A New Framework of Land Use Simulation for Land Use Benefit Optimization Based on GMOP-PLUS Model—A Case Study of Haikou

Abstract

Multi-scenario simulation and prediction of land use can provide guidance for the optimization of land use patterns. Combining the GMOP model with the PLUS model can better evaluate the influence of different land use strategies on the comprehensive benefits of land use and improve the scientificity of the simulation results. This study takes Haikou City as the research area. As the political, economic, and cultural center of Hainan Province, it is the highest urbanization area in Hainan Province and also the vane of the urban development of Hainan Province. Its development experience and model play an important leading role in the surrounding cities. The land use data of 2010, 2015, and 2020 were selected, and the spatiotemporal pattern of land use under the 2035 Business As Usual scenario (BAU), Economic Development scenario (ED), and Economic and Ecological Balanced Development scenario (EEB) was simulated based on the GMOP-PLUS model. The results show that: (1) The prediction results generally show the trend of the decrease in cultivated land and forest land and the increase in construction land, among which the expansion capacity of construction land is the strongest, and the forest land is more occupied, but the increase and decrease in land use types are different under different scenarios. (2) The three simulation scenarios all show the trend of economic benefit improvement and ecological benefit decline, which indicates that the primary objective of Haikou City’s future development remains focused on economic construction, with the potential compromise of ecological functions to accommodate urban expansion. (3) The comprehensive benefits of the region in the EEB scenario are significantly higher than those in the BAU and ED scenarios. The optimized land use structure is more balanced, the scale of urban expansion is limited, and the loss of important ecological land is reduced to a minimum, which is more in line with the current concept of sustainable development. The study can serve as a reference for the coordinated development of urban planning, land use management, and ecological environment in Haikou.

1. Introduction

All socio-economic activities of human beings are implemented directly or indirectly through land use. In recent years, the rapid economic and social development and the surge in population have led to a higher demand for natural resources by human beings. This has resulted in a gradual increase in the attention paid to the ecological and environmental problems arising from land use. Land use change is defined as the process of modifying the functions of land use within a specific area, from one or more original types to alternative types. As the impact of human activities on ecosystems has become more pronounced, there has been a growing interest in exploring ways to optimize land use patterns and safeguard ecosystem services through land use scenario simulation predictions.

The research field of land use spatial pattern simulation encompasses two principal aspects: quantitative prediction and spatial configuration. With regard to quantitative prediction, a number of techniques may be employed, including Markov chain (Markov) [1], system dynamics, and linear planning methods, etc. [2,3,4,5,6].Grey multi-objective planning model (GMOP) is a new type of method for assessing land management strategies, and it is an optimization system that is a cross-combination of grey prediction theory and multi-objective optimization function [1,7], which can be used to assess the impacts of different management strategies on ecosystem services and socio-economic factors [8]. The GMOP model links different land management strategies to their corresponding management objectives, uses integrated indicators of ecosystem services, production, and economy to assess the effectiveness of the strategies, and provides a scientific basis for policy formulation [9]. Commonly used in spatial configuration are cellular automata (CA) [10], CLUE-S model [11], FLUS model [12], etc., which have high prediction accuracy and reliability and have been widely used in different research fields. Nevertheless, these models exhibit certain deficiencies in the simulation process. For instance, they are incapable of analyzing the mechanism of land use change within a specific temporal framework during the simulation process. Additionally, they are unable to simulate the patch growth of multiple natural land use types at a fine scale [13]. To address these issues, a novel land use simulation model, the PLUS model, was developed by Prof. Liang Xun [14] and colleagues at Wuhan University. Studies have demonstrated that the PLUS model is a land use simulation model based on a planning support system. It is capable of refining the conversion patterns of different land use types and analyzing the triggers of land use changes through the use of conversion analysis and pattern analysis strategies. This enables more accurate prediction of future land use [15,16,17,18].

In order to better optimize the future land use pattern, some scholars began to combine land use simulation models with other models in recent years and developed composite models such as CA-Markov model [19], ANN-CA model [20], logistic-CA model [21], GMDP-FLUS model [22], MOP-PLUS model [23], and other composite models, which greatly improved the accuracy and prediction ability of the models and provide strong scientific support for regional sustainable development. Some scholars have coupled GMOP models with land use simulation models, such as the GMOP-CLUES model [19] and GMOP-SD model [24] to assess the impacts of different land use strategies on ecosystem services, and the simulation results are more scientific and accurate. Based on the respective advantages of the GMOP and PLUS models, the coupling of the GMOP model and PLUS model can carry out more scientific land use optimization in terms of quantitative structure and spatial layout.

As the political, economic, and cultural center of Hainan Province, Haikou City has experienced rapid economic development under the concept of sustainable development and the policies of “International Tourism Island” and “Hainan Free Trade Port.” However, the conflict between economic development and the ecological environment has become increasingly evident. The application of novel models, such as the GMOP-PLUS model, to simulate future land use changes, particularly throughout the Hainan province, remains relatively uncommon in the current literature. Consequently, this paper employs Haikou City as the study area. Utilizing the three land use data sets from 2010, 2015, and 2020, the GMOP and PLUS models are employed to project future land use patterns in Haikou City under various development scenarios in 2035. The objective is to enhance the ability to anticipate future changes in land use in Haikou City, thereby providing a foundation for informed urban planning, land use management, and the coordinated development of the ecological environment.

2. Materials and Methods

2.1. Study Area

Haikou City is located between 19°32′–20°05′ N latitude and 109°52′–110°32′ E longitude. The total land area of Haikou City is approximately 2297 km², comprising four districts: Meilan District, Qiongshan District, Longhua District, and Xiuying District (Figure 1). The terrain is high in the south and low in the north, tilting from southwest to northeast. The climate is tropical and oceanic, with abundant rainfall and water resources, and the dominant winds are mainly northeast and southeast.

2.2. Data Sources and Pre-Processing

2.2.1. Data Sources

(1)

Land Use Datasets

The land use data in this article are from the Resource and Environmental Science Data Platform (https://www.resdc.cn, accessed on 19 January 2022), which are 30 m resolution spatial distribution data in 2010, 2015, and 2020, respectively, and are divided into seven primary classifications (farmland, forest, grassland, watershed, urban and rural construction land, unutilized land and sea area) and 26 secondary classifications. The land use classification table of Haikou City was constructed by synthesizing the experience of ecosystem services and land use classification and referring to the classification method of Fu et al. [25] (

Table 1. Land use classification.

Land Use Classification	Secondary Classification of the Raw Land Use Data	Code
Cultivated land	Dryland	21
	Paddy field	22
Forestland	Forest	21
	Open forestland	23
	Other forestland	24
	Shrub	22
Grassland	High-cover grassland	31
	Medium-cover grassland	32
	Low-cover grassland	33
Water	Canal	41
Wetland	Reservoir pond	43
	Tidal flat	45
	Floodplain	46
	Sea area	99
Construction land	Townland	51
	Rural settlement	52
	Other construction land	53
Unutilized land	Sandy land	61

).

(2)

Other data sources

Natural drivers include elevation, slope, and slope direction data as well as temperature and rainfall data, obtained from the Geospatial Data Cloud and the National Meteorological Science Data Center; social drivers include population density, GDP, food production, and the distribution of municipal and public service facilities in the city. POI (Point of Interest) and road network data are obtained from the Baidu online map API interface and the OSM dataset, respectively. Population, GDP, and food production data were obtained from the Haikou City Statistical Bulletin, Statistical Yearbook, and socio-economic data from “the 14th Five-Year Plan for National Economic and Social Development of the People’s Republic of China”.

2.2.2. Data Pre-Processing

(1)

Pre-processing and local correction of raw land use data in combination with Google historical image data.

(2)

Produce land use driver data and rasterize some natural and economic driver panel data by inverse distance weighting method, sample point sampling, and Kriging interpolation to get 11 influence factor data (DEM, slope, slope direction, rainfall, temperature, GDP, population density, distance to river, distance to settlement, distance to road, distance to site).

(3)

The data were processed to harmonize the coordinate system, pixel size, and row and column numbers and converted to the Unsigned char format that is recognized by the PLUS model.

2.3. Research Methodology

This study proposed a comprehensive approach to construct a new framework of land use simulation for land use benefit optimization based on the GMOP-PLUS model. The research framework is shown in Figure 2.

2.3.1. GMOP-Based Land Use Scenario Setting Methodology

Grey Multi-objective Planning (GMOP) model is a novel method for evaluating land management strategies [1], which is an optimization system cross-combined by grey prediction theory and multi-objective optimization function. In this paper, we synthesize the GMOP model using gray prediction (GM(1, 1)) and multi-objective planning function (Multi-objective Planning) [26,27]. The model can fully consider the changes of parameters affecting the land use objective function at the predicted time or scenarios and solve the land use trade-off problem under different demands with the multi-objective planning function [28,29]. Three scenarios are set up according to the future development needs of Haikou City: (1) Business As Usual scenario (BAU), (2) Economic Development scenario (ED), and (3) Economic and Ecological Balanced Development scenario (EEB).

(1)

Variable setting

The selection of GMOP decision variables needs to be independent, accurate, and typical, while also conforming to the actual situation of the study area. The study, after comprehensively considering the land use planning policy of Haikou City and the subsequent accessibility of relevant parameter data, adopts Haikou City’s cultivated land ($X_1$), forestland ($X_2$), grassland ($X_3$), water ($X_4$), wetland ($X_5$), construction land ($X_6$), and unutilized land ($X_7$) seven land use types as decision variables.

(2)

Restriction setting

The total land use area constraint ensures that the sum of the areas of each land use type for the optimization scenario is equal to the total area of the study area (hm²).

$$X_1+X_2+X_3+X_4+X_5+X_6+X_7=228068.91$$

(1)

Population scale constraint: According to the 14th Five-Year Plan of Haikou City, Haikou City needs to allocate public services and infrastructure to serve the actual population scale of 4.5 million in 2035, so the number of population carried by the construction land in the region should be higher than the projected population scale, the expression is as follows:

$$A\times X_6\ge 4742800$$

(2)

In the formula, A is the average population density of 101.18 of construction land in Haikou City in 2035 predicted by GM(1, 1), and 4,742,800 is the total population in 2035 predicted by GM(1, 1).

Constraints on carbon emissions per unit of GDP: In order to achieve carbon peak and carbon neutrality, the government of Haikou City is actively promoting green development and transformation. Land as the second largest source of carbon emissions is directly related to the realization of Haikou City’s carbon peak goal in 2030 [29]. Carbon emissions per unit of GDP, as one of the indicators proposed in the 14th Five-Year Plan, can reflect the degree of optimization of regional energy structure and green industry transformation. Therefore, the carbon emissions per unit of GDP under the optimization scenario should be lower than the level in 2020. Research shows that the carbon emission capacity of cultivated land, forestland, grassland, water, wetland, and unutilized land remains basically unchanged [30,31] and, thus, can be set directly with reference to the carbon emission coefficients, which are 0.464, −5.052, −0.947, −0.41, −0.25, and −0.005, respectively, in t/hm² [32]. Carbon emissions from construction land are usually measured based on major energy consumption [33]. However, due to the lack of relevant data in the Haikou City Statistical Yearbook, the measurement method of Wei Yuan [29] was adopted, in which the approximate carbon emissions from construction land were obtained by multiplying the output value of the secondary and tertiary industries in the region and the energy consumption per unit of GDP, and then dividing it by the area of construction land in that year, to obtain the carbon emissions per unit area of construction land in each year. The specific formulas are as follows:

$$E_j={GDP}^{\prime }\times H\times K$$

(3)

$$GD{P}^{\prime }=GD{P}_2+GD{P}_3$$

(4)

$$E_i={\displaystyle \frac{E_j}{T}}$$

(5)

where $E_j$ is the total carbon emissions from construction land; $H$ is the energy consumption per unit of GDP (tons of standard coal/million yuan); $K={\displaystyle \frac{0.7476t}{t}}$, which is the carbon emission coefficient of coal consumption; $GD{P}_2$ and $G{DP}_3$ are the GDP of the secondary and tertiary industries. According to the above formula, the total carbon emissions of Haikou City in 2020 are calculated to be 355,503.99 tons. Additionally, the annual carbon emissions per unit of construction land from 2010 to 2020 have been determined. The energy consumption per unit of construction land is predicted to be 64.77 t/hm² in 2030 by GM(1, 1)². In summary, the carbon emission constraint formula is obtained as follows:

$$\left(0.464X_1-5.502X_2-0.947X_3-0.41X_4-0.25X_5+64.77X_6-0.005X_7\right)/G\le P$$

(6)

In the formula, $G$ denotes the total GDP of Haikou City in 2030, which is predicted by GM(1, 1), and is 70,469,643,300,000 Yuan. $P$ denotes the carbon emission per unit of GDP in 2020, which is 0.198 tons per million yuan.

Cultivated land security limit: The total amount of food produced on cultivated land should cover the needs of the planning projected population, in order to ensure food security. The expression is as follows:

$$X_1\times f_1\times f_2\times f_3\ge P\times f_4\times f_5$$

(7)

where $f_1$ is the projected grain production per unit area in 2035 (6.2 t/hm²), $f_2$ is the share of grain cultivation (0.26), $f_3$ is the replanting index, taking the average value from 2010 to 2020 (1), and $P$ is the projected total population in 2035 (4,742,800), $f_4$ represents the per capita food demand (0.52 t) as predicted by Xin et al. [34], and $f_5$ is the food self-sufficiency rate of Haikou City (0.25), and the above data projections are completed by GM(1, 1).

The total amount of forestland limit: According to “the Territorial Spatial Planning of Haikou City (2020–2035)”, the total amount of forestland in Haikou City in 2035 will be 82,215 hm². To summarize, the formula for the total amount of forestland limit is obtained as follows:

$$111726.54\ge X_2\ge 82215$$

(8)

Forest cover limitation: The “14th Five-Year Plan” for ecological environmental protection in Haikou City expects the forest cover to reach more than 39% in 2035. In this paper, the green equivalent method is used to calculate the forest cover rate, in which cultivated, forestland, and grassland are the land use types that satisfy the green equivalent, and their equivalent coefficients are 0.46, 1, and 0.49, respectively [35]. In summary, the forest cover constraint equation was obtained as follows:

$$0.46X_1+X_2+0.49X_3\ge 228068.91\times 39\%$$

(9)

Land use diversity limitation: Among all types of land use, grassland and unutilized land account for a relatively small proportion and have low ecological value, so they are often converted into construction land or cultivated land in the process of urban development. In order to safeguard the diversity of the urban landscape as well as to reserve for future development, this paper assumes that the sum of grassland and unutilized land in Haikou City in 2035 will account for at least 1% of the total area.

$$X_3+X_7\ge 228068.91\times 1\%$$

(10)

Watershed limitation: At present, Haikou City completely stops land reclamation, restricts the development of the shoreline along the Nandu River, and carries out ecological restoration projects for important shorelines. Therefore, from this, we assume that the watershed area in 2035 should be greater than or equal to the 2020 level.

$$X_4\ge 2804.31$$

(11)

Wetland limitation: At present, in order to build an international wetland city, Haikou City has carried out a lot of ecological protection and restoration work for wetland ecosystems. According to the research of Chen Shengtian [36], under the Business As Usual scenario (BAU) in 2035, the wetland area of Haikou City will be reduced by 12%. Therefore, this paper assumes that the fluctuation of wetland area in Haikou City in 2035 is within 10%, with the following formula:

$$11761.38\ge X_5\ge 11285.04$$

(12)

Limitations on the range of values: All land use variables must take on non-negative values.

$$X_i\ge 0,i=\mathrm{1,2},\dots ,6$$

(13)

(3)

Objective function setting

Business As Usual scenario (BAU) is a scenario that continues current land use trends without adding any constraints or influences. This scenario mainly utilizes the Markov chain model (Markov chain) to simulate the prediction based on the land use data from 2015 to 2020. According to the transfer probability between each land use in Haikou City from 2015 to 2020, the quantity of land use in Haikou City in 2035 is projected at an interval of 5 years on the basis of the current status of land use in 2020. Markov chain is a forecasting method based on probabilistic analysis of land use change patterns and trends, which uses a transfer probability matrix to predict future demand for land use types.

$$S_{(t+1)}=P_{ab}\times s_{\left(t\right)}$$

(14)

$S_{\left(t\right)}$ and $S_{(t+1)}$ represent the state of the land at the time t and t + 1, respectively; $P_{ab}$ is the state transfer probability matrix, i.e., land type a transfer to the land type b probability.

In order to enhance the accuracy of the simulation, this paper takes the 2010 land use as the base data and predicts it in five years, and the predicted data of each time serve as the base data of the next round and finally obtain the land use structure of Haikou City in 2035 under the prospect of natural development.

Economic Development scenario (ED) refers to the scenario that maximizes regional economic development as much as possible while meeting the constraints. With reference to previous studies and the actual economic situation of Haikou City, this paper characterizes the economic value of cultivated land, forestland, grassland, water, and wetland by the total output value of agriculture, forestry, livestock, and fishery in Haikou City, respectively, and the economic value of construction land by the total output value of secondary and tertiary industries. The annual output value of agriculture, forestry, fishery, livestock, and the secondary and tertiary industries in Haikou City was obtained by consulting the Haikou City Statistical Yearbook in previous years, and then the output value per unit area of each land use type was obtained from 2010 to 2020. Subsequently, GM(1, 1) was applied to statistically analyze the known data, and the economic value per unit area of each land use type was predicted to be obtained in 2035 (

Table 2. Economic value per unit area of land use type.

Related Industry	Land Use Type	Output per Unit Area ($ million/hm2)
Agriculture	Cultivated land	35.359
Forestry	Forestland	0.41
Livestock	Grassland	77.648
Fishery	Water	36.407
	Wetland	36.407
Secondary and tertiary industries	Construction land	1402.54
None	Unutilized land	0

). Therefore, the objective function of the ED scenario with maximized economic efficiency is as follows:

$$F_1\left(x\right)=max{\sum }_{j=1}^n{c}_j{x}_j$$

(15)

where $c_j$ is the output value per unit area of land use type j; $x_j$ is the area of land use type j.

Ecological and Economic Balanced Development scenario (EEB) refers to the search for the best land use solution between economic development and ecosystem protection to ensure that the benefits of both economic development and ecological protection are optimized, provided that the constraints are met. The formula is as follows:

$$max\left\{f_1\left(x\right),f_2\left(x\right)\right\}$$

(16)

This paper employs the concept of ecosystem service value to express the ecological benefits of the land. The evaluation of ecosystem services is based on the land ecosystem service value equivalent factor table by Xie in 2015 [1], which establishes a standard ecosystem service value equivalent factor equivalent to the average yield of farmland natural grain production economic value per hectare. The ecosystem service function comprises four categories: supply services, regulating services, support services, and cultural services. By consulting the statistical yearbook, the average grain output and average grain price of Haikou City from 2010 to 2020 were obtained, and the unit ecosystem service value of Haikou City was calculated to be 2936.24 hm². By GM (1, 1), the correction coefficient of social development, resource shortage degree, and unit grain output value in 2035 will be 0.8, 1.46, and 1.02, respectively. The ESV per unit area of each land use type is then calculated (

Table 3. ESV per unit area of land use types in Haikou City in 2035 (Unit: 10⁴ RMB).

Cultivated Land	Forestland	Grassland	Water	Wetland	Construction Land	Unutilized Land
1.384	6.926	6.924	23.89	18.292	−4.223	0.07

). The ESV value formula is obtained as follows:

$$F\left(x\right)=1.384X_1+6.926X_2+6.924X_3+23.89X_4+18.292X_5-4.223X_6+0.07X_7$$

(17)

2.3.2. PLUS Model-Based Land Use Simulation Modeling

The Patch Generation Land Use Simulation Model (PLUS) is a land use simulation model released by the China University of Geosciences (CUG) in 2021, which integrates the Land Use Expansion Analysis Strategy (LEAS) and the CA model based on multiple types of stochastic patch seeding and combines the stochastic seeding generation and threshold reduction mechanisms to ultimately realize the accurate simulation of land use [14,37].

(1)

Data preparation

The optimized land use quantitative structure derived from the GMOP model is imported into the PLUS model for spatial evolution simulation, which consists of three modules, namely, the ELE module (Extract land expansion), the LEAS module (Land expansion analysis strategy), and the CARS module (CA based on Multiple Random Seed).

Firstly, through the ELE module, this module can extract the land use change characteristics of the input data and output the land use expansion data. Secondly, through the LEAS module, the raster data of influence factors related to land use need to be prepared and processed in advance. In order to ensure that the data simulation is as close to reality as possible, the influence factors selected in this paper cover both social and natural factors, including DEM, slope, slope direction, rainfall, temperature, GDP, population density, distance to river, distance to settlement, distance to road, and distance to site, a total of 11 influence factor data. Finally, the land use spatial simulation results were generated by the CARS module.

(2)

Important parameter settings

Restricted conversion area setting: Based on the 2021 Haikou City Ecological Protection Red Line data, the Restricted Conversion Area file (Figure 3) was produced, and the protection area (0) and ordinary area were divided by assignment (1).

Transfer cost setting: According to the actual situation of land use in Haikou City and the experience of previous studies [35,38], the current social productivity can basically realize the conversion between any land type. However, considering the current strict protection policy of Haikou on the Nandu River and its branches, the conversion of waters to cultivated land and construction land is therefore restricted, and the remaining land types can be spatially exchanged with each other.

Domain weight setting: Domain weights indicate the spreading intensity of different land use types, taking values between 0 and 1. The closer the weight value is to 1, the stronger the expansion capacity of the land. In this study, the expansion capacity of each land use is calculated using dimensionless processing based on the data of 2015 and 2020, and the resulting parameters are shown in

Table 4. Neighborhood factor parameters.

Cultivated Land	Forestland	Grassland	Water	Wetland	Construction Land	Unutilized Land
0.438	0.512	0.051	−0.001	0.088	1	0

.

$$X_i^{\mathrm{*}}={\displaystyle \frac{X_i-X_{min}}{X_{max}-X_{min}}}(i=\mathrm{1,2},\mathrm{3,4},\mathrm{5,6})$$

(18)

where $X_i^{*}$ denotes the weight value and $X_i$ denotes the land use area; $X_{max}$ and $X_{min}$ denote the number of land uses with the largest area and the number of land uses with the smallest area, respectively.

(3)

Validation of model accuracy

To ensure the accuracy of the PLUS simulation, this study selects the land use data of 2010 and 2015 and repeats the previous steps to derive the land use simulation data of Haikou City in 2020 (Figure 4). The simulated data are compared with the actual land use data in 2020 to derive the overall accuracy and Kappa coefficient. Generally speaking, the closer the Kappa coefficient is to 1, the higher the precision, and the Kappa coefficient is more than 0.8, the simulation results are considered accurate and credible. In this study, the Kappa coefficient is 0.89 and the overall precision is 0.92, indicating that the simulation results are scientifically reliable.

3. Results and Analysis

3.1. Analysis of Spatial Patterns of Land Use in Simulation Scenarios

By overlaying the analysis of the expansion range of urban construction land under the three scenarios, the areas of greatest urban expansion are identified (Figure 5). By 2020, the primary urban area of Haikou City has been largely developed. In the future, urban expansion will primarily occur in the east and west directions, and the urban form will transition from a single-core structure to a multi-core development. The expansion of construction land in Meilan District and Xiuying District is the most pronounced among the administrative districts under the jurisdiction, while the expansion of Longhua District and Qiongshan District is relatively limited. The majority of the newly construction land in Meilan and Xiuying districts is situated in the northwest and southeast coastal areas of Haikou City. However, the construction land in Longhua and Qiongshan districts developed towards the south, with the majority of the land being used for township construction.

Figure 6 and Figure 7 present the simulation results of land use scenarios and the changes in major land uses in 2035. The figures illustrate the most notable changes in cultivated land, forestland, and construction land across the scenarios. The distribution of cultivated land is concentrated in the central and southern parts of Haikou City, and the cultivated land originally located in the northwestern part of Haikou City has been significantly degraded, particularly the cultivated land in the vicinity of the main urban area, which has largely disappeared. The cultivated land in the eastern region of Haikou City has been replenished to varying degrees. The most notable increase can be observed in the ED scenario. Forestland continues to constitute the majority of Haikou City, with the loss of forested areas occurring in a relatively scattered manner, primarily in the vicinity of cultivated land. The expansion of construction land is extensive in the northwest of Haikou City, with the entire northern part of the city along the coastline fully developed. The most significant expansion occurs under the ED scenario. Furthermore, the spatial consistency between the expansion of construction land and the reduction in cultivated land suggests that urban construction in Haikou City is primarily driven by the development of cultivated land, occurring predominantly in urban and rural settlements, the main urban areas, and the vicinity of emerging development zones. Wetland changes are relatively decentralized, with the majority transferring to construction land, cultivated land, and forestland. The transfer of wetland to construction land is concentrated in the northern part of Haikou City along the coastline and in the northwest corner of Xinhai Harbor. In contrast, the conversion to forestland and cultivated land is scattered in the central and southeastern parts of Haikou City. This may be related to the significant increase in agricultural activities in this area in the future. It has been demonstrated that agricultural production, the construction of water conservancy facilities, and other activities may result in the loss of original wetland habitats, thereby leading to localized degradation [39].

3.2. Analysis of Changes in Land Use Structure under Simulation Scenarios

Based on the results of MATLAB calculations, the data were imported into the PLUS model for spatial land use simulation, and the final land use simulation results were obtained for the BAU, ED, and EEB scenarios (

Table 5. Simulation results.

Land Use Type	BAU (hm2)	ED (hm2)	EEB (hm2)
Cultivated land	56,627.37	60,761.11	59,734.12
Forestland	101,500.11	89,885.87	102,592.32
Grassland	3439.26	2293.99	2317.58
Water	1858.59	2988.26	3030.97
Wetland	10,588.77	11,181.30	11,714.34
Construction land	48,610.08	60,932.44	48,663.53
Unutilized land	44.73	25.94	16.06

).

Table 6. Land use zoning area of Haikou under the simulated scenario (unit: hm²).

Year/ Scenario	Administrative District	Cultivated Land	Forestland	Grassland	Water	Wetland	Construction Land	Unutilized Land
2020	Longhua	7585.07	15,472.44	619.38	382.93	693.04	5430.63	95.3
	Meilan	18,152.76	19,510.71	733.89	1387.05	6515.54	9623.41	0
	Qiongshan	29,781.29	54,544.73	524.29	666.21	3039.27	4494.23	0
	Xiuying	12,774.09	22,187.98	1212.01	367.69	1687.19	10,404.86	0
	Total	68,293.2	111,715.86	3089.57	2803.89	11,935.03	29,953.13	95.3
BAU	Longhua	6054.97	13,949.60	692.13	256.06	575.56	8060.91	44.73
	Meilan	13,628.89	17,400.25	817.62	913.81	6076.74	15,858.98	0.00
	Qiongshan	30,102.26	50,225.48	585.13	442.94	2395.44	6562.86	0.00
	Xiuying	6841.25	19,924.79	1344.38	245.79	1541.03	18,127.33	0.00
	Total	56,627.37	101,500.11	3439.26	1858.59	10,588.77	48,610.08	44.73
ED	Longhua	6400.70	11,889.52	295.58	443.66	614.59	10,466.20	25.94
	Meilan	13,255.93	16,151.94	829.48	1418.94	6347.16	18,276.80	0.00
	Qiongshan	35,715.25	44,087.49	270.14	740.31	2618.05	9688.76	0.00
	Xiuying	5389.21	17,756.93	898.79	385.35	1601.50	22,500.67	0.00
	Total	60,761.10	89,885.88	2293.99	2988.26	11,181.30	60,932.44	25.94
EEB	Longhua	6398.71	14,089.55	385.14	490.68	643.33	8206.83	16.06
	Meilan	14,336.03	17,547.70	695.70	1383.96	6648.75	15,873.42	0.00
	Qiongshan	31,275.66	50,821.54	352.35	769.86	2747.86	6466.01	0.00
	Xiuying	7723.72	20,133.53	884.37	386.48	1674.40	18,117.27	0.00
	Total	59,734.13	102,592.32	2317.58	3030.97	11,714.34	48,663.53	16.06

illustrates the land use quantity structure of Haikou City administrative districts under the 2035 simulation scenario.

The cultivated land exhibited a declining trend in all the simulated scenarios. Among the scenarios, the BAU scenario exhibited the greatest loss, with a reduction of 11,673.36 hm². The ED scenario exhibited a slightly lower loss than the EEB scenario, with a reduction of 7539.62 hm². The EEB scenario exhibited the lowest loss, with a reduction of 6566.61 hm². With regard to administrative divisions, the greatest reduction in cultivated land is observed in the Xiuying District, followed by the Meilan and Longhua districts. Conversely, the Qiongshan District demonstrates a consistent increase in cultivated land.

The forestland in the simulation scenarios exhibited a declining trend, with the greatest reduction occurring in the periphery of the urban area and the central area of Haikou City. Among the scenarios, the ED scenario exhibits the greatest loss, with a reduction of 218,406.7 hm², while the BAU and EEB scenarios demonstrate a significantly lower loss than the ED scenario, with 102,264.3 hm² and 91,342.2 hm², respectively. With regard to administrative divisions, the loss of forestland is concentrated in Qiongshan District, with losses in Meilan District and Longhua District being approximately equal. The loss in Xiuying District is slightly higher than that in Meilan District and Longhua District.

Grassland maintained a decreasing trend under the modeled scenarios, but the scenarios varied greatly, with the loss of grassland in the ED and EEB scenarios being much higher than that in the BAU scenario. The ED scenario exhibited the greatest loss, with a decrease of 796.79 hm², followed by the EEB scenario, which exhibited a decrease of 773.2 hm², and the BAU scenario, which exhibited an increase of 348.48 hm². The areas of change in grassland are also significantly different across the scenarios. The loss of grassland under the ED and EEB scenarios is concentrated in the Xiuying and Longhua districts, while the increase in grassland under the BAU scenario is concentrated in the Xiuying and Longhua districts.

The water exhibits disparate trends across the simulation scenarios. The BAU scenario depicts a declining trend, accompanied by a loss of 945.72 hm². In contrast, the ED and EEB scenarios exhibit a marginal increase in the water, with an increase of 183.95 hm² and 226.66 hm², respectively. The expansion of the water is concentrated along the banks of the Nandu River, with the majority occurring within the Qiongshan District.

The extent of the wetland is projected to decline under the simulated scenarios, with the BAU scenario exhibiting the greatest loss, amounting to 1505.79 hm². This is followed by the ED scenario, which is estimated to result in a loss of 913.26 hm², and the EEB scenario, which is projected to result in a loss of 380.22 hm². The loss of wetland is concentrated in Meilan and Qiongshan districts, with Longhua and Xiuying districts experiencing a relatively lower degree of loss.

The construction land in the simulation scenarios exhibited a notable increase in size, with the ED scenario exhibiting the greatest growth, reaching 30,975.94 hm² and a total area of 60,932.44 hm². The scale of construction land in the BAU and EEB scenarios is relatively similar, with a total area of 48,610.08 hm² and 48,633.53 hm², respectively. In comparison with the 2020 scenario, the growth of construction land in these scenarios is 18,653.58 hm² and 18,677.03 hm², respectively. With regard to administrative divisions, the growth of construction land is concentrated in the Xiuying and Meilan districts, with the Xiuying District representing the primary growth area and the Meilan District the secondary growth area. The growth of construction land in Longhua District and Qiongshan District is relatively modest. In the BAU and EEB scenarios, the growth of Longhua District is slightly higher than that of Qiongshan District, while in the ED scenario, the growth of Qiongshan District is higher than that of Longhua District. The growth is slightly higher than that of Longhua District, but the overall construction levels of the two districts are roughly comparable.

The unutilized land shows a downward trend in the simulation scenario, with only a slight distribution in the Longhua District.

3.3. Analysis of Land Use Benefits under Simulation Scenarios

Figure 8, Figure 9 and Figure 10 illustrate the economic value of land use (ESV), as well as total carbon emissions, for Haikou City in 2020 and 2035 under each scenario. From an economic perspective, the ED scenario (883.395 billion yuan) is the most advantageous, followed by the EEB scenario (711.235 billion yuan) and the BAU scenario (709.417 billion yuan). In terms of meeting the planning objectives, the ED scenario is the most economically valuable, with an economic scale that far exceeds that of the BAU and EEB scenarios. As shown in the

Table 7. Economic value and ESV of land sources in Haikou from 2020 to 2035 (billion RMB).

Year/Scenario	Index	Cultivated Land	Grass Land	Water	Wetland	Construction Land	Unutilized Land	Grassland	Total
2020	Economic Value	241.50	4.58	24.00	10.21	44.03	4201.52	0.00	4525.85
	ESV	9.46	77.39	2.14	6.70	22.12	−12.65	0.00	105.16
BAU	Economic Value	200.23	4.16	26.71	6.77	38.55	6817.76	0.00	7094.17
	ESV	7.84	70.30	2.38	4.44	19.37	−20.53	0.00	83.81
ED	Economic Value	214.85	3.69	17.81	10.88	40.71	8546.02	0.00	8833.95
	ESV	8.41	62.26	1.59	7.14	20.45	−25.73	0.00	74.12
EEB	Economic Value	211.21	4.21	18.00	11.03	42.65	6825.26	0.00	7112.35
	ESV	8.27	71.06	1.60	7.24	21.43	−20.55	0.00	89.05

, in this scenario, cultivated land, wetland, and construction land contribute the most to economic growth, with the notable expansion of construction land becoming a key driver of high economic growth. However, construction land and cultivated land also have the most significant impacts on ecosystem services and carbon emissions. Furthermore, the scenario exhibits reduced economic efficiency, as evidenced by carbon emissions per unit of GDP, which stands at 39.95 tons, a figure that surpasses the BAU (37.43 tons) and EEB (37.21 tons) benchmarks. Conversely, the urban development patterns in both BAU and EEB scenarios are more moderate, and the overall land resource allocation is more similar. This demonstrates that the government’s development strategy is highly compatible with the concept of balanced economic ecology. Nevertheless, the net benefits of the EEB scenario are demonstrably superior to those of the BAU scenario. In the EEB scenario, the faster economic growth rate is maintained, while the scale of construction land is minimized, and the cultivated land, forestland, and wetland resources within the city are protected. This ultimately achieves a balance between the economic and ecological functions of land use and optimizes the value of ecosystem services and economic efficiency.

3.4. Influence of Diverse Driving Factors on LUCC

Land use changes in Haikou City are influenced by natural and social driving factors, which show different development probabilities in space. The probability of development and the influence of expansion drivers of different land use types in Haikou City was obtained through the Land Expansion Analysis Module (LEAS) of the PLUS model (Figure 11 and

Table 8. Influence of land use expansion drivers.

Type	Cultivated Land	Forestland	Grassland	Water	Wetland	Construction Land	Unutilized Land
Distance to attraction	0.07	0.12	0.14	0.05	0.14	0.16	0.04
DEM	0.07	0.06	0.08	0.22	0.19	0.06	0
GDP	0.1	0.19	0.08	0.11	0.11	0.1	0.03
Distance to settlement	0.09	0.08	0.06	0.04	0.08	0.08	0.08
Population density	0.12	0.09	0.1	0.06	0.09	0.13	0.1
Precipitation	0.16	0.13	0.11	0.04	0.08	0.09	0.02
Distance to river	0.07	0.06	0.09	0.23	0.11	0.09	0.17
Distance to road	0.1	0.09	0.1	0.09	0.09	0.15	0.57
Slope	0.05	0.06	0.05	0.02	0.03	0.02	0
Slope direction	0.06	0.06	0.01	0.01	0.02	0.01	0
Temperature	0.11	0.07	0.18	0.14	0.06	0.1	0

).

Changes in cultivated land are mainly influenced by rainfall, population density, and temperature. Comparison shows in Figure 12, it can be seen that the probability of cultivated land expansion is higher in areas with higher rainfall, lower population density, and moderate temperature conditions, and all three drivers are closely related to urbanization in the north. Rainfall and temperature are mainly affected by changes in the urban subsurface, with reduced evapotranspiration and heat dissipation, and the reduction in rainfall is accompanied by a significant urban heat island effect, which leads to a reduction in the capacity of localized cultivation. Population density mainly affects the distribution of cultivated land from the perspective of urban industrial structure. In general, areas with higher population density have a higher proportion of labor-intensive industries and a lower proportion of agriculture.

The change in forestland is mainly influenced by GDP, rainfall, and distance to attraction, and the probability of expansion is higher in areas with lower GDP, abundant rainfall, and far from the attraction. The economy of Haikou City is mainly based on the service industry, and the efficiency of forestry is relatively low, so the economic development of areas where forestland is concentrated is not high. Since forests can retain water and increase the humidity of the region, which in turn increases precipitation, it is generally believed that areas with high forest cover have more abundant rainfall, which is confirmed by the highly positive correlation between changes in forest cover and rainfall in this paper. The attraction usually implies the development of regional transportation and infrastructure, and these construction activities not only spatially encroach on part of the forest but also increase the human disturbance of the area, affecting the forest habitat.

The change in grassland is mainly affected by the distance to attraction, air temperature, and precipitation, and the probability of expansion is higher in the area near the attraction with abundant precipitation and moderate air temperature. Most of the natural grasslands in Haikou City are relatively scattered and are greatly affected by temperature and precipitation. The grasslands that are concentrated locally are mainly artificial grassland, including Mission Hills Scenic Area and Holiday Beach Scenic Area. These areas are tourist resorts with golf, so large areas of grass have been artificially created and planted.

Changes in water are mainly influenced by distance to the river, DEM, and GDP, with higher expansion probabilities in areas that are close to the river, have low DEM, and higher GDP. The water in this study mainly consists of canals, so it is highly consistent with the data on rivers in Haikou City. Elevation is the most critical factor influencing the river course and one of the determinants of channel change. In addition, the probability of river channel expansion is significantly higher in the main urban area with a higher GDP, which indicates that the urban area has achieved significant results in the coastal improvement of the Nandujiang River, a major river in the city, in recent years.

The expansion of construction land is mainly influenced by distance to road, distance to site, and population density, with a higher probability of expansion in areas with a dense road network, a large number of attractions, and a high population density. The direction of construction land expansion is mainly along the transportation network of urban roads, and with the increase in population density, the demand for construction land also increases sharply, thus promoting the development of construction land in relevant areas. Influenced by the well-developed service industry, the scale of construction around sites in Haikou City is also higher.

Overall, natural factors such as rainfall, temperature, DEM, and distance to rivers have the greatest influence. Complex topography and geomorphology are more restrictive to urban development and agricultural production, and these areas are more likely to be the focus of afforestation, cultivated land return to forest, and lake and wetland restoration. Social factors such as population density, GDP, and distance from roads and attractions have the greatest influence. Accelerating urbanization has led to massive migration of rural populations to cities, resulting in a slowdown in agricultural development and a surge in demand for land for urban development, as well as an increase in the degree of disturbance to natural ecosystems caused by human activities.

4. Discussion

4.1. An Optimization Method for Land Use Simulation Coupled with the GMOP-PLUS Model

Establishing an optimal land use allocation scheme is one of the basic tasks of land use planning, which usually needs to take into account the diverse development needs of cities [13]. However, a large number of land use simulation and prediction studies use a single model for simulation [40,41], which lacks integration with urban development policies and objectives, which often leads to the problem that the prediction is difficult to achieve the optimal results or fails to fit into the actual situation of the city. In this paper, we use the coupled GMOP algorithm and PLUS model to simulate the optimization of land use quantitative structure and spatial pattern from both top–down and bottom–up perspectives. The top–down process mainly considers the influence of macro policy on land resource allocation.

This approach takes into account the impact of macro policies on land resource allocation by incorporating urban development needs and restrictive indicators as constraints in the land optimization decision process through the GMOP algorithm. For the uncertain parameters in the constraints, the gray prediction model is used to supplement them. Then, the PLUS model is used to make reasonable spatial allocation of land use quantity optimization results according to the actual natural and social conditions of the region and to achieve the land use decision objectives by setting parameters such as conversion cost matrix. The PLUS model emphasizes the influence of nonlinear relationships on land use changes and is able to better identify and predict the evolution of natural patches. The study shows that the PLUS model, based on the land expansion analysis strategy and the principle of multi-type stochastic patch seeding, can better reflect the spatial and temporal difference patterns of LUCC. Therefore, compared with the traditional composite model, the GMOP model in the GMOP-PLUS model can be used to improve the overall efficiency of land use by determining the constraints, which can be used to allocate land resources to more efficient sites in simulation prediction, while the PLUS model allocates different land use types to spatial locations with high total probability under the constraints of the quantities. This method not only helps urban construction decision makers to foresee future land use changes and potential ecological and environmental impacts but also provides a scientific basis for policy formulation and implementation by generating a land use allocation scheme that maximizes comprehensive benefits.

4.2. Future Land Use Patterns

According to the simulation results, the future land use of Haikou City will increase in scale and speed year by year under continuous and strong anthropogenic influences, and the overall structure is unstable. The urban layout of Haikou City in 2035 will change from a single-core structure to a multi-core structure, and Changliu Cluster and Jiangdong New District will gradually become the new urban expansion centers with the expansion of the construction land to the east and west wings of the city. This trend is the same as the results of Wang Pai [42] and Wang Heng [43] and is also in line with the overall development strategy of “one main and three deputy” proposed in the 14th Five-Year Plan of Haikou City. The results show that the expansion hotspots under the three simulated scenarios are primarily attributable to the conversion of cultivated land to construction land. In comparison to other land uses, the loss of cultivated land resulting from future urban expansion is likely to be more serious. However, the impact of construction land is not confined to cultivated land, as the encroachment of construction land on cultivated land continues to intensify, the large-scale deforestation and expansion of land in some areas in order to achieve the balance of cultivated land occupation and replenishment will lead to a significant reduction in the area of forestland. Research has shown that the probability of cultivated land expansion in Haikou City is mainly affected by rainfall and temperature factors, while forestland plays an indispensable role in the regulation of the regional climate, and sacrificing forestry resources in exchange for an increase in the amount of cultivated land will not be beneficial to the long-term development of agriculture.

Therefore, it is important to strictly enforce land use controls to prevent potential urban sprawl in hotspot areas and to prohibit the encroachment of construction land on high-quality cultivated land. Where food security is threatened by the reduction in cultivated land, priority should be given to strengthening the protection and restoration of natural forest resources and improving low-quality and low-efficiency forests. In addition, in hotspots of urban expansion, priority should be given to improving the use of existing built-up areas, such as the redevelopment of shantytowns, urban villages, and abandoned factories [44], in order to avoid a “piecemeal” pattern of urban sprawl [45].

4.3. Recommendations for Optimizing Land Use Development Patterns

As the simulation results show, the ecological land in Haikou City will still suffer some loss in the future. The economy of Haikou City is mainly service-oriented [42], and the related industries are highly dependent on urban construction land [46]. The scale and agglomeration effect of the service industry will lead to the large-scale expansion of construction land, which will exacerbate the contradiction between the demand for local land functions and the decline of ecological functions [25]. Therefore, this paper constructs a balanced development scenario from the perspective of balancing ecological environment and economic development. Compared with the natural development scenario, the balanced development scenario obtained under the multi-objective planning framework significantly increases the economic benefits while preserving more ecological land and supports larger economic growth with lower land resource consumption, forming a trend of balancing economic benefits and ecological benefits. To realize this scenario, planners should focus on the optimal allocation of the carrying capacity of ecological services and economic development goals, aiming to build efficient and intensive urban living space, and reasonably control the decline of ecosystem services caused by the large-scale expansion of construction land, so as to achieve a win–win situation in terms of both economic and environmental benefits. However, the reduction in ecological land due to urban development is inevitable, and the function of ecological security barrier and sustainable economic development can only be fully realized by changing the land use development mode, building a green ecological industrial system, and promoting the diversified development of ecosystem structure and function.

4.4. Problems and Prospects

This study proposes a method to optimize land use simulation, but there are still shortcomings in the process of simulation. (1) In constructing the GMOP model, some uncertain parameters are predicted using the gray prediction model, which has a certain impact on the accuracy of the parameters due to the relatively high difficulty in obtaining certain socio-economic data and the difficulty in obtaining continuous or up-to-date statistics. In future research, it is hoped that the source of basic data can be optimized as much as possible, so that the parameters can be set more accurately. (2) In this paper, the first level of land use classification is used to assess the value of ecosystem services in Haikou City, which has some limitations in small- and medium-sized studies. With further refinement of the land use classification, it can reduce problems such as insufficient precision due to complex terrain or missing data, while improving the accuracy of the assessment. Subsequent studies should focus on subdividing the land use classification to more systematically reflect the internal structure and functional changes in regional ecosystem services. (3) Some of the driver data selected in this paper have low resolution or some grids, such as the watershed part, are missing, and this paper mainly adopts sample point collection and kriging interpolation to supplement the missing parts, but this method may also lead to the bias of data in local areas. With the continuous development of geographic information technology, subsequent studies can gradually replace these missing data.

5. Conclusions

This study integrated the GMOP model and PLUS model for land use simulation and conducts a multi-dimensional assessment of different simulation results from the perspectives of economic benefits, ecological benefits, and carbon emissions, to solve the problem of land use trade-offs under different needs. The Economic and Ecological Balance scenario was constructed by integrating anticipated development policies with the current status of the ecological environment.

(1)

In terms of spatial distribution, the urban spatial form of Haikou City in 2035 will undergo a significant transformation, shifting from a single-core structure to a multi-core structure. This will result in the emergence of new urban centers on the eastern and western peripheries of the main city. Cultivated land and forestland continue to be subjected to the most significant anthropogenic disturbances. Cultivated land has become the primary source of land for construction, resulting in the near-elimination of cultivated land in the northern part of Haikou City. In contrast, forestland has experienced a notable reduction in extent in the central and southeastern regions of Haikou City.

(2)

Quantitatively, the simulation scenarios indicate a decreasing trend for cultivated land, forested grassland, and wetland, while construction land exhibits significant growth. A comparison of the three simulation scenarios reveals that cultivated land, wetland, and water exhibit the greatest declines under the BAU scenario. Conversely, forested grassland demonstrates the highest loss, while construction land exhibits the most significant growth under the ED scenario.

(3)

A comparison of the land use benefits under the three simulation scenarios reveals that the ED scenario exhibits a significantly increased economic total, yet concomitantly, the ecological function is more severely degraded and the carbon emission is considerably higher. The expansion of construction land is the most significant in this scenario, with the area converted to construction land exceeding that in other scenarios for all land use types except water and wetland. The BAU scenario, which is based on historical trends, and the EEB scenario, which pursues economic and ecological synergistic development, are more closely aligned with each other. However, the EEB scenario is more economically and ecologically efficient, and the land resource allocation pattern is more conducive to the preservation of natural land types. With the exception of grassland, the conversion of all land types to construction land and cultivated land is effectively controlled. This results in a reduction in the fluctuation of the area of forestland, wetland, water, and unutilized land, thereby creating a more stable ecological environment.