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 (), forestland (), grassland (), water (), wetland (), construction land (), and unutilized land () 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 (hm
2).
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:
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/hm
2 [
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:
where
is the total carbon emissions from construction land;
is the energy consumption per unit of GDP (tons of standard coal/million yuan);
, which is the carbon emission coefficient of coal consumption;
and
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/hm
2 in 2030 by GM(1, 1)
2. In summary, the carbon emission constraint formula is obtained as follows:
In the formula, denotes the total GDP of Haikou City in 2030, which is predicted by GM(1, 1), and is 70,469,643,300,000 Yuan. 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:
where
is the projected grain production per unit area in 2035 (6.2 t/hm
2),
is the share of grain cultivation (0.26),
is the replanting index, taking the average value from 2010 to 2020 (1), and
is the projected total population in 2035 (4,742,800),
represents the per capita food demand (0.52 t) as predicted by Xin et al. [
34], and
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 hm
2. To summarize, the formula for the total amount of forestland limit is obtained as follows:
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:
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.
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.
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:
Limitations on the range of values: All land use variables must take on non-negative values.
- (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.
and represent the state of the land at the time t and t + 1, respectively; 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). Therefore, the objective function of the ED scenario with maximized economic efficiency is as follows:
where
is the output value per unit area of land use type
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:
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). The ESV value formula is obtained as follows:
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.
where
denotes the weight value and
denotes the land use area;
and
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.