Multi-Scenario Simulation Analysis of Land Use Based on Geographical Processes: A Case Study of Longhu Town, China

Abstract

To address the disconnect between macro-quantity planning and micro-spatial allocation at the township level during rapid urbanization, this study developed a coupled model framework based on Multi-Objective Planning (MOP) and the Future Land-Use Simulation (FLUS) model, using Longhu Town as a case study. First, economic and ecological benefit coefficients were calibrated via the Grey Prediction Model and equivalent factor method to define three scenarios: Economic Priority (EPS), Ecological Protection (EcPS), and Balanced Development (BDS). Second, an Artificial Neural Network (ANN) was employed to quantify driving factors, coupled with self-adaptive Cellular Automata (CA) for spatial allocation in 2030. The results indicate that: (1) The model exhibits high reliability for small-scale simulation, with a Kappa coefficient of 0.95 and a Figure of Merit (FoM) of 0.29. (2) Strategic orientations lead to distinct spatial differentiation: under the EPS, urban–industrial land expands significantly northwestward (+16.60%), causing fragmented erosion of cropland; the EcPS achieves a 5.27% increase in forest land and ecological restoration through strict quantitative constraints; the BDS realizes the synergy of urban clustering and ecological enhancement with a marginal urban increase (0.72%). (3) The eastern urban sectors and northeastern cropland belts are identified as future land-use conflict hotspots. The “quantity-space” collaborative optimization path proposed in this study provides a scientific basis and dynamic simulation tool for refined territorial spatial management at the township scale.

1. Introduction

Changes in land tenure, quantity, and spatial distribution not only affect human survival and the evolution of natural ecological environments globally but are also closely linked to the health and sustainability of global ecosystems. As China’s economy enters a stage of high-quality development, the growing demands of production and daily life have placed immense pressure on land resource utilization [1]. This intensified human activity has exacerbated ecological degradation, triggering a cascade of environmental issues. Consequently, the contradiction between human social development, territorial space, and natural ecosystems has become increasingly difficult to reconcile. The question of how to promote the healthy and rational utilization of land has thus emerged as a critical issue needing urgent resolution and remains a research hotspot in land-use-related fields.

Conducting multi-scenario simulation analysis for Longhu Town is a practical necessity for the construction of Zhengzhou as a National Central City. As the core of the Zhengzhou Metropolitan Area, Zhengzhou relies on Longhu Town—located in the south of the urban district—as a strategic gateway and pivotal hub. Against the backdrop of national strategies, it is imperative to conduct an in-depth study on the intrinsic mechanisms of its current land-use characteristics and spatiotemporal dynamic changes. Building upon the evaluation of previous land-use planning implementation, simulating future land-use structures and spatial distribution patterns under different scenarios is essential. These efforts will provide theoretical support for the future development of Longhu Town, facilitate the optimization of its land-use spatial structure, and advance the effective transmission of territorial spatial planning. Ultimately, this will better enable Zhengzhou to exert its radiating effect as a National Central City and the core of the metropolitan area.

At present, land use and management in China are confronted with multifaceted challenges. Research has indicated that systemic issues, such as irrational land-use structures and the persistent illegal occupation of cropland [2], remain prevalent, suggesting that the overall efficiency of land utilization requires further enhancement [3]. These problems are particularly acute at the township level, manifesting specifically as ambiguous property rights and fragmented agricultural operations. As the fundamental administrative unit for land use and management, the township government occupies a central position in the allocation of land resources [4]. By leveraging critical information and coordination capabilities, these local authorities play an indispensable intermediary role [5]. Consequently, the accurate prediction of future land-use development trends is regarded as a vital proactive strategy to circumvent or mitigate the aforementioned issues.

Regarding prediction methods, a diverse range of technical pathways has been developed within academia. Supported by robust data foundations, methods such as regression analysis, Markov chains, and multivariate statistics are frequently employed for precise forecasting [6]. To address land-use intensity, quantity, and long-term spatiotemporal dynamics, researchers have extensively applied a variety of models and methodologies. these include CLUE-S, CLUMondo [7], integrated data-driven and Monte Carlo models [6], and mixed-integer programming or discrete choice models integrated with Cellular Automata (CA) [8]. Furthermore, more advanced frameworks have emerged, such as the Patch-generating Land-Use Simulation (PLUS) model [9], the Artificial Neural Network-Cellular Automata (ANN-CA) model [10], hybrid Markov-CLUE-S models [11,12], and coupled System Dynamics (SD)-PLUS models [13]. More recently, comprehensive evaluation frameworks integrating SD, PLUS, and InVEST models have also been utilized to provide more holistic assessments [14].

Future scenario simulation transcends singular predictions, serving as a pivotal tool for actively exploring land-use patterns under divergent development trajectories. Researchers have designed multifaceted scenarios from various perspectives. For instance, S et al. established integrated SSP-RCP scenarios to predict land cover changes in South Korean administrative districts [15], while Y et al. adopted combined Shared Socioeconomic Pathways (SSPs) and Representative Concentration Pathways (RCPs) [13]. Additionally, Zhang et al. constructed natural growth and two types of ecological protection scenarios for Wuhan [16]; Jiang et al. simulated the evolution of “Production–Living–Ecological Space” (PLES) in Zhengzhou under various SSP scenarios [17]; and Lai et al. defined scenarios of natural development, rapid development, and ecological protection for Hainan Island [18]. Finally, Chen et al. focused on the impact of policy interventions by comparing strong and weak policy environments.

Among numerous simulation models, the Future Land-Use Simulation (FLUS) model has emerged as a prominent tool due to its capacity to integrate natural and socio-economic drivers for simulating Land-Use/Cover Change (LUCC). By coupling System Dynamics (SD) with Cellular Automata (CA), the FLUS model significantly enhances simulation precision, leading to its widespread adoption in both global and regional studies [19]. In China, the robustness of the FLUS model has been demonstrated across various geographical contexts, including metropolitan areas like Wuhan [20], Xuzhou [21], and Guangzhou [22], as well as critical ecological regions such as the Yangtze River Economic Belt [23], the coastal zones of Yancheng [24], the Huaihe River Basin [25], the Shule River Basin [26], and the Loess Plateau [27].

However, current applications of the FLUS model are predominantly concentrated at macro- and meso-scales, such as provincial, municipal, or watershed levels. Research specifically optimized for the micro-scale of townships remains relatively sparse, even though land-use conflicts at this level are often more concentrated and complex. Addressing this gap, the present study focuses on the township scale, employing the FLUS model to conduct land-use change simulations and multi-scenario analyses. Selecting this finer resolution is expected to minimize the risk of spatial detail distortion often caused by coarse data [28], thereby providing a more scientific and precise decision-making reference for refined land management and spatial planning at the grassroots level.

2. Materials and Methodology

2.1. Study Area

Longhu Town is situated in Henan Province, Central China, bordering Zhengzhou—the provincial capital—to the north. Spanning a total area of 98 km² and administering 28 administrative villages, it is the largest township in Xinzheng City by land area. The town is geographically positioned between 113°35′–113°44′ E and 34°33′–34°39′ N (Figure 1). Characterized by its strategic location and high transport accessibility, Longhu Town serves as a vital gateway for Zhengzhou’s southward expansion and is widely recognized as the “Southern Gate” of the city, with multiple national and provincial highways traversing its territory from north to south.

According to the Zhengzhou Metropolitan Area Master Plan, Longhu Town is designated as a regional trade and logistics hub, as well as a key base for education and scientific research. In recent years, by leveraging the development of the “China South City” (Huanancheng) comprehensive trade and logistics center and various large-scale cultural tourism projects, the town has experienced rapid growth in its economic and cultural industries. Consequently, it has been ranked among the “Top 100 Towns with Comprehensive Strength in China” for four consecutive years. The rapid urbanization and complex land-use structure make Longhu Town a representative case for studying land-use simulations at the township scale.

2.2. Data Sources and Preprocessing

2.2.1. Land-Use Classification and Data Integration

The study utilizes land-use data of the study area from 2015 to 2020 (

Table 1. Data types and sources.

Data Category	Data Name	Year	Source
Statistical data	Basic Economic Data of the Study Area	2015–2020	Zhengzhou Statistical Yearbook (2015–2020); Xinzheng Statistical Yearbook (2015–2020); Socioeconomic statistics provided by Longhu Town Government (2015–2020)
Land use data	Land use data	2015–2020	Data provided by the Xinzheng Municipal Bureau of Natural Resources and Planning (2015–2020)
Natural Environment Data	DEM elevation	2020	Geospatial Data Cloud, Computer Network Information Center, Chinese Academy of Sciences https://www.gscloud.cn
	Slope	2020
	Slope direction	2020
	Water Area Spatial Data	2020
Transportation Network Data	Distance to a highway, Distance to a railway	2020	Geospatial Data Cloud, Computer Network Information Center, Chinese Academy of Sciences https://www.gscloud.cn
	Distance to a town center	2020
Socio-economic Data	GDP density	2020	Geospatial Data Cloud, Computer Network Information Center, Chinese Academy of Sciences https://www.gscloud.cn
	Population density	2020

). Due to the implementation of the Third National Land Survey in China in 2017, significant discrepancies existed in land-use classification standards between different years. To ensure temporal consistency, this study engaged in institutional consultations with the Municipal Natural Resources Bureau and the Township Land Management Office. Based on these consultations and specific research requirements, the land-use categories were reclassified and unified while strictly adhering to national standards and definitions. Consequently, the land use in the study area was categorized into nine distinct types: cropland, garden land, forest land, urban–industrial–mining land, scenic and special-use land, transportation and water conservancy land, village land, grassland, and water and tidal flats, as detailed in Appendix A

Table A1. Land-Use Reclassification System and Description.

ID	Reclassified Category	Original Category in GB/T 21010-2017 [41]	Description of Reclassification Logic
1	Cropland	0101 Paddy Field	Direct Adoption: Retains the primary category “Cultivated Land” and its sub-categories from the national standard without merging.
		0102 Irrigated Land
		0103 Dry Land
2	Garden Land	0201 Orchard	Direct Adoption: Retains the primary category “Garden Land” and its sub-categories from the national standard without merging.
3	Forest Land	0305 Shrub Land	Direct Adoption: Retains the primary category “Forest Land” and its sub-categories from the national standard without merging.
		0307 Other Forest Land
4	Urban-Industrial-Mining Land	0701 Urban Residential Land	Functional Aggregation: Due to the difficulty in precisely delineating “urban” and “isolated mining” boundaries in township data, and their shared non-agricultural construction function, these are merged into a single category.
		0601 Industrial Land
		0602 Mining Land
5	Scenic and Special-Use Land	0906 Scenic Facilities Land	Scale-Dependent Simplification: At the township scale, these land types are small and fragmented. Statistical significance is limited if treated separately, hence they are aggregated.
		090 Special Use Land
		0903Corrections Land
6	Transportation and Water Conservancy Land	1003 Highway Land	Morphological Aggregation: Given that transportation lines and water facilities are often mixed or hard to distinguish precisely on township maps, and both function as linear or point-based infrastructure, they are merged to simplify analysis.
		1004 Urban/Town Road
		1006 Rural Road
		1109 Hydraulic Structure Land
7	Village Land	0702 Rural Homestead	Strategic Preservation: As the core carrier of rural residential functions, this category is crucial for township-level research and is thus retained as an independent class.
8	Grassland	0404 Other Grassland	Direct Adoption: Retains the primary category “Grassland” and its sub-categories from the national standard without merging.
9	Water and Tidal Flats	1104 Pond Surface	Ecological Distinction: To highlight the ecological attributes of natural water bodies and wetlands, these are separated from artificial hydraulic facilities and merged into a single ecological category.
		1102 Lake Surface
		1106 Inland Mudflat

.

2.2.2. Spatial Preprocessing and Standardization

To facilitate spatial modeling, all vector datasets were converted into raster format using ArcGIS 10.8 and resampled to a uniform spatial resolution of 30 m × 30 m. To ensure spatial consistency across all layers, the CGCS2000_3_Degree_GK_Zone_38 coordinate system was adopted. All raster data underwent rigorous clipping and mosaicking processes to ensure identical spatial extents, row–column dimensions, and precise spatial alignment.

2.3. Research Design and Technical Framework

Taking the prominent issues of land use in China as a starting point, this study systematically reviews the principles and application progress of land-use prediction models. Following the logic of “Macro-problem identification—Scale-gap positioning—Research-theme establishment,” we define the specificity and urgency of land-use issues at the township level. Addressing the research gap in small-scale (township) land-use prediction, this study establishes the theme: “Geographical Process-based Multi-scenario Simulation Analysis of Land Use: A Case Study of Longhu Town.” The overall workflow adheres to a core technical route of “Data support—Model coupling—Scenario simulation—Accuracy verification—Optimization suggestions,” detailed as follows (Figure 2):

Firstly, multi-source baseline data—including current land-use status, master planning, physical geography, socio-economics, transportation accessibility, and ecological constraints—were systematically collected. Through standardization and data cleaning, coordinate systems and spatial resolutions were unified to construct a standardized research database.

Subsequently, a coupled framework integrating the Multi-Objective Planning (MOP) and Future Land-Use Simulation (FLUS) models was adopted. Specifically, the MOP model focuses on the dual objectives of economic and ecological benefits. By setting three differentiated development scenarios and incorporating constraints such as macro-policy control and sustainable land use, the optimal area structure for various land-use types in the target year was calculated. Simultaneously, the FLUS model identifies key driving factors (topography, transport accessibility, ecological, and economic factors). Through Artificial Neural Network (ANN) training, the influence weights of these factors were quantified to calculate the spatial probability of occurrence for each land-use type. Parameter calibration was completed by integrating the MOP output, land-type transition rules, and expansion factors.

Based on the coupled model framework, the spatial distribution patterns of land use in Longhu Town for the target year under the three scenarios were generated. Simulation validity was then verified using the Kappa coefficient and Figure of Merit (FoM).

Finally, a multi-dimensional visualization and scenario comparison were conducted on the validated simulation results. From the perspectives of land-use structure optimization, spatial layout adjustment, and scenario path selection, targeted and actionable recommendations for the future development of the township are proposed. This study aims to provide robust decision support for the refined management and sustainable development of land use at small scales.

2.4. Land-Use Demand Prediction: Multi-Objective Planning (MOP)

To determine the optimal land-use structure under various future development pathways, this study first constructed a Multi-Objective Planning (MOP) model. The core of this model lies in translating macro-strategic scenarios into quantifiable mathematical constraints to collaboratively optimize two often-conflicting objectives: economic benefits and ecological benefits.

2.4.1. Parameter Prediction Based on the Grey Prediction Model (GM(1,1))

The objective functions in the MOP model (economic and ecological benefits) depend on the value coefficients per unit area of each land-use type. To obtain these coefficients for the target year (2030), the Grey Prediction Model GM(1,1) was employed. This model is particularly suitable for small data sequences with exponential trends, fitting the practical situation of limited historical data at the township scale.

The specific steps were as follows: First, based on historical data from Longhu Town (2015–2020), the equivalent factor tables for economic and ecological benefit coefficients per unit area were calculated. Subsequently, these historical time series were used as inputs for the GM(1,1) model implemented on the MATLAB R2024b to predict the two types of benefit coefficients for 2030.

Prior to modeling, a rigorous step-by-step ratio test was conducted on the time-series data from 2015 to 2020. The results indicate that the ratios for all indicators fall strictly within the admissible coverage interval, confirming that the data exhibits the quasi-exponential behavior required for GM(1,1) modeling. Furthermore, the Posterior Variance Ratio (C) and Small Error Probability (P) for all land-use types meet high-precision standards (detailed accuracy metrics are provided in Appendix A

Table A2. Data Validity Check, Fitting Accuracy of the GM(1,1) Model, and Predicted Benefit Coefficients for 2030.

Benefit Type	Factor	Data Suitability Check	Model Accuracy (Posterior Variance C)	Model Reliability (Error Prob. P)	2030 Predicted Value
Economic	Cropland	Passed	0.21	0.95	203
	Garden Land	Passed	0.25	0.98	1.9
	Forest Land	Passed	0.15	0.94	8.6
	Urban-Industrial-Mining Land	Passed	0.13	0.91	1073
	Scenic and Special-Use Land	Passed	0.21	0.94	261
	Transportation and Water Conservancy Land	Passed	0.24	0.92	1023
	Village Land	Passed	0.22	0.99	165
	Grassland	Passed	0.14	0.99	1.2
	Water and Tidal Flats	Passed	0.17	0.98	2.1
Ecological	Cropland	Passed	0.28	0.95	1.5
	Garden Land	Passed	0.21	0.94	1.5
	Forest Land	Passed	0.19	0.95	5.3
	Urban-Industrial-Mining Land	Passed	0.26	0.94	0.2
	Scenic and Special-Use Land	Passed	0.24	0.97	0.3
	Transportation and Water Conservancy Land	Passed	0.25	0.98	0.2
	Village Land	Passed	0.25	0.95	0.1
	Grassland	Passed	0.27	0.91	2.2
	Water and Tidal Flats	Passed	0.22	0.99	8.5

). The resulting predicted coefficients serve as the core parameters for constructing the objective functions in the MOP model.

It is important to note that the 2015–2020 period was deliberately selected as the calibration window based on specific temporal considerations. This timeframe aligns with China’s “13th Five-Year Plan,” representing a critical transitional phase characterized by the intersection of rapid urbanization and ecological civilization policies [29]. Moreover, recent studies have demonstrated that new-type urbanization exerted a significantly stronger economic radiative effect on surrounding areas specifically during the 2015–2020 period [30]. For peri-urban areas like Longhu Town, capturing these recent “high-frequency fluctuations”—which reflect the latest trends in urbanization and economic transformation—thereby offers higher predictive validity for the near future than using long-term historical averages, which may dilute these emerging structural signals. This methodological choice is further supported by recent findings from Li, T et al. [31] and Li, D et al. [32].

2.4.2. Model Construction and Scenario Settings

• Decision Variables: In this study, nine land-use types are defined as decision variables: S1 (cropland), S2 (garden land), S3 (forest land), S4 (urban–industrial–mining land), S5 (scenic and special-use land), S6 (transportation and water conservancy land), S7 (village land), S8 (grassland), and S9 (water and tidal flats).
• Objective Functions: The economic benefit objective aims to maximize the sum of the area of each land-use type multiplied by its respective economic coefficient. The ecological benefit objective follows the same logic.

Based on the availability of township-scale statistics and following consultation with local authorities, economic coefficients were defined as follows: cropland is proxied by the total output value of agriculture minus tea and fruit; garden land uses tea and fruit output; forest land corresponds to forestry output; urban–industrial–mining and transportation and water conservancy land are proxied by the output value of secondary and tertiary industries; and scenic and special-use land is linked to tourism income. For types with low economic returns (e.g., unused land), coefficients were excluded from the economic objective. Ecological coefficients were assigned based on the Ecosystem Service Value (ESV) equivalent factor table published by Xie et al. [33].

• Constraints: The constraint system integrates multiple policy and natural rigid limitations, categorized into three types: total land area constraints, planning target constraints, and intensive development constraints (

  Table 2. Constraints for the Multi-Objective Planning (MOP) model.

  Constraint Category	Factor	Expression (Unit: hm2)	Explanation
  Total Area	Total Land Area	S1 + S2 + ……. S9 = 9800	Total area consistency
  Planning Targets	Cropland	666.97 ≤ S1 ≤ 806.76	Mandatory targets
  	Garden Land	2110.03 ≤ S2 ≤ 2353.31	Mandatory targets
  	Forest Land	1720.42 ≤ S3 ≤ 2006.79	Mandatory targets
  	Urban-Industrial-Mining Land	2948.45 ≤ S4 ≤ 4709.17	Mandatory targets
  	Scenic and Special-Use Land	54.30 ≤ S5 ≤ 386.82	Mandatory targets
  	Transportation and Water Conservancy Land	656.43 ≤ S6 ≤ 667.68	Mandatory targets
  	Village Land	978.20 ≤ S7 ≤ 998.8	Mandatory targets
  	Grassland	47.57 ≤ S8 ≤ 101.09	Mandatory targets
  	Water and Tidal Flats	117 ≤ S9 ≤ 161	Mandatory targets
  Intensive Dev	Coordination Index	0.15 ≤ S6/S4 ≤ 0.25	Development patterns
  	Urban-Rural Ratio	0.28 ≤ S7/S4 ≤ 0.34	Development patterns
  	Dev. Intensity	0.46 ≤ (S4 + S6 + S7)/9800 ≤ 0.55	Development patterns

  ). The setting of constraint intervals for decision variables follows a “Status Quo–Inertia Dual Benchmark” principle. Specifically, the upper and lower bounds for each variable are defined by two key benchmarks: (1) the actual area in 2020, representing the policy baseline of ‘status quo’ (i.e., zero growth or zero loss); and (2) the inertial projected value for 2030, calculated based on the average historical change rate from 2015 to 2020, representing the boundary of natural evolution driven by market forces. The MOP model seeks the optimal solution within the feasible region bounded by these two values. Furthermore, the values in Table 2 are presented with two decimal places as they are directly derived from the precise statistics of the GIS vector database. This precision is maintained to ensure strict mass balance with the total study area, preventing cumulative errors caused by rounding.

Scenario Weights: Three representative scenarios were established: Economic Priority Scenario (EPS), Ecological Protection Scenario (EcPS), and Balanced Development Scenario (BDS). By adjusting the weights of the two objectives (

Table 3. Objective weights and core constraints across three scenarios.

Scenario	Economic Weight	Ecological Weight	Core Focus
EPS	0.8	0.2	Maximizing land utilization efficiency within minimal regulatory limits.
EcPS	0.2	0.8	Prioritizing ecological security within strict ecological carrying capacities.
BDS	0.5	0.5	Balancing development and protection through refined spatial governance.

) and reflecting policy intensity in the constraints, the model calculates the optimal land-use allocation for 2030. It should be noted that the weight combinations (0.8/0.2, 0.2/0.8, 0.5/0.5) are not intended to be precise empirical measurements of current policies. Instead, they serve as archetypal scenarios selected from the Pareto optimal set. The high weight of 0.8 is deliberately set to simulate the boundary effects of strong policy interventions, thereby amplifying the spatial differentiation under conflicting development goals. The 0.5/0.5 ratio serves as a theoretical neutral baseline. This distinct gradient setting allows for a clearer identification of spatially sensitive areas in the simulation.

2.4.3. Solving and Output

The MOP model was solved using the LINGO 19.0 optimization solver. LINGO 19.0 was selected for its high-efficiency global optimization capabilities in linear, non-linear, and integer programming, making it ideal for the multi-variable and complex constraint system in this study.

By inputting the corresponding target weights and constraint sets for each scenario, the global solver sought exact solutions for the integrated objective function. The core outputs represent the optimal planned areas for each land-use type in Longhu Town for 2030. These structured results were then imported into the FLUS model as the quantitative control framework to drive the spatial allocation process.

2.5. Spatial Land-Use Simulation: The FLUS Model

To translate the quantitative land-use structures optimized via Multi-Objective Planning (MOP) into high-resolution spatial distribution patterns, this study employed the Future Land-Use Simulation (FLUS) model for spatial allocation. By integrating Artificial Neural Networks (ANN) with Self-adaptive Cellular Automata (CA), the FLUS model effectively characterizes the spatial competition of land use under various driving forces. The ANN training process essentially quantifies the non-linear coupling between driving factors and land-use transitions within complex geographical evolutionary processes, thereby providing the dynamical support necessary for high-precision spatial simulation.

2.5.1. Driving Factors and Probability of Occurrence

Based on a systematic review of existing literature [34,35,36,37,38], land-use change is driven by multi-dimensional factors: physical geography, socio-economics, and proximity. Given the micro-scale of this township study, we prioritized data availability and consistency at the local level.

Eight driving factors were selected: elevation, slope, aspect, distance to major roads, distance to the township government, distance to water bodies, GDP per unit area, and population density. The ANN was trained on historical data to quantify the influence weights of these factors, generating initial suitability probabilities for each land-use type across the study area.

2.5.2. Coupling Mechanism and Spatial Iteration

A key design of this study is the bidirectional closed-loop coupling of the MOP and FLUS models, creating a feedback framework where “macro-objectives guide micro-evolution”:

• Top-level Quantity Input: The optimal area structures for each scenario (from Section 2.4 serve as the total spatial allocation targets for the FLUS-CA module.
• Scenario Implementation: The transition cost matrix and neighborhood factor parameters are the primary tools for implementing scenario-specific logic.

• Cost Matrix: A binary matrix (0 or 1) defining permitted (1) or prohibited (0) transitions between land types.
• Neighborhood Factors: Ranging from 0 to 1, these quantify the relative expansion intensity of each land type. Higher values indicate a stronger competitive advantage in the spatial neighborhood.

The specific settings for the transition rules and neighborhood weights across the EPS, EcPS, and BDS scenarios are summarized in Appendix A

Table A3. Land-use transition cost matrices under three scenarios.

EPS
	S1	S2	S3	S4	S5	S6	S7	S8	S9
S1 *	1	1	1	0	0	0	1	1	1
S2	1	1	1	0	0	0	1	1	1
S3	1	1	1	0	0	0	1	1	1
S4	1	1	1	1	1	1	1	1	1
S5	1	1	1	1	1	1	1	1	1
S6	0	0	0	1	1	1	0	0	0
S7	1	1	1	0	0	0	1	1	1
S8	1	1	1	0	0	0	1	1	1
S9	1	1	1	0	0	0	0	0	1
EcPS
	S1	S2	S3	S4	S5	S6	S7	S8	S9
S1 *	1	1	0	1	1	1	1	0	0
S2	1	1	0	1	1	1	1	0	0
S3	1	1	1	1	1	1	1	1	0
S4	0	1	0	1	1	1	1	0	0
S5	0	1	0	1	1	1	1	1	0
S6	0	1	0	1	1	1	1	0	0
S7	0	1	0	1	1	1	1	0	0
S8	1	1	0	1	1	1	1	1	0
S9	1	1	1	1	1	1	1	1	1
BDS
	S1	S2	S3	S4	S5	S6	S7	S8	S9
S1 *	1	1	0	1	1	1	1	1	1
S2	1	1	1	1	1	1	1	1	1
S3	1	1	1	1	1	1	1	0	0
S4	1	1	0	1	1	1	1	1	1
S5	0	1	1	1	1	1	1	1	1
S6	1	1	0	1	1	1	1	0	0
S7	1	1	1	1	1	1	1	1	0
S8	0	1	1	1	1	1	1	1	1
S9	1	1	1	1	1	1	1	1	1

* S1 (cropland), S2 (garden land), S3 (forest land), S4 (urban–industrial–mining land), S5 (scenic and special-use land), S6 (transportation and water conservancy land), S7 (village land), S8 (grassland), and S9 (water and tidal flats).

and

Table A4. Neighborhood factor parameters under three scenarios.

Land-Use Type	S1	S2	S3	S4	S5	S6	S7	S8	S9
EPS	0.2	0.2	0.5	0.9	0.8	0.9	0.2	0	0
EcPS	0.2	0.2	1	0.5	0.8	0.5	0.2	0.8	0.8
BDS	0.2	0.2	0.5	0.8	0.8	0.8	0.2	0.2	0.2

.

• Bottom-up Spatial Competition: The self-adaptive CA module integrates suitability probabilities, neighborhood effects, and scenario-based transition costs. Under the constraint of the total area targets, the model performs multiple iterations to reach a spatial equilibrium.

This framework ensures that the “top-down” policy optimizations from the MOP are spatially manifested through “bottom-up” dynamics, resulting in patterns that satisfy both macro-strategy and local spatial logic.

2.6. Model Calibration and Validation

To ensure the reliability of the 2030 projections, the FLUS model underwent rigorous historical back-testing. We used 2015 as the base year to simulate the land-use pattern of 2020 based on contemporary driving factors. The simulated results were then compared with actual 2020 land-use observations.

The calibration process involved fine-tuning inertia coefficients and transition costs to ensure the simulated spatial patterns and statistical features closely matched reality. Accuracy was quantitatively assessed using the Kappa coefficient and the Figure of Merit (FoM) index.

Once these indicators reached the accepted validity thresholds, the parameters were deemed reliable for reflecting the land-use change mechanisms of Longhu Town. These verified high-precision parameters were then applied to all future scenario simulations for 2030, ensuring that the coupled framework rests on a scientifically robust foundation.

3. Results

3.1. Model Calibration, Mechanism, and Accuracy Validation

3.1.1. Historical Backtracking Calibration and Accuracy Verification

To ensure the reliability of the future land-use simulation, this study first performed historical back-testing and accuracy validation on the FLUS model. The Kappa coefficient and Figure of Merit (FoM) were adopted as the core evaluation metrics. The Kappa coefficient measures the categorical consistency between the simulation and actual observations, while the FoM index focuses on the spatial accuracy of cells that undergo transitions. Based on existing literature, thresholds of Kappa > 0.75 and FoM > 0.15 (or 0.25 depending on scale) are generally considered acceptable.

Validation Metrics for Model Accuracy:

• Kappa Coefficient: This metric was calculated based on a global pixel-by-pixel comparison between the simulated 2020 land-use map and the actual observation, reflecting the overall consistency of the macroscopic spatial pattern. The formula is expressed as follows:

Kappa = (P0 − Pc)/(1 − Pc)

(1)

where P0 is the observed proportionate agreement (i.e., the ratio of correctly simulated pixels to the total number of pixels), and Pc is the expected proportionate agreement by chance.

• Figure of Merit (FoM) Index: Recognizing that the Kappa coefficient is often biased upward by the large proportion of unchanged “persistent” pixels in short-term (e.g., 5–10 years) township-scale simulations, this study introduced the FoM index to specifically evaluate the model’s accuracy in simulating the dynamic “changed areas.” The formula is given by:

FoM = B/(A + B + C + D)

(2)

where B represents the areas where actual changes were correctly simulated (Hits), while A, C, and D represent various spatial allocation errors, specifically corresponding to Misses, False alarms, and Wrong hits, respectively.

The specific procedure involved utilizing the Artificial Neural Network (ANN) and Cellular Automata (CA) modules within the GeoSOS-FLUS V2.4 software. Taking the 2010 land-use baseline and its corresponding driving factors as inputs, the ANN module calculated the transition probabilities for each land category, and parameters were calibrated to simulate the land-use pattern for 2020. By employing the model’s built-in validation function, the simulated 2020 results were compared against the actual 2020 land-use data (Figure 3). The validation yielded a Kappa coefficient of 0.95 and a Figure of Merit (FoM) of 0.29. Both metrics significantly exceeded established thresholds, demonstrating the model’s high-precision capability in reproducing historical land-use dynamics. This result confirms the robustness of the parameter system and provides a solid methodological foundation for projecting 2030 land-use scenarios using 2020 as the base year.

3.1.2. Driving Forces and Transition Mechanisms for Future Projections

To elucidate the internal logic of the multi-scenario projections and avoid a “black-box” simulation, this section presents the intermediate results generated using 2020 as the base year. First, the driving factors—including topography (elevation, slope), demographic (population), and economic (GDP) metrics—were spatialized and standardized (Figure 4). Subsequently, the ANN module of the FLUS model utilized these factors to generate the spatial probability of occurrence for each land-use type for the 2030 simulation (Figure 5).Taking the Urban-Industrial-Mining land (S4) probability map as an example, it intuitively reflects the potential suitability of various grid cells for conversion into construction land under current socio-economic and natural constraints. High-probability areas (warm tones) are significantly clustered along major transportation axes, around the urban core, and in regions with high GDP density. This spatial distribution aligns closely with classic land-use change theories, indicating that the model successfully learned and quantified the influence of core drivers on urban expansion. Such transparent intermediate outputs enhance the interpretability of the simulation process and provide a mechanistic basis for the subsequent multi-scenario spatial patterns.

3.2. Evolution of Land-Use Structure Under Multiple Scenarios

Based on the Multi-Objective Planning (MOP) model solved via MATLAB R2024b and LINGO 19.0, this study obtained the optimized quantitative land-use structures for the study area in 2030 under three development scenarios (

Table 4. Simulation results and changes in land-use quantity structure under multiple scenarios in 2030.

Land-Use Type	2020 Area (hm2)	EPS (2030)			EcPS (2030)			BDS (2030)
		Area	Change	%	Area	Change	%	Area	Change	%
S1 Cropland	806.73	666.97	−139.76	−17.32	806.7	−0.03	0	806.7	−0.03	0
S2 Garden Land	2353.32	2110.03	−243.29	−10.34	2343.21	−10.11	−0.43	2343.31	−10.01	−0.43
S3 Forest Land	1720.42	1720.42	0	0	1811.11	+90.69	+5.27	1762.48	+42.06	+2.44
S4 Urban-Industrial-Mining Land	2948.45	3437.83	+489.38	+16.60	2948.45	0	0	2969.6	+21.15	+0.72
S5 Scenic and Special-Use Land	54.3	54.3	0	0	88.23	+33.93	+62.48	64.3	+10.00	+18.41
S6 Transportation and Water Conservancy Land	656.43	667.68	+11.25	+1.71	656.43	0	0	656.43	0	0
S7 Village Land	998.26	978.2	−20.06	−2.01	981.3	−16.96	−1.7	979.09	−19.17	−1.92
S8 Grassland	101.09	47.57	−53.52	−52.94	47.57	−53.52	−52.94	101.09	0	0
S9 Water and Tidal Flats	161	117	−44	−27.33	117	−44	−27.33	117	−44	−27.33

). Overall, driven by established constraints and scenario objectives, each pathway exhibits distinct structural differentiation: land-use types with lower economic or ecological outputs were reduced, while high-benefit types expanded, thereby achieving the optimal comprehensive benefits predefined for each scenario.

Under the Economic Priority Scenario (EPS), the land-use structure is significantly biased toward economic benefit objectives. Compared with the 2020 baseline data, urban–industrial–mining land (S4) and transportation and water conservancy land (S6) increased by 16.60% (+489.48 hm²) and 1.71% (+11.25 hm²), respectively, serving as the primary spatial carriers for regional economic growth. Conversely, cropland (S1) experienced the most substantial shrinkage, decreasing by 17.32% (−139.76 hm²), while grassland (S8) and water and tidal flats (S9) also saw sharp declines exceeding 27%. Although this structure maximizes economic output in the short term, it risks the loss of cropland resources and damage to the ecological foundation, posing challenges to long-term regional sustainability.

Under the Ecological Protection Scenario (EcPS), ecological benefit objectives dominate, leading to a profound restructuring of land use. Forest land (S3) and scenic and special-use land (S5) emerged as the primary gainers, with areas increasing by 5.27% (+90.69 hm²) and 62.48% (+33.93 hm²), respectively, significantly strengthening the scale and function of ecological spaces. To accommodate this expansion, grassland and water and tidal flats were both heavily compressed (each decreasing by approximately 27.33%). The expansion of various construction lands was strictly restricted, largely maintaining their current status. This structure clearly reflects the spatial trade-offs under the “ecology first” strategy, prioritizing ecological conservation while securing basic economic needs.

Under the Balanced Development Scenario (BDS), the BDS seeks a middle ground between development and conservation, exhibiting the most moderate and inclusive land-use changes. Among construction lands, only urban–industrial–mining land showed a slight growth of 0.72% (+21.15 hm²). Regarding ecological land, forest and scenic and special-use land increased by 2.44% and 18.41%, respectively—growth rates that lie between the two aforementioned radical scenarios. Cropland and garden land were almost entirely preserved, with negligible rates of change. Area reductions were primarily concentrated in water and tidal flats (−27.33%). This structure achieves the collaborative optimization of both economic and ecological benefits at the quantitative level, reflecting the trade-offs and equilibrium central to the philosophy of sustainable development.

3.3. Spatial Pattern Simulation and Comparison Under Multiple Scenarios

Leveraging the spatial simulation capabilities of the GeoSOS-FLUS V2.4 software, this study generated the land-use distribution patterns for the study area in 2030 under three development scenarios (Figure 6). The simulation results exhibit pronounced spatial differentiation across scenarios. The spatial reorganization of various land categories clearly reflects the divergent spatial logics driven by the strategic orientations of “Economic Priority,” “Ecological Protection,” and “Balanced Development.” The spatial reorganization across scenarios reflects the spatial responses of geographical evolutionary processes, governed by socio-economic drivers under specific policy constraints.

3.3.1. Spatial Expansion and Encroachment Under the Economic Priority Scenario (EPS)

Under the EPS, urban-industrial-mining land (S4) demonstrates a prominent trend of agglomeration and expansion toward the northwest. A substantial portion of the original construction land in the east is phased out, with the spatial center of gravity shifting northwestward to form more intensive, contiguous growth zones integrated with newly added transportation and water conservancy land (S6). This process is accompanied by extensive encroachment on peripheral lands: cropland in the southeast and northeast exhibits marked fragmented recession, partially converting into construction or garden land. Garden land (S2) itself undergoes spatial reorganization, shifting from a dispersed to a concentrated pattern. Simultaneously, forest and scenic and special-use lands in the central region are squeezed, with some areas transitioning into garden or construction land. This pattern clearly illustrates the outcome of space competition oriented toward economic growth, where high-output land types expand and aggregate while ecological and agricultural spaces are segmented and eroded.

3.3.2. Spatial Protection and Restoration Under the Ecological Protection Scenario (EcPS)

Under the EcPS, the preservation and expansion of ecological space become dominant. Forest land (S3) coverage expands significantly, forming a more connected ecological matrix. Meanwhile, scenic and special-use land (S5) exhibits concentrated development in the eastern sector. To achieve these ecological goals, development is strictly constrained: urban-industrial-mining land in the east recedes noticeably, with its spatial demand redirected to the west. Village land (S7) is reconfigured through agglomeration in the northern and central regions, and the reclaimed space is partially converted into ecological land. While cropland remains generally stable, sporadic eco-conversions (e.g., returning farmland to forest) occur in the north. This pattern reflects the spatial implementation of the “Ecology First” strategy, enhancing regional ecological functions through construction land contraction, layout optimization, and systematic restoration of ecological spaces.

3.3.3. Spatial Trade-Offs and Optimization Under the Balanced Development Scenario (BDS)

The spatial pattern under the BDS is characterized by compromise and optimization. Development and protection reach a spatial equilibrium: while urban–industrial–mining land moderately shifts and concentrates in the northwest, no large-scale recession occurs in the east. The spatial morphology of cropland (S1) and garden land remains largely stable, with only minor conversions occurring at the peripheries. Ecological construction focuses on quality enhancement, with forest land showing steady “infill” growth and scenic-use land undergoing minor adjustments. Rural spaces are optimized via agglomeration in the north and center; specifically, some central villages are converted to forest, while eastern villages are integrated into urban-industrial clusters. This pattern achieves a precise spatial alignment and synergy between development and protection, providing a feasible spatial blueprint for long-term regional sustainability.

4. Discussion

4.1. Dialogue with Existing Research

The spatial patterns revealed in this study—characterized by “northwestern agglomeration” of construction land and “southeastern fragmentation” of cropland—contrast with the homogenized expansion observed by Men and Pan (2023) [35] at the watershed scale. This disparity validates their hypothesis regarding the scale-dependency of driving factor effects. By utilizing the MOP-FLUS coupling framework, we quantified the non-linear trade-offs between urban expansion and ecological loss. This deepens the observations of Zhang et al. (2022) [37]; while they noted conflicts between multiple objectives, our research further clarifies how these conflicts manifest spatially through specific geographical processes of marginal encroachment and contiguous protection.

Furthermore, our simulation identifies “Economic–Ecological Equilibrium” as the optimal sustainable path for Longhu Town. This echoes the conclusions of Yu et al. (2024) [38] in coastal wetland research: extreme scenarios of either protection or development trigger significant system vulnerabilities, whereas moderate, managed co-evolution is key to regional resilience. Finally, the identification of the eastern industrial conflict zone and the northeastern cropland sensitivity zone provides precise targets for spatial governance, operationalizing the framework of Wu et al. (2022) [34] regarding the spatial heterogeneity of climate and development impacts.

4.2. Interpretations of Deep Mechanisms

4.2.1. Spatial Maneuvering Between Market and Policy Forces

Under the Economic Priority Scenario (EPS), the pattern of northwestern urban clustering and marginal cropland encroachment is a direct spatial mapping of capital’s profit-seeking nature. In the absence of strong constraints, land development naturally gravitates toward areas with optimal locational conditions (e.g., proximity to the urban core and transport axes), revealing the “inertial expansion” mechanism of market-driven land use. Conversely, the Ecological Protection Scenario (EcPS) demonstrates how intensive policy intervention reshapes spatial logic. By polarizing the transition cost matrix and neighborhood weights, we simulated the “transmission-response” process of rigid constraints like Ecological Red Lines. Here, ecological value logic overrides economic utility, reconstructing the spatial pattern toward functional integrity.

4.2.2. The “Core–Periphery” Effect in Land Competition

The spatial evolution revealed by FLUS simulations is not a homogeneous phenomenon but exhibits a pronounced “core–periphery” geographical process, reflecting the spatial expansion logic driven by metropolitan radiation [39]. As a highly competitive land category, urban-industrial land demonstrates significant path dependency and agglomeration effects; it prioritizes “infilling” around existing built-up areas via neighborhood effects and extends axially along transportation corridors. In contrast, the recession of cropland and ecological land predominantly occurs at these expansionist “peripheral frontiers,” revealing their systemic vulnerability in spatial competition. The significance of this study lies in quantifying these competitive dynamics through multi-scenario comparisons: policy interventions (e.g., increasing the transition costs of ecological land) can substantially recalibrate competitive weights, shifting these land types from “passive retreat” to “stabilization or proactive expansion,” thereby optimizing the overall spatial structure.

4.2.3. Feedback Mechanisms of Regional Strategy at the Township Scale

As a constituent of the Zhengzhou Metropolitan Area, Longhu’s simulation results reflect the localization of regional strategies. The shift of the construction land’s center of gravity toward the Zhengzhou main city in all scenarios confirms that regional polarization is the fundamental driver of township evolution. Township land-use change is a product of regional exogenous forces and local endogenous governance, suggesting that its future is not a simple linear extrapolation but can be regulated through local institutional tools.

4.2.4. Systematic Trade-Offs from “Quantity Balance” to “Spatial Efficiency”

The optimal quantity structure from the MOP model is not mechanically implemented in FLUS; instead, it undergoes complex spatial competition. This reveals a tension between macro-quantity balance and micro-spatial efficiency. For instance, the drastic urban growth in the EPS necessitates the encroachment on high-quality cropland, leading to a “spatial mismatch” of food production and ecological services. The Balanced Development Scenario (BDS) essentially mitigates this mismatch through refined spatial rules (e.g., guiding industrial transfer and village agglomeration), maximizing overall territorial efficiency.

4.3. Methodological Contributions

The MOP-FLUS framework bridges the long-standing disconnect between “scale” and “layout” in land-use planning.

Synergy Disclosure: Unlike standalone MOP (lacking spatial logic) or FLUS (lacking strategic guidance), this coupling reveals non-linear trade-offs. In the BDS, for example, a mere 0.72% quantity increase achieves high spatial efficiency via “infill growth” strategies.

Precision Governance: It allows planners to “pre-test” the spatial consequences of quantitative targets, providing a robust methodology for precise territorial management.

4.4. Validation of Temporal Representativeness and Long-Term Evolutionary Trajectories

This study primarily calibrated the model parameters based on land-use change data from 2015 to 2020. We acknowledge that this five-year span is relatively short and may carry the risk of failing to fully capture long-term evolutionary patterns of land use, potentially leading to a misjudgment of trends for 2030. To validate the robustness of the simulation results, we incorporated additional historical time-series data (2009, 2018, and 2020) and conducted a retrospective verification of the region’s long-term spatial logic by calculating the Standard Deviational Ellipse (SDE). It should be noted that, while the land-use datasets for 2009 and 2018 explicitly included the “Construction Land” category, the 2020 dataset did not directly employ this terminology. To ensure data consistency for the longitudinal analysis, the “Urban–Industrial and Mining Land; Scenic and Special-Use Land; Transportation and Water Conservancy Land” category in the 2020 dataset is treated as equivalent to “Construction Land” in this discussion.

The calculation results (see Appendix A

Table A5. Gravity Center Coordinates and SDE Rotation Angles of Construction Land (2009, 2018, and 2020).

Year	Gravity Center (X)	Gravity Center (Y)	Rotation Angle (°)
2009	38,506,630.20	3,835,105.63	304
2018	38,506,400.00	3,835,300.00	310
2020 *	38,506,200.00	3,835,500.00	315

* Area statistics for 2009 and 2018 refer to the explicitly labeled “Construction Land” category. For 2020, due to differences in classification standards, the “Urban–Industrial and Mining Land; Scenic and Special-Use Land; Transportation and Water Conservancy Land” category is aggregated and treated as “Construction Land” to ensure comparability across the time series. The rotation angle is measured clockwise from true north.

and Figure A1) indicate that, based on the data from 2009, 2018, and 2020, the gravity center of construction land in Longhu Town exhibited a pronounced and continuous migration trend from “southeast to northwest.” This trajectory profoundly reflects the sustained radiative influence and gravitational pull of Zhengzhou’s main urban area on Longhu Town. Notably, the simulated gravity center of urban-industrial and mining land for 2030 continues this northwestward migration path, showing no significant directional deviation. This high consistency across the “History–Status Quo–Future” spatial trajectory demonstrates that, despite the relatively short modeling window, the model has successfully captured the core driving mechanisms dominating regional development (the urban sprawl pressure extending from the north). Consequently, the projected spatial pattern for 2030, derived from the 2015–2020 data, aligns with the long-term geographical evolutionary process and possesses high credibility.

4.5. Limitations and Future Directions

Although the “Optimization–Simulation” coupling framework constructed in this study has demonstrated high simulation accuracy at the township scale and provided valuable spatial decision support, several limitations remain, pointing directions for future research.

Potential Bias in Long-term Predictions Driven by Short-term Data Baselines: While the Standard Deviational Ellipse (SDE) analysis successfully validated the consistency of the spatial evolution trend (specifically, the northwest expansion), limitations regarding the quantitative prediction of land use must be acknowledged. Constrained by data availability, the future quantitative demands in this study were extrapolated using the GM(1,1) model based on the short-term time series from 2015 to 2020. This methodological approach inherently assumes that the urbanization rate and land conversion logic established during the “13th Five-Year Plan” period will continue in a linear or quasi-exponential manner. However, long-term projections derived from short baselines carry inherent risks, as they may not fully capture potential structural breakpoints triggered by future macroeconomic shifts or disruptive policy interventions. Consequently, the projected area values and spatial configurations for 2030 should be interpreted as reference scenarios indicating relative trends rather than absolute, precise predictions. Future research should aim to acquire multi-period historical data spanning longer decades to recalibrate the demand prediction module, thereby minimizing this extrapolation uncertainty.

Static Assumptions of Driving Factors: The use of annual static data for GDP and population ignores the dynamic nature of development, such as real-time traffic or emerging industrial zones. Future research should incorporate spatiotemporal sequence data (e.g., nighttime lights, mobile signaling) to enhance the model’s foresight.

Depth of Ecological Process Simulation: The current model treats ecological benefits as linear functions of area. Future studies should couple this framework with ecological models like InVEST or SWAT to transition from “area constraints” to “ecological function constraints,” maximizing ecosystem services.

Abstraction of Micro-Agent Behaviors: The model overlooks the differentiated decisions of households and enterprises. Incorporating Agent-Based Modeling (ABM) would link macro land-use change with micro-adaptive behaviors, enhancing the model’s power to explain complex human–environment interactions [40].

5. Conclusions

This study employed a coupled MOP-FLUS model framework to simulate and optimize land-use patterns in Longhu Town for the year 2030 under three distinct development pathways. The primary conclusions are as follows:

5.1. Core Findings and Optimal Pathway Identification

The research demonstrates that land-use evolution in rapidly urbanizing townships is a profound trade-off between economic drivers and ecological constraints. Among the simulated pathways, the Balanced Development Scenario (BDS) is identified as the optimal strategic choice for Longhu Town. Unlike the EPS, which secures industrial growth at the cost of a 17.32% loss in high-quality cropland, or the EcPS, which might stifle economic vitality, the BDS achieves a synergistic equilibrium. It accommodates a moderate 0.72% urban expansion while facilitating a 2.44% increase in forest land. This confirms that sustainable development in peri-urban areas is attainable through “infill” optimization rather than blind extensive sprawl.

5.2. Methodological and Theoretical Contributions

The primary contribution of this study is the construction of an operational “Evaluation–Simulation–Optimization” framework specifically tailored for the township scale.

Methodological Value: By coupling the quantitative optimization of MOP with the spatial dynamics of FLUS, this study successfully decodes the non-linear interactions between macro-targets and micro-allocations, effectively addressing the long-standing “quantity–space disconnect” in single-model applications.

Theoretical Significance: The study deepens the understanding of the “asymmetric spillover” mechanism in satellite towns. It confirms that small-scale land-use changes are jointly governed by metropolitan core radiation and policy-driven transition costs, providing a theoretical benchmark for research on similar peri-urban townships globally.

5.3. Planning Implications and Policy Recommendations

Based on the findings, we propose the following strategic recommendations for Longhu Town and similar administrative units:

• Adopt the Balanced Development Model: Planning authorities should transition from “single-growth” targets to a BDS-aligned framework, prioritizing the protection of the northeastern cropland belts while guiding industrial clustering toward the northwest.
• Implement Differentiated Zonal Governance: Targeted policies are required to resolve core spatial conflicts. In eastern conflict zones, we recommend “functional optimization and stock renewal” strategies to integrate ecological functions through urban renewal. In cropland-sensitive areas, a refined management approach focusing on “total volume stability, quality improvement, and layout adjustment” should be enforced to strictly protect contiguous high-quality agricultural land.
• Establish a “Diagnosis–Simulation–Evaluation” Dynamic Support System: The MOP-FLUS framework should be integrated as a routine tool for territorial spatial planning. By institutionalizing regular scenario simulations, planners can pre-evaluate the spatial impacts of various strategies, facilitating a shift from static “indicator control” to dynamic “scenario guidance and adaptive management.”