Characteristics of Spatial and Temporal Evolution of Coastal Wetland Landscape Patterns and Prediction Analysis—A Case Study of Panjin Wetland, China

Abstract

The Panjin Wetland is a complex ecosystem comprising coastal and inland wetland. It has an important function in wetland conservation and ecological. For this research, we quantified the landscape type changes in Panjin Wetland from 1992–2022, and analyzed the interaction between the combined PLUS and InVEST models to predict the future evolution of spatial and temporal patterns of habitat quality (HQ) and landscape patterns in Panjin Wetland. The results showed the following: (1) The change in natural wetland area from 1992 to 2022 generally showed a decreasing trend, the landscape patterns showed a trend of fragmentation. In 2032, the ecological protection scenarios showed an increase of 79.51 km² of natural wetland, while the other scenarios showed a decrease. (2) In 2022, the average habitat quality score is 0.441, and in 2032, the average habitat quality scores in the natural development scenarios, farmland protection scenarios, ecological protection scenarios, and economic development scenarios are 0.427, 0.448, 0.438, and 0.416, respectively. (3) The outcomes of this study offer insights into the sustainable management and rational development of coastal wetland, thereby contributing to the existing body of knowledge in the field of coastal wetland research, particularly in the areas of land planning and forecasting.

1. Introduction

Wetlands are of significant ecological, economic, and social value. Coastal wetlands, situated at the interface between marine and terrestrial environments, play a pivotal role in maintaining regional ecological security, environmental health, biodiversity, and sustainable development [1,2,3]. In recent years, the relentless advance of urbanization, industrialization, and agricultural development has severely disrupted and encroached upon wetland resources, leading to increasingly serious degradation and destruction [4,5,6,7]. Understanding the macro-patterns of wetland landscapes and their changing dynamics is foundational for conservation research and serves as an important means for regional wetland ecological assessment, restoration effect evaluation, and comprehensive management studies [8,9,10].

Landscape patterns are the most significant expression of changes in landscape types, and many scholars have employed land allocation models to explore changes in wetland landscape patterns, aiming to unveil the evolution patterns of these landscapes [11,12,13,14]. The allocation of land resources is studied based on a dual-scale perspective of time and space, and the establishment of models serves as a quantitative method for researching Land Use/Cover Change (LUCC) [15,16]. Since the last century, scholars both domestically and abroad have constructed numerous research models, primarily encompassing land quantity prediction models and land spatial simulation models. The former emphasizes a top-down approach to studying changes in land quantity structure over time, including Markov models, linear programming models, and System Dynamics (SD) models [17,18]. As static models, Markov and linear programming models find it challenging to reflect the interrelationships among various factors influencing LUCC. Conversely, SD models leverage computer simulation technology to effectively simulate the dynamic causal relationships between LUCC and driving factors, thus being widely applied in LUCC research across different scales. Land use spatial simulation models focus on studying the characteristics of land use distribution patterns from a bottom-up perspective, utilizing spatial scales [19]. These models include Cellular Automata (CA), CLUE-S (Conversion of Land Use and Its Effects at Small Regional Extent), FLUS (Future Land Use Simulation Model), and PLUS (Patch-generating Land Use Simulation Model) [20,21,22,23,24,25,26,27]. The PLUS model skillfully simulates the dynamic evolution of different landscape types at the patch level, effectively addressing the uncertainties inherent in LUCC development driven by the interplay of socio-economic factors and the natural environment. In summary, both the FLUS model and the PLUS model stand as mainstream research methodologies for simulating landscape demands and spatial changes. However, the FLUS model imposes stringent requirements on the consistency of certain input data attributes, whereas the PLUS model not only offers enhanced data compatibility but also integrates a novel multi-type random patch seed, allowing for a more nuanced simulation of the changing trends of diverse landscape types at the patch level.

Panjin Wetland is located in the core of Liaohe River Delta. The ground is covered with deep silt deposits, which is a special surface formed under the alternating effects of river alluvium and marine deposition, constituting a coastal wetland complex ecosystem characterized by extensive reed swamp. Reed swamps are wetland habitats characterized by the dominance of reeds. They are often rich in biodiversity, hosting a wide range of plant and animal life, and contribute to the overall health and balance of the ecosystem. Because of its typical and important ecosystem, it is also included in the list of important wetlands, which are of great value in providing ecological services such as water resource regulation and biological habitat regulation. The research on the Panjin Wetland encompasses a number of different areas of study, including the conservation of bird resources, the influence of economic development, the impacts of climate change upon the ecological environment, the dynamic changes that occur within the wetland, and the effect of human activities on the ecosystem [28,29,30,31]. These studies reveal the changes that have occurred in the Panjin Wetland’s ecological environment and highlight the need for its protection. However, despite extensive research by many scholars, this study is the first multi-scenario simulation of the future ecological pattern evolution and change trend of Panjin Wetland on this basis, revealing the health risk of the wetland ecological environment, and assessing the distribution and future trend of the habitat quality (HQ) of Panjin Wetland in 2032 using the InVEST model, which will provide comprehensive and multifaceted scientific support for the “Proposal for the Conservation and Ecological Restoration of Panjin Wetland”. In order to slow down the degradation trend and improve the ecological environment of the wetland, it is necessary to strictly implement the protection boundaries defined in the Panjin Territorial Spatial Master Plan (2021–2035), and provide ecological protection by implementing the “green mountains are golden mountains” policy in the process of urban development. Scenario simulation results provide data support for rationally controlling the urban scale, optimizing the ecological compensation mechanism, and implementing the Panjin Cultured Waters and Mudflats Plan (2018–2030) and ecological restoration policies such as returning farmland to forest and farmland to wetland. To elucidate the evolutionary patterns and driving factors of the Panjin Wetland ecosystem and to delve deeper into the characteristics of landscape changes, this study incorporates a comprehensive array of socio-economic factors and climate change elements in a PLUS prediction model. Using 2002 to 2022 as the historical simulation period, it forecasts the spatial-temporal changes in landscape for Panjin Wetland in 2032 under various scenarios. Concurrently, the HQ parameter was introduced to predict the prospective evolution of spatial and temporal patterns of HQ in Panjin Wetland through the coupled PLUS and InVEST models, and to analyze the interaction of landscape patterns.

2. Materials and Methods

2.1. Study Area

The Panjin Wetland is situated in the southernmost region of the Liaohe Delta, within the province of Liaoning, at the confluence of the Daliao River, the Shuangtaizi River, and the Daling River. Panjin Wetland is rich in wetland types, including reed swamp, beach and river, and other wetland types. The Liaohe Delta Reed Wetland is the largest typical warm temperate coastal estuarine wetland in Asia, formed by the alluvial deposits of the Liaohe River and Daling River, and has the largest reed wetland in Asia, including the “Red Beach”, which is considered one of the wonders of the world [32,33]. Its geographical position is between 121°30′–122°31′ east longitude and 40°45′–41°27′ north latitude. The administrative division includes Panshan and Dawa counties, with a total area of 3920.5 km², highest elevation 88 m, lowest elevation 0.3 m, no mountains, and a coastal wetland complex ecosystem. The highest annual temperature in the Panjin Wetland reaches 36.4 °C, while the lowest annual temperature stands at 30.8 °C. The annual evaporation ranges from 1392 to 1705 mm, and the average wind speed is between 2.4 to 4 m/s. Panjin Wetland is rich in biological resources. There are 124 species of fish, 267 species of birds (45 of which are under national first- and second-class protection), and 229 species of plants. For example, the coastal wetlands of Panjin Wetland are an important stopover, wintering, and breeding area for red-crowned cranes on the East Asia–Australasia Flyway. The waters at the mouth of the Liao River are an important breeding ground for Western Pacific spotted seals. Figure 1 displays the geographical position of the Panjin Wetland.

2.2. Data Processing

The data sources consist of 30 m resolution remote sensing imagery downloaded from the geospatial data cloud for the years 1992, 2002, 2012, and 2022. Through methods such as image preprocessing, conversion of polygon to raster format, and mask extraction within a GIS platform, the information extraction for the fourth phase of landscape in the Panjin Wetland was established. As per the “National Remote Sensing Land Use Cover Classification System” and references from relevant literature, the landscape types of the Panjin Wetland are categorized into seven types: farmed lake, reed swamp, residential land, paddy field, dry land, beach, and river.

Table 1. Imaging information from 1992 to 2022.

Year	Satellite	Sensors	Date	Orbital No.	Cloud	Resolution
1992	Landsat 5	TM	10.2	120/31, 120/32	17.67	30 m
2002	Landsat 5	TM	9.14	120/31, 120/32	0.25, 0.48	30 m
2012	Landsat 7	ETM+	8.16	120/31, 120/32	0.07, 0.05	30 m
2022	Landsat 8	OLI	2.25	120/31, 120/32	2.29, 0.57	30 m

is the imaging information.

The driving factors used in the analysis of driving mechanisms and the prediction of wetland evolution in this study are as follows (

Table 2. Description of driving factors and data sources.

Driving Factor	Spatial Resolution	Data Sources
Population (POP)	1000 m	https://www.resdc.cn (accessed on 26 August 2024)
GDP	1000 m
Normalized difference vegetation index (NDVI)	30 m
Night lighting (NPP)	30 m
Average annual temperature (TEM)	1000 m
Average annual precipitation (PRE)	1000 m
Average annual ground temperature (GST)	1000 m
Average annual relative humidity (RHU)	1000 m
Annual sunshine duration (SSD)	1000 m
Elevation (DEM)	30 m	http://www.dsac.cn (accessed on 26 August 2024)
Railroads	300 m	https://www.webmap.cn (accessed on 26 August 2024)
River	300 m

, Figure 2):

2.3. Method

In this study, Panjin Wetland was taken as the research object, and the landscape classification database of Panjin Wetland was established by image interpretation, based on which the spatio-temporal evolution of Panjin Wetland was analyzed by landscape transfer and landscape patterns index from 1992 to 2022, and the PLUS model was applied to simulate the landscape types under different scenarios of Panjin Wetland in 2032 using the historical data simulation from 2002–2022, then combined with the InVEST model to evaluate the habitat quality of Panjin Wetland under different scenarios. Figure 3 shows the flowchart of the study. By simulating multiple scenarios, the distribution characteristics of landscape types and the quality of wetland habitats can be predicted, thus providing scientific support for optimizing landscape patterns and realizing the synergistic development of economy and ecology in the study area.

2.3.1. Landscape Dynamic Change Model

The Landscape Dynamic Change Model is a model designed to quantify the dynamic alterations occurring in the landscape. The Landscape Dynamic Change Model can be measured in two dimensions, spatial and temporal, where the spatial dimension refers to the movement from one region to another and the temporal dimension refers to the changes at different points in time. The Landscape Dynamics Index can assist us in achieving a better comprehension of modifications in landscape patterns for better forecasting and management [34]. The formula is as follows:

$$K={\displaystyle \frac{w_b-w_a}{w_a}}\times {\displaystyle \frac{1}{T}}\times 100\%$$

where w is the area of the wetland type in years a and b; T represents the duration of the research period from year a to year b; and the value of K denotes the mean annual rate of variation of a specific wetland category from period a to b.

2.3.2. Landscape Transfer Matrix

The Landscape Transfer Matrix is a method that can quantitatively describe changes in landscape type, which can help us to better understand and predict trends in landscape type change [35].

$$S_{ij}=\left[\begin{array}{ccc}{S}_{11}& \cdots & S_{1n}\\ ⋮& \ddots & ⋮\\ S_{n1}& \cdots & S_{nn}\end{array}\right]$$

where S is the landscape area; n is the quantity of landscape patterns types prior to and subsequent to the transfer; and i and j are the landscape classifications at the conclusion of the initial research phase.

2.3.3. Landscape Patterns Index

The distribution of landscape elements of different shapes and sizes is revealed by landscape patterns, which reflect the results of the combined action of natural and anthropogenic factors [36]. In this investigation, patch density (PD), landscape shape index (LSI), landscape patch index (LPI), Shannon’s diversity index (SHDI), Shannon’s evenness index (SHEI), and aggregation index (AI) were used to investigate the degree to which the landscape patterns of Panjin Wetland have evolved.

2.3.4. InVEST Model

The InVEST Model is a system designed to assess the functionality of ecosystem services and their associated economic value, with the objective of providing support for the management and decision-making processes related to ecosystems [37,38]. The Habitat Quality Module generates data regarding land cover and biodiversity threat elements using an integrated assessment methodology that takes into account the comparative impact weights of threats, the relative susceptibility coefficients of habitat types, and the spacing between habitat grids and sources of threats [39]. The calculation formula is as follows.

$$i_{rxy}=1-\left({\displaystyle \frac{d_{xy}}{d_{r max}}}\right)$$

$$i_{rxy}=exp\left(-\left({\displaystyle \frac{d_{xy}}{d_{r max}}}\right)d_{xy}\right)$$

$$D_{xj}={\sum }_1^r{\sum }_1^y\left({\displaystyle \frac{w_r}{{\sum }_{r=1}^n{w}_r}}\right)\times r_y\times i_{rxy}\times {\beta }_x\times S_{jr}$$

$$Q_{xj}=H_{xj}\times \left[1-\left({\displaystyle \frac{D_{xj}^2}{D_{xj}^2+k^2}}\right)\right]$$

where $i_{rxy}$ is the effect of the threat factor r in the grid y on the grid x; $d_{xy}$ denotes the direct distance connecting the grid x and y; $d_{r max}$ denotes the peak action distance of the threat r; $D_{xj}$ is the habitat deterioration of the grid x in the landscape type j; $w_r$ is the weight of different threat factors; $r_y$ is the threat factor strength; ${\beta }_x$ is the level; $S_{jr}$ is the sensitivity of different habitats to different threat factors; $Q_{xj}$ is the HQ of grid x in landscape type j; $H_{xj}$ is the habitat fitness of x in landscape type j; and k is a half-full sum constant. The quality of a habitat is quantified using a value ranging from 0 to 1, with higher values signify superior HQ.

2.3.5. PLUS Model

The PLUS model includes two modules, LEAS (land expansion analysis strategy) and CARS (CA based on multiple random seeds), and provides two Markov chain and linear regression methods for the simulation of changes in landscape type among distinct contexts [40].

The PLUS model calculates the potential expansion probability of each wetland type based on the LEAS module and applies CARS to imitate future wetland landscape changes among distinct contexts. The mathematical expression for this process is as follows.

$$P_{i,k}^d(x)={\displaystyle \frac{{\displaystyle {\sum }_{n=1}^{M}I=\left[h_n(x)=d\right]}}{M}}$$

where d value is 1 or 0; if d is 1, it implies that other landscape categories are converted into k landscape types; if d is 0, it implies that landscape categories are converted into other types that do not involve k landscape categories; x is the vector constituted by driving factors; I denotes the indicator function associated with the decision tree; h_n (x) represents the forecasted landscape category when the decision tree is n; P denotes the probability of expansion of landscape types of the spatial unit of class k at i; and M denotes the overall number of decision trees.

In light of the landscape analysis conducted in Panjin Wetland spanning from 1992 to 2022, the temporal and spatial variation patterns of landscape transformation were synthesized, the impelling factors underlying landscape alteration were distilled based on the data furnished by ecological monitoring, and the driving factors have the characteristics of accuracy, quantifiability, and uniformity, and then a landscape simulation model of Panjin Wetland is constructed, which will provide a new way of landscape simulation and image comparison in the future. Twelve driving factors were selected, including: elevation (DEM), average annual precipitation, air temperature, average annual soil temperature (GST), average annual relative humidity (RHU), annual sunshine hours (SSD), distance to water system, distance to railway, night light, NDVI, population density (POP), and GDP density. LEAS random forest parameter realization settings: sampling method: set randomly selected sampling points, use the random forest algorithm to achieve training on the data of every kind of landscape, and obtain the transformation rules of the expansion law of different types of landscape; sampling rate: set to 1%; number of regression trees: set to 20; number of parallel threads: set to 2, to enhance the running velocity. On the basis of the CARS model, the following simulation parameters were implemented: domain range: set to 3; count of parallel threads: set to 4; attenuation coefficient of decreasing threshold: set to 0.9; probability of random patch seed: set to 0.1; seed ratio: set to 0.0001.

3. Results

3.1. Spatial and Temporal Changes in the Panjin Wetland

Regarding the area alteration, during the period from 1992 to 2002, the wetland area declined sharply on account of economic expansion and rapid population increment. The area of natural wetland shows a decreasing and then increasing trend from 2002 to 2022, while the area of artificial wetland shows an increasing and then decreasing trend.

As shown in Figure 4, in terms of landscape type transfer changes, from 1992 to 2022, the most prominent change has been the transition of residential land, particularly marked by the conversion of paddy field to residential land, which expanded by 332.67 km² from 1992 to 2002, while the conversion of dry land to residential land accounted for 31.5%. From 2002 to 2012, 32% of farmed lake was transformed into paddy field. Additionally, from 2012 to 2022, 95.23 km² of dry land was converted to residential land, representing 23.0%, while 42.49 km² of farmed lake transitioned to beach, accounting for 28.6%. The results of the study showed that between 1992 and 2002, although the policy of prioritizing economic development promoted the encroachment of agricultural land and the expansion of aquaculture farms, the area of retreat increased after 2002 due to the introduction of local environmental protection policies such as “retreat and return to wetland” and “enclosure of the sea to build up fields”, as well as the increased awareness of ecological protection among the local populace.

3.2. Analysis of the Changing Landscape Patterns in the Panjin Wetland

From

Table 3. Landscape patterns index value.

Year	Patch Density (PD)	Landscape Shape Index (LSI)	Landscape Pattern Index (LPI)	Shannon’s Diversity Index (SHDI)	Shannon’s Evenness Index (SHEI)	Aggregation Index (AI)
1992	34.42	155.49	28.15	1.50	0.77	85.28
2002	41.72	189.08	10.16	1.70	0.87	82.07
2012	43.19	189.86	6.52	1.71	0.88	81.99
2022	39.28	211.71	14.13	1.77	0.91	80.00

, it can be observed that the PD index has shown an upward trend from 1992 to 2012, reflecting a fragmentation trend, but between 2012 and 2022, the degree of fragmentation has slightly improved. From 1992 to 2022, both the LSI and SHEI indices have gradually increased, indicating that the various landscape types of the Panjin Wetland are developing in an increasingly irregular and uneven manner. In terms of the LPI index, both paddy field and reed swamp remain the dominant landscape features of the Panjin Wetland, with the LPI index for built-up areas consistently rising from 1992 to 2012. Consequently, based on the changes in LPI values, the dominance of larger landscape patches has gradually increased between 2012 and 2022, with a decrease in the influence of human interference, resulting in a more stabilized landscape pattern. The SHDI index has continuously increased from 2012 to 2022, while the AI index has steadily decreased from 85.28 in 1992 to 80.00 in 2022, indicating that the quantity of small patches within the Panjin Wetland landscape augmented while their degree of aggregation worsened.

3.3. Driving Force Analysis and Multi-Scenario Modelling of Panjin Wetland Development

In accordance with the simulation parameter configurations of the CARS module in the PLUS model, the 2022 landscape data was simulated based on the 2012 landscape data of Panjin Wetland, and 5% of the raster cells were extracted from the simulated 2022 landscape map to calculate the Kappa coefficient and conduct the model accuracy test. Accuracy test results show that the Kappa coefficient is 0.764 and total accuracy is 0.875, more than 0.75. Test results show that PLUS model simulation data is more satisfying, and the selected driving factors and set parameters can be used as driving factors for each landscape type, which indicates that the simulation results of the relevant parameters of the CARS module can meet the accuracy requirements of landscape simulation, and can be used to represent the landscape simulation of Panjin Wetland. It shows that the simulation results of the relevant parameters of the CARS module can meet the accuracy requirements of landscape simulation, and can be used to reflect the development of landscape changes in Panjin Wetland. Figure 5 shows the probability map of landscape type suitability. Based on previous experience, this paper will be divided into four scenarios for simulation and comparison [41,42].

3.3.1. Natural Development Scenarios

Based on the values of various parameters, we simulated and predicted the landscape patterns of the Panjin Wetland in 2032 under natural development scenarios. Without considering the influence of other factors, we assume that the landscape patterns from 2022 to 2032 will evolve without significant constraints, in alignment with the changes observed in the landscape patterns from 2012 to 2022. The settings will reflect the results of natural evolutionary trends based on existing policies, using the landscape classification distribution map from the baseline year of 2022. The Markov chain model from the Demand Prediction module was employed for forecasting, yielding the landscape patterns of the Panjin Wetland in 2032 within a framework of minimal disturbance under natural development conditions.

As illustrated in Figure 6 and

Table 4. Landscape type area from 2022 to 2032 (natural development scenarios) (km²).

Year	Residential Land	Dry Land	River	Paddy Field	Beach	Farmed Lake	Reed Swamp
2022	819.19	349.24	215.31	1307.48	269.82	435.95	523.54
2032	856.48	364.71	207.09	1232.91	289.78	499.90	469.65

, under the natural development scenarios, areas of residential land, dry land, beach, and farmed lake all witnessed a notable expansion. The increase in residential land necessitates the conversion of some regions within paddy field and river. Consequently, the paddy field in 2032 has seen a reduction, with small patches of paddy field within residential zones having been redeveloped. Similarly, the properties of reed swamp, beach, and farmed lake have undergone certain changes, leading to a more centralized and regular distribution of their patches.

3.3.2. Farmland Protection Scenarios

In the area of farmland protection, the focus is primarily on the implementation of enhanced protection measures for agricultural resources, with strict controls on the expansion of residential land that encroaches on cultivated fields. In accordance with the “General Plan for Land and Space in Panjin City (2021–2035)”, we aimed to manage the development of arable land resources on the basis of natural development. An overlapping analysis of paddy and dry land for the years 1992, 2002, 2012, and 2022 was performed, taking the areas classified as cropland (paddy or dry land) in all four years as a constraint condition, which means long-term stable cropland (Figure 7), thus limiting the conversion of paddy field and dry land. The probability of conversion of farmland to residential land and farmed lake was reduced by 70% and 50%, respectively, yielding an overview of the distribution of wetland landscape patterns in the Panjin Wetland under the farmland protection scenarios in 2032.

In the forecasting process, the establishment of long-term stable arable land conversion restrictions has relatively safeguarded agricultural resources. As illustrated in

Table 5. Landscape type area from 2022 to 2032 (farmland protection scenarios) (km²).

Year	Residential Land	Dry Land	River	Paddy Field	Beach	Farmed Lake	Reed Swamp
2022	819.19	349.24	215.31	1307.48	269.82	435.95	523.54
2032	726.53	360.13	198.75	1368.07	299.85	479.62	487.58

, the areas of dry land and paddy field have increased by 10.89 km² and 60.59 km², respectively, while the quantity of arable land converted to other landscapes has diminished to a certain extent, particularly in terms of the area transitioning to residential land, which has shown a notable downward trend. Compared to data from 2022, there is a slight increase in the amount of arable land, indicating that agricultural resources have been protected to some degree. The area of farmed lake has risen by 43.67 km² compared to 2022; however, in contrast to the area of farmed lake under natural development scenarios, it has declined by 20.28 km². The areas of residential land, river, and reed swamp have seen reductions of 92.66 km², 16.56 km², and 35.96 km² respectively compared to 2022, while beach areas have increased by 30.03 km². Relative to the results of natural development scenarios, only residential land and paddy field have exhibited substantial changes, with a decrease in residential land area of 129.95 km² and an increase in paddy field area of 135.16 km². These results indicate that the expansion of residential land was restrained under the farmland protection scenario.

3.3.3. Ecological Protection Scenarios

In the context of ecological conservation, the primary objective is to achieve ecological restoration by strengthen implementing environmental policies, such as reforesting and restoring wetland. The constraints include the designation of the National Nature Reserve as an ecological protection restricted area (Figure 8) to restrict the conversion of ecological reserve land into other land categories and enhance the safeguarding of ecological land assets. A stringent control will be established to prevent the conversion of beach, reed swamp, and river into residential land, paddy field, and dry land types of non-natural wetland. Reducing the probability of converting mudflats, reed swamp, and rivers into residential areas by 70%, and into paddy field and dry land by 40%, resulted in the distribution pattern of the landscape in Panjin Wetland in 2032 under the ecological protection scenarios.

By setting up ecological protection areas, ecological land resources have been relatively preserved, resulting in a decrease in the amount of land transferred, and ensuring that the area of ecological land remains intact and protection efforts are further deepened. In comparison with 2022, the residential land area has expanded by 15.93 km², although it has decreased by 21.35 km² compared to the natural development scenarios. Since 2022, the areas of paddy field and farmed lake have decreased by 115.78 km² and 0.28 km², respectively, while the areas of dry land, river, beach and reed swamp have increased by 20.61 km², 37.04 km², 41.50 km², and 0.97 km², respectively. Nevertheless, compared with the 2032 outcomes of the natural development scenarios, the alteration in paddy field area has diminished by 41.21 km², and the river area has augmented by 45.26 km². The expanse of dry land has seen minimal change, with a slight increase of 5.15 km², and the extent of farmed lake has declined by 64.24 km², while beach and reed swamp have increased by 21.54 km² and 54.87 km², respectively. This evidently demonstrates that within the ecological protection scenarios, the growth rate of residential land has slowed down, whereas the expansion rate of natural wetland has risen remarkably.

3.3.4. Economic Development Scenarios

In economic development, production and living needs drive growth in demand for residential and arable land. Constraints prohibit the conversion of residential land and farmed lake. The likelihood of transforming farmed lake, beach, and reed swamp into residential land, paddy field, and dry land is increased by 5%, 5%, and 10%, respectively, while the transformation of residential land, paddy field, and dry land into other landscapes (excluding river) is reduced by 40%. This aims to ascertain the distribution of the landscape patterns in Panjin Wetland in 2032 under the economic development scenarios.

Dry land and paddy field have shrunk by 8.70 km² and 32.30 km², respectively, compared to 2022, while residential area increased by 117.65 km². When juxtaposed with the natural development scenarios, the expansion of residential zones has encroached upon both dry land and paddy field, while the burgeoning farmed lake have intruded into the reed swamp. In comparison to the year 2022, the structure of landscape types reveals an increase of 9.54 km² in beach; however, when contrasted with the natural development scenarios, a reduction of 10.43 km² is observed in beach areas. Additionally, river and farmland have decreased by 1.304 km² and 3.595 km², respectively, while reed swamp area has diminished by 95.52 km² since 2022. Relative to the results projected for 2032 in the natural development scenarios, the change in dry land has diminished by 24.17 km², whereas paddy field areas have increased by 42.27 km². River areas have experienced no significant fluctuation, merely seeing a slight increase of 8.05 km². Conversely, the area of farmed lake has diminished by 64.24 km² and reed swamps have declined by 42.63 km². These findings underscore that under the context of economic development, the acceleration of residential land expansion will inevitably impact the conservation of natural wetland.

3.4. Habitat Quality

Grounded on the landscape patterns of 2032 under the four scenarios projected combined with the InVEST model to derive the HQ in Panjin Wetland, as demonstrated in Figure 9, compared to the average HQ score of 0.441 in 2022, the average HQ scores for the natural development scenarios, the farmland protection scenarios, the ecological protection scenarios, and the economic development scenarios were 0.427, 0.448, 0.438, and 0.416, respectively. The HQ of the index was partitioned into five levels of low, lower, medium, higher, and high values, with the intervals of [0–0.2), [0.2–0.4), [0.4–0.6), [0.6–0.8), [0.8–1], and area statistics were performed as demonstrated in

Table 6. Habitat quality area and percentage under multiple scenarios in Panjin Wetland in 2032.

Habitat Quality	Natural Development Scenarios		Farmland Protection Scenarios		Ecological Protection Scenarios		Economic Development Scenarios
	Area/km2	Percentage/%	Area/km2	Percentage/%	Area/km2	Percentage/%	Area/km2	Percentage/%
low	850.74	22	716.71	18	828.21	21	934.82	24
lower	863.89	22	837.66	21	803.75	21	782.57	20
medium	1561.13	40	1706.19	44	1566.21	40	1603.44	41
higher	83.43	2	79.50	2	98.12	3	81.88	2
high	561.16	14	580.29	15	624.07	16	517.65	13

. Under the natural development scenarios, without the implementation of strong protection measures, allowing production, life, and ecology to develop naturally, the overall trend of HQ in the study area is declining, in which compared to 2022, the low value area and lower value area of HQ have increased, respectively, by 45.23 km² and 80.60 km², and the area of medium value area and high value area have decreased, respectively, by 66.76 km² and 53.48 km²; under the farmland protection scenarios, the overall HQ remains stable, with the extent of medium and high HQ areas increasing by 145.06 km² and 19.13 km², respectively, and the area of low HQ areas decreasing by 134.03 km²; under the ecological protection scenarios, the overall HQ decreases slightly, but the degree of environmental degradation is relatively small compared to the natural development scenarios, and the distribution area of medium, higher, and high value zones increases, with the area increasing by 5.08 km², 14.69 km², and 62.90 km², respectively, while the area of low and lower value zones decreases by 22.53 km² and 60.14 km², respectively. In the economic development scenarios, areas with low HQ increased significantly, by 84.07 km², and were concentrated in the center of the city with rapid economic development. Overall, under different scenarios, the HQ of the farmland protection scenarios improved, the HQ of the ecological protection scenarios remained stable, and the HQ of the other scenarios deteriorated, mainly because the farmland protection scenarios restricts the conveyance of paddy and dry land to residential land in Panjin Wetland, and the ecological protection scenarios restricts the conveyance of wetland to non-wetland, which leads to the conclusion that the HQ is more seriously affected by the expansion of urban area and population growth. Therefore, future development should appropriately limit the enlargement of residential land and strengthen the enforcement of ecological protection measures to promote sustainable development.

4. Discussion

The study shows that the Panjin Wetlands are still in the stage of metropolitan expansion, encroaching on agricultural and ecological land. In the farmland protection scenarios, the low value of HQ is greatly reduced, which changes the median value and leads to an increase in the average value of HQ, mainly due to the peculiar situation of the Panjin Wetlands, with too large a proportion of dry land and paddy field. Under the ecological protection scenarios, the areas of low and lower values of HQ are reduced, and the areas of medium, higher, and high values are increased, which is optimized to a certain extent, meets the requirements of sustainable development, and is more in line with the ideals and goals of healthy and sustainable development, and is therefore the basis for optimizing the spatial layout of Panjin Wetland and delineating the pattern of protected areas. Landsat imagery with a resolution of 30 m was used as the data source for this study. The lack of multi-temporal land feature and detailed characteristics results in the indistinctness of certain landforms, thereby causing deviations in the evolution characteristics and patterns of wetland. Support Vector Machines, Neural Networks, and Hierarchical Classification Discriminant Methods were employed to extract and classify wetland information from the same data source, followed by a longitudinal comparison of classification accuracy, ultimately constructing an optimal classification scheme for wetland information extraction and classification. Although this research is grounded in social, economic, and environmental factors, such as landscape fragmentation and patch characteristics, the data acquisition constraints have led to the reliance solely on publicly available data from the official website of Panjin City and other platforms, thus limiting the scope of the evaluation results. This study applied the PLUS model to forecast the alteration of the landscape patterns in Panjin Wetland, with simulations considering only four distinct scenarios. The consideration of various development scenarios and predictive conditions is not sufficiently comprehensive. Urban development is a complex outcome of the interplay of multiple factors; therefore, further refinement and expansion of constraints and limitations, in conjunction with policy factors, is necessary.

5. Conclusions

This research, using remote sensing technology and geographic information systems for data processing, has created a classification database of the wetland landscape. Building upon this foundation, the study conducted an analysis of the distribution characteristics and evolution of landscape patterns, integrating data on driving factors. Various constraints were set under different scenarios, predicting and analyzing the future distribution of landscape patterns and HQ in Panjin Wetland using the PLUS model and the InVEST model. The following conclusions were obtained:

(1) From the perspective of the changes in landscape area types, between 1992 and 2022, the residential land area has expanded by 594.45 km², and the paddy field area has reduced by 508.56 km². The area of farmed lake shows an upward trend, increasing by 308.90 km², whereas the area of reed swamp has consistently declined, totaling a decrease of 472.89 km². Dry land has exhibited a trend of rise followed by a fall, while beach areas have shown fluctuations of rise, fall, and then rise again.

(2) Compared to the landscape area in 2022, in terms of urban expansion, only the farmland protection scenarios experienced a decrease of 92.66 km², while the ecological protection scenarios, natural development scenarios, and economic development scenarios experienced an increase of 15.93 km², 37.29 km², and 117.65 km², respectively. Compared to the natural development scenarios, the economic protection scenarios alone witnessed a decrease of 44.01 km² in natural wetland. In contrast, the farmland protection scenarios and the ecological protection scenarios saw increases of 19.65 km² and 121.65 km^2, respectively, primarily due to their suppression of urban expansion and the consequent enlargement of riverine areas, mudflats, and reed swamp.

(3) In 2022, the average HQ score is 0.441, and the average HQ scores in the natural development scenarios, farmland protection scenarios, ecological protection scenarios, and economic development scenarios are 0.427, 0.448, 0.438, and 0.416, respectively.