for vegetation, leading to
the conversion of water and mudflats into vegetation coverage (Fig. 5A). Specifically, aridification has shortened the inundation
duration of the dish-shaped lake, leading to water translation to vegetation (Fig. 8C). In the delta wetland, the extended exposure time
of the lake bottom facilitated plant growth, leading to a rapid increase in migration of vegetation towards mudflats and water coverage
(Fig. 8D). However, this occupying poses risks of shrinkage and degeneration for the lake.

4.4. Implications for lake management

In this study, we demonstrate the advantages of using machine learning, in processing large-scale remote sensing data for wetland
monitoring. Compared to traditional classification approaches, such as threshold-based or unsupervised methods, the machine

10
F. Bu et al. Global Ecology and Conservation 59 (2025) e03585

Fig. 9. A) Historical increase in the total capacity of dams; and B) Correlations between the annual sediment load of the Poyang Lake and the
Capacity in the basin dams.

Fig. 10. Geomorphological evolution of the transverse section. A. Hukou; B. Datong; C. Difference in elevation between the lowest part of the
Hukou and Datong sections.

learning-based approach enhances classification accuracy, especially in distinguishing vegetation, mudflats, and water bodies across
complex spatiotemporal patterns (Amani et al., 2017; Long et al., 2021).
Wetlands in Poyang Lake have experienced significant degradation as a result of intensified human activities, changes in sediment
and water supply, and hydrodynamic changes (Feng et al., 2016; Mei et al., 2018; Zeng et al., 2024). These cumulative pressures have

11
F. Bu et al. Global Ecology and Conservation 59 (2025) e03585

Fig. 11. A) Average annual input and output discharge of the lake and their difference; B) Average annual input and output sediment load of the
lake and their difference.

Fig. 12. Numbers of days in different conditions. A) Date of entering the dry season; B) Number of days below the drought level.

put the remaining wetland at great risk. Our findings contribute to understanding the interplay between hydrological regulation,
sediment dynamics, and wetland evolution. The observed vegetation expansion and mudflat loss result from both climate-induced
hydrological shifts and human interventions, particularly dam construction and sand mining. These insights help refine models of
wetland succession under anthropogenic influence. For effective lake management, adaptive strategies are required to mitigate
wetland loss, including regulating sand mining, optimizing water level control, and restoring degraded areas. The integration of
remote sensing and machine learning into decision-making will enhance long-term conservation efforts and sustainable development
of lake ecosystems.

5. Conclusions

Poyang Lake is the largest freshwater lake in China and holds significant importance as part of the global wetland ecosystem. The
vegetation and mudflat in the wetlands exhibit considerable variations, and this research aims to provide a comprehensive assessment
of these temporal and spatial variations, as well as identify possible drives of wetlands evolution in Poyang Lake. The main conclusions
are as follows.
1) From 1987–2023, the total area of mudflats in Poyang Lake gradually decreased by 62 %. Meanwhile, wetland vegetation

12
F. Bu et al. Global Ecology and Conservation 59 (2025) e03585

expanded by 36 %, increasing from 1035.32 km² in 1987 to 1411.99 km² in 2023. This growth was characterized by lakeward mi­
grations and accretions, predominantly occurring in 1987–2013.
2) From 1987–2023, the phenomenon of mudflat loss and vegetation increase was significant in Poyang Lake. Mudflat loss was
most severe in the northern lake zone (64.46 % reduction), followed by the central (67.15 %) and western (58.35 %) zones. Vegetation
expansion was most pronounced in the central lake zone (11-fold increase, 8.31 km² to 1107.80 km²) and the northern zone
(53.79 km² increase), whereas the western zone remained stable. The southern and eastern zones also saw moderate increases (23 %
and 33 %, respectively). These shifts resulted in a net loss of 441.71 km² in mudflat area and a net gain of 305.07 km² in vegetation
cover, reflecting significant land-use transitions from exposed mudflats to vegetated wetlands.
3) Human activities emerged as the dominant driver of wetland evolution in Poyang Lake. The TGD and regional reservoirs have
reshaped the hydrological regime, generating both mudflat reduction and vegetation expansion. Additionally, increased rainfall,
particularly after 2013, has contributed to higher water coverage, further facilitating vegetation expansion.

Declaration of Competing Interest

The authors declare that there is no conflicts of interest regarding the submitted paper of “Machine learning-based mapping
wetland dynamics of the largest freshwater lake in China”.

Acknowledgments

This study was supported financially by the National Social Science Foundation Major Project of China (23&ZD105).

Data availability

Data will be made available on request.

References

Adhinata, F.D., Wahyono, Sumiharto, R., 2024. A comprehensive survey on weed and crop classification using machine learning and deep learning. Artif. Intell. Agric.
13, 45–63.
Amani, M., Brisco, B., Afshar, M., Mirmazloumi, S.M., Mahdavi, S., Mirzadeh, S.M.J., Huang, W., Granger, J., 2019. A generalized supervised classification scheme to
produce provincial wetland inventory maps: An application of Google Earth Engine for big geo data processing. Big Earth Data 3 (4), 378–394.
Amani, M., Salehi, B., Mahdavi, S., Granger, J.E., Brisco, B., Hanson, A., 2017. Wetland classification using multi-source and multi-temporal optical remote sensing
data in Newfoundland and Labrador, Canada. Can. J. Remote Sens 43, 360–373.
Aslam, R.W., Naz, I., Shu, H., Yan, J., Quddoos, A., Tariq, A., Davis, J.B., Al-Saif, A.M., Soufan, W., 2024. Multi-temporal image analysis of wetland dynamics using
machine learning algorithms. J. Environ. Manag. 371, 123123.
Bai, J., Guan, Y., Liu, P., Zhang, L., Cui, B., Li, X., Liu, X., 2020. Assessing the safe operating space of aquatic macrophyte biomass to control the terrestrialization of a
grass-type shallow lake in China. J. Environ. Manag. 266, 110479.
Ballut-Dajud, G., Betanzo-Torres, E.A., Herazo, L.C.S., 2025. Diagnosis of the vegetation cover in the wetlands of La Caimanera Swamp, Colombia and Casitas
Wetland, Mexico by means of Landsat and Sentinel-2A images during last four decades. Results Eng. 25, 104211.
Chandrasekar, K., Sesha Sai, M.V.R., Roy, P.S., Dwevedi, R.S, 2010. Land Surface Water Index (LSWI) response to rainfall and NDVI using the MODIS Vegetation Index
product. Int. J. Remote Sens. 31 (15), 3987–4005.
Chen, J., Li, Y., Shu, L., Fang, S., Yao, J., Cao, S., Zeng, B., Yang, M., 2023. The influence of the 2022 extreme drought on groundwater hydrodynamics in the
floodplain wetland of Poyang Lake using a modeling assessment. J. Hydrol. 626, 130194.
Chen, X., Qiao, Q., McGowan, S., Zeng, L., Stevenson, M.A., Xu, L., Huang, C., Liang, J., Cao, Y., 2019. Determination of geochronology and sedimentation rates of
shallow lakes in the middle Yangtze reaches using 210Pb, 137Cs and spheroidal carbonaceous particles. Catena 174, 546–556.
Dai, Z., Mei, X., Darby, S.E., Lou, Y., Li, W., 2018. Fluvial sediment transfer in the Changjiang (Yangtze) river-estuary depositional system. J. Hydrol. 566, 719–734.
De Leeuw, J., Shankman, D., Wu, G., De Boer, W.F., Burnham, J., He, Q., Yesou, H., Xiao, J., 2010. Strategic assessment of the magnitude and impacts of sand mining
in Poyang Lake, China. Reg. Environ. Change 10 (2), 95–102.
Feng, L., Han, X., Hu, C., Chen, X., 2016. Four decades of wetland changes of the largest freshwater lake in China: Possible linkage to the Three Gorges Dam? Remote
Sensing of Environment 176, 43–55.
Feng, L., Hu, C., Chen, X., Cai, X., Tian, L., Gan, W., 2012. Assessment of inundation changes of Poyang Lake using MODIS observations between 2000 and 2010.
Remote Sens. Environ. 121, 80–92.
Feng, L., Hu, C., Chen, X., Zhao, X., 2013. Dramatic inundation changes of China’s two largest freshwater lakes linked to the Three Gorges Dam. Environ. Sci. Technol.
47 (17), 9628–9634.
Ge, Y., Zhang, Q., Dong, X., Yang, X., 2021. Revealing anthropogenic effects on lakes and wetlands: Pollen-based environmental changes of Liangzi Lake, China over
the last 150 years. Catena 207, 105605.
Gehring, T.M., Blass, C.R., Murry, B.A., Uzarski, D.G., 2020. Great Lakes coastal wetlands as suitable habitat for invasive mute swans in Michigan. J. Gt. Lakes Res. 46
(2), 323–329.
Heale, R., Twycross, A., 2015. Validity and reliability in quantitative studies. Evidence-based nursing 18 (3), 66–67.
Hosseiny, B., Mahdianpari, M., Brisco, B., Mohammadimanesh, F., Salehi, B., 2022. WetNet: a spatial–temporal ensemble deep learning model for wetland
classification using sentinel-1 and sentinel-2. IEEE Trans. Geosci. Remote Sens. 60, 1–14.
Huang, W., 2019. Machine-learning phase prediction of high-entropy alloys. Acta Mater. 169, 225–236.
Huang, A., Liu, X., Peng, W., Dong, F., Han, Z., Du, F., Ma, B., Wang, W., 2023. Spatiotemporal heterogeneity of inundation pattern of floodplain lake wetlands and
impact on wetland vegetation. Sci. Total Environ. 905, 167831.
Huete, A., Didan, K., Miura, T., Rodriguez, E.P., Gao, X., Ferreira, L.G., 2002. Overview of the radiometric and biophysical performance of the MODIS vegetation
indices. Remote Sens. Environ. 83 (1–2), 195–213.
Jahangeer, J., Zhang, L., Tang, Z., 2024. Evaluating wetland hydrological performance under three different conservation programs in Nebraska, United States, during
2018–2021. JAWRA J. Am. Water Resour. Assoc. 60 (1), 132–147.
Jiang, W., Yuan, L., Wang, W., Cao, R., Zhang, Y., Shen, W., 2015. Spatio-temporal analysis of vegetation variation in the Yellow River Basin. Ecol. Indic. 51, 117–126.
Jing, X., Leigh, L., Teixeira Pinto, C., Helder, D., 2019. Evaluation of RadCalNet output data using landsat 7, landsat 8, sentinel 2A, and sentinel 2B sensors. Remote
Sens. 11 (5), 541.

13
F. Bu et al. Global Ecology and Conservation 59 (2025) e03585

Karki, S., Thandar, A.M., Uddin, K., Tun, S., Aye, W.M., Aryal, K., Kandel, P., Chettri, N., 2018. Impact of land use land cover change on ecosystem services: a
comparative analysis on observed data and people’s perception in Inle Lake, Myanmar. Environ. Syst. Res. 7 (1), 25.
Kayranli, B., Scholz, M., Mustafa, A., Hedmark, Å., 2010. Carbon storage and fluxes within freshwater wetlands: a critical review. Wetlands 30 (1), 111–124.
Kumar, P., Dobriyal, M., Kale, A., Pandey, A.K., 2021. Temporal dynamics change of land use/land cover in Jhansi district of Uttar Pradesh over past 20 years using
LANDSAT TM, ETM+ and OLI sensors. Remote Sens. Appl.: Soc. Environ. 23, 100579.
Lai, X., Shankman, D., Huber, C., Yesou, H., Huang, Q., Jiang, J., 2014. Sand mining and increasing Poyang Lake’s discharge ability: a reassessment of causes for lake
decline in China. J. Hydrol. 519, 1698–1706.
Li, G., Wang, X.T., Yang, Z., Mao, C., West, A.J., Ji, J., 2015. Dam-triggered organic carbon sequestration makes the Changjiang (Yangtze) river basin (China) a
significant carbon sink. J. Geophys. Res.: Biogeosci. 120 (1), 39–53.
Liang, Y., Dong, B., Li, P., Zhang, K., Gao, X., 2023. Prediction of overwintering crane population in Poyang Lake wetland based on RS and regression Model, China.
Ecol. Indic. 149, 110183.
Liu, Y., Li, J., Yan, D., Chen, L., Li, M., Luan, Z., 2023. Typical vegetation dynamics and hydrological changes of Dongting Lake wetland from 1985 to 2020. Ecohydrol.
Hydrobiol. 24 (4), 910–919.
Long, X., Li, X., Lin, H., Zhang, M., 2021. Mapping the vegetation distribution and dynamics of a wetland using adaptive-stacking and Google Earth Engine based on
multi-source remote sensing data. Int. J. Appl. Earth Obs. Geoinf. 102, 102453.
Lou, Y., Dai, Z., Long, C., Dong, H., Wei, W., Ge, Z., 2022. Image-based machine learning for monitoring the dynamics of the largest salt marsh in the Yangtze River
Delta. J. Hydrol. 608, 127681.
Lu, M., Zou, Y., Xun, Q., Yu, Z., Jiang, M., Sheng, L., Lu, X., Wang, D., 2021. Anthropogenic disturbances caused declines in the wetland area and carbon pool in China
during the last four decades. Glob. Change Biol. 27 (16), 3837–3845.
Lv, T., Liu, W., Zhang, X., Yao, L., 2021. Spatiotemporal evolution of the green efficiency of industrial water resources and its influencing factors in the Poyang Lake
region. Phys. Chem. Earth, Parts A/B/C. 124, 103049.
Mabwoga, S.O., Thukral, A.K., 2014. Characterization of change in the Harike wetland, a Ramsar site in India, using landsat satellite data. SpringerPlus 3 (1), 576.
Mahato, S., Pukhrambam, G., Joshi, P.K., 2023. Damming effects on hydrological abundance and eco-hydrological alteration in upstream wetlands of Eastern
Himalaya. J. Clean. Prod. 418, 138089.
Mahmood, R., Jia, S., 2019. Assessment of hydro-climatic trends and causes of dramatically declining stream flow to Lake Chad, Africa, using a hydrological approach.
Sci. Total Environ. 675, 122–140.
Mei, X., Dai, Z., Du, J., Chen, J., 2015. Linkage between Three Gorges Dam impacts and the dramatic recessions in China’s largest freshwater lake. Poyang Lake. Sci.
Rep. 5 (1), 18197.
Mei, X., Dai, Z., Du, J., Cheng, J., 2024. Three Gorges Dam enhanced organic carbon burial within the sediments of Poyang Lake. China Catena 238, 107859.
Mei, X., Du, J., Dai, Z., Du, J., Gao, J., Wang, J., 2018. Decadal Sedimentation in China’s Largest Freshwater Lake, Poyang Lake. Geochem., Geophys., Geosyst. 19 (8),
2384–2396.
Okotto-Okotto, J., Raburu, P.O., Obiero, K.O., Obwoyere, G.O., Mironga, J.M., Okotto, L.G., Raburu, E.A., 2018. Spatio-temporal impacts of lake victoria water level
recession on the fringing Nyando Wetland, Kenya. Wetlands 38 (6), 1107–1119.
Peng, H., Xia, H., Shi, Q., Chen, H., Chu, N., Liang, J., Gao, Z., 2022. Monitoring spatial and temporal dynamics of wetland vegetation and their response to
hydrological conditions in a large seasonal lake with time series Landsat data. Ecol. Indic. 142, 109283.
Pike, R.J., Rozema, W.J., 1975. Spectral analysis of landforms. Ann. Assoc. Am. Geogr. 65 (4), 499–516.
Radmanesh, F., Esmaeili-Gisavandani, H., Lotfirad, M., 2022. Climate change impacts on the shrinkage of Lake Urmia. J. Water Clim. Change 13 (6), 2255–2277.
Ren, L., He, Y., Zhu, J., Liu, Y., Cai, W., Yang, W., Zhao, J., Wu, C., 2024. Analysis on driving factors and prediction of dynamic change of Poyang lake area. Energy
Rep. 12, 495–503.
Reynaud, A., Lanzanova, D., 2017. A global meta-analysis of the value of ecosystem services provided by lakes. Ecol. Econ. 137, 184–194.
Sterner, R.W., Keeler, B., Polasky, S., Poudel, R., Rhude, K., Rogers, M., 2020. Ecosystem services of Earth’s largest freshwater lakes. Ecosyst. Serv. 41, 101046.
Tao, S., Fang, J., Ma, S., Cai, Q., Xiong, X., Tian, D., Zhao, X., Fang, L., Zhang, H., Zhu, J., Zhao, S., 2020. Changes in China’s lakes: climate and human impacts. Natl.
Sci. Rev. 7 (1), 132–140.
Tucker, C.J., 1979. Red and photographic infrared linear combinations for monitoring vegetation. Remote Sens. Environ. 8, 127–150.
Wang, M., Mao, D., Wang, Y., Xiao, X., Xiang, H., Feng, K., Luo, L., Jia, M., Song, K., Wang, Z., 2023. Wetland mapping in East Asia by two-stage object-based Random
Forest and hierarchical decision tree algorithms on Sentinel-1/2 images. Remote Sens. Environ. 297, 113793.
Wang, Y., Molinos, J.G., Shi, L., Zhang, M., Wu, Z., Zhang, H., Xu, J., 2019. Drivers and changes of the Poyang Lake Wetland ecosystem. Wetlands 39 (S1), 35–44.
Wang, S., Zhang, L., Zhang, H., Han, X., Zhang, L., 2020. Spatial–temporal wetland landcover changes of Poyang lake derived from landsat and HJ-1A/B Data in the
dry season from 1973–2019. Remote Sens. 12 (10), 1595.
Weiming, C., Yao, X., 2011. Extracting and spatial analysis for lacustrine landforms and Lake-Wetlands in China. Procedia Environ. Sci. 10, 1920–1925.
White, E.E., Ury, E.A., Bernhardt, E.S., Yang, X., 2022. Climate change driving widespread loss of coastal forested wetlands throughout the North American Coastal
Plain. Ecosystems 25 (4), 812–827.
Williams, M.R., King, K.W., 2020. Changing rainfall patterns over the Western Lake Erie Basin (1975–2017): effects on tributary discharge and phosphorus load.
Water Resour. Res. 56 (3) e2019WR025985.
Woolway, R.I., Kraemer, B.M., Lenters, J.D., Merchant, C.J., O’Reilly, C.M., Sharma, S., 2020. Global lake responses to climate change. Nat. Rev. Earth Environ. 1 (8),
388–403.
Wu, X., Lv, M., Jin, Z., Michishita, R., Chen, J., Tian, H., Tu, X., Zhao, H., Niu, Z., Chen, X., Yue, T., Xu, B., 2014. Normalized difference vegetation index dynamic and
spatiotemporal distribution of migratory birds in the Poyang Lake wetland, China. Ecol. Indic. 47, 219–230.
Wünnemann, B., Yan, D., Jiang, Z., Chen, G., 2024. Holocene process-based hydroclimate evolution coupled with human behaviours in Dian Lake basin. Southwest
China Catena 237, 107771.
Xi, Y., Peng, S., Ciais, P., Chen, Y., 2021. Future impacts of climate change on inland Ramsar wetlands. Nat. Clim. Change 11 (1), 45–51.
Xiang, H., Xi, Y., Mao, D., Mahdianpari, M., Zhang, J., Wang, M., Jia, M., Yu, F., Wang, Z., 2023. Mapping potential wetlands by a new framework method using
random forest algorithm and big Earth data: a case study in China’s Yangtze River Basin. Glob. Ecol. Conserv. 42, e02397.
Xiao, X., Hollinger, D., Aber, J., Goltz, M., Davidson, E.A., Zhang, Q., Moore, B., 2004. Satellite-based modeling of gross primary production in an evergreen needleleaf
forest. Remote Sens. Environ. 89 (4), 519–534.
Xu, H., 2006. Modification of normalised difference water index (NDWI) to enhance open water features in remotely sensed imagery. Int. J. Remote Sens. 27 (14),
3025–3033.
You, H., Liu, Y., Lei, P., Qin, Z., You, Q., 2023. Segmentation of individual mangrove trees using UAV-based LiDAR data. Ecol. Inform. 77, 102200.
Yuan, Y., Yan, D., Wang, H., Wang, Q., Weng, B., 2014. Quantitative assessment of drought in a lacustrine wetland based on a water balance model. Nat. Hazards 70
(1), 693–703.
Zeng, J., Qiu, J., Wu, Z., Liu, X., Li, Y., 2024. Impact of the Three Gorges Dam on hydrological connectivity and vegetation growth of Poyang Lake floodplain, China.
J. Hydrol. 631, 130831.
Zhang, Z., Jin, G., Tang, H., Zhang, S., Zhu, D., Xu, J., 2022. How does the three gorges dam affect the spatial and temporal variation of water levels in the Poyang
Lake? J. Hydrol. 605, 127356.
Zhang, D., Liao, Q., Zhang, L., Wang, D., Luo, L., Chen, Y., Zhong, J., Liu, J., 2015. Occurrence and spatial distributions of microcystins in Poyang Lake, the largest
freshwater lake in China. Ecotoxicology 24 (1), 19–28.
Zhang, Y., Su, Y., Liu, Z., Sun, K., Kong, L., Yu, J., Jin, M., 2018. Sedimentary lipid biomarker record of human-induced environmental change during the past century
in Lake Changdang, Lake Taihu basin, Eastern China. Sci. Total Environ. 613–614, 907–918.

14
F. Bu et al. Global Ecology and Conservation 59 (2025) e03585

Zhang, X., Wan, W., Fan, H., Dong, X., Lv, T., 2024. Temporal and spatial responses of landscape patterns to habitat quality changes in the Poyang Lake region, China.
J. Nat. Conserv. 77, 126546.
Zhao, F., Feng, S., Xie, F., Zhu, S., Zhang, S., 2023. Extraction of long time series wetland information based on Google Earth Engine and random forest algorithm for a
plateau lake basin – a case study of Dianchi Lake, Yunnan Province, China. Ecol. Indic. 146, 109813.
Zhao, C., Qin, C.-Z., Wang, Z., Mao, D., Wang, Y., Jia, M., 2022. Decision surface optimization in mapping exotic mangrove species (Sonneratia apetala) across
latitudinal coastal areas of China. ISPRS J. Photogramm. Remote Sens. 193, 269–283.
Zheng, L., Wang, X., Li, D., Xu, G., Guo, Y., 2021. Spatial heterogeneity of vegetation extent and the response to water level fluctuations and micro-topography in
Poyang Lake, China. Ecol. Indic. 124, 107420.
Zhong, Y., Xue, Z., Davis, C.C., Moreno-Mateos, D., Jiang, M., Liu, B., Wang, G., 2022. Shrinking habitats and native species loss under climate change: a multifactorial
risk assessment of China’s Inland Wetlands. Earth’S. Future 10 (6) e2021EF002630.

15