Land-Use Change and Driving Force Analysis of Wetland in Poyang Lake Based on Remote Sensing

Poyang Lake is one of the largest freshwater lakes in China and the only two remaining lakes in the middle reaches of the Yangtze River. It is an internationally important wetland with important ecological value. In recent years, due to the influence of natural and human factors, the ecology of Poyang Lake International Wetland has been greatly damaged. Using relevant research models and methods to study and predict land-use change can provide a scientific basis for relevant departments to manage land and has important demonstration significance for biodiversity conservation and ecological function restoration of wetlands with important ecological value. Based on Landsat remote sensing images from 1986 to 2020, land-use information data were obtained by supervised classification interpretation, and five land-use type maps with an interval of about 8 years were generated by ENVI and ArcGIS software. The land-use change of Poyang Lake wetland was analyzed by using land-use transfer matrix and land-use dynamic attitude. The land-use distribution in 2030 was predicted by the FLUS model. Based on the gray correlation model, the driving force of land-use change was discussed. The results show that the wetland land-use types in Poyang Lake will change greatly from 1986 to 2020: construction land, mudflat, paddy field, and dry land changed significantly and showed an increasing trend. The area of water and grassland decreased on the whole, and the transferred area was large, which was mainly transferred to construction land, paddy field, and dry land. The area of woodland increased slowly, but the change range was not large from the perspective of dynamic attitude. The change of wetland area of reed flat decreased first and then increased, and the overall land-use change was relatively gentle. Through the FLUS model prediction, it is found that the water area of the study area will be greatly reduced in 2030, the area of woodland and reed flat will be greatly increased, and the ecological situation will gradually improve. According to the gray correlation analysis, the total population and annual precipitation are the main driving factors of land-use change in the Poyang Lake wetland. This study studies the land-use change and driving force from the regional wetland scale, which can provide a theoretical basis for the future international important wetland protection and land resource management of Poyang Lake. At the same time, it provides a reference for further regional wetland habitat protection and ecological network construction.


Introduction
Wetland is an ecosystem with unique functions.As a special land-use type, it plays an important role in human production activities.It provides abundant natural resources for human beings and is one of the environments for human survival (Mao, 2016).Wetland land-use types are abundant, and it is of great significance to study the changes in wetland land-use types for wetland protection and ecological management (Han, 2017).However, since the twentieth century, the decrease in wetland areas has led to the decline of wetlands' ecological value and service function.Human disturbance and climate change have affected wetlands in recent decades (Guo et al., 2020;Ren et al., 2016).Wetland resources were being developed and utilized continuously, and the land-use pattern of wetlands has changed significantly (Cui et al., 2019;Jian et al., 2020).The environmental and ecological problems of wetlands have attracted great attention around the world (Tao et al., 2021;Zhu et al., 2021).Researchers are looking for rational use of wetlands to promote the harmonious coexistence of people and wetlands (César et al., 2014;Mustafa et al., 2007).In this context, an increasing number of studies have focused on land use within wetlands (Zhang et al., 2019;Zhou et al., 2019).Therefore, the study of land-use change and driving force analysis for long-time series of wetland ecosystems are of profound significance for maintaining the healthy development of the wetland ecological environment.
At present, domestic and foreign scholars have carried out a lot of research on wetland ecological risk or vulnerability and analyzed the spatial and temporal changes and spatial clustering of wetland ecological risk and vulnerability.Li et al. established an index system through two aspects: the degree of external harm and the vulnerability of the wetland itself.Through long-term data and multiple indicator factors, the wetland ecological risk in six periods from 1990 to 2015 was comprehensively evaluated, and the risk change and vulnerability characteristics among wetland risk sources were comprehensively analyzed (Li et al., 2020).Yang et al. established the ecological environment vulnerability assessment model of the Three Gorges Reservoir area, used the improved k-means clustering algorithm to classify the ecological environment of the study area, and evaluated the vulnerability of the wetland ecological environment, providing a scientific and accurate method for the management and sustainable development of urban wetlands (Yang et al., 2021).Manob et al. used subjective KBRM method and objective PSR model to explore the temporal and spatial changes of wetland ecological risk and analyzed the spatial clustering of wetland ecological risk through spatial autocorrelation, which was conducive to the spatial management of wetland ecological risk and sustainability (Manob et al., 2022).However, most of these studies focus on the analysis of wetland ecological risk and ecological vulnerability, ignoring the importance of land-use and land-cover (LULC) data in wetland landscape pattern and degradation research, and the selection of driving factors is less studied (Wan et al., 2011;Mondal et al., 2017).Land-use change can affect hydrology, water quality, regional climate, and biodiversity.Human activities are the most important factors affecting wetland land-use change, and agricultural activities are the main human activities that cause wetland environmental pollution (Xu et al., 2022).Therefore, this study analyzes the long-term sequence of land-use change in Poyang Lake by Envi and ArcGIS software and uses the gray correlation model to analyze the driving force factors, which plays a direct positive role in the protection of the wetland ecosystem.
Poyang Lake is the largest freshwater lake in China, which plays an important role in regulating the water level of the Yangtze River, improving the local climate, and maintaining the ecological balance of the surrounding areas (Tian et al., 2021).Debela et al. studied the habitat suitability of antiformal in vegetation, tidal flats, water bodies, and sandy land in Poyang Lake (Debela et al., 2021).Fang analyzed the status quo of Poyang Lake wetland water quality and its correlation with the surrounding land use at different scales (Fang, 2020).Wang et al. analyzed the changes in the ecosystem and water area in the Poyang Lake region through remote sensing monitoring (Wang et al., 2019).Zhang used the ca-Markov model to predict the coupling degree of land intensity and ecological security in the Poyang Lake watershed in 2020 (Zhang, 2018).Previous studies have focused on the lack of long-term changes in land-use types in the Poyang Lake Basin.Wang et al. interpreted the land-use types of Laoha River Basin by using remote sensing images in 1985, 1995, 2006, 2015, and 2019 and explored the driving forces of wetlands through principal component analysis.They found that human factors were the main cause of wetland landscape fragmentation in the Laoha River Basin (Wang et al., 2021).Taking Caizihu wetland in Anhui Province, China as the research object, Peng Na interpreted satellite remote sensing images in 1999, 2004, 2011, and 2017 and used geographic detectors and correlation analysis to quantitatively explore the influence and importance of economic expansion, fishery economic development, and human growth on wetland evolution.The study found that the landscape area changed significantly, the area of natural wetlands decreased significantly, and the fragmentation of the wetland landscape gradually intensified, which was affected by human activities and the social economy together (Peng, 2021).Therefore, based on the remote sensing images of Poyang Lake wetland from 1986 to 2020 and the data on wetland land-use change, this paper analyzed the change rate, transfer direction, and driving factors of each land-use area in Poyang Lake wetland and analyzed the impact of natural factors and human activities on the land-use area change in Poyang Lake: to provide reference data for the harmonious development of the human-land relationship and human-bird relationship in Poyang Lake wetland, to provide suggestions for wetland protection and management, and to achieve the goals of sustainable economic development of Poyang Lake wetland, "frequent visit of migratory birds," and harmonious coexistence between human and nature.

Study Area
Poyang Lake National Nature Reserve (115° 47′ E-116° 45′ E, 28° 22' N-29° 45' N) is located in the northern part of Jiangxi Province, China, covering Yongxiu, Xingzi, and Xinjian counties, with Wucheng Town of Yongxiu as the center.The reserve is composed of nine lakes: Beng Lake, Dacha Lake, Dahu Lake, Sha Lake, Changhu Lake, Zhonghu Lake, Xiang Lake, Meihu Lake, and Zhushi Lake.In addition, it includes more than a dozen lakes and grasslands nearby, with a total area of 22400 hm 2 .Based on the limits of Poyang Lake National Nature Reserve, the study area selected in this paper extends about 1 km around (about 5 km to the southeast, including Vida Dou Lake) according to the topography and the distribution of rivers and lakes, with a total area of 42000 hm 2 (Fig. 1).Poyang Lake wetland in Jiangxi province has a subtropical monsoon climate with great seasonal variations.From 1986 to 2020, the annual average temperature of the Poyang Lake wetland was 17.9 ℃, the highest temperature in 2007 was 18.1 ℃, and the lowest temperature in 2012 was 17.8 ℃.The average annual precipitation was 1550.1 mm, with a minimum precipitation of 967 mm in 2006 and a maximum precipitation of 3250.5 mm in 1998.

Data Source and Processing
Five remote sensing images of Poyang Lake were selected from 1986 to 2020, including 1986, 1994, 2002, 2010, and 2020 (there is a 10-year gap between 2010 and 2020).The basic image data came from the Geospatial Data Cloud Platform (http:// www.gsclo ud.cn/).In 1986In , 1994In , and 2002, the satellite parameters and types were Landsat4-5, TM, with a spatial resolution of 30 m.In 2010 and 2020, the satellite parameters and types were Landsat8 and OLITIRS, with a spatial resolution of 15 m.To facilitate the unification of data format and data processing and analysis, classification results were uniformly converted to a spatial resolution of 30 m after the completion of interpretation.ENVI is used for atmospheric correction, image enhancement, image fusion, image clipping, and other data preprocessing.
Shengjin Lake and Poyang Lake are typical inland lake wetlands in China, and Xu et al. divided the Shengjin Lake wetland landscape into eight categories when studying the land landscape changes in Shengjin Lake.They are dry land, woodland, paddy field, construction land, reed flat, grassland, and mudflat (Xu et al., 2023).By referring to the relevant research of domestic scholars, the classification of land-use status, and the characteristics of Poyang Lake wetland land types, the land-use types in the study area are divided into eight types: water, construction land, paddy field, dry land, woodland, grassland, reed flat, and mudflat.In the selection of images, to ensure the accuracy of the selection of images, mainly from November to February of the next year, on the one hand, it prevents the image from being disturbed by cloudy and rainy weather, and at the same time, it prevents the water level rising in the wet season from obscuring other land-use types.On the other hand, the change range of land-use type is small, and the image accuracy is not affected.Different ground objects have different spectral features, which are reflected in different colors, textures, and shapes of remote sensing images.According to the color and texture differences of the five remote sensing images of the Poyang Lake wetland and combined with the land type characteristics of Poyang Lake wetland, eight land-use types were mapped to the images one by one.Based on the identification of image classification signs, combined with the investigation of the land-use status of Poyang Lake wetland, the ENVI software supervised classification support vector machine method and was used to supervise the classification of five Poyang Lake wetland remote sensing images, and the landuse grid maps in 1986, 1994, 2002, 2010, and 2020 were obtained.Using the conversion tool and statistical analysis in ArcGIS, the land-use type maps of Poyang Lake wetland in five periods were obtained.Remote sensing images may produce errors during supervised classification.To ensure the accuracy of classification results, ENVI software was used to conduct confusion matrix analysis and calculation and kappa coefficient value to verify the accuracy of results.Through calculation, the overall accuracy of Poyang Lake wetland in 1986, 1994, 2002, 2010, and 2020 is 88.36, 89.98, 89.98, 89.65, and 89.96, respectively, and the kappa coefficients are 0.8690, 0.8731, 0.8745, 0.8720, and 0.8721, respectively.Then NDVI is introduced to enhance the classification accuracy.NDVI has been widely used in vegetation growth monitoring, vegetation spatial distribution density, and other aspects, is sensitive to different soil backgrounds of vegetation, and is the best indicator of vegetation information.The classification accuracy after 1 3 introducing NDVI is shown in Table 1.After adding the NDVI index, the classification accuracy of water, paddy field, woodland, dry land, grassland, reed flat, mudflat, and construction land increased by 1.6%, 3.3%, 5.4%, 4.3%, 5.48%, 1.85%, 2.35%, and 3.1%, respectively.It meets the requirements of the study and shows the reliability and authenticity of the classification results and provides reasonable data support for further research.

LULC Classification
At present, LULC classification mainly includes the maximum likelihood method, random forest algorithm, neural network algorithm, support vector method, and so on.Maximum likelihood classification has the advantage of being simple and convenient and classified according to the theory of BAYES theorem and other prior knowledge fusion, the density distribution function can effectively and clearly explain the classification results, but this kind of method is suitable for use in a band less multiband data, classification of time delay, and the classification of time with the increase in band information into a quadratic growth.The random forest algorithm obtains the features of the land image, establishes the land feature set based on the Boruta feature selection algorithm, and completes the land classification by combining various classification algorithms.However, this method does not enhance the acquired land image, which leads to low user accuracy and overall accuracy of classification results and reduces the classification effect of this method.The classification accuracy of neural network algorithm is high, and the parallel distributed processing ability is strong, but it needs a lot of parameters, which cannot observe the learning process between, and the output result is difficult to explain.In this paper, a support vector machine (SVM) was used for LULC classification.SVM can improve the clarity of images, obtain the details of remote sensing images through image enhancement processing, and then improve the classification effect and planning effect of the method.The support vector machine can improve the image sharpness and obtain the details of remote sensing images through image enhancement processing and then improve the classification effect and planning effect of the method.

Land-use Transfer Matrix Analysis
The land-use transfer matrix is mainly used to analyze the transfer rate and direction between different land-use types in different periods and to analyze the internal correlation and change trend among land-use types.Transfer matrix formula (1) is as follows: P represents the area transferred by land-use type; N represents the classification number of land-use types in the (1) study area, i,j (i,j are 1, 2, 3…Integer of n) represents the area before and after the transfer of a certain land-use type; Pij represents the number of land-use types converted from Class I land-use type to Class j land-use type.

Land-use Change Rate Analysis
The single land-use dynamic attitude represents the change in the number of land-use types in the study area within a certain time range.Its specific expression is as follows: In the formula, K represents the change of dynamic attitude of land-use type in the study area; Ua and Ub represent the number of land-use types in the early and late stages of the study area; T is the duration of the study; Ub-Ua represents the amount of land-use type change in the study area (negative value means decreased area and a positive value means increased area).

Simulation of Land-use Change Based on the FLUS Model
At present, the commonly used models of land-use prediction are gray prediction model, logistic regression model, cellular automata model, CLUE-S model, and so on (Liu et al., 2017).Most of these models cannot take into account the impact of quantity and time and space on land-use at the same time.As a new model developed in recent years, the FLUS model integrates an artificial neural network algorithm and roulette wheel selection mechanism on the basis of a cellular automata model (Li et al., 2022).It can simulate future land-use change by integrating a variety of natural and socioeconomic factors through a period of land-use data.The simulation results have high similarity and accuracy with reality (Feng and Chen, 2023;Liang et al., 2018).FLUS model can simulate the long-term spatial evolution trajectory of various types of land use.Through the coupling between top-down system dynamics (SD) and bottomup cellular automata (CA), an adaptive inertia and competition mechanism is established in the CA model, which improves the simulation and prediction capabilities (Wang, 2021).The Markov chain module of the FLUS model was used to predict the future land-use pixel.Based on the two periods of land-use data of 2010 and 2020, the land-use data for 2030 are predicted, and the formula is: S t and S t+1 represent the state of land at time t and time t + 1, respectively, and P ij represents the probability of land type transformation at time t.

Gray Correlation Analysis
(1) Calculation of correlation coefficient.
The correlation coefficient between X 0 (k) and Δ i (k) ∈ (0, ∞) is called the resolution coefficient.The smaller is, the greater the resolution is.Generally, the value of is (0,1), which is usually set as = 0.5.
(2) Calculate the correlation degree.The correlation degree between the comparison sequence Xi and the reference sequence X0 is reflected by N correlation coefficients, and the correlation degree between Xi and X0 can be obtained by calculating the average value 0i :

Land-use Quantity Change
Envi software was used to obtain the TM images of the Poyang Lake protective area in five phases (Fig. 2). (3) ArcGIS, Excel, and other software were used to make a statistical analysis of the land-use data of Poyang Lake wetland in five phases in 1986, 1994, 2002, 2010, and 2020, and the area of each category in this region was obtained (Fig. 3).Based on this, the variation trend of the area of each class in each period is analyzed.
It can be seen from Figs. 2 and 3 that, among the landuse areas of Poyang Lake wetland from 1986 to 2020, the total area of water and grassland was large and the change range was small, but the overall area decreases continuously.The construction land area has been increasing and growing rapidly, mainly due to the influence of population increase and the urbanization process.The fluctuation of the paddy fields and dry land area was mainly affected by population factors and the policy of returning farmland to forests and lakes.The area of Woodland kept rising and increased rapidly from 2002 to 2010, which was mainly influenced by the policy of returning farmland to the forest by the local government.The change of reed flat area firstly decreased, then increased, and then decreased, which can be divided into three stages: The first stage, 1986-2002, was natural reed flat and was always in the declining stage; in the second stage, from 2002 to 2010, artificial planting and natural growth co-existed.The area of reed planted continuously increased, and the area of reed flat increased rapidly.In the third stage, there was a slight decrease from 2010 to 2020.The change in mudflat was mainly influenced by natural factors such as annual total precipitation and water level change, and the interannual change was great.

Land-use Type Area Transfer
In this section, ArcGIS software was used to merge and superimpose land-use data in different periods to obtain land-use transfer matrices in two periods and establish land-use transfer matrices from 1986 to 2020.
From Table 1, it can be seen that from 1986 to 2020, the areas of construction land, paddy field, dry land, woodland, grassland, and reed flat were mainly transferred to the water area, 36.63 hm 2 , 122.36 hm 2 , 65.51 hm 2 , 50.38 hm 2 , 3298.76 hm 2 , and 514.98 hm 2, respectively.The area of water had a large base, and the above land type was adjacent to Water.With the joint action of natural and policy factors in the past 34 years, the location of the surrounding water area had changed somewhat.The area of water and mudflat mainly shifted to grassland, with 3902.68 hm 2 and 131.21 hm 2, respectively.According to the field investigation, it can be found that grassland was close to water and mudflat.When the location of water and mudflat began to shift to other directions, the soil conditions were extremely suitable for the development of grassland.Although some 1 3 Fig. 2 Classification of wetland land-use types in Poyang Lake of the construction land, paddy fields, and dry land used by residents for production and living had been transferred out, it can be seen from Table 1 that the area of grassland turned into construction land was as high as 680.23 hm 2 , the area of water and grassland turned into paddy field was 561.69 hm 2 , and the area of water and grassland turned into dry land was 439.31 hm 2 .Compared with 1986, the area of construction land, paddy fields, and dry land has doubled, with a large area growth rate.With the increase in population, construction land, paddy field, and dry land, as important carriers of residents' production and life, tend to be rationally utilized and developed under the guidance of the local protection bureau and relevant administrative departments.

Land-use Dynamic Attitude Change
According to formula (2), the dynamic attitude of single land-use type in Poyang Lake wetland from 1986 to 2020 can be obtained.
From the perspective of the dynamic attitude of single land use (Fig. 4a), water, woodland, grassland, and reed flat had little change from 1986 to 2020.The change range of reed flat from 2002 to 2010 was 3.67%, which was mainly due to the policy support and guidance of the development of the reed industry.Construction land has been increasing from 1986 to 2020 with a growth rate of 7.04%.From 1986 to 2020, the dynamic rates of paddy field and dry land were 2.08% and 1.58%, respectively.From 2002 to 2010, the change rate of mudflats was 13.66%, indicating that the area of mudflats changed the most during this period, with a total growth rate of 4.17% from 1986 to 2020.
From 1986 to 2020 (Fig. 4b), construction land kept increasing, and it was the land category with the biggest change, followed by the area of Mudflat, and the change range was next to construction land.Paddy field, dry land, and woodland were increasing, while the area of water, grassland, and reed flats were decreasing.Population increase, economic development, and environmental policy were all factors affecting land-use change.

Land-use Prediction
The predicted land-use data in 2020 were compared with the actual value in 2020, and the overall accuracy and Kappa coefficient were calculated.The closer the value of overall accuracy and Kappa coefficient was to 1, the better the simulation accuracy was.When the kappa coefficient was greater than 0.8, it indicates that the simulation accuracy of the model was statistically satisfactory.Through accuracy verification, it can be seen that the overall accuracy is 0.931 and the kappa coefficient is 0.823, which achieved high simulation accuracy, indicating that the FLUS model has good applicability in this study.No restriction factor was set in the CA simulation process, and all ground classes are converted to each other according to the rules of the conversion cost matrix.
According to the simulation results (Table 2, Fig. 5), the water area of Poyang Lake decreased significantly in 2030, from 14,536.60 hm 2 in 2020 to 9868.95 hm 2 in Fig. 3 Data of land type area of Poyang Lake wetland in five stages 1 3 2030, with a decrease of 4667.65hm 2 .Secondly, the reed flat increased from 1372.37hm 2 in 2020 to 4016.07hm 2 in 2030, an increase of 2643.7hm 2 .Woodland increased from 1155.85 hm 2 in 2020 to 3527.46 hm 2 in 2030, an increase of 2371.61 hm 2 .Due to the gradual intensification of human activities, the construction land area showed an increasing trend, with a total increase of 719.95 hm 2 in 2030 compared with 2020.The areas of paddy field, grass flat land, and dry land showed a decreasing trend, while mudflat land showed a slightly increasing trend.As one of the main habitats of rare wintering cranes, reed flat has high ecological value.Woodland is the habitat and main food source of some rare species.In general, the significant increase in the area of reed flat and woodland makes the ecological status of Poyang Lake wetland gradually better in 2030.At the same time, it is necessary to pay attention to the expansion area of construction land in the central and western parts of the study area.Driving Forces of Land-use Change The driving forces of land-use change mainly use qualitative and quantitative methods.In the qualitative analysis, three natural driving forces and three social-economic driving forces were selected to explore the impact of each driving factor on the land-use change of Poyang Lake wetland.The quantitative analysis mainly selects the gray correlation analysis model, calculates, and obtains the correlation degree of driving factors of land-use change in Poyang Lake wetland by selecting driving factors, finds the driving factors of each type of change, and finally obtains the leading driving factors of eight land-use types.

Qualitative Analysis
The natural factors, mainly include three factors.
(1) Topographic factors: The land-use layout of Poyang Lake wetland is affected by topographic and geomorphic features.
There are three hillsides in the Poyang Lake wetland, namely Dingjia Mountain, Ji Mountain, and Yang Mountain.The terrain is high in the middle, low around, high in the west and low in the east, and the overall terrain is low.The study area can be roughly divided into mountain and hilly areas, low plain areas, and lake water areas.
(2) Climatic factors: For climatic factors, the study area has a subtropical monsoon climate, with high temperatures and rainy summer and cold and dry winter, presenting a scene of "abundant water and dry water."The average annual precipitation days of Poyang Lake wetland reached 143 days, mainly from June to September, during which the water level area of Poyang Lake rose, the water surface area reached the maximum, and the total precipitation increased with the increase of annual precipitation days, the annual average precipitation total from 1986 to 2020 was 1135.55 mm.(3) Hydrological factors: the impact of hydrological factors on land-use change is mainly manifested in water level and sediment concentration.Poyang Lake wetland is influenced by the Ganjiang River, Xiushui River, and Poyang Lake, and is one of the lakes that regulate the water level of the Yangtze River.The flood from the middle and upper reaches of the Yangtze River flows into Poyang Lake, which leads to the rapid hydrological rise of Poyang Lake.The hydrological elements of Poyang Lake are one of the important driving factors of land change in the study area.Social and economic factors, mainly include three factors.(1) Population: the number population directly affects the degree and efficiency of land use.As human beings engage in production and life, they are bound to transform the land.The pressure of population growth on land increases and more land is needed to meet the needs of their development.Studies show that population growth is in direct proportion to the area of construction land and cultivated land.
(2) Economic factors: They play an important role in the process of land-use change.GDP can reflect the degree of development and utilization of land resources in a region and is also an important indicator to measure the level of economic development in a region.In the process of economic development, the industrial structure is constantly optimized and adjusted to improve the construction of social infrastructure, and the land-use structure also changes accordingly.(3) Policy factors: Policy factors play a driving role in land-use change and are important factors affecting land-use structure in the region.The introduction of policies can optimize the layout of industrial structures, promote the change of landuse mode, and drive the coordinated development of the regional economy.Under the policy control and influence of the government, the land-use types of Poyang Lake wetlands are constantly changing.

Quantitative Analysis
The land-use change of Poyang Lake wetland is affected by many factors, not only by the natural environment but also because of the change in human use.Since the qualitative analysis of the driving forces of land-use change cannot identify the leading driving forces of land-use change in Poyang Lake wetland, the gray correlation analysis model is used to conduct a quantitative analysis of the driving forces of land-use change based on the principles of scientific nature, correlation with wetland change, and accessibility of data.Find the most important driving factors affecting land-use type change.
Combined with the actual situation of the Poyang Lake wetland and the research needs, the driving factors of this paper are determined as follows: Natural factors include total annual precipitation (X1), average annual temperature (X2), average annual water level difference (X3), etc. Social and economic factors include total population (X4), the total output value of primary industry (X5), the total output value of secondary industry (X6), the total output value of tertiary industry (X7), the total grain output value (X8), the total output value of fishery (X9), etc.The data came from the Jiangxi Provincial Statistical Yearbook and Yongxiu County Statistical Yearbook, and the economic data of the Poyang Lake wetland were obtained after modification, and the correlation degree was calculated (Table 3).
It can be seen from Tables 4 and 5 that each type of landuse change had its leading driving factors, among which the main driving factors of water change and reed flat change were the annual mean water level difference, and the gray correlation coefficient was 0.4934 and 0.4980, respectively.The main driving factor for the change in construction land, paddy field, and woodland was the total population, and the gray correlation coefficient was 0.4774, 0.4806, and 0.4931, respectively.Dry land was mainly driven by the total output value of the primary industry, and the gray correlation coefficient was 0.4745.The gray correlation coefficients of grassland and mudflat areas were 0.4937 and 0.4796, respectively.
The land-use type was not affected by a single driving factor, but by many driving factors at the same time (Di, 2019).Among the natural factors, the leading driving factor of grassland and mudflat was the total annual precipitation, but the annual mean water level difference can also cause changes in the area of grassland and mudflat.In social and economic factors, the change in the population there was a significant effect on the various land-use type.Taking reed flat as an example, in its natural state, the area of reed flat was only affected by climate and precipitation.However, in the land-use type area table of Poyang Lake wetland from 1986 to 2020, reed flat decreased from 1986 to 2002.But, since 2010, the area of reed flat began to increase, mainly due to the introduction of emerging industries and characteristic planting, policy support, and guidance for the development of the reed industry.And relevant management departments gradually realize the importance of wetland protection and economic development in parallel, strengthen the protection of wetland resources, and promote the harmonious coexistence of man and nature.In general, population change and climate factors were the main driving factors of land-use change in the Poyang Lake wetland.

Discussion
The Influence of Various Driving Factors on the Wetland Landscape of Poyang Lake Land-use change is affected by many factors, and the response of ecosystem change to land-use/change (LUCC) is still elusive.In particular, wetland ecosystems are highly sensitive to climate change and human activities (Cui et al., 2021;Wei et al., 2020).By analyzing the present situation of the Poyang Lake wetland, this paper provided the scientific basis for how to better protect and restore the ecological function of the wetland.Poyang Lake wetland has become the focus of economic development because of its unique geographical location and resource characteristics.Due to population change and climate disturbance, the wetland landscape of Poyang Lake has undergone significant changes.Poyang Lake wetland is mainly composed of water and grassland, followed by reed flat and woodland, construction land and arable land account for a relatively small proportion.From 1986 to 2020, water and grassland were the two most important land-use types, but their area was decreasing.On the whole, the area of woodland showed a slowly rising trend.Mudflat fluctuates greatly due to hydrology, total annual precipitation, and other natural factors.Reed flat appeared the trend of the first decline and then rise.The area of construction land was on the rise year by year, and the rising rate was fast, mainly due to the increase of land for infrastructure construction and development.The area of cultivated land made up of paddy fields and dry land was rising.The wetland ecosystem was gradually fragmented, weakening the energy storage capacity of the wetland, leading to changes in the hydrological conditions of the whole wetland, and weakening the ecological benefits of the wetland.The increase of construction land and the decrease of reed banks led to the decrease of natural wetland area in the Poyang Lake area, and the destruction of the habitat of birds and other organisms, affecting the living environment of birds.Poyang Lake wetland was affected by precipitation and population, so it was urgent to take restoration measures for ecological protection and restoration.

Global Sensitivity Analysis and Potential Limiting Factors
In this study, the global sensitivity analysis of LUCC and various driving factors in Poyang Lake was carried out.This analysis can assign the uncertainty of change to the driving factors, describe the change of the predicted area, and supplement the above gray correlation analysis (Chu-Agor et al., 2011;Servadio and Convertino, 2018).We show the Sobol first-order index as the relative sensitivity of each parameter (Table 6).The Sobol index helps to determine the main driving factors of the output.The sum of the Sobol index is almost 1, reflecting that the total variance can be completely explained by the individual variation of the measured parameters (Hoque et al., 2020(Hoque et al., , 2021)).On the other hand, the interval does not provide any residual variance for the modeling of the variation, indicating that there is no significant interaction between the interval and the total variance.
The Sobol method realizes the sensitivity analysis of the input variables of the decision tree regression model.Firstly, the train _ test _ split function was used to divide the input variables and output variables, and the Decision-TreeRegressor model was used to fit the training set.Then, the input variables were sampled using the saltelli.samplefunction in the SALib library, and the sampled values are used as test data.The trained decision tree model was used to predict the parameter sampling values.Then, the Sobol sensitivity analysis of the prediction results was performed using the sobol.analyzefunction in the SALib library, and the first-order sensitivity index of each input variable was obtained.The sensitivity analysis method can help researchers identify which input variables have the greatest impact on the output variables, so as to better understand the behavior of the model and make corresponding improvements.The first-order sensitivity index represents the contribution of an independent variable to the dependent variable.The larger the value, the greater the influence of the independent variable on the dependent variable.The following is an analysis of the influence of each independent variable on the dependent variable: The first column (dependent variable water): The firstorder sensitivity indexes of total annual precipitation and annual average water level difference are larger, which are 0.639 and − 0.1202, respectively, indicating that these two independent variables have a greater impact on water, and the annual average water level difference is negatively correlated with water.The second column (dependent variable construction land): the first-order sensitivity index of the total output value of the fishery is the largest, 0.493, followed by the total population, 0.2941, indicating that the total output value of the fishery and total population may be the main factors affecting construction land.The third column (dependent variable paddy field): THE firstorder sensitivity index of the total annual precipitation and the total output value of the primary industry is larger, which is 0.284 and − 0.1604, respectively, while the first-order sensitivity index of other independent variables is relatively small.The total annual precipitation and the total output value of the primary industry are the main factors affecting the paddy field, and the total output value of the primary industry is negatively correlated with the paddy field.The fourth column (dependent variable dry land): we found that the first-order sensitivity index of the total output value of the fishery and the total output value of the primary industry is larger, which is 0.284 and 0.1204, respectively, while the first-order sensitivity index of other independent variables is relatively small.The total output value of the fishery and the total output value of primary industry are the main factors affecting dry land.The fifth column (dependent variable woodland): the first-order sensitivity index of total output value of fishery and total population is relatively large, which is 0.57 and − 0.2594, respectively, while the first-order sensitivity index of other independent variables is relatively small.This suggests that the total output value of fishery and total population may be the main factors affecting forest land.The sixth column (dependent variable grassland): The first-order sensitivity index of total grain output value and total annual precipitation is the largest, which is 0.1621 and 0.2592, respectively, while the first-order sensitivity index of other independent variables is relatively small.Total grain output value and total annual precipitation are the main factors affecting grassland.The seventh column (dependent variable reed flat): The firstorder sensitivity index of the total output value of the fishery and annual average water level difference is larger, which is 0.4755 and 0.2726, respectively, which is 0.4755, indicating that the total output value of the fishery and annual average water level difference may be the main factors affecting reed beach.The eighth column (dependent variable mudflat): the first-order sensitivity index of the total annual precipitation is the largest, 0.311, and the total annual precipitation is the main factor affecting the mudflat.The research results correspond to the above gray correlation analysis.The first-order sensitivity index of Sobol, the driving factor with the highest gray correlation degree, is also high.This paper combines gray correlation degree and the Sobol model to provide an effective reference for the next wetland protection and management of Poyang Lake.

Limitations of the Research and the Direction of Future Work
Different human activities affect ecosystem services through land-use change, which in turn affects the ecological security pattern (Li et al., 2023).In this study, based on the current situation of wetland field investigation, we refined the land-use types to eight, avoiding subjectivity.However, there are still many opportunities and challenges in this study.Firstly, in the prediction of land use, we found that the area of water, paddy field, grassland, and dry land in Poyang Lake wetland showed a decreasing trend, among which the water area changed most dramatically.The area of reed flat land, woodland, construction land, and mud beach land showed an increasing trend, and the reed flat land changed most dramatically.However, due to differences in time and space, ecological restoration measures for important drivers lag behind in improving ecosystems, which brings many uncertainties to future benefits.With the continuous occurrence of extreme events in the Poyang Lake area, these conditions are undoubtedly the key to the implementation of long-term ecological protection and observation (Zou et al., 2023).Secondly, due to the limited research on land-use prediction in different scenarios of international important wetlands, this study does not involve multi-scenario land-use prediction such as natural growth, cultivated land protection, and ecological protection.Finally, the study of ecological security pattern has an important impact on the ecological restoration of wetlands (Gao et al., 2021), which can effectively visualize the spatial heterogeneity and movement characteristics of rare species (Huang et al., 2021).Recent research results show that based on the circuit theory, it can play a positive role in the protection of rare white-headed cranes in internationally important wetlands.In the future research, it is necessary to expand the research from the aspects of multi-scenario land-use prediction, ecosystem services and ecological security pattern.

Conclusion
Poyang Lake wetland is a typical area of the Yangtze River wetland, and the research results can also be used in all ecological planning and protection of the Yangtze River wetland.Based on remote sensing image interpretation, this paper analyzes the spatiotemporal changes of wetland landscape patterns and their driving mechanisms, to provide a scientific basis for wetland regional planning and ecological construction.To better protect the natural ecology and habitat protection of Shengjin Lake wetland and reduce human disturbance, the following suggestions are put forward for the future protection of Poyang Lake: Build overall planning layout: The area of Poyang Lake wetland is large, and there are many cross-counties and cities.The management of each region is not coordinated, management loopholes are frequent, and there is no unified overall plan.The existing local planning is diverse, there are insufficient management, landuse planning lag, and other problems.According to the general strategic direction and goal of Poyang Lake wetland, based on the actual situation of nature, economy, and society of the wetland, the development, utilization, management, and protection of land resources are coordinated regionally, and multi-party relations are coordinated, to construct the overall planning layout, and promote the sustainable development of land use and the comprehensive development of the regional economy.
Raise awareness of wetland protection.
Poyang Lake wetland residents are mostly farmers with low education levels and weak awareness of environmental protection.In the process of agricultural production, they consider more economic benefits and make extensive use of wetland resources, ignoring ecological benefits.To avoid this situation, on the one hand, it is necessary to strengthen people's ecological protection publicity, publicity, education is an indispensable part of the cause of nature protection, through the wetland protection publicity film, wetland protection brochures, and other simple forms to enhance people's awareness of environmental protection; on the other hand, it can develop the study tour market, build a new industrial system of "tourism + education + situational experience," design research tour products with the dimension of ecological protection, realize the organic combination of "learning" and "tourism," and enhance tourists' ecological protection consciousness based on satisfying tourists' play.Finally, by creating jobs and creating environmental protection groups, we can attract residents to join us and turn former environmental vandals into today's conservationists.
Improve laws and regulations on ecological progress.
Based on the land use of Poyang Lake wetland, the analysis of the problems existing in the process of combining with the sustainable development of the Poyang Lake wetland land use and the concept of ecological civilization into the economy, environment, and development plan, follow under the premise of ecological civilization construction and development planning, to further improve in Poyang Lake wetland planning of ecological civilization construction policy system.

Fig. 1
Fig. 1 Geographic location of the study area

Fig. 4
Fig. 4 Dynamic attitude of land use (a).Dynamic attitude of land use (b)

Fig. 5
Fig. 5 The distribution map of land-use types in Poyang Lake in 2030

Table 2
Land-use structure change of Poyang Lake Wetland in 2020 and 2030 (Unit: hm 2 )

Table 3
Driving factor data of Poyang Lake wetland