Demarcation of Future Urban Rigid and Elastic Development Boundaries of the City of Haikou

Abstract

Accelerated urbanization both promotes the rapid development of social economy and leads to a series of disadvantages, such as the excessive consumption of resources, environmental pollution, and food security threats. It is thus necessary to reasonably demarcate future urban development boundaries. Therefore, both the external supply and the elastic space of urban internal development need to be considered. In the present study, the current urban boundaries were first identified. Then, the urban rigid and elastic development boundaries in the next three decades were obtained by employing the minimal cumulative resistance (MCR) and CA-Markov models. Lastly, some suggestions were put forward for the implementation of future urban development boundaries. The results were the following. (1) The areas of the current urban boundaries of Haikou in 2000, 2010, and 2020 were 93.71, 124.26, and 260.41 km², respectively. (2) By using the MCR model, the urban rigid development boundaries of Haikou in 2030, 2040, and 2050 were 361.27, 480.17, and 505.22 km², respectively. (3) By using the CA-Markov model, the areas of urban elastic development boundaries in 2030, 2040, and 2050 were 381.86, 483.95, and 536.06 km². (4) The increased elastic expansion space of urban development of Haikou while meeting the rigid constraint conditions in 2030, 2040, and 2050 was 20.59, 3.78, and 30.84 km², respectively. (5) Suggestions need to be put forward on the implementation of future urban development boundaries from the aspects of technology, policy, and management. The results of demarcating the urban rigid and elastic development boundaries can not only prevent the excessive urban expansion and ensure the orderly, efficient and sustainable development of the city, but also more effectively protect important ecological resources, which could provide quantitative reference and decision-making basis for regional territorial space planning.

1. Introduction

Over the past 40 years of reform and opening up, China’s Global Domestic Product (GDP) has grown rapidly from 367.87 billion CNY in to 101.6 trillion CNY. One of the important manifestations of China’s rapid economic development is the continuous acceleration of urbanization. Urbanization refers to the process of social, economic, and regional spatial-structure changes caused by the continuous increase in urban population as the agricultural population gathers in cities during the development of human society. Since the reform and opening up, as the capital city of Hainan Province, the process of urbanization of Haikou has also been accelerating, and the scale of urban built-up land has expanded rapidly. From 1978 to 2018, the urban population of Haikou increased from about 709,000 to 2.3 million, the urbanization rate increased to 78.64%, and the area of urban built-up expanded to 151.6 km² [1]. Although urbanization is necessary to realize modernization, accelerate industrial transformation, and promote regional coordinated development, the rapid development of urbanization also leads to a series of disadvantages. On the one hand, the urban sprawl brings urban issues such as traffic jams, high housing prices, and environmental deterioration [2]. On the other hand, the continuous expansion of urban land leads to the extensive development and utilization of land resources, and a large number of high-quality cropland and ecological land areas are transformed into built-up land, resulting in frequent changes in land use and land cover, which leads to an increasingly fragile and unstable ecosystem, and induces a series of ecological and environmental problems [3]. Therefore, the scientific and reasonable demarcation of future urban development boundaries can control the infinite expansion of cities, coordinate the contradiction between urban ecological protection and spatial growth demand, and ensure the sustainable development of cities [4]. The demarcation of future urban development boundaries is now the focus of China’s territorial space planning [5]. China has introduced a series of policies and measures. For example, in 2019, China issued a policy document titled “Guidelines on Designating and Implementing Three Lines of Control in Territorial Space Planning”, which called for demarcating urban development boundaries in accordance with the requirements of intensive, moderate, and green development to prevent urban sprawl and leave room for future development. In 2021, the “Master Plan for Territorial Space of Haikou City (2020–2035)” was issued, which also advocated that future urban development boundaries should be reasonably delimited on the basis of the current situation of urban development, combined with suitability evaluation, considering the resource carrying capacity, population distribution, and economic layout.

Urban development boundaries, also known as urban space growth boundaries, were first proposed in Howard’s theory of the “pastoral city” [6]. Chinese scholars believe that an urban development boundary is a comprehensive line which aims at maximizing social, economic, ecological, and other comprehensive benefits, and gives consideration to urban construction land development and ecological protection. It must have both rigid and elastic characteristics. A rigid boundary is areas that cannot be occupied by urban built-up land, the ecological bottom line that cannot be exceeded by urban expansion, and the upper limit of ecological environment capacity and resource carrying capacity, which ensures the safety of the natural resources and ecological system of the region [7]. It focuses on the supply of external conditions, such as natural, social, and economic conditions, for urban development, that is, various resistance factors affect urban development. The elastic boundary is the maximal expansion capacity to meet the reasonable layout of urban land for some time to come. It is also the development line of urban internal demand, which is conducive to the intensive, economical, and efficient use of land resources [8]. It pays more attention to the spatial development inside the city and represents the reserved elastic space for urban development in the future. Theoretically, the elastic boundary is dynamic, and can be evaluated and adjusted according to the spatial needs of urban development at different stages, while the rigid boundary is more permanent.

Research on the demarcation of urban development boundaries has a long history. Scholars have also explored many research methods for the demarcation of elastic and rigid boundaries. On the basis of the principle of reserving elastic space for urban expansion, the demarcation of elastic boundaries is mainly determined by combining quantity and spatial distribution. Researchers mainly build models to study the dynamic changes of urban space and predict the elastic boundaries. For example, Moghadam et al. [9], Goodarzi et al. [10], Dadashpoor et al. [11], and Zhang et al. [12] used the CA, CLUE-S, and SLEUTH models to simulate land use changes and predict urban growth. Based on the DDM and DIM models, Tayyebi et al. constructed a UGBM model to demarcate the urban boundary of the Tehran metropolitan area [13]. Rigid boundaries of urban growth mainly aim at protecting the important ecological resources of the city and determine the “rigid” boundaries that cannot be crossed by urban growth. Some scholars directly selected important ecological land such as nature reserves, drinking water land, or concentrated patches of woodland within the region as rigid growth boundaries, which was usually not scientific [7]. Some scholars established an index evaluation system and adopted spatial analysis methods such as ecological suitability evaluation and ecological sensitivity analysis to screen ecological dominant patches and demarcate regional rigid boundaries. For example, Zhong et al. demarcated the urban development boundary through the evaluation model of ecological space occupancy [14]. Ren et al. studied the demarcation of urban development boundaries of Jiayuguan in 2020 and 2030 on the basis of remote-sensing images and the MCE-CA model of geographic optimization simulation system [15]. In addition, the minimal cumulative resistance (MCR) model is widely used to predict the rigid boundaries of cities. Bai demarcated the rigid urban development boundaries of Simao District, Pu’er City, Yunnan Province, based on the MCR model [16]. Yi et al. used the MCR model to delimit the boundary of construction land expansion in Jiangxi Province [17].

Therefore, previous studies either focused only on controlling urban development and protecting important natural resources and ecological space in the region, demarcating urban boundaries from the perspective of “rigidity”, or focused on the spatial development of cities themselves, demarcating urban development boundaries from the perspective of “elasticity”. The above research of the demarcation of urban development boundaries in theory and practice has progressed and can provide important support for this study. However, the demarcation of urban development boundaries needs the integrated use of a rigid- and elastic-boundary delimitation method. It is necessary to identify the permanent “rigid” bottom line that can control urban development, protect regional important natural resources and ecological space, maintain the stability of the ecosystem, and identify the dynamic “elastic” boundary of urban development that can predict the space suitability for urban expansion in the next stage, and meet the needs of urban development in quantity and space [7,18]. There are few studies on jointly delimiting the urban elastic and rigid growth boundaries.

This study compensated for the shortcomings of the above research. Based on the suitability evaluation, prediction model, and the GIS spatial analysis method, the urban development boundaries of Haikou in the next few decades were demarcated from the perspective of elasticity and rigidity. The objectives of this study were the following: (1) on the basis of Landsat remote-sensing images, the current urban boundaries of Haikou city in 2000, 2010, and 2020 were identified; (2) based on the current urban boundaries, the MCR model was used to define the development boundaries of Haikou from the perspective of rigidity; (3) the CA-Markov model was employed to define the development boundaries from the perspective of elasticity; (4) suggestions were put forward according to the results of rigid and elastic boundaries. The result of demarcating urban boundaries can not only prevent excessive urban expansion, and ensure the orderly, efficient, and sustainable development of the city, but also protect important ecological resources more effectively. It is of great practical significance to promote the healthy development of Haikou and the harmony and stability of society.

2. Materials and Methods

2.1. Study Area

In this study, Haikou was chosen as the study area. Haikou is the capital city of Hainan Province, the fulcrum city of the national “One Belt and One Road” strategy and the core city of the Hainan Free Trade Port [19]. It extends from 19°31′44″ N to 20°04′52″ N and 110°07′40″ E to 110°42′39″ E, and is located at the north of Hainan Island, separated from the continental plate by the Qiongzhou Strait (Figure 1). Haikou is composed of the northern Hainan, Haidian, and Xinbu islands. It includes the Xiuying, Longhua, Qiongshan, and Meilan districts, with a total area of 3145.93 km², divided into 2284.49 km² of land area and 861.44 km² of sea area. The length of the coastline is about 140 km. The terrain of Haikou is relatively gentle, and the overall trend is high in the south and low in the north. The annual average temperature is 23.8 °C, the annual accumulated precipitation is about 1800 mm, and the annual average sunshine duration exceeds 2000 h [20]. Nandu River, the longest river in Hainan Island, runs through the middle of Haikou.

2.2. Data Sources and Preprocessing

The data in this study included Landsat remote-sensing images and land cover data in 2000, 2010, and 2020, terrain data, the spatial distribution data of roads and railways, and the spatial distribution data of population density and GDP covering Haikou.

Landsat remote-sensing images covering Haikou came from Landsat TM images in 2000, Landsat ETM images in 2010, and Landsat OLI images in 2020 downloaded from the Geospatial Data Cloud platform (http://www.gscloud.cn/, accessed date: 13 November 2021). The spatial resolution was 30 m for all. When downloading the images, cloud coverage had to be less than 10%. After the pretreatment of radiometric calibration and atmospheric correction, the near-infrared (NIR), red, and green bands were given red, green, and blue, respectively, to synthesize false color images.

Land cover data covering Haikou came from the global land cover dataset provided by the GlobeLand30 platform. This dataset is an important achievement of the Global Land Cover Remote Sensing Mapping and Key Technologies research project of the National High Technology Research and Development Program of China (863 Program), including global land cover types in 2000, 2010, and 2020 (http://www.globallandcover.com, accessed date: 14 November 2021) [21]. Spatial resolution was 30 m. When verifying the accuracy of land cover data in 2010, 80 maps were extracted from 853 maps around the world, and more than 150,000 samples were set up. The overall accuracy was 83.50%, and the kappa coefficient was 0.78. In 2020, the overall accuracy was 85.72%, and the kappa coefficient was 0.82. The land cover types in the dataset were cultivated land, woodland, grassland, shrub land, wetland, water body, tundra, artificial surface, bare land, glaciers, and permanent snow. In this study, the land cover types covering Haikou were divided into six classes, namely, cultivated land, woodland, grassland, wetland, water body, and artificial surface.

Terrain data were derived from the Digital Elevation Model (DEM) data provided by the Shuttle Radar Topography Mission (SRTM) system with a spatial resolution of 90 m. Slope data were further processed by ArcGIS.

Spatial distribution data of roads and railways were derived from the vector spatial distribution data of roads and railways provided by the platform of Resources and Environment Science and Data Center (https://www.resdc.cn/, accessed date: 1 December 2021), and further processed by IDRISI Selva software into the distance data of roads and railways.

Spatial distribution data of population density in 2020 came from the global population density dataset released by the Worldpop platform (https://www.worldpop.org/, accessed date: 23 November 2021). The spatial distribution data of GDP came from the Resources and Environment Science and Data Center (https://www.resdc.cn/, accessed date: 27 November 2021). As the spatial distribution data of GDP in 2020 had not yet been produced, the spatial distribution data of GDP in 2015 were used instead.

2.3. Methods

2.3.1. Identification of Current Urban Boundaries of Haikou

In this study, we identified the current urban boundaries of Haikou. First, using land cover data of 2000, 2010, and 2020 as the data sources, ArcGIS software was employed to extract the artificial surface (i.e., built-up land) patches in Haikou of 2000, 2010, and 2020 in sequence. Second, as the built-up land included urban, rural residential, industrial, and mining land, the Landsat remote-sensing images of 2000, 2010, and 2020 were used as reference images to screen the artificial surface patches of the three years and obtain the urban patches of Haikou. Third, the urban envelope surfaces of Haikou in 2000, 2010, and 2020 were determined as the current urban boundaries of Haikou. Fourth, we calculated the area inside the current urban boundaries of Haikou, and then identified the direction of urban expansion from 2000 to 2020.

2.3.2. Demarcation of Future Urban Rigid Development Boundaries

First, the MCR model was used in this study to define future urban rigid development boundaries of Haikou. The MCR model was first proposed by Dutch ecologist Knaapen in 1992 [22]. Its essence was the minimal resistance to be overcome from the source to the target location, which reflected a kind of accessibility [23]. After modification by Chinese scholars, this model was mostly employed for the distribution suitability of rural settlements [24,25] and ecological security pattern construction [26,27] in many studies. As the rigid boundary is the area that cannot be occupied by urban construction land, the ecological bottom line that cannot be exceeded by urban expansion, and the upper limit of ecological environment capacity and resource carrying capacity, urban development is affected by many resistance factors. Therefore, the MCR model is now also mostly used to demarcate future urban rigid development boundaries. The calculation formula is as follows:

$$\mathrm{MCR}=f_{\min }{\displaystyle \sum }\left(D_{ij}\times R_i\right)$$

(1)

where MCR is the minimal cumulative resistance, Dij represents the spatial distance from source j to grid cell i, Ri represents the resistance coefficient of landscape unit i to space movement, f represents the positive correlation between minimal cumulative resistance and ecological process, ∑ represents the distance and resistance accumulation across all units between source j and unit i, and min means that unit i takes the minimal cumulative resistance for different sources.

In this study, the MCR model was used to determine the urban rigid development boundaries under existing natural conditions, and social and economic, and traffic-location factors according to the process of determining expansion sources, analyzing expansion resistance factors, constructing expansion resistance surfaces, and dividing expansion suitability grades. First, the spatial distribution of built-up land in Haikou in 2020 was taken as the expansion source. Then, we selected the resistance factors of urban expansion, including three kinds of natural condition factors (land cover classification, elevation, and slope), two kinds of traffic location factors (distance from highway and distance from railway), two kinds of socioeconomic factors (population density and GDP), divided the grades, assigned the scores and weights, constructed the resistance surface of urban expansion, and delimited five suitability grades of urban expansion (very suitable, suitable, relatively suitable, unsuitable, very unsuitable). Then, the regions of “very suitable” and “suitable” grades that could be used for built-up land development were determined. Lastly, combining with the area change of urban built-up land in Haikou from 2000 to 2020 and future available land for built-up land development, the outer envelope surfaces in 2030, 2040, and 2050 were determined as the urban rigid development boundaries. The process of demarcating future urban rigid development boundaries of Haikou with the MCR model is shown in Figure 2.

2.3.3. Demarcation of Future Urban Elastic Development Boundaries

In this study, the CA-Markov model, which combined the Markov and cellular-automata models, was used to simulate future land cover scenarios of Haikou and demarcate future urban elastic development boundaries of Haikou. Simulating future land use and cover scenarios is the process of calculating land use and cover changes over a known period and extrapolating them to future land use and cover changes [28,29,30,31]. Since the elastic boundary is the maximal expansion capacity to meet the reasonable layout of urban land in a certain period of time in the future, urban development is paid more attention. This is the development line of urban internal demand. The CA-Markov model is based on the law of urban quantity and spatial development in the past period of time and simulates the law of urban development in the future on the basis of various driving factors and limiting factors. Therefore, it is suitable for demarcating future urban elastic development boundaries.

Using the CA-Markov model, according to process of making the rules, simulating, verifying the accuracy, andpredicting the results, the urban elastic development boundaries of Haikou in the next three decades (2030, 2040, and 2050) were demarcated by simulating the spatial location of urban built-up land in Haikou. First, using land cover data of Haikou in 2000 and 2010, and on the basis of analyzing the driving factors (including terrain, traffic location, and socioeconomic factors) and limiting factors (spatial distribution of water body) of land cover conversion, the suitability images of land cover conversion were made to simulate the land cover scenario of Haikou in 2020. Then, the simulated land cover scenario in Haikou in 2020 was compared with the actual land cover data in 2020 to verify the accuracy of the CA-Markov model. Next, land cover scenarios of Haikou in 2030, 2040, and 2050 were simulated. Lastly, artificial surface patches were extracted and scattered, broken patches were removed, and patches with a high aggregation degree were retained. The outer envelope surfaces of urban built-up land in 2030, 2040, and 2050 were determined as future urban elastic development boundaries. The process of demarcating future urban elastic development boundaries of Haikou with the CA-Markov model is shown in Figure 3.

2.3.4. Suggestions on Implementation of Future Urban Development Boundaries

According to the demarcation results of urban rigid and elastic development boundaries of Haikou in the next three decades, this study puts forward some suggestions on the implementation of future urban development boundaries in Haikou from the three aspects of technology, policy, and management.

3. Results

3.1. Identification of Current Urban Boundaries of Haikou

Taking land cover data of Haikou in 2000, 2010, and 2020 as data sources, artificial surface (i.e., built-up land) patches were extracted by using ArcGIS to obtain the spatial distribution of built-up land in Haikou in the three years (Figure 4). From 2000 to 2020, the area of built-up land in Haikou showed a trend of continuous increase. According to the statistics, the areas of built-up land in Haikou in 2000, 2010, and 2020 were 120.39, 134.35, and 333.76 km², respectively.

Although built-up land in Haikou covered the urban built-up area to a certain extent, there was still redundant land use information belonging to industrial and mining land, and the built-up land of surrounding villages and towns, which could not extract the complete and closed boundary of the urban built-up area. Therefore, by referring to Landsat remote-sensing images covering Haikou in 2000, 2010, and 2020, and combining them with the spatial distribution of built-up land in Haikou, the outer envelope surfaces of urban built-up area in Haikou in 2000, 2010, and 2020 were obtained by manual visual interpretation (Figure 5). According to the statistics, the urban built-up areas of Haikou were 93.71, 124.26, and 260.41 km², respectively, and the increase rates for urban built-up areas of Haikou were 3.05 km² per year and 13.62 km² per year, respectively, in the two decades.

Figure 5 shows that, in 2000 and 2010, most of the urban built-up land was distributed in the Longhua and Qiongshan districts, presenting a long and narrow coastal line in the west and a massive distribution in the east, and the main part of the urban expansion trend from 2000 to 2010 was southward. This was consistent with the fact that the density of buildings in the main urban area of Haikou increased and the scope expanded in this decade. From 2010 to 2020, the urban built-up area of Haikou rapidly expanded from the coastal zone to the inland, making the main part of urban built-up area rapidly expanded to the Xiuying and Meilan districts. The reason was that Haikou proposed to develop a new urban area centered on the town of Haixiu in 2013, and planned to focus on the construction of commercial, medical, and service facilities to improve the urban functions of Xiuying district and drive the development of the surrounding regions. Therefore, the construction of the new government office area and the Hainan International Convention and Exhibition Center in the coastal area in the northwest of Haikou (northwest of the town of Changliu) began, which started the boom of the development of Changliu, leading to the built-up land around it growing rapidly in the following years. At the same time, in 2010, Haikou actively accelerated the construction of the east coast and the new district of Jiangdong. Therefore, the urban built-up area which was originally only distributed on the western side of the Nandu River spread rapidly to the east side, connecting with Meilan Airport to become a new urban area.

3.2. Demarcation of Future Urban Rigid Development Boundaries of Haikou Based on MCR Model

3.2.1. Analysis of Resistance Factors of Future Urban Development in Haikou

In this study, seven factors were selected as the resistance factors of future urban development in Haikou from the three aspects of natural condition, traffic location, and social and economic factors, including land cover types, elevation, slope, distance from highways and railways in 2020, population density in 2020, and GDP in 2015. In general, the lower the elevation is, the lower the slope, the more convenient the traffic, the greater the population density, the more developed the economy, the less the resistance of urban expansion, and the more suitable for urban development. Conversely, the greater the resistance to urban expansion is, the more unsuitable for urban development it is. Thus, the grades and scores of natural condition factors were determined by referring to existing studies, and transportation factors, population density, and GDP were graded and scored by Natural Breaks method (

Table 1. Grades, scores, and weights of resistance factors.

Resistance Factors	Grades	Scores	Weights	Resistance Factors	Grades	Scores	Weights
Land cover types	Artificial surface	5	0.1866	Distance from the railway/m	0–4947.03	5	0.0780
	Cropland	4			4947.03–10,805.35	4
	Woodland and grassland	3			10,805.35–17,184.72	3
	Wetland	2			17,184.72–23,433.29	2
	Water body	1			>23,433.29	1
Elevation/m	0–40	5	0.1645	Population density/(People per km2)	<1898	1	0.1650
	40–80	4			1898–6655	2
	80–120	3			6655–16,614	3
	120–160	2			16,614–41,185	4
	>160	1			>41,185	5
Slope/°	0–2°	5	0.1645	GDP/(10,000 yuan per km2)	<2715	1	0.1650
	2–6°	4			2715–5708	2
	6–15°	3			5708–10,102	3
	15–25°	2			10,102–23,355	4
	>25°	1			>23,355	5
Distance from the highway/m	0–886.40	5	0.0778
	886.40–1894.76	4
	1894.76–3097.42	3
	3097.42–4664.01	2
	>4664.01	1

). The higher the scores are, the less the resistance of urban expansion is. Next, the Analytical Hierarchy Process (AHP) was used to determine the weights of seven resistance factors (

Table 2. Analytic Hierarchy Process matrix.

	Land Cover Types	Elevation/m	Slope/°	Distance from the HIGHWAY/m	Distance from the Railway/m	Population Density/(People per km2)	GDP/(10,000 Yuan per km2)
Land cover types	1	——	——	——	——	——	——
Elevation/m	1/2	1	——	——	——	——	——
Slope/°	1/2	1	1	——	——	——	——
Distance from the highway/m	1/3	1/2	1/2	1	——	——	——
Distance from the railway/m	1/3	1/2	1/2	1	1	——	——
Population/ (People per km2)	1	1	1	2	2	1	——
GDP/(10,000 yuan per km2)	1	1	1	2	2	1	1

).

3.2.2. Suitability Evaluation of Future Urban Development in Haikou

First, the above seven resistance factors were classified according to the grades in Table 1 and are shown in Figure 6. Then, the resistance surface was obtained by calculating the weights of the seven resistance factors. Next, the Cost Distance tool in ArcGIS software was used to combine the expansion source, built-up land of Haikou in 2020 with the resistance surface to obtain the accumulated cost of urban expansion. Lastly, the natural breaks method was used to classify the suitability grades of future urban development in Haikou according to the cumulative resistance values: very suitable, suitable, relatively unsuitable, unsuitable, and very unsuitable (Figure 7). The area and proportion of five suitability grades are shown in

Table 3. Area and proportion of suitability grades.

Levels	Area (km2)	Rates (%)
Very suitable	702.34	33.34
Suitable	584.00	27.72
Relatively suitable	449.79	21.35
Unsuitable	263.89	12.53
Very unsuitable	106.43	5.05

.

Combined with Figure 6 and Figure 7 and Table 3, regions that were very suitable and suitable for urban development were mainly located in the surrounding areas of the original built-up land and the distribution areas of the nearby cropland, and areas with low elevation, small slope, developed transportation, and large population density and GDP, which were 702.34 and 584.00 km², respectively. The rates were 33.34% and 27.72%, respectively. The regions that were very unsuitable and unsuitable for urban development were mainly located in the areas of woodland, grassland, and wetland, and the areas with high elevation, large slope, small population density, and less developed economy, accounting for 12.53% and 5.05%, respectively.

3.2.3. Future Urban Rigid Development Boundaries of Haikou

According to the results of demarcating future urban development suitability grades in Haikou, the regions of very suitable and suitable grades can be supplied with built-up land expansion. We selected very suitable and suitable areas for urban development, selected patches with high aggregation degree and large area, and eliminated scattered and broken patches. Then, we combined with the distribution of urban built-up area in Haikou in 2020 and the expansion of urban built-up areas from 2000 to 2020 and determined future urban built-up areas in Haikou in 2030, 2040, and 2050. The outer envelope surfaces were used as future urban rigid development boundaries (Figure 8). Figure 8 shows that the urban built-up areas of Haikou were divided into two main areas by Nandu River. Based on the MCR model, the urban built-up areas of Haikou would continue to expand from 2030 to 2050, mainly showing a trend to expand southward along the east coast. In addition, the main areas of urban built-up land on the left also showed the trend to expand southward.

Figure 9 shows changes in urban built-up land and internal area of development boundaries in Haikou from 2030 to 2050 simulated by the MCR model. In 2030, 2040, and 2050, the areas of urban built-up land in Haikou are 343.26, 403.88, and 422.59 km², respectively, and the areas of urban rigid development boundaries are 361.27, 480.17, and 505.22 km², respectively. In the next two decades, the areas of urban rigid development boundaries would increase by 118.90 and 25.05 km², respectively. As the urban built-up area of Haikou was 260.41 km² in 2020, the reserved space for urban development by 2030, 2040, and 2050 would be 100.86, 219.76, and 244.81 km², respectively, according to the rigid development boundaries.

3.3. Demarcation of Future Urban Elastic Development Boundaries of Haikou Based on the CA–Markov Model

3.3.1. Simulation of Land Cover Scenario of Haikou in 2020

First, the Markov module in IDRISI Selva software was used to calculate the transition matrices of various land cover types in Haikou from 2000 to 2010, including the transition area matrix and transition probability matrix (

Table 4. Land cover transition probability matrix from 2000 to 2010.

		2010
		Cropland	Woodland	Grassland	Wetland	Water Body	Artificial Surface
2000	Cropland	0.6306	0.3424	0.0111	0.0003	0.0048	0.0107
	Woodland	0.0170	0.8073	0.0965	0.0029	0.0323	0.0440
	Grassland	0.0001	0.4488	0.4025	0.0156	0.0052	0.1279
	Wetland	0.1398	0.2246	0.0901	0.4866	0.0443	0.0145
	Water body	0.1075	0.2827	0.0176	0.0181	0.5645	0.0096
	Artificial surface	0.0547	0.1397	0.0154	0.0000	0.0062	0.7840

). The proportional error of normal distribution was set to 0.15.

When the CA-Markov model was used to simulate future land cover scenarios of Haikou, if the transition area matrix and transition probability matrix generated by Markov module were only used to simulate the quantity and spatial distribution of each land cover type, it would lead to low simulation accuracy. Therefore, it was necessary to determine the conversion rules of each land cover type according to various driving factors and limiting factors to produce suitability images. Driving factors are various factors that promote or inhibit the mutual transformation of various land cover types, including the elevation and slope of Haikou, the spatial distribution of population density in 2010 and 2020, the spatial distribution of GDP in 2010 and 2015, and the distance from railways and highways. Limiting factors indicate that the spatial distribution area cannot be transferred to other land cover types. Since it was difficult for water bodies to transfer to some land cover types, the spatial distribution of water bodies in Haikou was selected as the limiting factor in the four land cover scenario simulations. The driving factors and limiting factors of mutual transformation of various land cover types are shown in

Table 5. Driving factors and limiting factors of land cover scenario simulations of Haikou in 2020.

Land Cover Types	Driving Factors						Limiting Factors
Cropland	Elevation, decrease	Slope, decrease	——	——	Distance from highway, decrease	Distance from railway, decrease	Water body in 2010
Woodland	Elevation, increase	Slope, increase	——	——	Distance from the highway, decrease	Distance from the railway, decrease	Water body in 2010
Grassland	Elevation, decrease	Slope, decrease	——	——	Distance from the highway, increase	Distance from the railway, increase	Water body in 2010
Wetland	Elevation, decrease	Slope, decrease	——	——	Distance from the highway, increase	Distance from the railway, increase	——
Water body	——	——	——	——	——	——	——
Artificial surface	Elevation, decrease	Slope, decrease	Population density, increase	GDP in 2010, increase	Distance from highway, increase	Distance from railway, increase	Water body in 2010

.

Next, the AHP method was used to determine the weights of the driving factors according to the impact on land cover types transformation (

Table 6. Weights of driving factors for land cover type transformation.

Driving Factors	Elevation	Slope	Population Density in 2010	GDP in 2010	Distance from the Highway	Distance from the Railway
Cropland	0.1223	0.4236	—	—	0.2270	0.2270
Woodland	0.2000	0.4000	—	—	0.2000	0.4000
Grassland	0.3333	0.3333	—	—	0.1667	0.1667
Wetland	0.3750	0.3750	—	—	0.1250	0.1250
Artificial surface	0.2678	0.2679	0.1692	0.1333	0.0810	0.0810

). Since water bodies are generally difficult to transfer to some land cover types, water body transformation was not affected by the above driving factors. Therefore, in this study, the suitability map of water body transformation was replaced by the probability map of water body transformation generated by Markov module. Lastly, the MCE module was used to obtain the suitability maps of land cover type transformation.

Lastly, using the CA-Markov model, based on the transition area matrix generated by Markov module and the suitability maps generated by MCE module, the land cover scenario of Haikou in 2020 was simulated. The simulated land cover scenario and actual land cover data in 2020 are shown in Figure 10. The land cover scenario of Haikou in 2020 simulated by the CA-Markov model was generally consistent with the actual data. However, the area change of artificial surface and cropland was greatly affected by artificial subjective factors, so the areas of artificial surface and cropland were underestimated to a certain extent. The area of grassland was overestimated to a certain extent.

3.3.2. Accuracy Verification of the CA-Markov Model

The calculation of the kappa index can be used to verify the consistency between simulated results and reference data, so it is widely used in land use change simulation and the accuracy verification of remote sensing images. The range of the kappa index is from −1 to 1. In general, when 0.75 ≤ kappa ≤ 1, the consistency between simulation results and reference data is high. When 0.50 ≤ kappa < 0.75, consistency is moderate. If kappa < 0.5, consistency is poor [32,33].

In this study, the Crosstab module of IDRISI Selva software was used to test the consistency level of the simulated land cover scenario and actual land cover data in Haikou in 2020. The calculated Kappa index was 0.79, indicating that the simulation result of the model had high similarity with the actual data, and simulation accuracy was high. Therefore, the model could be used to simulate the land cover scenarios of Haikou in the next three decades (2030, 2040, and 2050).

3.3.3. Future Urban Elastic Development Boundaries of Haikou

The land cover scenarios of Haikou in 2030, 2040, and 2050 simulated by the CA-Markov model are shown in Figure 11. Artificial surface (i.e., built-up land) patches were extracted from simulated land cover scenarios in three years and broken and scattered patches were removed. Patches with high aggregation were retained as urban built-up land patches, and the outer envelope surfaces of urban built-up land were drawn as urban development boundaries (Figure 12). The simulation results of the CA-Markov model and the MCR model were similar. From 2030 to 2050, the density of urban built-up land in Haikou would increase gradually, and the scope of urban built-up area would also expand continuously, which was also mainly reflected in the southward extension along the east coast and the southward expansion of the main area of urban built-up land on the left.

However, the areas of urban built-up land and elastic boundaries of Haikou during 2030–2050 simulated by the CA-Markov model were both larger than the MCR model (Figure 13). The areas of urban built-up land in Haikou in 2030, 2040, and 2050 simulated by the CA-Markov model were 345.04, 404.78, and 424.26 km², respectively. The areas within the urban elastic development boundaries were 381.86, 483.95, and 536.06 km², respectively. In the next two decades, the urban elastic development boundaries would increase by 102.09 and 52.11 km², respectively. As the urban built-up area of Haikou was 260.41 km² in 2020, the reserved space for urban development by 2030, 2040, and 2050 would be 121.45, 223.54, and 275.65 km², respectively, according to the elastic development boundaries.

3.4. Suggestions on Implementation of Future Urban Development Boundaries in Haikou

According to the demarcation results of future urban development boundaries of Haikou, the areas of the elastic boundaries were 20.59, 3.78, and 30.84 km² more than the rigid boundaries in 2030, 2040, and 2050, respectively, representing the increased elastic expansion space of urban development in the next three decades while meeting the rigid constraint conditions. Some suggestions on the implementation of future urban development boundaries were put forward in Haikou from the aspects of technology, policy, and management.

(1) From the aspect of technology, the demarcation of urban development boundaries should be combined with various types of planning, such as land use, urban, and rural planning. At the same time, the demarcation of urban development boundaries should be refined and amended according to various materials and documents. For example, the specific demarcation of rigid boundaries needs to refer to vector maps and text data of ecological protection land, while the demarcation of elastic boundaries requires understanding various built-up land indicators such as urban land and rural homestead.

(2) From the aspect of politics, the implementation of urban development boundaries needs a strong legal guarantee. The ecological compensation mechanism within the rigid growth boundaries should also be improved during the implementation of the rigid development boundaries. In addition, the evaluation mechanism of the implementation effect of urban development boundaries should be actively developed.

(3) From the aspect of administration, when implementing the management of urban development boundaries, it is necessary to assign responsibilities to all levels of government.

4. Discussion

Although this study obtained clear results of demarcating future urban rigid and elastic development boundaries, there are still some deficiencies that need to be further explored. For example, in future studies, more policy factors need to be considered and relevant data need to be collected to optimize the results of demarcating urban development boundaries.

First, this study suggested that policy factors should be considered in the implementation of future urban development boundaries, but policy documents concerning territorial spatial planning or urban planning issued by the central or all levels of government should also be fully considered in future studies on demarcating urban development boundaries. For example, in the policy documents of “The Letter on accelerating the rectification and supplement of permanent basic farmland and demarcation of urban development boundary” and “the Master Plan for Territorial Space of Haikou City (2020–2035)”, the demarcation of urban development boundaries should give consideration to “two control lines” of permanent basic farmland and ecological protection areas, so permanent basic farmland and ecological conservation redline should be theoretically restricted areas for urban expansion. In particular, they should be used as resistance factors in the demarcation of urban elastic development boundaries. However, since it is difficult to obtain the spatial distribution of permanent basic farmland and ecological conservation redline in Haikou, which cannot be applied to this study, it is necessary to strengthen the search of relative data in the future, so as to optimize the results of demarcating urban development boundaries. As policies and regulations are in constant change, research on the demarcation of future urban development boundaries should follow future policy changes and constantly adjust the research results.

5. Conclusions

To demarcate future urban development boundaries, we should not only consider the rigid requirements of natural resources and social and economic conditions for urban development, but also consider the elastic space of urban internal development. Therefore, the MCR and CA-Markov models were used to demarcate the rigid and elastic development boundaries of Haikou in the next three decades, and some suggestions were put forward for the implementation of the urban development boundaries in the future. According to the identification of the current urban boundaries, the areas of current urban boundaries of Haikou in 2000, 2010, and 2020 were 93.71, 124.26, and 260.41 km², respectively. The trend of urban expansion from 2000 to 2020 was that the main part of the city expanded southward and rapidly expanded inland along the coastal zone. This was very consistent with urban development planning and the actual development direction of Haikou.

In this study, the MCR model was used to demarcate urban rigid development boundaries. The urban rigid development boundaries of Haikou in 2030, 2040, and 2050 would be 361.27, 480.17, and 505.22 km², respectively, which represented the upper limit of ecological environment capacity and resource carrying capacity.

By using the CA-Markov model, the areas of urban elastic development boundaries in 2030, 2040, and 2050 would be 381.86, 483.95, and 536.06 km², respectively, representing the development line of urban internal demand for some time to come.

The areas of the elastic boundaries were 20.59, 3.78, and 30.84 km² more than the rigid boundaries in 2030, 2040, and 2050, respectively, which were the increased elastic expansion space of urban development while meeting the rigid constraint conditions.

This study put forward some suggestions on the implementation of future urban development boundaries from the aspects of technology, policy, and management. For example, the demarcation of urban development boundaries should be combined with planning. Strong legal guarantee is needed. Assigning responsibilities to all levels of government is necessary.