Optimizing Estuarine Aquatic–Terrestrial Ecotone Landscapes Under Economic–Ecological Trade-Offs: Evidence from the Pearl River Delta

Abstract

Under the dual pressures of rapid urbanization and climate change, urban expansion in high-density estuarine urban agglomerations has intensified economic–ecological trade-offs in the aquatic–terrestrial ecotone, necessitating land-use planning that reconciles economic growth with ecological protection. Here, we integrated linear programming with the CLUE-S model and incorporated marine–terrestrial integration objectives and typical natural disturbance factors. With this approach, a landscape pattern simulation framework capable of jointly optimizing ecological and economic benefits was developed. The framework was applied to the estuarine aquatic–terrestrial ecotone of the Pearl River Delta. This study drew on a land-use dataset, landscape dynamics, socioeconomic and biophysical drivers, and regional planning constraints to conduct simulation experiments under alternative development scenarios. The model achieved a Kappa coefficient of 0.904. From 2010 to 2020, built-up land expanded rapidly and encroached on ecological space. Simulations indicated that the natural evolution scenario increased fragmentation and ecological conflicts despite economic gains, whereas the sustainable development scenario balanced expansion and protection, reduced forestland fragmentation, safeguarded key ecological spaces, and improved ecological benefits while maintaining economic growth. Ecological benefits in the coastal aquatic–terrestrial ecotone from −0.2 to 0 km increased by 283.3%. The framework embeds land-use dynamics and spatial constraints, providing decision support for territorial spatial planning and ecological security pattern optimization.

1. Introduction

Ecotones are transitional zones where the environmental boundaries of species and communities overlap, serving as interfaces between distinct landscape or vegetation types [1,2]. They are highly sensitive to climate change and human disturbance [3,4]. Their stability and resilience depend strongly on external disturbances [5]. Accordingly, tracking ecotone change is vital for understanding global change and for anticipating human–environment trajectories [6]. The aquatic–terrestrial ecotone represents one of the three major ecotone types worldwide. It features intensive exchanges of matter and energy and provides functions such as intercepting surface runoff, removing pollutants, providing habitats, and enhancing landscape heterogeneity [7]. Furthermore, these areas concentrate population, economic activity, and energy consumption [8]. Approximately 60% of global economic output is concentrated in estuaries [9]. Favorable topography and transport accessibility attract population, industry, and cities to estuarine areas. These areas often merge into large urban agglomerations. Under the combined pressures of global climate change and intensified human activities, the continued expansion of built-up land and the rapid growth of pollutant emissions have encroached upon and fragmented ecological land in the aquatic–terrestrial ecotone. This process has exacerbated landscape fragmentation and degraded ecosystem services, resulting in increasing ecological and environmental stress [10,11,12]. Accordingly, the estuarine aquatic–terrestrial ecotone has become a focal issue for mega-urban agglomerations worldwide. Given the tension between ecological conservation and socioeconomic development during urbanization, it is crucial to elucidate the mechanisms of landscape pattern evolution and to develop optimized spatial allocations for estuarine aquatic–terrestrial ecotones in high-density urban agglomerations under alternative future scenarios. Such efforts can support coordinated regional economic–ecological development.

Landscape pattern change is a central topic in land-use change research [13]. Increasingly, studies combine remote sensing and GIS with field surveys and sampling analyses to examine the causes, processes, and consequences of such change [14,15,16,17]. Previous work has concentrated on agro-pastoral and forest-grassland ecotones [18,19]. Although interest in aquatic–terrestrial ecotones is growing, existing studies have largely concentrated on wetlands, riparian zones, and floodplains [2,20,21]. The exploration of the estuarine aquatic–terrestrial ecotone has not yet been reported. Ecotones exhibit diverse landscape types, pronounced heterogeneity, and complex drivers. Therefore, landscape pattern simulation is needed to clarify landscape change and the integrated mechanisms shaped by natural environmental and socioeconomic factors [22]. Among commonly used models, the Markov model, cellular automata (CA), and their variants (CLUE-S, FLUS, and System Dynamics models) each offer distinct strengths [23,24,25]. Notably, CLUE-S integrates empirical land-use driver relationships, a dynamic competition module, and multi-scenario spatial prediction. It provides strong capability for visualizing land-use transitions and for assessing planning schemes. It can simulate and predict both the quantities and spatial allocation of land use, and global applications attest to its applicability [26,27]. Nevertheless, a salient limitation of CLUE-S applications is their dependence on exogenous demand inputs. This dependence constrains their capacity to support fully integrated planning [26]. Integrating aspatial optimization models, such as linear programming with CLUE-S spatial simulation, mitigates this limitation. Linear programming determines optimal land-use quantities based on policy objectives, whereas CLUE-S translates these demands into spatially explicit patterns through cell-level allocation. In recent years, integrated linear programming and CLUE-S frameworks have been mainly applied to terrestrial urban expansion or watershed-scale land-use optimization and spatial allocation [28,29,30]. However, existing applications have rarely accounted for marine–terrestrial interaction processes and associated governance constraints in the estuarine aquatic–terrestrial ecotone. Therefore, this study developed an integrated linear programming-CLUE-S framework for the estuarine aquatic–terrestrial ecotone, in which marine–terrestrial coordinated governance targets are operationalized as solvable constraints and typical natural disturbance factors are incorporated. This framework supports dynamic simulation of landscape pattern evolution and scenario assessment in a highly heterogeneous ecotone.

Existing studies on urban agglomerations have made substantial progress in land-use change, ecosystem service assessment, and identification of ecological security patterns [31,32,33,34]. However, ecological and economic objectives are often characterized by trade-offs, making it difficult to define synergistic targets and translate them into implementable spatial allocation schemes. Therefore, economic–ecological coordination research is needed to generate actionable trade-offs within a unified framework and to deliver land-use optimization pathways under alternative scenarios and constraints, thereby providing operational decision support [35,36]. As one of the world’s most densely populated, economically developed, and rapidly growing regions, the Pearl River Delta urban agglomeration faces persistently high demand for construction land in the context of industrial upgrading, population agglomeration, and major infrastructure development in the Guangdong–Hong Kong–Macao Greater Bay Area, which poses severe threats to regional ecological security [37,38]. Compared with other urban agglomerations or estuarine regions, the Pearl River Delta features a high-density river network, where three rivers converge, and the system drains to the sea through eight estuaries, forming an extensive estuarine aquatic–terrestrial ecotone interface. Tidal forcing and flood risk further enhance ecosystem sensitivity and spatial heterogeneity. As a global manufacturing center and an export-oriented economic hub, the Pearl River Estuary has experienced seawater quality persistently below Class IV standards and continued coastal mangrove shrinkage due to intensive urban development and high population density, leading to marked declines in regional ecological resilience and environmental carrying capacity.

As a key zone for high-quality, cross-habitat development and governance along the “river–lake–‘dike–pond’–ocean” continuum, the estuarine aquatic–terrestrial ecotone of the Pearl River Delta requires empirical research on balancing economic development and ecological protection. Accordingly, this study proposed two questions. What are the characteristics and causes of landscape pattern evolution in the estuarine aquatic–terrestrial ecotone? How can a land-use optimization framework solve the economic–ecological trade-offs between environmental protection and economic development in densely populated urban agglomerations of estuarine regions? To address these questions, we identify drivers of landscape pattern evolution and priority ecological protection areas in the estuarine aquatic–terrestrial ecotone of the Pearl River Delta. We develop a dynamic landscape simulation framework that integrates CLUE-S with linear programming. Guided by ecological protection and economic development objectives, we design landscape simulation scenarios to compare the performance of landscape configurations, thereby building an evidence base for optimizing landscape patterns and supporting policy for the estuarine aquatic–terrestrial ecotone. Research in this region also provides valuable insights into regional coordinated development and cross-regional ecological governance in China. The proposed framework is transferable to high-density urban agglomerations facing ecological risks, offering a feasible pathway to balance ecological protection and economic development.

2. Materials and Methods

2.1. Overview of the Study Area

The estuarine aquatic–terrestrial ecotone of the Pearl River Delta lies along the central–southern coast of Guangdong Province, China (113°8′–114°36′ E, 21°50′–23°9′ N) and covers 18,600.52 km². It forms an important part of the Guangdong–Hong Kong–Macao Greater Bay Area. Hills and mountains bound the east, river channels meander to the west, and the south connects to the South China Sea through eight major estuarine outlets. The study extent is defined by landward and seaward boundaries. The landward boundary follows the administrative borders of Guangzhou (Haizhu, Panyu, and Nansha Districts), Zhongshan, Zhuhai, Dongguan, and Shenzhen in Guangdong Province, and the Hong Kong Special Administrative Region and Macao Special Administrative Region. Due to the numerous islands and sea-level rise, which have altered the land area over time, a fixed seaward boundary is delineated for dynamic landscape simulation. Specifically, the minimum bounding rectangle that encompasses all islands is used (Figure 1).

The estuarine aquatic–terrestrial ecotone of the Pearl River Delta hosts extensive mangroves, coastal wetlands, estuarine shoals, and a dense water network. It provides high-quality habitat for aquatic and hygrophilous herbaceous and woody plants and for mangrove species, as well as for fish, birds, and amphibians. It also underpins flood regulation and storage, water purification, and biodiversity maintenance. Since the Bay Area development plan was proposed in 2010, the Pearl River Delta has gradually shifted from extensive growth toward coordinated regional governance. In 2020, the Guangdong–Hong Kong–Macao Greater Bay Area development plan was fully rolled out, further promoting a strategic shift toward marine–terrestrial integration and improving land-use efficiency. The Guangdong–Hong Kong–Macao Greater Bay Area is a world-class urban agglomeration. Rapid urbanization, industrialization, and infrastructure expansion demand large areas of land. Development pressures drive encroachment on and fragmentation of the aquatic–terrestrial ecotone. They intensify landscape fragmentation, degrade habitats, reduce biodiversity, and worsen water quality. Consequently, the estuarine aquatic–terrestrial ecotone of the Pearl River Delta combines high urbanization, critical ecological functions, and complex human–environment tensions. Landscape pattern optimization in this region is both typical and of practical value. It informs how developed regions reconcile protection with development and supports a resilient territorial spatial pattern.

2.2. Data

2.2.1. Land-Use Classification

Given the coastal setting of the Pearl River Delta estuarine aquatic–terrestrial ecotone, we followed the Convention on Wetlands of International Importance Especially as Waterfowl Habitat, which defines wetlands as areas of marsh and peatland, whether natural or artificial, permanent or temporary, with static or flowing water that is fresh, brackish, or saline, including marine waters no deeper than 6 m at low tide. Accordingly, this study defined the estuarine aquatic–terrestrial ecotone as a functional interface between aquatic and terrestrial ecosystems, characterized by high productivity and biodiversity. Inland boundaries are diffuse, whereas the seaward boundary is delineated by the nearshore 6 m isobath, corresponding to waters no deeper than 6 m at low tide [39].

Using the CNLUCC dataset in 2010 and 2020 (https://www.resdc.cn/, accessed on 26 November 2024), land use in the estuarine aquatic–terrestrial ecotone of the Pearl River Delta was reclassified into seven categories: grassland, forestland, rivers and lakes, reservoirs and ponds, cropland, built-up land, and ocean. Because the CNLUCC dataset does not distinguish between subtypes of nearshore shallow-sea wetlands, to ensure consistency between the land-use classification system and the functional boundary of the study area, we extracted the 6 m isobath from a 1:100,000 coastal nautical chart and performed vectorization and coordinate spatial reference in GIS. Subsequently, we reclassified the nearshore shallow-sea area located shoreward of the isobath as the “nearshore 6 m isobath zone”.

2.2.2. Driving Factors

Sixteen driving factors were selected based on field surveys, a literature review, and consultations with experts in urban and rural planning, land resources management, ecology, and geography. The factors reflected the geomorphic and hydrological sensitivity of the estuarine aquatic–terrestrial ecotone and the development intensity of the urban agglomeration. They were organized into two dimensions: natural environment and socioeconomic factors. The selection rationale and data sources are provided in Table S1. The natural environment factors included elevation, slope, long-term average temperature, long-term average precipitation, flooding impact, typhoon impact, soil organic carbon content, soil texture, the Normalized Difference Vegetation Index (NDVI), and distance from rivers. The socioeconomic factors included distance from main roads, population density, distance from built-up land, distance from protected farmland, gross domestic product (GDP), and points of interest (POIs). For 1 km socioeconomic datasets such as GDP and population density, we performed nearest-neighbor resampling with value replication rather than interpolation-based downscaling to avoid introducing spurious high-resolution variability.

2.3. Methods

Figure 2 presents the methodological framework. This study focused on the estuarine aquatic–terrestrial ecotone of the Pearl River Delta and identified key drivers of landscape pattern evolution in 2010 and 2020. An assessment of ecological conservation importance was built from ecosystem service function importance and ecological sensitivity, and priority ecological protection areas were delineated as spatial constraints. Multi-scale binary logistic regression was used to fit relationships between driving factors and land-use types, and ROC curves were used to select the optimal simulation scale. Using 2010 as the baseline year, the CLUE-S model was used to pre-simulate land use in 2020. Model robustness and accuracy were evaluated using the Kappa coefficient. The year 2030 corresponds to the conclusion of the 15th Five-Year Plan period and serves as a critical planning horizon for assessing future land-use demands and evaluating alternative spatial allocation pathways. Two 2030 scenarios were then designed. The natural evolution scenario in 2030 estimated land-use demand using a Markov chain model based on transition rules from 2010 to 2020. The sustainable development scenario in 2030 integrated a gray multi-objective optimization (GMOP) model with priority ecological protection areas to determine demand by considering economic and ecological benefits. Finally, landscape pattern indices, economic and ecological benefits, and the encroachment intensity on priority ecological protection areas were compared across scenarios. Based on these comparisons, this study proposed landscape pattern optimization and zoning control strategies.

2.3.1. Ecological Conservation Importance Assessment

Considering the study area’s ecological context, the importance of ecosystem service functions was assessed across four aspects: water conservation, soil conservation, coastal protection, and biodiversity maintenance. Ecological sensitivity was assessed using two factors: soil erosion and coastal erosion (

Table 1. Indicators of ecological conservation importance assessment.

Type	Indicator	Calculation Method		References
Ecosystem service function importance	Water conservation	$WC=F_{pre}-R-ET$	(1)	[40]
		WC denotes water conservation; Fpre is the multi-year mean precipitation; R is surface runoff; ET is evapotranspiration.
	Soil conservation	$SC=R\times K\times S\times C\times L$	(2)	[41]
		SC denotes soil conservation; R is rainfall erosivity; S is slope; K is soil erodibility; C is vegetation cover; L is slope length.
	Coastal protection	We followed China’s Guidelines for Resource Environment Carrying Capacity and Territorial Spatial Development Suitability Evaluation. Extremely important coastal protection zones are coasts with high authenticity and integrity that require priority protection. High-function zones include mangroves, shelterbelts, tidal flats, and other natural shoreline types. Medium-function zones include dike–pond coasts and bedrock coasts. Low-function zones include artificial hard coasts and sandy coasts.		[42]
	Biodiversity maintenance	$Q={NPP}_{mean}\times F_{pre}\times F_{tem}\times (1-Falt)$	(3)	[43]
		Q denotes biodiversity maintenance; NPPmean is the multi-year mean net primary productivity; Fpre is the multi-year mean precipitation; Ftem is the multi-year mean air temperature; Falt is altitude.
Ecological sensitivity	Soil erosion	$SE=\sqrt[4]{R\times K\times LS\times C}$	(4)	[42]
		SE denotes soil erosion sensitivity; R is rainfall erosivity; K is soil erodibility; LS is topographic relief; C is vegetation cover.
	Coastal erosion	Coastal erosion sensitivity is classified by coastal geomorphology: low for artificial hard revetments, bedrock coasts, and dike–pond coasts; medium for shelterbelts, mariculture areas, mangroves, and other natural shoreline types; and high for tidal flats, major estuarine zones, and sandy coasts [44].		[42]

and Figure S1). The two assessments were overlaid to derive ecological conservation importance, which was classified into five levels using the natural breaks method. The high and sub-high classes were designated as priority ecological protection areas.

2.3.2. CLUE-S Model

The CLUE-S model allocates landscape patterns by iteratively updating land-use change variables until the allocated area matches land-use demand, thereby simulating the next period’s landscape pattern [27].

$${TPROP}_{i,u}=P_{i,u}+{ELAS}_u+{ITER}_u$$

(5)

TPROP_i,u denotes the total probability that cell i is assigned to land-use type u. P_i,u is the suitability probability of cell i for type u estimated by binary logistic regression. ELAS_u is the conversion elasticity parameter set according to land-use evolution characteristics and represents conversion cost. ITER_u is the competition factor for type u, which is automatically updated during iterative simulation.

1.

Weighting of driving factors

This study applies the analytic hierarchy process (AHP) with the 1–9 scale to score pairwise comparisons among the 16 driving factors. Values 1, 3, 5, 7, and 9 indicate equal, moderate, strong, very strong, and extreme importance, respectively. Reciprocals indicate the opposite comparison. The weights of all factors are then calculated (Table S2).

2.

Spatiotemporal scale settings

The baseline year is set to 2010, with a 10-year time step. Five cell sizes are tested: 40 × 40 m, 45 × 45 m, 50 × 50 m, 55 × 55 m, and 60 × 60 m. The optimal simulation scale is determined using the maximum ROC value [45].

3.

Land-use conversion rules

The expansion weight reflects the expansion capacity of each land-use type and ranges from 0 to 1. Values closer to 1 indicate stronger expansion capacity. These weights are determined from land-use evolution in 2010 and 2020 and from landscape pattern indices (Table S3).

2.3.3. Model Accuracy Validation

The Kappa coefficient measures overall simulation accuracy. It quantifies agreement between land-use simulations and observations. This study uses the Kappa coefficient to evaluate the accuracy of the dynamic landscape simulation [24,46].

$$Kappa={\displaystyle \frac{P_O-P_C}{P_P-P_C}}$$

(6)

$$P_O={\displaystyle \frac{s}{n}}$$

(7)

$$P_c={\displaystyle \frac{a_1\times b_1+a_2\times b_2+......a_i\times b_i}{n\times n}}$$

(8)

P_o is the observed proportion of correctly simulated land-use types. P_c is the expected proportion correct due to chance agreement. P_p reflects the true proportion correct within the actual classes [45]. n is the total number of samples. s is the number of correctly classified samples. a_i is the number of true samples of land-use type i. b_i is the number of samples classified as type i. Kappa ranges from 0 to 1. Larger values indicate higher simulation accuracy. Values above 0.70 indicate ideal agreement.

2.3.4. Scenario Settings

Two scenarios are defined based on the study area’s socioeconomic trajectory and relevant protection policies for the aquatic–terrestrial ecotone: the natural evolution scenario and the sustainable development scenario. The natural evolution scenario assumes no intervention in the development process. The land-use transition rule matrix (Table S4) and regional constraint remain consistent with the pre-simulation stage. Land-use demand is projected linearly using a Markov chain model based on the trend from 2010 and 2020. The sustainable development scenario jointly optimizes economic and ecological benefits using the GMOP model. Eight land-use categories are used as decision variables. X₁ through X₈ denote grassland, forestland, rivers and lakes, reservoirs and ponds, the nearshore 6 m isobath area, cropland, built-up land, and ocean, respectively. Ecological and economic benefits serve as objective functions, and constraints on decision variables are imposed to formulate the GMOP model. The model projects optimize land-use demand under coordinated economic–ecological development. Transfers from rivers and lakes and from forestland to non-ecological land-use types are restricted, and land-use conversions within priority ecological protection areas are constrained (Table S5).

1.

Markov chain model

The Markov chain model computes land-use demand for an equal future period from land-use status data over a given period and a land-use transition probability matrix [24,26]. These results are not affected by land-use policies. The model serves as a non-spatial module of CLUE-S to forecast the area demand of each land-use type. Combined with the CLUE-S spatial module, it enables spatial visualization of quantity changes by type.

$$S_{t+1}=P\times S_t$$

(9)

S_t+1 is the land-use demand area in 2030, S_t is the land-use area in 2020, and P is the land-use transition probability matrix.

2.

Gray multi-objective optimization (GMOP) model

The GMOP model integrates multi-objective linear programming with gray forecasting theory (GM(1,1)). By constructing objective functions and constraints, it yields the 2030 areas of each land-use type under the sustainable development scenario [47].

• Objective system construction

Considering the study area’s socioeconomic conditions, the equivalent value table of ecosystem services by Xie et al. is calibrated using the economic value generated by unit-area grain yield, with details in [28]. Economic benefits are estimated using the industrial gross output value by land-use type. Based on municipal statistical yearbooks in 2010 and 2020, livestock, forestry, agriculture, and fishery gross output values are assigned to grassland, forestland, cropland, and water areas, respectively. Water areas include rivers and lakes, reservoirs and ponds, the ocean, and the nearshore 6 m isobath area. The combined output of the secondary and tertiary industries is assigned to built-up land. For each year, output values are divided by the corresponding land-use area in each year to obtain per-unit-area gross output. The GM(1,1) model is then used to forecast the 2030 economic benefit coefficient for each land-use type. The economic benefit objective function (units: 10,000 CNY per hm²) is formulated as follows:

$$maxJ1=1.2X_1+0.02X_2+{0.17X}_3+1.75X_4+{0.41X}_5+15.91X_6+3201.72X_7+1.28X_8$$

(10)

Ecological benefits are represented by ecosystem service value. For each land-use type, ecosystem service values from the calibrated equivalent value table are summed to derive an ecosystem service value coefficient (Table S6), which is used to construct the ecological benefit objective function. The ecological benefit objective function (units: 10,000 CNY per hm²) is formulated as follows:

$$maxQ1=45.44X_1+54.90X_2+{366.78X}_3+366.78X_4+{151.90X}_5+19.21X_6+366.78X_8$$

(11)

X₁ through X₈ denote the areas of grassland, forestland, rivers and lakes, reservoirs and ponds, the nearshore 6 m isobath area, cropland, built-up land, and ocean, respectively.

• Constraint settings

Constraints bound the ranges of decision variables in land-use demand projections, ensuring that each land-use type satisfies both economic development and ecological protection; principles are shown in

Table 2. Constraints on decision variables.

Type	Basis	Formula
Total area	The sum of all the areas of all the decision variables equals the total study area.	${\sum }_i^8{X}_i=\mathrm{1,860,052}$
Forestland	The natural background is fragile, and degradation of forestland readily triggers soil erosion and desertification. Following the Outline of Guangdong Territorial Greening Plan, some low-quality and low-efficiency forestland is allowed to undergo structural optimization and functional restoration within ecological land, whereas conversion of forestland to built-up land or grassland is strictly prohibited. Accordingly, this study constrained forestland area change to prevent forestland loss and to ensure a net increase.	X2 > 287,137.75
Rivers and lakes	The river chief system and the lake chief system have been fully implemented. Rivers and lakes are effectively protected, and their area should continue to increase.	X3 > 44,908.25
Reservoirs and ponds	Municipal policy documents indicate that the area of reservoirs and ponds supporting aquaculture and planting will continue to decrease. Dike–pond agriculture is an important agricultural heritage and should be protected. Therefore, the rate of decrease in reservoirs and ponds should be moderated.	84,121.47 < X4 < 94,950.25
The nearshore 6 m isobath area	China’s 14th Five-Year Plans for Ecological Civilization and for the Marine Economic Belt call for major ecological restoration of key estuaries and nearshore wetlands and biodiversity. Therefore, the rate of decrease in the nearshore 6 m isobath area should be moderated and may achieve net growth.	X5 > 209,418.75
Cropland	Permanent basic farmland zones are delineated, and farmland protection is strict. Therefore, the rate of decrease in cropland will decline.	158,676.64 < X6 < 167,843
Built-up land	The Pearl River Estuary is a key marine-industry cluster and a core Belt and Road economic zone. Therefore, the growth rate of built-up land should be controlled, and encroachment on ecological land-use types should be restricted.	392,944.25 < X7 < 508,446.24

.

2.3.5. Benefit Comparison

To analyze dynamic differences in spatial distribution and temporal sequences among landscape elements, FRAGSTATS 4.2 was used to compute landscape pattern changes at the landscape and class levels (Table S7). At the landscape level, seven metrics were used: NP, PD, LPI, LSI, AI, SHDI, and SHEI. At the class level, six metrics were used: NP, CA, ED, LPI, PAFRAC, and SPLIT. To assess benefit changes under the two scenarios for three aquatic–terrestrial zones, namely rivers and lakes, reservoirs and ponds, and coasts, this study created buffers on both sides of the shoreline at 0.2 km intervals. Distances were defined as negative seaward and positive landward (Figure S2). Economic and ecological benefits were compared across buffers to identify key ranges where development should be restricted.

3. Results

3.1. Model Validation Results

The average ROC across land-use types was highest at 50 × 50 m (0.9335; Table S8). This indicates the strongest explanatory power of the drivers at this scale, so 50 × 50 m was selected as the optimal simulation scale. Comparing the 2020 pre-simulation with the 2020 observed land-use map, grassland accuracy was 69.7%, while all other land-use types exceeded 85%. The Kappa coefficient was 0.904, indicating high overall accuracy and reliable results. Constant parameters were held fixed, and non-constant parameters were configured for the 2030 scenario simulations of landscape pattern. These results provide a solid basis for analyzing landscape pattern evolution in the estuarine aquatic–terrestrial ecotone of the Pearl River Delta.

3.2. Landscape Pattern Evolution Characteristics

3.2.1. Dynamic Changes in Land-Use Types

Comparing 2010 and 2020 (Figure 3a,b), built-up land expanded rapidly. Cropland and forestland decreased in the east, and reservoirs and ponds decreased in the west. Cropland, forestland, reservoirs and ponds, and the nearshore 6 m isobath area were mainly converted to built-up land. Some reservoirs and ponds were converted to rivers, lakes, and the nearshore 6 m isobath area (Figure S3). In the natural evolution scenario in 2030 (Figure 3c), built-up land expands substantially, and reservoirs and ponds, cropland, forestland, and the nearshore 6 m isobath area continue to shrink. Expansion of built-up land is stronger in areas closer to reservoirs and ponds and to the nearshore 6 m isobath area. This pattern is evident in northern Zhongshan (Figure 3e) and in the coastal ecotone of Hong Kong and Shenzhen (Figure 3f). Compared with the natural evolution scenario, the sustainable development scenario in 2030 constrains the expansion of built-up land (Figure 3d). Forestland, the nearshore 6 m isobath area, rivers and lakes increase slightly, whereas grassland, reservoirs, ponds, and cropland exchange with almost every other land-use type.

3.2.2. Dynamic Changes in Landscape Pattern

At the landscape level (Figure 4), AI increased, while NP, PD, LPI, LSI, SHDI, and SHEI declined between 2010 and 2020. Relative to 2020, NP, PD, and LPI rise in both 2030 scenarios. The increase is larger in the natural evolution scenario in 2030 and indicates greater fragmentation and stronger dominance of built-up land patches. In the natural evolution scenario, AI and SHDI decrease, and LSI increases, implying more regular shapes and lower diversity. The sustainable development scenario in 2030 shows the opposite pattern, with higher aggregation and improved connectivity and diversity.

At the class level (Figure 5), NP and ED decreased most sharply for reservoirs and ponds, forestland, cropland, and built-up land in 2020. Built-up land alone showed increases in CA and LPI, whereas PAFRAC decreased. This pattern suggests progressive encroachment of fragmented forestland, cropland, reservoirs, and ponds by more regular built-up land, thereby reducing overall fragmentation. In the natural evolution scenario in 2030, built-up land exhibits the largest increases in CA and NP. For the ocean and the nearshore 6 m isobath area, LPI declines while PAFRAC increases. ED decreases for all classes except grassland and built-up land. In the sustainable development scenario in 2030, forestland has the largest increase in CA and NP. ED increases, PAFRAC and SPLIT decrease, and the LPI of built-up land declines slightly.

3.3. Comparison of Built-Up Land Encroachment on Priority Ecological Protection Areas

Grades of ecological conservation importance increase from inland to nearshore (Figure 6a). High and sub-high classes are concentrated along Pearl River tributaries and estuaries and in mountainous areas with high vegetation cover, accounting for 68.47% of the study area. Within priority ecological protection areas, dominant land-use types are ocean, forestland, and the nearshore 6 m isobath area, accounting for 44.18%, 17.99%, and 16.43%, respectively. Encroachment by built-up land under the natural evolution scenario is widely dispersed with small clusters (Figure 6b). Compared with 2020, the encroachment share increases by 1.9 percentage points. The hotspots are mainly distributed in the river and lake ecotone in each city (Figure 6b-1), at the estuaries of Pearl River tributaries (Figure 6b-2), and in the coastal ecotone along Shenzhen and Hong Kong (Figure 6b-3). These locations are economically developed areas. This pattern indicates stronger conflicts between built-up land and priority ecological protection areas and a risk to regional ecological security. Under the sustainable development scenario, strict conversion limits prevent encroachment within priority ecological protection areas.

3.4. Economic and Ecological Benefit Comparison

Both the natural evolution scenario and the sustainable development scenario increase economic benefits in the study area (

Table 3. Statistics of economic and ecological benefits (units: ×10⁶ ten-thousand CNY).

Type	2020		Natural Evolution Scenario		Sustainable Development Scenario
	Economic Benefits	Ecological Benefits	Economic Benefits	Ecological Benefits	Economic Benefits	Ecological Benefits
Grassland	0.16	0.55	0.04	0.65	0.03	0.49
Forestland	0.01	6.32	0.01	6.19	0.01	6.59
Rivers and lakes	0.01	6.60	0.01	6.71	0.01	6.68
Reservoirs and ponds	0.11	13.95	0.15	12.36	0.16	13.81
The nearshore 6 m isobath area	0.06	12.75	0.08	12.56	0.09	12.84
Cropland	1.64	1.29	2.39	1.16	2.40	1.16
Built-up land	798.55	0.00	1356.59	0.00	1279.15	0.00
Ocean	0.52	92.94	0.81	93.08	0.81	92.96
Total	801.06	134.40	1360.08	132.70	1282.66	134.54

). However, ecological benefits diverge. Under the natural evolution scenario, ecological benefits decrease by 1.26% and pose risks to regional ecological security. Under the sustainable development scenario, overall ecological benefits rise by 0.11%, with marked increases for forestland, rivers and lakes, and the nearshore 6 m isobath area. Therefore, the sustainable development scenario supports economic growth while maintaining steady ecological protection and delivers a win-win outcome for economic and ecological benefits.

In the natural evolution scenario in 2030, economic benefits within buffers on both sides of the aquatic–terrestrial ecotone boundary for rivers and lakes, reservoirs and ponds, and coasts increase markedly relative to 2020, whereas ecological benefits generally decline (Figure 7). Within the immediately adjacent buffers, defined as −0.2–0 km seaward and 0–0.2 km landward, average growth in economic benefits reaches 36,542.67% and 33.16%, while average ecological benefits fall by 6.88% and 0.02%. These patterns indicate strong and near-edge development pressure under the natural evolution scenario. By contrast, in the sustainable development scenario, economic benefits in buffers for rivers and lakes and for reservoirs and ponds are only slightly above 2020 levels, whereas coastal buffers increase by 57.47% on average. Ecologically, reservoir and pond buffers decline by 9.51% within −0.2–0 km, whereas coastal buffers rise by 283.30% over the same range. Across 0.2–1 km from each ecotone boundary, ecological benefits for all three ecotones remain essentially unchanged from 2020. Overall, the sustainable development scenario maintains regional economic growth while improving ecological functions in key coastal zones, achieving coordinated economic–ecological gains.

4. Discussion

4.1. Model-Scale Selection and Simulation Accuracy Underpin This Study’s Reliability

City-scale studies commonly use coarse resolutions. Typical cell sizes include 100 × 100 m, 200 × 200 m, 500 × 500 m, and 1000 × 1000 m [46,48,49,50]. In comparison, 50 × 50 m was identified as the optimal scale. This likely reflects the higher spatial heterogeneity and pronounced edge effects of the estuarine aquatic–terrestrial ecotone. At this resolution, the driving factors capture land-use change mechanisms more effectively. Although grassland accuracy is slightly lower at 69.7%, this is likely attributable to its transitional nature and frequent confusion with forestland and cropland. Grassland misclassification may bias ecosystem service value estimates, and a slight overestimation may occur when grassland is misclassified as forestland. Because grassland accounts for only about 1–2% of the study area, the resulting influence on total regional ecological benefits and relative scenario comparisons is limited. The overall Kappa coefficient is 0.904, indicating that the driving factor system—incorporating marine–terrestrial interaction characteristics—effectively captures landscape pattern evolution in the estuarine aquatic–terrestrial ecotone under combined human activities and natural processes. This performance demonstrates the model’s applicability in the Pearl River Delta and provides a robust foundation for future scenario projections.

4.2. The Evolution of Landscape Patterns in Estuarine Urban Agglomerations Clearly Reflects an Intense Interplay Between Human Activities and Natural Processes

Landscape pattern evolution results from the combined effects of natural environments and human activities [51]. Comparing 2010 and 2020, the study area exhibited rapid urbanization. Built-up land became the dominant patch, and its regular and clustered expansion directly fragmented cropland, forestland, and water bodies. This aggregated urban construction was an unsustainable homogenization process at the expense of natural ecological space. The main reason was that agricultural production in the Pearl River Delta estuarine area was largely replaced by secondary and tertiary industries. As a result, cropland, forestland, reservoirs, ponds, and the nearshore 6 m isobath area were converted to built-up land, and built-up dominance was strengthened. Although ecological protection gained attention in 2010 and 2020, ecological space grew more slowly than built-up expansion. The resulting benefits were smaller than losses from urban growth, similar to patterns in estuarine ecotones in the Pearl River Delta and the Yangtze River Delta, China, Cambodia, and Malaysia [52,53,54,55]. These patterns indicate that economic–ecological trade-offs are reflected in landscape pattern change [56,57].

Accelerating urbanization altered fragmentation and weakened the dominance of cropland and forestland. Outward expansion from urban cores to suburbs reduced fragmentation and concentrated patterns in core districts of Guangzhou, Shenzhen, and Dongguan. Discontinuous growth around new development centers in peripheral areas produced dispersed expansion and encroachment on ecological land, for example, along the western Shenzhen and eastern Zhongshan coasts, thereby increasing fragmentation of the ecotone. Expansion of built-up land led to sustained declines in water conservation, soil conservation, coastal protection, and biodiversity maintenance. Overall, fragmentation in a typical ecotone is closely tied to urbanization. Encroachment of built-up land on ecological space, and the mismatch and overlap between productive space and ecological spaces, are key drivers of declines in ecosystem service functions [58,59,60].

4.3. Scenario-Based Comparison of Landscape Pattern and Ecological Conflicts

Regional landscape patterns are closely linked to human activities. Scenario simulation provides a useful lens and analytical framework for assessing human-induced impacts on landscape patterns. The contrast between the two 2030 scenarios further highlights the importance of coordinating ecological protection with urban expansion. Under the natural evolution scenario, we assume that land-use change follows existing expansion trends and that nearshore ecological control constraints are not further strengthened. Driven by port deepening and upscaling, improved shipping and logistics systems, and the agglomeration of port-related industries, demand for shoreline resources and coastal reserve space continues to increase. This demand promotes the allocation of newly added built-up land along deep-water shorelines and adjacent areas and locally led to expansion toward nearshore shallow-sea functional zones. Spatially, built-up land expansion intensifies encroachment into and consolidation within the estuarine aquatic–terrestrial ecotone. Patch shapes become more regular, edge effects increase, and landscape diversity declines, accompanied by impaired continuity of ecological space, indicating elevated risks of ecological function degradation in the estuarine aquatic–terrestrial ecotone. Under the sustainable development scenario, newly added built-up land is constrained by stricter ecological bottom-line requirements and spatial control measures, and expansion is limited through intensive land use and structural optimization. On the one hand, built-up land is preferentially allocated through infill development and redevelopment within and around existing concentrated construction areas. More new development is placed in areas with better infrastructure conditions and higher development suitability. This pattern reduces encroachment on sensitive estuarine and coastal zones and key ecological spaces, thereby lowering ecological conflict and fragmentation risks. On the other hand, forestland recovery and quality improvement are promoted through degraded forest restoration and the construction of ecological buffer belts and corridors around urban areas. These measures strengthen the continuity and integrity of the blue–green network and improve landscape connectivity and diversity, demonstrating the key role of human interventions in maintaining landscape sustainability.

Overlay analysis of land-use dynamics and the spatial distribution of ecological protection importance enables a direct assessment of the impacts of built-up land expansion on the ecological security pattern, elucidating the spatial distribution and drivers of conflicts. Core ecological protection areas are spatially concentrated in estuarine and coastal zones, which provide habitat maintenance, water purification, flood storage and regulation, and shoreline protection. However, their relatively flat terrain and the concentration of ports and port-related industries also make them major frontiers of economic expansion. Under the natural evolution scenario, in the absence of strengthened nearshore ecological bottom-line constraints, newly added built-up land is preferentially allocated to estuarine and coastal belts with high development suitability, inevitably overlapping scattered ecologically important areas. This produces a widely dispersed encroachment pattern with small clusters and increases the conflict by 1.9%. These results suggest that, in high-intensity development regions such as the Pearl River Delta, spontaneous development tends to threaten regional ecological security by sacrificing key ecological spaces. In contrast, the sustainable development scenario applies strict ecological redline control and embeds rigid constraints into land-demand optimization and spatial allocation, thereby providing stronger protection for forestland and other ecological spaces and effectively restraining built-up land expansion. This reduces direct conflicts with core areas and achieves concurrent gains in economic and ecological benefits. Overall, these findings highlight the role of forestland in supporting ecological security in the estuarine aquatic–terrestrial ecotone [61] and demonstrate the relevance and effectiveness of related policies in reconciling development with protection. Scenario simulation clarifies economic–ecological trade-offs and shows that scientific spatial planning can achieve coordinated gains, especially in fragile and critical coastal ecotones.

4.4. Policy Implications Based on Landscape Pattern Simulation and Optimization

For future optimization and land-use planning, development pathways are decisive. Based on conflict hotspots and benefit responses identified by scenario simulations, priority attention focuses on estuaries and coastal zones and the river and lake ecotone along inland rivers and lakes. We should promote integrated optimization and reallocation of industry, space, population, resources, and the environment across marine–terrestrial subsystems. When functional zones conflict in marine–terrestrial spatial coordination, ecotone integrity and sustainable use should be prerequisites. Ecological rationality should take precedence over economic and technical rationality [62]. This study proposes a three-level zoning control suggestion for ecotone management based on buffer distance and conflict intensity.

Level 1 is the nearshore core belt and a strict control zone. It covers the water-side belt of −0.2–0 km and the land-side belt of 0–0.2 km, and it overlays the shallow-sea functional zone within the nearshore 6 m isobath. As the most conflict-sensitive area, it prioritizes maintaining the integrity of the intertidal zone, nearshore wetlands, and critical aquatic ecological process spaces. Level 2 is the buffer and restoration belt and a strict control zone. It focuses on the 0.2–1 km range and aims to reduce fragmentation and enhance ecological resilience through pond-to-tidal-flat conversion, natural shoreline restoration, and reconstruction of waterfront vegetation belts. Level 3 is the general coordination belt. It covers the remaining areas and seeks to reduce the disruption of ecological network connectivity by promoting compact and clustered land-use configurations and embedding blue–green networks into spatial layouts. This zoning scheme is grounded in the nearshore aggregation of conflicts. It helps prioritize the restoration of key nodes and improve landscape connectivity among river and lake waters, forestland, grassland, and nearshore wetlands, thereby supporting the construction of a blue–green ecological security network in the estuarine aquatic–terrestrial ecotone of the Pearl River Delta.

4.5. Implications and Limitations

Understanding landscape pattern change underpins the explanation of land-use and land-cover change and supports simulation, prediction, and management [63]. The CLUE-S can be integrated with external models for dynamic landscape simulation. The integrated framework with a Markov chain model and GMOP to forecast land-use demand is applicable to similar studies. Landscape classes defined by characteristics of the estuarine aquatic–terrestrial ecotone of the Pearl River Delta provide strong explanatory power for ecotone dynamics but are not suited to vegetation-cover-type evolution [64]. CLUE-S captures patch-scale dynamics more precisely [26] and is well-suited to contexts with severe fragmentation of reservoir and pond patches and strong variation in river width. This study quantifies the trade-offs between ecological and economic benefits and identifies conflict hotspots within the estuarine aquatic–terrestrial ecotone, along with their distance-decay patterns. These findings provide empirical support for computable incremental development boundaries, optimized land-use structures, and cross-scenario comparisons. Furthermore, we propose zoned management and ecological restoration strategies for high-conflict nearshore zones, which help minimize encroachment on key ecological spaces and mitigate associated risks of ecological degradation while preserving essential regional development benefits. In addition, by promoting intensive use of built-up land, redevelopment of underutilized areas, and optimization of spatial configurations in port-adjacent and urban regions, our results offer actionable spatial guidance for the high-quality development of port logistics and port-related manufacturing under ecological constraints.

The limitations of this study should also be stated. Because the original socioeconomic data had limited spatial resolution, nearest-neighbor resampling with value replication may not represent fine-scale spatial heterogeneity. This may reduce the explanatory power for local pattern changes and may affect simulation results. CLUE-S typically requires at least 1000 patches per single landscape class, so this study relied on first-level classes. Future research can narrow the study extent and scale, target hotspots, use high-resolution remote sensing, and subdivide classes by ecotone gradient position. The simulations did not include climate change. Sea-level rise and extreme events can reshape land-use patterns. For benefit prediction, we mainly extrapolated trends from historical data and did not incorporate uncertainties such as future global economic conditions and inflation. Interactions among administrative units were not modeled. Within the Guangdong–Hong Kong–Macao Greater Bay Area, close regional linkages mean that metropolitan and agglomeration development can substantially influence landscape pattern change.

5. Conclusions

This study integrates linear programming with the CLUE-S model to develop a land-use optimization and spatial allocation framework that coordinates ecological protection and economic development. Scenario simulations were conducted for the estuarine aquatic–terrestrial ecotone of the Pearl River Delta. The main conclusions are as follows.

(1)

At the 50 × 50 m scale, the CLUE-S simulation for the 2020 land-use pattern shows high agreement with observations, with a Kappa coefficient of 0.904, indicating that the framework can be used to simulate pattern evolution in a complex aquatic–terrestrial ecotone. Landscape pattern change in the study area is mainly driven by rapid built-up land expansion. Encroachment of built-up patches on cropland, forestland, and water bodies is the main driver of landscape reorganization and ecological function degradation.

(2)

Although the natural evolution scenario increases economic benefits by 69.79%, it intensifies conflicts between built-up land and core ecological protection areas. In contrast, under planning constraints and ecological control, the sustainable development scenario avoids encroachment into high-importance ecological protection areas, improves overall ecological benefits, and maintains economic growth, with notable benefits for coastal ecological functions.

(3)

Embedding rigid constraints, including ecological redlines and shoreline controls, into land-use structure optimization and spatial allocation can curb unregulated built-up land expansion, particularly encroachment into ecologically sensitive areas such as estuaries and coastal zones, and guide more intensive development or orderly expansion at urban fringes. This approach can effectively alleviate conflicts between urban expansion and ecological protection and provides operational technical support and reference for land-use planning and classified management in rapidly urbanizing urban agglomerations.

All figures in this study were created by the authors using ArcMap 10.8.