Evolution and Prediction of Urban Fringe Areas Based on Logistic–CA–Markov Models: The Case of Wuhan City

Abstract

The urban fringe is the transitional area from rural form to urban form, and it is also the urban space reserve land in the Territorial Spatial Plan. However, few researchers predict its overall evolution and guide the implementation of the Territorial Spatial Plan. This study attempts to explore the dynamic evolution law of urban fringe, analyze its driving factors, predict its future development, and put forward management suggestions for the implementation of the Territorial Spatial Plan. In this paper, the land use data of Wuhan in 2000, 2010 and 2020 are applied to delimit the urban fringe area of Wuhan by means of a sliding t-test. Fifteen driving factors are selected from three dimensions, natural factors, socio-economic factors and traffic accessibility, and brought into the Logistic model to explore the driving factors of its spatial evolution. The CA–Markov model is used to predict the fringe area of Wuhan in 2035. The results show that the transformation of rural hinterland into urban fringe is obviously affected by the distance from railway stations, highways, commercial centers and urban main roads. It is predicted that the outer boundary of Wuhan’s fringe area in 2035 will be basically the same as the planned urban development boundary. In order to realize the intention of land space planning, the development and construction of the northwest of the Huangpi District, the East Lake Scenic Area, and the west side of the Jiangxia District should be restricted. From the perspective of the evolution of the fringe area, this paper puts forward some management suggestions for the implementation of the Territorial Spatial Plan and makes a beneficial attempt in theory and method to understand the development characteristics of the fringe area and promote the implementation of the Territorial Spatial Plan.

1. Introduction

With the advancement of urbanization, the city continues to expand outward, gradually eroding rural Hinterland, the green ecological space is significantly reduced, and the contradiction between man and nature is constantly prominent [1,2]. The urban fringe, located at the junction of the urban core and rural hinterland, is a transitional area in the process of transforming the rural form into the urban form, and it is also the urban space reserve land in the Territorial Spatial Plan, which has three characteristics: fuzziness, dynamics and uncertainty. Because of its special geographical location and attributes, it has become one of the most intense and complicated areas in the process of urbanization [3].

Under the Territorial spatial planning system, China is implementing the most stringent farmland and ecological protection system, and the expansion of urban construction land is strictly restricted. The urban fringe, as the most drastic and complex area of urban change, is facing increasing pressure on ecological protection. At the same time, the urban fringe, as the reserve space for urban development, is included in the Territorial Spatial Plan, and it is also a very precious construction land during the development period of urban land stock in China, so it is necessary to study its development law in depth to provide support for making scientific development plans [3,4,5,6]. In this context, the government and planning practitioners need to accurately grasp the boundary of its scope, the changing law of geographic spatial structure and its future development trend so as to provide a policy basis for ecological protection and efficient development of construction land in urban fringe areas.

Based on the land use data of Wuhan in 2000, 2010 and 2020, this paper analyzes the change in Impervious Surface Ratio by using a moving t-test and delineates the urban fringe of Wuhan by combining the analysis of landscape flocculation and population density, which is the basis of this study. Through the dynamic degree model and geographic spatial structure transfer matrix, this paper explores the dynamic evolution process of the Wuhan fringe and grasps the macro-law of the dynamic evolution of fringe. Through the Logistic–CA–Markov combination model, this paper analyzes the driving force of urban fringe evolution and the evolution trend of the urban fringe, explores the evolution law of the urban fringe, forecasts its future, and puts forward guidance and control suggestions for the urban fringe based on the Territorial Spatial Plan and urban development status, so as to help promote the effective implementation of the Territorial Spatial Plan and promote the healthy and sustainable development of the city.

The rest of this paper is organized as follows: Section 2 introduces the literature review; Section 3 introduces the urban fringe division method, urban fringe spatio-temporal evolution analysis method and urban fringe prediction analysis method adopted in the study; Section 4 provides the description of the research area and related data; Section 5 analyzes the results of the model and gives some planning suggestions. Section 6 is a summary and discussion.

2. Literature Review

This paper provides a description of relevant literature from three aspects: the scientific and quantitative delineation of the urban fringe, the selection of driving factors for spatial pattern evolution, and research methods for urban spatial prediction models.

2.1. Scientific and Quantitative Delineation of the Urban Fringe

From the empirical discriminant method to the single-factor discriminant method and the multi-factor comprehensive discriminant method, the methods for delineating the urban fringe are constantly updated and improved. The multi-factor comprehensive discriminant method can more comprehensively reflect the multidimensional and complex characteristics of the urban fringe, thereby making the delineation of the urban fringe more reasonable. Initially, Western scholars defined the urban fringe as a specific range extending outward from the urban core, and L.H. Russwurm considered the urban fringe to be a continuous ring-shaped area of about 10 km located between urban areas and rural hinterlands [7]. The academic community in Japan uses various defining indicators centered around population to divide the central city and the peripheral area, such as population density and population flow ratio (commuting rate, school commuting rate) [8]. Zhou selected three indicators, namely, Impervious Surface Ratio, landscape flocculation index and population density, from physical, landscape and population perspectives to determine the urban fringe range of Xi’an City [9]; Xiong determined the urban fringe area of Wuhan according to the night lighting data of NPP-VIIRS and the moving t-test method [10]; Dai et al. identified the urban fringe range of Jiangyin City by using wavelet transform to detect multiple characteristic indicators of the urban fringe [11]; Zhang et al. determined the Beijing urban fringe range through analysis of urban land distribution and mutation theory, and conducted an analysis of urban expansion characteristics in terms of quantity, type, direction, and intensity [12]; Wang and Zhang used multi-criteria decision-making to determine the urban fringe range of Guangzhou, and conducted an analysis from the perspectives of quantity, morphology and dynamic degree [13]; Fan et al. describe and define the spatial extent of the semi-urbanized areas in Hangzhou City by calculating “land fragmentation” and “proportion of arable land” using land use coverage data [14]. Yang et al. delineated the urban fringe range of Jiangyin City and Zhangjiagang City using a new method comprised of three main parts: comprehensive urbanization index, mutation point detection, and constraint-based boundary recognition. The feasibility of the new method was determined through a comparative analysis of the results obtained under different constraint conditions [15]. Due to the complexity of the urban fringe, researchers have conducted comprehensive evaluations from multiple perspectives to determine its spatial range, with land use type, landscape form and population being the main factors considered.

2.2. Selection of Driving Factors for Spatial Pattern Evolution and Research Methods

On the driving factors, economic, political, transportation and natural environmental factors have different degrees of influence on the evolution of urban fringe areas. On the economic factors, Wu and Ma and Sun et al. believe that economic factors, including industry and social infrastructure, are the determining factors for the expansion of urban fringe areas [16,17]. In terms of planning policies, Chang et al. and Zhou believe that urban planning and government policies are the main factors influencing the development of urban fringe areas [9,18]. As for transportation layout, Zhang suggests that the main roads of a city will impact the development of urban fringe areas [19]. Regarding natural geographical factors, Zhang et al. and Zhou argue that factors such as terrain, geology and hydrology often limit the scope of urban expansion [9,12].

In terms of quantitative analysis of driving factors in urban development, scholars both domestically and internationally mainly use methods such as multiple linear regression and logistic regression models. Li et al. identified the dominant driving factors for the temporal evolution of urban construction land through multiple linear regression and correlation analysis to be economy, population and industry [20]; Zohreh Shahbazian analyzed the land use change in Iran’s Mehran Plain using a logistic regression model and found that factors such as distance to roads, distance to water wells, distance to streams, and distance to residential areas have a negative impact [21]; Chen et al. analyzed the driving mechanism of the dynamic change in the landscape pattern of land use in Fuzhou City using a logistic regression model and found that the conversion of construction land is mainly influenced by population factors. It can be seen that when exploring the driving factors of urban land change, scholars often use multiple linear regression and logistic regression models for quantitative analysis. The scope of these studies is relatively fixed, and there is little literature that conducts quantitative analysis on the dynamic urban fringe evolution [22].

2.3. Application of Urban Spatial Forecasting Models

In predicting urban construction land and land use change, domestic and foreign scholars use models such as ANN–CA, FLUS and Markov to predict urban changes. Liu and Chen analyzed the expansion pattern of urban construction land in Zhengzhou City from 2013 to 2018 using remote sensing interpretation data and used the ANN–CA model to predict the spatial expansion of Zhengzhou City in 2025 [23]; Li et al. believed that the CA–Markov coupled model integrates the characteristics of spatial pattern evolution simulation and long-term quantitative prediction of the two models, which can overcome the deficiencies of a single model and can predict temporal and spatial changes more accurately and stably [24,25,26]. However, Wang believed that the traditional CA–Markov model has certain deficiencies and needs to simulate and predict the land use situation in the Beibei District, Chongqing City, through the multi-criteria evaluation (MCE) model [27]. Nourqolipour R et al. used multi-criteria evaluation to successfully simulate the landscape changes in the oil palm plantation in Kuala Langat, Malaysia [28]. Li et al. coupled the Logistic regression model with the CA–Markov model to simulate the spatio-temporal changes in land use in the Jiaodong Peninsula, which performed well; in addition [29], Wang et al. simulated the spatial pattern development of land use in Shenzhen City under multiple scenarios using the FLUS–Markov model [30]. Cheng et al. used the Logistic–CA–Markov model to predict the land use changes in Pingchang County, proving that the Logistic–CA–Markov model can effectively simulate the land use change process from both temporal and spatial perspectives, with high credibility of the prediction results [31]. It can be seen that the current focus of most urban spatial change predictions is on the changes in land use, and models such as ANN–CA, FLUS, and Markov are also commonly used in land prediction studies, among which the Logistic–CA–Markov model has higher accuracy in predicting land use evolution.

To alleviate the problems of chaotic development in the urban fringe that arise during rapid development, it is necessary to conduct an in-depth analysis of the development of its urban fringe and provide policy support for the future development of the urban fringe. Wuhan, as one of the new first-tier cities in Central China, has developed rapidly in the past 20 years, with significant increases in economy, population and urban construction land. Its urban fringe area has undergone relatively drastic changes in natural resources and socio-economic aspects. Choosing the urban fringe of Wuhan as a case study is highly representative and beneficial for providing a demonstration effect for cities of the same level. Researching and predicting the dynamic evolution of the urban fringe in Wuhan can further reveal the laws of spatial expansion in Wuhan and guide the further regulation of the urban fringe in Wuhan. It can also provide support for the implementation of the Territorial Spatial Plan and improve research on the prediction of urban geographic spatial structures. This paper takes the years 2000, 2010 and 2020 as the research objects in Wuhan City and delineates the urban fringe area based on the Impervious Surface Ratio, Landscape Flocculation Index and Population. It analyzes the evolution process of the urban fringe in Wuhan from the two dimensions of dynamic degree and urban territorial structural transition. The driving forces of the urban fringe in Wuhan are quantitatively analyzed from the three perspectives of natural environment, socio-economic factors and transportation accessibility. Based on the data of the urban fringe area of Wuhan City in 2020, the spatial pattern of the urban fringe in Wuhan in 2035 is predicted using the Logistic–CA–Markov model. A comparison is made between the prediction results and the urban construction boundary of the Wuhan City 2035 Territorial Spatial Plan to provide suggestions for the implementation strategy of the Territorial Spatial Plan in Wuhan.

In summary, there is currently no unified theory or method for delineating the scope of the urban fringe. Among various delineation methods, the comprehensive discriminant method using multiple factor indicators is more accurate. Multiple regression models, such as multivariate linear regression and logistic regression, are commonly used for qualitative and quantitative analysis of driving factors in urban land use change and landscape pattern change. According to

Table 1. Advantages and Disadvantages of Commonly Used Prediction Models.

Model Category	Commonly Used Models	Advantages	Disadvantages
Quantitative prediction model	Markov	It has an absolute advantage in predicting the quantitative changes in the urban geographic spatial structure over a longer period of time	It is unable to reflect the changes in the distribution pattern of urban geographic spatial structure
	Logistic	Quantitative analysis of the correlation between driving forces and urban fringe changes enables the prediction of the probability of urban fringe emergence	It only reflects the changes in quantity and is unable to reflect the spatial distribution
Spatial prediction model	CA	It is capable of long-term forecasting and simulating complex spatial changes	There are limitations in detail prediction
	CLUE-S	It accurately simulates small-scale spatial pattern changes while simultaneously simulating multiple variations	Neglecting the possibility of transforming non-dominant geographic spatial structures
	FLUS	It is capable of simulating discontinuous land use changes	It is difficult to reflect the spatial differences in the changes in geographic spatial structure in different locations

, in terms of urban land development prediction models, the Markov model, cellular automaton (CA) model, and their coupled model CA–Markov are more commonly used. It is evident that logistic regression models and CA–Markov models are more mature in urban development analysis and prediction. By combining logistic regression models, CA models, and Markov models, the Logistic–CA–Markov model is obtained, which not only simulates complex spatial systems but also has the advantage of predicting long-term changes, greatly enhancing prediction accuracy. The prediction of urban geographic spatial structure is not driven by a single influencing factor but rather a result of complex interactions among various factors such as the natural environment, socio-economic factors and transportation accessibility. The changes in urban geographic spatial structure are manifested in terms of both quantity and spatial distribution. Based on this, this study utilizes the Logistic–CA–Markov model to analyze and predict the driving forces of the urban fringe.

3. Research Methodology

In this study, Wuhan City is taken as the research area, and multiple data sources are used as the basis. The moving t-test of the Impervious Surface Ratio, the Landscape flocculation index, and the population density gradient analysis method are used to determine the urban fringe range of Wuhan City from 2000 to 2020. Based on this, the urban geographic spatial structure transfer matrix and the urban geographic spatial structure type transition matrix are used to analyze and predict the urban fringe of Wuhan City during the 20-year period. Finally, based on the multiple periods of the urban fringe range in Wuhan, the evolution pattern of the urban fringe in Wuhan, and multiple sources of data, the Logistic–CA–Markov model is used to predict the urban fringe of Wuhan in 2035. Figure 1 shows the Technical Route.

3.1. Identification and Analysis Methods of Urban Fringe

In the “Edge Community L.H. Russwurm’s Model,” L.H. Russwurm divided the city from the center to the fringe areas into four parts: urban core, urban fringe, urban influence area and rural hinterland. Due to the rapid development of Chinese cities, it is difficult to clearly distinguish the urban influence area and the urban fringe, as they are relatively similar. In this study, while referring to L.H. Russwurm’s theory [13], the actual situation of Wuhan was taken into account, and Wuhan City was divided into three parts: urban core, urban fringe and rural hinterland. According to the research methods of Zhou [9], the initial delimitation of the urban fringe was determined by the Impervious Surface ratio and Landscape Flocculation Index and then corrected by the population density index to ultimately determine the urban fringe range.

3.1.1. Impervious Surface Ratio and Moving t-Test

The Impervious Surface Ratio in the urban core is high due to extensive urban and architectural construction, with values reaching as high as 0.8–0.9. Conversely, in rural areas, the Impervious Surface Ratio is very small, typically not exceeding 0.2. However, in the urban fringe, which is the combination of the urban core and rural areas, the Impervious Surface Ratio often exhibits mutation points. By extracting these mutation points, the boundary of the urban fringe can be initially identified.

The moving t-test is a commonly used statistical method to test data mutation by examining the significant differences in the mean values of random samples [5,32,33]. For a sequence with n samples, setting time point a as the baseline with corresponding t value t_a, if the values on the t curve exceed this baseline, it indicates the presence of mutation data at that point. The expression is as follows:

$$t=\frac{\overline{x_1}-\overline{x_2}}{S\sqrt{\frac{1}{n_1}+\frac{1}{n_2}}}$$

(1)

where

$$S=\sqrt{\frac{n_1{s}_1^2+n_2{s}_2^2}{n_1+n_2-2}}$$

(2)

The equation obeys the degrees of freedom $\mathsf{\gamma }={\mathrm{n}}_1+{\mathrm{n}}_2-2$ of the distribution. Among them, the two sequences before and after the reference point at time a are ${\mathrm{x}}_1$ and ${\mathrm{x}}_2$, the corresponding sample lengths are ${\mathrm{n}}_1$ and ${\mathrm{n}}_2$, and the variances are ${\mathrm{s}}_1^2$ and ${\mathrm{s}}_2^2$, respectively.

3.1.2. Landscape Flocculation Index

The land use in the urban core is usually more regular and orderly, while in rural areas, the land use is more complete and single, mainly consisting of agricultural land and forests. The landscape disorder in these two types of areas is usually low. The urban fringe is the transitional area between the urban core and the rural area, with complex urban core landscapes and single rural landscapes. Various types of land use are interlaced and loosely arranged in this area, resulting in a higher landscape flocculation index. Therefore, due to the different expressions of the landscape flocculation index in different regions, it can be used as an indicator for delineating the urban fringe.

The landscape flocculation index is a physical quantity that measures the complexity and balance of a system. A balanced and orderly system has a low landscape flocculation index, while a chaotic and complex system has a high landscape flocculation index. The formula for the landscape flocculation index is as follows [9]:

$$W=-{\displaystyle {\displaystyle \sum }_{i=1}^n}XiLNXi$$

(3)

where $W$ represents the landscape flocculation index value, $x_i$ represents the proportion of a certain type of land use in a grid unit per unit area, and $i$ represents the number of land use types in the grid unit.

3.1.3. Population Density

In urban land, the population density of the urban core is high, while that of rural areas is low. The urban fringe is the transitional area between the two, and the population density shows a gradient change. Based on the changing pattern of population density, the urban fringe can be identified [34]. In this study, the population density data published by WorldPop was used to verify the preliminary delineation of the urban fringe in the first two steps. The overlapping areas determined as the urban fringe by both the Impervious Surface Ratio mutation point and the landscape flocculation index were directly identified as the urban fringe. For the non-overlapping areas, the population density was used to determine whether they were the urban fringe. If they meet the urban population density threshold, they are considered as the urban fringe; otherwise, they are not considered as the urban fringe. This method takes into account the changes in impervious surface area, landscape disorder and population density, and finally determines the urban fringe.

3.2. Spatial-Temporal Evolution Analysis Method for Urban Fringe

The degree and spatial transfer analysis are commonly used to study the changes in land use types. This paper applies them to the changes in the urban geographic spatial structure, which can reflect the changes and transfers of the urban core, urban fringe, and rural hinterland more clearly.

3.2.1. The Dynamic Degree of Urban Geographic Spatial Structure Changes

The dynamic degree model of land use can quantify the intensity of changes in the quantity of various land uses within a region over a period of time. This paper introduces it into the calculation of the changing intensity of the urban geographic spatial structure in order to quantify the changing intensity of the urban core, urban fringe and rural hinterland. The expression is as follows:

$$K=\frac{u_b-u_a}{u_a\times t} \times 100\%$$

(4)

where $t$ means study period and $K$ means degree of dynamic change of a particular urban geographic spatial structure form in this study area during this period $t$; when the $K$ value is positive, it indicates that the area of this type of urban geographical structure increases during the study period $t$, and when the $K$ value is negative, it means that its area decreases.

3.2.2. Urban Geographic Spatial Structure Type Transition Analysis

The land use transition matrix can quantitatively describe the conversion areas between different land use types over a certain period of time. This paper introduces the land use transition matrix into the analysis of urban geographic spatial structure transition. The expression is as follows:

$$S_{ij}=\left[\begin{array}{cccc}{S}_{11}& S_{12}& \cdots & S_{1n}\\ S_{21}& S_{22}& \cdots & S_{2n}\\ ⋮& ⋮& ⋮& ⋮\\ S_{n1}& S_{n2}& \cdots & S_{nn}\end{array}\right]$$

(5)

where $n$ is the urban geographic spatial structure type; $i$ means urban geographic spatial structure type at the beginning of the study period, $j$ means urban geographic spatial structure type in the study period at end; $S_{ij}$ means the area of an urban geographic spatial structure type transferred from period $i$ to period $j$.

3.3. Urban Fringe Prediction Analysis Method

The principle of the Logistic–CA–Markov model in predicting urban geographic spatial structure is as follows: in the Logistic model of each type of geographic spatial structure, driving factors from the natural environment, socio-economic factors and transportation accessibility are introduced for calculation. The driving factors that affect the transformation of each type of geographic spatial structure are selected, and their driving degree is quantified. The driving factors selected by the Logistic model are transformed into probability maps for each type of geographic spatial structure through a multi-criteria evaluation (MCE) model. At the same time, the probability maps of each type are combined to form a probability atlas, which is used as the transformation rules in the CA model. The defined urban geographic spatial structure is operated by the Markov model to obtain the transition rules of the Markov model. By integrating the above steps, the changing trend of urban geographic spatial structure can be simulated.

3.3.1. Logistic Regression Model

The logistic regression model is a binary classification model used to explore the relationship between multiple independent variables and a binary dependent variable. Its basic idea is to transform the relationship between independent variables and dependent variables into a probability relationship and use the logistic function to model the probability. The advantage of logistic regression is that the independent variables do not need to follow a normal distribution and can be continuous or discrete. The logistic regression model has better advantages than OLS in binary variables, and it is widely used in driving force research. In this study, various driving forces were analyzed based on this model to study their impact on the evolution of urban fringe [35]; the expression is:

$$log\left(\frac{P_i}{{1-P}_i}\right){=\alpha +\beta }_1{X}_1{+\beta }_2{X}_2+\cdots {+\beta }_n{X}_n+\epsilon$$

(6)

where $P_i$ is the probability of transforming other urban geographic spatial outcomes into urban geographic spatial structure $i$; $X_n$ is the driving factor; ${\beta }_n$ is the correlation coefficient; $\epsilon$ is the residual. The binary logistic regression was performed by the software SPSS 20.0, and the Hosmer–Lemeshow (HL) index could test the model fit, and when the HL value was greater than 0.05, the simulation fit was good. Otherwise, the model fit was bad.

Zhang et al. analyzed the driving forces of the urban fringe from aspects such as elevation slope, population growth, economic development and government policies [12]. Che Tong [36] considered population density, per capita GDP, and distance to city and county centers as possible driving forces. The natural environment includes elevation and slope, which are the basic conditions for the evolution of urban geographic spatial structure. Flat areas with good geological conditions are more suitable for construction activities than mountainous and hilly areas, so natural environment factors such as topography directly affect the direction and speed of urban expansion. The social environment includes whether it is within the overall planning and population, socio-economic factors and government policy directions are the leading factors for urban expansion. The concentration of population, industrial development, and new district construction continuously stimulate the expansion of the city. Transportation accessibility includes distance to primary roads, freeway entrance/exit, and other factors that promote the expansion of the urban core area and drive construction activities along these routes, playing a guiding role in the urban expansion. Location factors include the distance to commercial centers and government centers. Areas close to commercial and government centers can influence the city’s expansion through the attraction of people. Indicators are usually selected from the natural environment, socio-economic factors, transportation accessibility and location to serve as possible driving factors.

3.3.2. CA Model

The Cellular Automata (CA) model is a rule-based discrete-space model that divides space into several cells, each cell having its own state. By defining transition rules between cells, the CA model simulates the transition process of the system between different states [37]. When dealing with complex systems, the advantage of using cellular automata models is that they can divide space into several cells and define transition rules between cells based on historical data and other factors. Cells represent each grid in the geographic spatial structure raster, indicating the state of the cell. The CA model can be described by the following equation:

$$S_{(t+1)}=f(S_{\left(t\right)}, N)$$

(7)

where $S$ is a finite and discrete set of states in the cell;$f$ is the transition rule of the cell; $N$ is the neighborhood of the cell; $t$, $t+1$ are different moments in time at which the system exists.

The CA model has the ability to handle spatial computation, but it usually only considers single and linear factors as driving factors, neglecting the combined effects of multiple factors and the spatial heterogeneity within cells. Therefore, it is necessary to replace its transition rules. In this study, the logistic regression model is used to analyze the natural environment, socio-economic, and transportation accessibility driving factors that affect the geographic spatial structure evolution. By combining the multi-criteria evaluation (MCE) model of the Analytic Hierarchy Process to assign values to each driving factor, a geographic spatial distribution probability map is obtained. Then, the Collection Editor tool is used to combine the spatial distribution probability maps of various geographic spatial structures to form a geographic spatial structure probability atlas. Finally, the probability atlas is imported into the CA model to replace its original transition rules and eliminate its limitations.

3.3.3. Markov Model

The Markov model is a probability model based on state transitions, which describes the probability of a system transitioning between different states. The Markov model can be used to predict future states and analyze and optimize the state transition process of a system [38]. This article uses a Markov model to predict the future development of geographic spatial structure. Its main features are stability and the absence of lag effects, which means that the future state at time t + 1 only depends on the current state at time t and is independent of any previous states. Stability is a characteristic attribute of Markov processes, referring to their tendency to exhibit convergence to a steady state over a long period of time. Given that the dynamic evolution of urban geographic spatial structure possesses the characteristics of Markov stability and the absence of lag effects, it can be effectively used to predict the future trend of quantity development in the geographic spatial structure. Its mathematical expression is as follows:

$$S_{\left(t+1\right)}=P_{ij}{S}_{\left(t\right)}$$

(8)

where $S_{\left(t+1\right)}$ is the state of the urban geographic spatial structure type at the future moment $t+1$; $S_{\left(t\right)}$ is the state of the urban geographic spatial structure type at the current moment $t$; $P_{ij}$ is the state transfer probability matrix of the geospatial structure type.

The Markov model usually assumes that the transition probabilities in space are uniformly distributed, ignoring spatial heterogeneity, which means that for the same spatial sample, different regions have different probability structures, lacking the ability to handle spatial evolution. The advantage of the coupled CA–Markov model is that it combines the advantages of the CA model in cell interaction and spatial connectivity, the advantages of historical data evolution, and the advantages of the Markov model in analyzing the specific quantities of transformation between different spatial types based on probability and transition matrix theory, thereby exploring the changes in geographic spatial structure in both spatial and quantitative dimensions.

3.3.4. Logistic–CA–Markov Construction and Verification

Based on the geographic spatial structure data of Wuhan in 2000, 2010, and 2020, this study uses the multi-criteria evaluation (MCE) model based on the logistic regression model to generate a geographic spatial structure probability atlas as the conversion rule and runs the CA–Markov model on the urban fringe of Wuhan using the IDRISI 17.0software platform for prediction. First, the scientificity of the model is verified. Taking the 2010 urban fringe data of Wuhan as the base period, the simulation of the urban fringe changes in 2020 using the geographic spatial structure transfer matrix from 2000 to 2010 and various geographic spatial structure probability atlases is performed. The Kappa coefficient is used for verification. If the Kappa coefficient is greater than 75%, the prediction is considered valid. After the prediction passes the Kappa coefficient test, the urban fringe of Wuhan in 2035 is predicted.

(1)

Prediction of the urban fringe of Wuhan in 2020

Firstly, the geographic spatial structure data of 2000 and 2010 are imported through the Markov module, and the distribution ratio error is set to 0.1. The geographic spatial structure transfer matrix and geographic spatial structure type transition matrix from 2000 to 2010 are obtained, and then the CA–Markov function is used to take 2010 as the basic image for prediction. Taking the probability atlas of the Wuhan geographic spatial structure in 2010 as the transformation rule of CA and importing the transfer matrix of the Wuhan geographic spatial structure from 2000 to 2010, the number of cycles is 10, and by choosing the 5 × 5 filter, the forecast map of Wuhan urban fringe in 2020 is obtained by simulation.

Compare the forecast map of the Wuhan urban fringe in 2020 with the designated map of the Wuhan urban fringe in 2020. If the Kappa coefficient is greater than 75%, the next prediction will be made;

(2)

Prediction of the urban fringe of Wuhan in 2035

Based on the feasibility of the model, the geographic spatial structure of Wuhan in 2035 is predicted. Firstly, the geographic spatial structure data of Wuhan City in 2020 is taken as the forecast starting years, and the geographic spatial structure transfer matrix from 2010 to 2020 is selected in the transfer matrix. Import the Urban geographic spatial structure Probability Atlas in 2020 as the conversion rule, set the filter to 5 × 5, input 15 for the number of cycles once a year, and obtain the geographic spatial structure data of Wuhan in 2035.

4. Study Area Overview and Data

4.1. Overview of the Study Area

Wuhan is one of the representative cities in central China. It consists of thirteen administrative districts with seven main urban areas and six remote urban areas. The total area of Wuhan is 8494 km² (Figure 2). From 2000 to 2020, Wuhan City developed rapidly, the core area of the city continued to expand, and the urban built-up area almost quadrupled (http://tjj.wuhan.gov.cn/tjfw/tjgb/202001/t20200115_841037.shtml) (accessed on 10 January 2023), the urban population increased from 8.3126 million to 12.3265 million, and the urbanization rate increased to 84.56% (http://tjj.wuhan.gov.cn/tjfw/tjgb/202104/t20210425_1675623.shtml) (accessed on 10 January 2023). During its rapid urbanization process, Wuhan City, like other cities in China, has encountered a considerable number of urban challenges. The areas at the forefront of urban expansion, commonly referred to as urban fringe areas or urban–rural interface zones, have been confronted with issues including fragmented land use, suboptimal land use efficiency, and degradation of landscape ecology [39,40,41,42]. In order to solve the above problems, guide and restrict urban expansion, this paper analyzes the evolution characteristics of the urban fringe area in Wuhan, explores the relevant factors that promote its evolution, and predicts the future expansion of the urban fringe area while providing data reference for urban development. As shown in Figure 2, the yellow part in the upper left corner is Hubei Province, and the red part in the lower left corner is Wuhan City.

4.2. Data Source and Processing

This study incorporates various data sets, including land use, digital elevation model (DEM), precipitation, temperature, Wuhan City’s overall planning scope, gross domestic product (GDP), population density, road network and points of interest (POI) data, spanning from 2000 to 2020. The land use data is obtained from ZENODO; DEM data and GDP kilometer grid data are sourced from the Geospatial Data Cloud platform (http://www.gscloud.cn/) (accessed on 13 January 2023); precipitation and temperature data are acquired from the National Earth System Science Data Center (http://www.geodata.cn) (accessed on 13 January 2023); population data is retrieved from WorldPop (https://www.worldpop.org/) (accessed on 13 January 2023); Wuhan City’s road network data is derived from OpenStreetMap (https://www.openstreetmap.org) (accessed on 13 January 2023); POI data is gathered through web scraping from AMap. To ensure data accessibility and completeness, three periods (2000, 2010, and 2020) were selected for analysis. All the data were projected onto the WGS_1984 coordinate system using ArcGIS 10.5, creating a comprehensive spatial database for Wuhan City’s fringe area from 2000 to 2020, as shown in

Table 2. Data Sources and Processing.

Data Name	Data Content	Data Source	Data Processing
Wuhan City Land Use Data (2000–2020)	CLCD, Wuhan 30 m resolution land use data from 2000–2020	ZENODO (https://doi.org/10.5281/zenodo.8176941) (accessed on 13 January 2023)	ArcGis 10.5
Wuhan City Topographic Data	DEM30 m resolution digital elevation data	Geospatial Data Cloud (http://www.gscloud.cn/) (accessed on 13 January 2023)	ArcGis 10.5 surface analysis
Precipitation	Annual average precipitation data at 1 km resolution in Wuhan from 2000 to 2020	National Earth System Science Data Center (http://www.geodata.cn) (accessed on 13 January 2023)	ArcGis 10.5 image metric statistics
Temperature	Annual average temperature data at 1 km resolution in Wuhan from 2000 to 2020	National Earth System Science Data Center (http://www.geodata.cn) (accessed on 13 January 2023)	ArcGis 10.5 image metric statistics
Wuhan City Master Plan	Wuhan City Master Plan Scope	Wuhan Natural Resources and Planning Bureau (http://zrzyhgh.wuhan.gov.cn/) (accessed on 13 January 2023)	ArcGis 10.5 drawing elements
GDP	Wuhan City kilometer grid GDP data 2000–2020	Data Center for Resource and Environmental Sciences, Chinese Academy of Sciences (http://www.resdc.cn/) (accessed on 13 January 2023)	ArcGis 10.5 displays partition statistics in a table
Population density	Spatial distribution data of population from 2000 to 2020	WorldPop (https://www.worldpop.org/) (accessed on 13 January 2023)	ArcGis 10.5
Wuhan City Road Network Data	Wuhan City trunk road network data from 2000 to 2020	OpenStreetMap (https://www.openstreetmap.org) (accessed on 13 January 2023)	ArcGis 10.5 Euclidean distance
POI Data	Spatial data points of freeway entrances/exits, railway stations, government centers and commercial centers in Wuhan from 2000 to 2020	Gaode Map (https://lbs.amap.com/) (accessed on 13 January 2023)	ArcGis 10.5 Euclidean distance

.

5. Results and Analysis

5.1. Urban Fringe Area Identification

This paper identifies the urban fringe of Wuhan City based on the Impervious Surface Ratio, Landscape Flocculation Index and population Density with the aim to improve the scientificity of urban fringe delineation. When considering only the difference in urban and rural Impervious Surface Ratio, it ignores the reflection of population development, GDP, and other socio-economic indicators on urban–rural relationships, making it difficult to characterize the complex characteristics of the urban fringe in terms of population, land use, landscape and economy. Therefore, the composite delineation of the urban fringe based on multiple factors can largely avoid the limitations caused by the single-factor index in delineating the urban fringe.

5.1.1. Extraction of Impervious Surface Ratio Index

This paper uses the moving t-test to extract the points where the Impervious Surface Ratio in Wuhan City undergoes a sudden change. The operation steps include designing sampling points, visualizing mutation points, and extracting mutation points to depict boundaries.

(1)

Designing spatial sampling points

Taking the traditional business district of downtown Wuhan as the city center, 360 radiation rays are drawn at an angle of 1° from the center, and 180 concentric rings are drawn outward from the center at an interval of 500 m to form a ring radiation grid that covers the entire study area. The entire grid has a total of 64,377 spatial sampling points. Overlaying it with the city boundary of Wuhan, we obtain 36,564 data points within the city boundary. Next, we extract the Impervious Surface Ratio at each data point and look for mutation points on each radiation ray. The above process is shown in Figure 3.

(2)

Obtaining mutation points

The data points are divided into 360 groups based on the radiation rays, and the data is imported into Matlab for moving t-test with a step size of 5 to obtain mutation points. Partial experimental results of the data set are shown in Figure 4. The points where the lines deviate from the yellow or purple dashed lines indicate mutations at those points. The corresponding spatial points on the remote sensing image are the mutation points of the Impervious Surface Ratio.

(3)

Extraction of mutation points and delineation of urban fringe boundaries

In fact, there are multiple mutation points on each ray. In order to determine the mutation points between the urban core and urban fringe, as well as between the urban fringe and rural areas, manual discrimination is required in combination with land use, satellite image data, etc. To maintain the continuity and integrity of land use, eliminate abnormal mutation points, and finally identify the mutation points in 360 radial directions. Connecting them in sequence outlines the urban fringe boundary, as shown in Figure 5.

5.1.2. Extraction of Landscape Flocculation Index

In the originally obtained land cover data, land use types are divided into seven categories: farmland, forest, Impervious Surface, unused land, water bodies, shrubs and grassland. In order to improve the identification accuracy, land uses such as farmland, forest and grassland, which have obvious rural hinterland characteristics, are merged into agricultural land. A 500 m × 500 m fishnet covering the entire Wuhan area is established to calculate the proportion of each land use in each grid and then calculate the landscape flocculation index within each grid. From Figure 6, it can be seen that there are significant spatial differences in the Landscape Flocculation Index within the Wuhan urban area. In the built-up area, the Impervious Surface area accounts for a large proportion, leading to a lower landscape flocculation index. In rural areas, large areas of farmland show a lower Landscape Flocculation Index. Both of them vary between 0 and 0.7. The urban fringe and areas along the main river channels have a higher landscape flocculation index. After multiple experiments, a large area with entropy values of 0.7 and above surrounding the main urban area is taken as the range of the urban fringe in Wuhan, based on the judgment of the landscape flocculation index, as shown in Figure 6.

5.1.3. Population Density Determination and Urban Fringe Range

There are discrepancies between the population data provided by WorldPop and the official population census data. First, these discrepancies are corrected. According to WorldPop’s grid data, the population in Wuhan City in 2000, 2010 and 2020 were 8.6376, 9.6003, and 11.4912 million, respectively. The official published population census data for the fifth, sixth, and seventh population censuses of Wuhan City (http://tjj.wuhan.gov.cn/tjfw/tjgb/) (accessed on 10 February 2023) were 8.3126, 9.7853, and 12.3265 million, respectively. The correction factors for the three years (population census data/WorldPop population data) were calculated as 0.9623, 1.0193, and 1.0726. The WorldPop data was multiplied by the corresponding correction factors using a grid calculator to obtain the corrected population. The WorldPop data has a precision of 100 × 100 square meters grid, and high-precision data can minimize the overestimation of population density gradients caused by a large proportion of undeveloped land in the urban fringe. Then, based on the corrected WorldPop grid data, the population density contour lines of Wuhan City were plotted. After repeated comparisons with land-use data and other relevant information and multiple adjustments and corrections, it was determined that when the population density is greater than 3000 people/km², it is considered urban core; when the population density is approximately 1000–3000 people/km², it is considered urban fringe; when the population density is less than 1000 people/km², it is considered rural areas. Based on this, the urban fringe range of Wuhan City will be validated in the following sections.

5.1.4. Determining the Urban Ringe Range

The urban fringe ranges determined in Section 5.1.1 and Section 5.1.2 are spatially overlaid. For the overlapping areas where the conclusions of both indicators are consistent, they are directly designated as the urban fringe of Wuhan City. For the non-overlapping areas, where only one indicator is designated as urban fringe, the population density threshold is used for screening. Areas with population density between 1000–3000 people/km² are designated as urban fringe, and areas with population density outside this range are not classified as urban fringe. The final urban fringe ranges of Wuhan City for the years 2000, 2010 and 2020 are shown in Figure 7.

5.2. Dynamic Evolution of Urban Fringe Areas

5.2.1. Dynamic Degree Analysis of Urban Fringe

The areas of the urban core, urban fringe, and rural hinterland of Wuhan City from 2000 to 2020 were calculated to analyze the overall evolution characteristics of the urban geographic spatial structure of Wuhan City in the past 20 years.

From

Table 3. Statistics on the Area of Wuhan’s Urban Geographic Spatial Structure.

Projects	2000	Amount of Change	2010	Amount of Change	2020	Change of Dynamic Degree/%
						2000–2010	2010–2020
Urban core area (km2)	253.573	230.267	483.84	260.735	744.575	9.081	5.389
Urban fringe area (km2)	474.351	390.631	864.982	355.122	1220.104	8.235	4.106
Rural hinterland (km2)	7853.823	−621.573	7232.25	−615.182	6617.068	−0.791	−0.851
Registered population (10,000 people)	804.81	173.73	978.54	254.11	1232.65	2.16	2.60
GDP (billion)	1207	4308.76	5515.76	10,100.3	15,616.06	35.70	18.31

, it can be seen that from 2000 to 2010, the urbanization process in Wuhan developed rapidly. The urban core area increased from 253.573 km² to 483.840 km², with an average annual dynamic degree of 9.081%. During this period, Wuhan developed rapidly, with a large number of urban fringes transforming into urban cores. The urban fringe area increased from 474.351 km² to 864.982 km², with an average annual dynamic degree of 8.235%. The urban fringe continuously expanded outward and spread to the rural hinterland. The area of rural hinterland decreased from 7853.823 km² to 7232.925 km², with an average annual dynamic degree of 0.791%. Under the rapid development of Wuhan, a large number of rural areas were transformed into urban fringes due to the increase in construction land, leading to an increasing degree of urbanization in Wuhan. The registered population grew rapidly from 8.0481 million to 9.7854 million, with an average annual increase of 173,700 people. Compared with other provincial capital cities, the growth rate was relatively fast, matching the expansion of the city’s scale. The urban GDP also grew rapidly during the same period, increasing from 1207 to 5515.76 billion, with an average annual dynamic degree of 35.7%, which is in line with the characteristics of China’s rapid economic and urbanization development from 2000 to 2010.

In summary, between 2000 and 2010, the urban core area in Wuhan experienced a 90.80% increase in its urban core area, while the urban fringe area witnessed an increase of 82.35%. Meanwhile, the growth rate of the registered population stood at 21.59%, and the GDP grew at an impressive rate of 358.89%. These figures suggest that Wuhan experienced rapid economic growth, as it attracted significant investment and witnessed substantial industrial development. However, the relatively low population growth rate may suggest limited appeal for population migration or inflow or the presence of population outflow. Regarding area increase, both the urban core and urban fringe areas witnessed substantial growth, whereas the rural hinterland area experienced a decline. The growth rates of the urban core and urban fringe were relatively similar, both exceeding 80%. This indicates that Wuhan has experienced significant urban expansion and land development, leading to an increase in the availability of urban land. However, a comparison between the growth rate of the urban population and the GDP growth rate suggests that urban construction land primarily serves the development of the urban economy, whereas the growth rate of the urban population does not align with the pace of urban expansion. This reflects Wuhan’s urbanization model during that time, which was mainly characterized by “land urbanization”.

From 2010 to 2020, the urban core area increased to 744.575 km², with an average annual dynamic degree of 5.389%. In this stage, the dynamic degree decreases, indicating that the urban core area continues to increase, but the growth rate gradually slows down. In 2010, Wuhan’s overall urbanization reached 77.07%, exceeding 60% and reaching the later stage of the Northam curve of urbanization development, conforming to the trend of a gradual slowdown in urbanization rate during this period. The urban fringe area increased to 1220.104 km², with a dynamic degree of 4.106%. The dynamic degree decreased, indicating that the expansion speed of the urban core has slowed down, and Wuhan has gradually entered a stage of stable development, with the urban fringe area steadily increasing. The rural hinterland area continued to decrease to 6617.068 km², with a dynamic degree of 0.851%. Although the dynamic degree decreased, the rural hinterland area continues to decrease, and the urban core and urban fringe are still expanding overall. The growth rate of the population is increasing compared to the previous stage, with an average annual increase of 254,100 people and an average annual dynamic degree of 2.6%. In terms of population growth, Wuhan ranks among the top in the same period as provincial capital cities. Although the expansion speed of the city has slowed down, the attractiveness of the city to the population has increased, which is somewhat related to Wuhan’s gradual emphasis on the city’s ecological environment in its urban planning from 2010 to 2020. In addition, although Wuhan’s average annual GDP increment has increased, growing by 101.003 billion yuan per year, its overall growth rate has gradually slowed down, with the average annual dynamic degree decreasing from 35.7% to 18.31%.

In summary, from 2010 to 2020, the urban core area of Wuhan increased by 53.89%, and the urban fringe area increased by 41.06%. At the same time, the growth rate of the population is 25.97%, slightly higher than the previous period, but the GDP growth rate is 183.12%, significantly lower than the previous period. This indicates that Wuhan experienced relatively stable economic growth during this period while population growth was relatively fast. The rapid population growth may reflect the strengthening attractiveness and economic vitality of Wuhan, attracting more of the population to migrate. From the perspective of area increase, the growth rates of the urban core and urban fringe have relatively slowed down, at 53.89% and 41.06%, respectively. This indicates that during this period, Wuhan still carried out a certain degree of urban expansion and land development to meet the growing needs of the economy and population. Comparing the growth rates of the population and GDP, it can be observed that the increase in urban construction land is mainly for the development of the urban economy, but it also alleviates some of the relationship with household registration. This reflects that, during this period, the urbanization model of Wuhan has gradually transitioned from “land urbanization” to “population urbanization”.

5.2.2. Urban Fringe Spatial Change Analysis

The results of the transfer matrix calculation for the urban of Wuhan from 2000 to 2020 are shown in

Table 4. Percentage Shift in Urban Geographic Spatial Structure Change.

Name	2000–2010				2010–2020
	Urban Core	Urban Fringe	Rural Hinterland	Total	Urban Core	Urban Fringe	Rural Hinterland	Total
Urban Core	99.235	0.765	0.000	100	98.967	1.033	0.000	100
Urban Fringe	41.013	58.987	0.000	100	30.412	69.336	0.252	100
Rural Hinterland	0.480	7.426	92.094	100	0.037	8.508	91.455	100

,

Table 5. Transfer Matrix of Urban Core, Urban Fringe, and Rural Hinterland in Wuhan from 2000 to 2010 (Unit: km²).

	2010 Urban Core	2010 Urban Fringe	2010 Rural Hinterland	Total
2000 Urban Core	251.633	1.940	0.000	253.573
2000 Urban Fringe	194.545	279.806	0.000	474.351
2000 Rural Hinterland	37.662	583.237	7232.925	7853.823
Total	483.840	864.982	7232.925	8581.747

and

Table 6. Transfer Matrix of Urban Core, Urban Fringe and Rural Hinterland in Wuhan From 2010 to 2020 (Unit: km²).

	2020 Urban Core	2020 Urban Fringe	2020 Rural Hinterland	Total
2010 Urban Core	478.841	4.999	0.000	483.840
2010 Urban Fringe	263.056	599.747	2.179	864.982
2010 Rural Hinterland	2.678	615.358	6614.889	7232.925
Total	744.575	1220.104	6617.068	8581.747

and Figure 8.

During the period of 2000–2010, 0.765% of the urban core transformed into urban fringe, with no region converting to rural hinterland. Moreover, 41.013% of the urban fringe transformed into urban core, with no region converting to rural hinterland. Additionally, 0.48% of the rural hinterland transformed into urban core, and 7.426% transformed into urban fringe.

From 2010 to 2020, 1.033% of the urban core transformed into urban fringe, with no region converting to rural hinterland, 30.412% of the urban fringe transformed into urban core, and 0.252% transformed into rural hinterland. Moreover, 0.037% of the rural hinterland transformed into urban core, and 8.508% transformed into urban fringe.

In the subsequent decade, the speed of urban fringe transforming into urban core significantly decreased, indicating a slowdown in the expansion rate of Wuhan’s urban core. The proportion of direct transformation from rural hinterland to urban core has greatly reduced, and urban development follows the general pattern of rural hinterland—urban fringe—urban core, which better conforms to the general laws of urban development.

5.3. Forecasting of Urban Fringe

5.3.1. Analysis of Driving Factors of Urban Fringe in Wuhan Based on Logistic Regression Model

In this paper, the transformation types of urban geographic spatial structure are assigned values, and the results are shown in

Table 7. Assignment of Urban Geographic Spatial Structure Transformation Types.

Type of Transformation of Urban Urban Geographic Structure	Urban Core	Urban Fringe	Rural Hinterland
Urban Core	-	-	-
Urban Fringe	1	2	-
Rural Hinterland	-	3	4

. The logistic regression model is used with the relevant factors as independent variables to determine the impact of each factor on the transformation of urban geographic spatial structure type. Since some transformation types of urban geographic spatial structure have very few observations, this paper only discusses the driving factors for the transformation from urban fringe to urban core and from rural hinterland to urban fringe.

Taking into account the availability of relevant research and data factors, 15 relevant factors are selected as independent variables: ① natural environment factors: Slope, Slope direction, Elevation, Precipitation, Temperatures, Water body, Distance to water body; ② socio-economic factors: Population, GDP, whether within the overall planning scope; ③ transportation accessibility factors: Distance to primary road, Distance to the railway station, Distance to the highway exit, Distance to commercial center, Distance to government center. Using ArcGIS to process the independent variables and dependent variables, samples are obtained. The sample size and distribution range are shown in

Table 8. Sample Size and Distribution Range Included in the Binary Logistic Regression Model.

Year	Conversion Type	Sample Size	Distribution Range
2000–2010	1	608	2000 Urban fringe area
	2	8200	2000 Rural Hinterland
2010–2020	1	106	2010 Urban Fringe
	2	7530	2010 Rural Hinterland

. The samples are included in the binary logistic regression model. In the regression analysis, “backward-conditional” is selected to remove independent variables with smaller effects on the dependent variable and improve the fit of the model. Due to the absence of GDP grid data in 2020, the impact of GDP data is only discussed for the period of 2000–2010. After regression analysis, the HL values of all models are greater than 0.05, indicating good simulation fitting effects and that the simulation results can be adopted.

(1)

Analysis of the driving factors for the transformation from urban fringe to urban core

The regression results of the relevant factors for the transformation between the urban fringe and urban core are shown in

Table 9. Regression Analysis Results of the Driving Forces for the Change in Urban Fringe.

Independent Variable		2000–2010		2010–2020
Category	Factor	β	Exp(β)	β	Exp(β)
Natural environment Factors	Elevation	0.001	1.001	-	-
	Slope	−0.344	0.70	-	-
	Slope direction	-	-	-	-
	Precipitation	−2.428	0.088	-	-
	Temperatures	−2.045	0.12	-	-
	Distance to water body	0.015	1.015	0.006	1.006
	Water body	0.003	1.003	0.001	1.001
Socio-economic factors	Whether within the scope of the master plan	0.34	2.544	0.1	2.64
	GDP	0.001	1.001	-	-
	Population	0.002	1.002	−0.001	0.
Transportation accessibility	Distance to primary road	−0.571	0.565	−0.056	0.46
	Distance to the railway station	0.356	1.428	-	-
	Distance to freeway entrance/exit	0.274	1.316	−0.211	0.81
	Distance to commercial center	−0.10	0.87	−0.116	0.8
	Distance to government center	-	-	−0.237	0.78

Note: “-” indicates the amount of change in the regression analysis p > 0.05, which is not statistically significant and is not included in the regression equation.

.

According to Table 9, there are 13 significant driving forces for the transformation of the urban fringe to the urban core in Wuhan City from 2000 to 2010. The most influential factor is “Whether within the scope of the” with a dominance ratio of 2.544. The next influential factors are “Distance to the railway station”, “Distance to freeway entrance/exit”, “Distance to water body”, “Water body”, “Population”, “Elevation”, “GDP”, “ Distance to commercial center”, “Slope”, “Distance to primary road”, and “Temperatures” with dominance ratios of 1.428, 1.316, 1.015, 1.003, 1.002, 1.001, 1.001, 0.897, 0.709, 0.565, and 0.129, respectively. The factor with the smallest influence is “Precipitation” with a dominance ratio of 0.088.

From 2010 to 2020, there were eight significant driving forces for the transformation from the urban fringe to the urban core in Wuhan. The most influential factor was whether it was within the scope of the master plan, with an odds ratio of 2.694, followed by distance to the water body, water body, population, distance to the primary road, distance to the commercial center, and distance to freeway entrance/exit. Their odds ratios were 1.006, 1.001, 0.999, 0.946, 0.89, and 0.81, respectively. The least influential factor was distance to government center, with an odds ratio of 0.789.

Factors including natural environment, socio-economic factors, and transportation accessibility all had impacts on the transformation of the urban fringe. From 2000 to 2020, whether within the scope of the master plan was the most significant driving force for the transformation in Wuhan. This indicates that urban development and construction in this period were strictly implemented according to the urban master plan. National and regional policies and plans played a crucial role in the city’s economic development and land use [43]. In the 21st century, Wuhan, as a developed city in central China and a pilot zone for resource-saving and environmentally friendly societies, has experienced rapid economic development under the support of policies [44]. The urban fringe continued to expand outward, and the overall quality of Wuhan City has been continuously improved.

Additionally, in the period from 2000 to 2010, besides whether within the scope of the master plan, transportation accessibility factors such as distance to the railway station and freeway entrance/exit also had significant impacts on the transformation from the urban fringe to the urban core. This indicates that transportation accessibility played a relatively important role in the transformation process during this period.

Wuhan City, known as the city of a hundred lakes, has experienced the encroachment and filling of a large number of water bodies into construction land and widespread water pollution due to economic development and population increase in recent years [45,46]. In 2014, Wuhan City revised the “Wuhan Lake Protection Regulations” and implemented the strictest measures to protect and enhance the ecological landscape function of lakes [47,48]. In the period from 2010 to 2020, besides whether within the scope of the master plan, distance to water bodies and regional water bodies were the main factors influencing the transformation from the urban fringe to the urban core. This indicates that a good water-side ecological environment promotes the transformation from urban fringe to urban core. It also suggests that the city’s development in this period relied more on the requirements of environmental quality rather than the convenience of transportation, transitioning from relying on transportation and economy to pursuing ecological environment and landscape quality.

(2)

Analysis of Driving Factors in the Transformation from Rural Hinterland to Urban Fringe

The regression results of the related factors for the transformation between rural hinterland and urban fringe are shown in

Table 10. Regression Analysis Results of Driving Forces for Rural Hinterland Transformation.

Independent Variable		2000–2010		2010–2020
Category	Factor	β	Exp(β)	β	Exp(β)
Natural environment Factors	Elevation	-	-	-	-
	Slope	-	-	-	-
	Slope direction	-	-	−0.065	0.937
	Precipitation	-	-	3.357	28.696
	Temperatures	−8.538	0.00004	−6.241	0.002
	Distance to water body	-	-	−0.009	0.1
	Water body	−0.001	0.999	-	-
Socio-economic factors	Whether within the scope of the master plan	−0.867	0.42	−3.084	0.046
	GDP	−0.003	0.997	-	-
	Population	−0.003	0.997	-	-
Transportation accessibility	Distance to primary road	0.078	1.081	0.04	1.041
	Distance to the railway station	0.188	1.207	0.039	1.04
	Distance to freeway entrance/exit	-	-	0.554	1.739
	Distance to commercial center	0.025	1.026	0.456	1.578
	Distance to government center	0.054	1.056	−0.111	0.895

Note: “-” indicates the amount of change in the regression analysis p > 0.05, which is not statistically significant and is not included in the regression equation.

.

Based on Table 10, it can be seen that from 2000 to 2010, there were eight significant driving forces for the transformation of rural hinterland to urban fringe in Wuhan City. The most significant factor is Distance to the railway station, with an odds ratio of 1.207, followed by Distance to primary road, Distance to government center, Distance to commercial center, Water body, GDP, Population, and Temperatures, with odds ratios of 1.081, 1.056, 1.026, 0.999, 0.997, 0.997, and 0.00004, respectively. The least influential factors are Temperatures and Whether within the scope of the, with odds ratios of 0.00004 and 0.42. From 2010 to 2020, there were 10 significant driving forces for the transformation of rural hinterland to urban fringe in Wuhan City. The most significant factor is Precipitation, with an odds ratio of 28.696, followed by Distance to freeway entrance/exit, Distance to commercial center, Distance to primary road, Distance to the railway station, Distance to water body, Slope direction, Distance to government center, and Whether within the scope of the, with odds ratios of 1.739, 1.578, 1.041, 1.04, 0.991, 0.937, 0.895, and 0.046, respectively. The least influential factor is Temperatures, with an odds ratio of 0.002.

In the transformation process of rural hinterland, Distance to the railway station, Distance to freeway entrance/exit, Distance to commercial center, and Distance to primary road are the main driving forces. It can be seen that the urbanization of Wuhan City’s rural hinterland is greatly influenced by infrastructure, with transportation factors being the most significant. An efficient and convenient transportation network can reduce the temporal and spatial distance between urban and rural areas, expand the economic radiation circle of the main urban area, promote rural construction, and accelerate urban–rural integration [49].

Due to the large amount of construction in the urban core, its formation process is strictly guided and constrained by the urban master plan. On the other hand, the construction in the urban fringe is much less than in the urban core and is not subject to strict regulation. Its formation process is relatively less influenced by urban planning. The transformation from rural hinterland to urban fringe and its connection with transportation and economy are more closely related, with transportation and economic factors having a more significant impact, while natural environment factors do not show obvious influence.

5.3.2. Prediction of Wuhan City Urban Fringe Based on Logistic–CA–Markov Model

The spatial prediction result of Wuhan City in 2020 is shown in Figure 9. The Image Calculator tool in IDRISI was used to perform calculations between the predicted map of the Wuhan City urban fringe in 2020 and the segmented map of the Wuhan City urban fringe in 2020 to obtain the simulation accuracy result. According to the calculation result, the Kappa coefficient is 85.87%, indicating high reliability and ideal data results.

(1)

Geographic spatial structure analysis

The 2020 urban fringe forecast map of Wuhan City conforms to the characteristics of rapid development in Wuhan City from 2010 to 2020, with the city continuously expanding outward. The transformation from rural hinterland to urban fringe conforms to the conclusion in Section 5.3.1, where the city has experienced development around commercial centers, primary roads and railway stations. This is represented as a spotted urban fringe and a linear continuous urban fringe in the southeast in Figure 9.

(2)

Comparison between forecast results and planning objectives

The “Wuhan City Urban Master Plan (2009–2020)” (source: http://zrzyhgh.wuhan.gov.cn/xxfw/ghzs/202007/t20200707\_1396424.shtml, hereinafter referred to as the planning) (accessed on 11 March 2023) has played an important role in promoting sustainable development strategies and improving urban functions in Wuhan City from 2009 to 2020. The planning has mainly formed a multi-level and network-like urban system. Relying on the compound traffic corridor composed of urban expressway, skeleton trunk road and rail transit, six urban spatial development axes are constructed in the direction of Yangluo, Baoyi, Zhifang, Changfu, Hanjiang and Panlong. By integrating the new city and the new city groups that develop in conjunction with it, six new city groups (Figure 10) are formed, including the east, southeast, south, southwest, west and north.

By comparing the prediction results in this paper with the planning map, it can be concluded that the 2020 urban fringe forecast range in Wuhan City generally overlaps with the overall planning boundary, but there are differences in certain areas. These differences can be divided into two categories. The first category includes areas where the predicted range exceeds the planning boundary, such as the northwest mountainous area of Huangpi, Donghu Scenic Area, Qingling Lake and Qinglongshan Forest Park. In these areas, the urban fringe occupies ecological land, which is mostly located near water bodies and mountains, with dense commercial points, urban primary roads, and a certain population density, as well as favorable natural environment, socio-economic, and transportation accessibility conditions. They have the potential to be transformed into urban cores. However, the planning aims to consciously control such areas and strengthen the management of the planning to prevent the encroachment of urban fringe on ecological land. The consideration of the environment has gradually increased, and the path of urbanization in Wuhan City has gradually shifted from “quantity” growth to “quality” development. In addition, there are scattered urban fringes in various districts, which are mostly located in the original towns of Wuhan City. These areas have certain infrastructure and socio-economic conditions, some of which are located in areas with good ecological environments, such as water bodies, and are also close to ancient cultural sites or scenic areas, which are driving factors not considered in this study. These areas have potential for future development and can serve as reserve land for the next round of urban land planning.

The second type of area is the area beyond the planned boundary. Examples include Xiangkou and Suohe in Caidian District and Jiujie in Xinzhou District. The insufficient GDP, population, commercial centers, and primary roads in these areas have resulted in them not being designated as urban fringes despite being within the overall planning scope. The formulation of the plan does not match the local natural environment, socio-economic conditions, and transportation accessibility, indicating that these areas need more policy and resource support to implement the plan as intended (Figure 11).

The predicted results of the geographic spatial structure of Wuhan City in 2035 are shown in Figure 12.

(1)

Analysis of geographic spatial structure

The future urban geographic spatial structure of Wuhan City will continue to see a continuous rise in urban fringes, while rural hinterlands will continue to decline. It is predicted that by 2035, the area of urban fringes will continue to increase by 2691.72 km², occupying a larger proportion of the total area. In contrast, the area of rural hinterlands will shrink to 5152.87 km², and its proportion will continue to decrease. This development will lead to the continuous expansion of the city, putting greater pressure on the rural hinterlands.

(2)

Comparison of forecast results with planning goals

“The Wuhan City Territorial Spatial Plan (2021–2035) “(Source: http://zrzyhgh.wuhan.gov.cn/bsfw_18/ghpqgs/ghcags/202107/t20210713_1737015.shtml) (accessed on 11 March 2023) is the concrete implementation of the new strategies and requirements of the national, Hubei province, and Wuhan City’s “Fourteenth Five-Year Plan”. It promotes the implementation of Wuhan City’s Territorial Spatial Plan, advances the coordinated development of the Wuhan urban circle, and facilitates the high-quality development of the Yangtze River Economic Belt. Wuhan City’s Territorial Spatial Plan (2021–2035) focuses on seven aspects: optimizing the spatial pattern of land, respecting the natural environment, anchoring the ecological framework of clear mountains and beautiful waters, showcasing its charm, and building a world-renowned waterfront cultural city, among others. The scope covers the entire Wuhan area of 8569 square kilometers (Figure 13).

According to the prediction results, the expansion trend of the urban fringe and the planned urban space boundary are consistent, and their expansion ranges are highly similar. The 2035 urban fringe of Wuhan City has a similar outline to the planned urban space boundary, overall indicating that the predicted results include the planning scope. This is consistent with the conclusion in Section 5.3.1 that the main factors affecting urban development align with the overall planning, demonstrating that the planned urban space is in a matching state with Wuhan City’s natural environment, socio-economic conditions, transportation accessibility and other elements. In this prediction, there is no occurrence of the first type of situation where the planning scope exceeds the prediction scope.

The areas with the largest discrepancies between the prediction results and the plan are in the northwest of Huangpi District, the East Lake Scenic Area, and the west side of Jiangxia District. The northwest side of Huangpi District is primarily the Mulan Mountain Scenic Area, Mulan Ancient Gate Scenic Area, and Qingliangzhai Scenic Area. This region has numerous mountains and water bodies, with an excellent ecological environment. There are a few primary roads passing through the area, and along these roads, there are a large number of villages with some basic infrastructure to accommodate tourists. However, this area lacks commercial points. It has a beautiful ecological environment and good socio-economic conditions but average transportation accessibility, indicating its potential for conversion into the urban fringe. The East Lake Scenic Area is located at the East Lake, Yanxi Lake, Yandong Lake, Moshan Mountain, Longjiao Mountain and Jilong Mountain. There are several primary roads and highways passing through, and it has a considerable population base. The ecological environment, transportation accessibility, and socio-economic conditions in this area are all excellent; thus, it is designated as the urban fringe in the prediction. Lastly, the area near Yehu Lake in the northwest of Jiangxia District is close to water bodies with average ecological resources. However, it is a population gathering area for villages such as Hushan Village, Heli Village, Yaowan Village, and Tongsheng Village. This region has convenient transportation within the area, a relatively large population, and some commercial points, making it the most likely area among the three regions to develop into an urban fringe. Overall, these areas all have the conditions to be transformed into urban spaces, which is why the phenomenon of the urban fringe occupying the urban ecological space appears in the prediction results. However, the new round of the Territorial Spatial Plan places more emphasis on the construction of the overall ecological system of Wuhan City, aiming to create a “two axes, two rings, six wedges and multiple corridors” ecological framework. One of the goals is to make Wuhan City a world-renowned waterfront city. Based on this, many ecological protection policies have been formulated, such as protecting natural water networks and natural mountain bodies. The consideration of the natural environment dimension is the main reason for the disparity between this part of the area and the predicted map. However, this reflects Wuhan City’s continuous transformation towards a “quality” oriented urbanization development direction. In order to achieve the goals of the plan, strict control over the population growth around Huangpi District and Donghu Scenic Area is needed to prevent their development into population accumulation centers. In addition, regional commercial growth should be limited to prevent it from developing into urban spaces, and stricter control over the ecological space should be strengthened. As for the northwest area of Jiangxia District, it has excellent comprehensive conditions and can be identified as a key development area in the next round of the Territorial Spatial Plan.

Furthermore, there are varying degrees of sporadic urban fringe development in Xinzhou District, Jiangxia District, Hannan District, and Huangpi District, which are beyond the scope of the plan. These areas are generally located near water bodies and mountains, with a certain population base and high transportation accessibility, thus possessing the conditions for urban development. In the planning process, it is necessary to control the influx of population and the development of commerce in order to prevent them from becoming urban spaces (Figure 14).

The predicted results reflect the development trend of urban fringe without major intervention, and the overall planning reflects more of a vision. Where the forecast is different from the planning, we should pay special attention to it in future development and guide it according to the development intention of the national land space planning.

6. Conclusions and Discussion

This paper takes Wuhan City as the research object and studies the dynamic evolution and driving forces of the urban fringe of Wuhan City from 2000 to 2020 based on three phases of CLCD land use data and WorldPop population density data. The range of the urban fringe of Wuhan City from 2000 to 2020 is determined from three aspects: the Impervious Surface Ratio, landscape flocculation index and population. The changes in the Wuhan City urban fringe over 20 years are analyzed from the perspectives of urban fringe dynamic degree and geographic spatial structure transfer matrix. The driving factors influencing the changes in the urban fringe are quantitatively analyzed using the logistic regression model, and the evolution process and rules of the urban fringe are summarized. Finally, Wuhan City’s multiple periods of urban fringe range, the evolution rules of the urban fringe, and multiple sources of data are comprehensively predicted using the Logistic–CA–Markov model for the urban fringe of Wuhan City in 2035. The results show that:

(1)

By comprehensively using multiple sources of data, the urban fringe can be accurately delineated. The urban fringe is a transitional zone between urban and rural areas and has the dual attributes of both. Based on this, specific indicators of cities and rural areas can be combined with the detection of mutations and other methods to divide its range. The Impervious Surface Ratio, landscape flocculation index, and population density are attributes with significant differences between urban and rural areas. By detecting mutations based on these attributes, the geographic spatial structure of Wuhan City in 2000, 2010, and 2020 can be visualized. The results show that Wuhan City has shown a distinct “urban core—urban fringe—rural hinterland” layered structure in its development over the past 20 years;

(2)

The geographic spatial structure of Wuhan City has evolved in a spreading circle pattern. In 2000, the area of the urban fringe was 474.351 km²; in 2010, it expanded to 864.82 km², and by 2020, it had increased further to 1220.104 km². In contrast, the area of rural hinterland decreased from 7883.823 km² in 2000 to 6617.068 km² in 2020. Overall, there is a transformation from the urban fringe to the urban core and from the rural hinterland to the urban fringe. The dynamic degree of change for the urban core and urban fringe initially increased and then decreased, while the dynamic degree of change for the rural hinterland continued to increase. This indicates that from 2000 to 2010, Wuhan City was in an accelerated urbanization stage with rapid development, and the urban fringe extended deep into the rural hinterland. From 2010 to 2020, Wuhan City continued to develop rapidly but gradually stabilized. Low-density residential areas near the urban core were gradually included in the urban development scope, and the urban fringe “filled in” and developed into the urban core;

(3)

In the past 20 years, the transformation from the urban fringe to the urban core has been significantly influenced by the urban master plan. The ratio of urban fringe within the urban master plan scope transforming into the urban core is higher, indicating that under the strict urban construction control system in China, Wuhan City has implemented the urban master plan effectively in the past twenty years. In addition, from 2000 to 2010, urban transportation facilities have had a significant impact on the urban fringe; the impact of fringe conversion to urban core is more obvious. From 2010 to 2020, the Distance to water bodies and the number of water bodies in the region had a significant influence on the conversion of urban fringe to urban core. This indicates that in the first decade, the transformation of the urban geographic spatial structure is more easily influenced by transportation and economy, while in the second decade, the construction of the urban core focuses more on the pursuit of good landscape ecology. The driving force of urban construction has gradually shifted from a simple focus on transportation and economy to the pursuit of environmental quality, which is in line with the general law of urban development. It is recommended that urban managers, when hoping for the transformation from urban fringe to urban core, should pay attention to improving environmental quality. From 2000 to 2020, the transformation from rural Hinterland to urban fringe was significantly influenced by factors such as transportation facilities and Distance to commercial centers. The development and construction of rural areas are clearly driven by transportation and have not yet shifted to the pursuit of environmental quality;

(4)

By integrating multi-source data, multi-period urban fringe data, and the evolution law of urban fringe, the Logistic–CA–Markov model can effectively quantify and visualize the future range of urban fringe. The predicted map of Wuhan’s urban fringe in 2035 is basically consistent with the development trend of urban space in the 2035 Territorial Spatial Plan of Wuhan, with slight differences in the expansion area. It is predicted that the Wuhan urban fringe will continue to expand outward, with an estimated increase of about 1471.62 km², accounting for about 14% of the total area of Wuhan. At the same time, the area of rural Hinterland will further shrink, and its proportion will decrease to about 31%. This development trend indicates that urbanization will continue to advance, which will bring greater pressure and challenges to rural Hinterland. The expansion of the urban fringe will also bring more environmental and social problems. Therefore, urban managers need to take effective measures to control the speed and direction of urban expansion, protect the ecological environment and agricultural resources of rural Hinterland, and promote coordinated urban–rural development;

(5)

The prediction of urban fringe based on the Logistic–CA–Markov model has positive significance for rational resource allocation, macro-control, optimization of urban construction planning, and enhancement of the environmental protection function of the Territorial Spatial Plan. Urban planning is a conscious process of systematic analysis and decision-making. Planners improve the quality of decision-making by enhancing their understanding of various aspects of the problem and ensuring that established goals can be achieved in the future through a series of decisions. For the implementation of the 2035 Territorial Spatial Plan in Wuhan, planning should also limit the inflow of population from some ecologically beautiful areas with relatively developed transportation to prevent them from developing into urban spaces encroaching on urban ecological land. Secondly, for areas with good ecological environment, high transportation accessibility and certain socio-economic foundations, they should be considered as urban development land for the next round of planning.

Influenced by various factors, this study mainly followed the methods mentioned in the references when delineating the urban fringe, lacking new innovations. In the selection of driving factors for the evolution of urban fringe, due to the limited availability of some data, the factors used in the study are not comprehensive, lacking other factors that affect the evolution of the urban geographic spatial structure, such as historical and cultural conservation areas, original residential areas and so on. Due to the complexity of factors affecting urban development, it is also difficult to interpret the results of predictions. The possible reasons for the predicted results in this study can be further analyzed in depth.

In future research, we hope to further deepen the following aspects: firstly, more objective criteria for determining the urban fringe should be studied, and a more systematic system for delimiting and testing the urban fringe should be established. Secondly, combined with the new forecasting theory, more forecasting methods should be analyzed to improve the accuracy of urban fringe forecasting. Thirdly, further research should be conducted on the intrinsic mechanisms that affect urban development, the interrelationship between the development of the urban fringe and the urban master plan discussed, and reference recommendations for the formulation and implementation of the urban master plan from the perspective of the urban fringe provided.