Urban expansion simulation method based on gravity field and geographically weighted regression cellular automaton
Through the cellular automata method based on gravity field and geographic weighted regression, combined with Northum curve and multi-dimensional driving force, the problem that existing technology is difficult to fully consider the phased laws and spatial heterogeneity of urban agglomerations is solved, and a more realistic one is achieved and accurate urban expansion simulation.
Patent Information
- Application Number
- CN202311784086.7
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2023-12-22
- Publication Date
- 2024-12-03
- Estimated Expiration
- 2043-12-22
AI Technical Summary
The existing urban expansion simulation method is difficult to fully consider the phased laws of urban agglomerations, spatial heterogeneity, interactions between cities and ecological protection planning constraints, resulting in the simulation effect being inauthentic and accurate enough.
Using a cellular automata method based on gravity field and geographic weighted regression, a simulation model and micro-layout model of urban expansion scale are constructed, combining Northum curve and multi-dimensional driving force to simulate the urban expansion process of urban agglomerations.
The urban expansion process of urban agglomerations is realized more realistically, and the characteristics of urban agglomerations are fully considered, and the accuracy and reliability of the simulation are improved.
Smart Images

Figure FDA0005054943510000011 
Figure FDA0005054943510000014 
Figure FDA0005054943510000016
Abstract
Description
Technical Field
[0001] The invention belongs to the technical field of land planning, and relates to a method for simulating urban expansion, and in particular to a method for simulating urban expansion based on a gravity field and a geographically weighted regression cellular automaton. Background Art
[0002] Cellular automata (CA) is the most widely used model for intelligent prediction and simulation. Due to the simplicity, flexibility, intuitiveness and the ability to integrate time and space, the CA model is widely used in urban spatial system evolution simulation, urban land use simulation and urban expansion simulation research. Many scholars have tried to improve the CA model and have produced a variety of urban expansion application models. Li Xia et al. [1] The particle swarm algorithm, ant colony algorithm, genetic algorithm and artificial neural network algorithm were introduced into the CA model, and the GeoSOS system was developed to contribute to the dynamic simulation of intelligent urban space. [2] proposed a model of urban dynamic evolution; Clarke et al. [3] Proposed the SLEUTH model; Yang Liangjie et al. [4] The proposed object-based urban evolution (GeoObject-Urban) model. If urban agglomerations are used as the objects of CA simulation, it is necessary to base on the expansion characteristics and laws of urban agglomerations, taking into account the spatial differences in multiple aspects such as the functional attributes and interactive relationships of cities and even cells. [5-6] The expansion of urban agglomerations is a complex system evolution with the comprehensive effects of multi-dimensional processes. The improvement and optimization of a single level has very limited effect on enhancing the applicability of CA in the simulation of urban agglomeration expansion. [7] At present, most domestic and foreign studies regard urban agglomerations as a whole or treat cities as individuals, failing to comprehensively consider their connections and differences, and it is difficult to take into account the characteristics of spatial heterogeneity. [8-9] .
[0003] The references mentioned in this article are as follows:
[0004] [1] Li Xia, Ye Jiaan, Liu Xiaoping, Yang Qingsheng. Geographic simulation system: cellular automata and multi-agents[M]. Beijing: Science Press, 2007.
[0005] [2] Zhou Chenghu, Ouyang, Ma Ting, Qin Biao. Theoretical discussion on CA model for geographical system simulation[J]. Progress in Geography. 2009, 28(6): 833-838.
[0006] [3]Clarke, KC, Hoppen, S., Gaydos, L. (1997) A self-modifying cellularautomaton model of historical urbanization in the San Francisco Bayarea. Environment and Planning B-Planning&Design 24, 247-261.
[0007] [4] Yang Liangjie, Xue Chongsheng. Research on urban cellular automaton model based on geographical entities[J]. Journal of Beijing Institute of Technology (Social Sciences Edition). 2009, 11(1): 80-84.
[0008] [5] Qiao Jigang, He Jinqiang. Cellular automata and urban expansion simulation based on regional division[J]. Geography and Geographic Information Science. 2009, 25(3): 67-70.
[0009] [6] Li Shengquan, Hu Peng, Yan Weiyang. Division of urban influence range based on weighted Voronoi diagram [J]. Journal of Wuhan University (Engineering Edition), 2004, 37(1): 94-97.
[0010] [7] Fang Chuanglin, Mao Hanying, Ye Danian, Liu Shenghe, Bao Chao. (2016). Theory and methods of optimizing the spatial pattern of urban development in China. Science Press.
[0011] [8] Fang Chuanglin, Bao Chao, Ma Haitao. (2016b). 2016 China Urban Agglomeration Development Report. Science Press.
[0012] [9]Chunyang He, Norio Okada et al.Modeling dynamic urban expansion processes incorporating a potential model with cellular automata.Landscapeand Urban Planning.2008, 86:79-91. Summary of the invention
[0013] In order to solve the above technical problems existing in the background technology, the present invention provides an urban expansion simulation method based on gravity field and geographically weighted regression cellular automaton, which can more realistically simulate the urban expansion process of urban agglomerations.
[0014] In order to achieve the above object, the present invention adopts the following technical solution:
[0015] A method for simulating urban expansion based on gravity field and geographically weighted regression cellular automaton, characterized in that: the method for simulating urban expansion based on gravity field and geographically weighted regression cellular automaton comprises the following steps:
[0016] 1) Determine the transition probability value of non-urban cell i;
[0017] 2) Determine the town size of city j;
[0018] 3) According to the transition probability value of the non-urban cell i obtained in step 1) and the urban scale of city j determined in step 2), the state of cell i in city j during iteration is constructed, and the urban expansion is simulated according to the state of cell i in city j during iteration.
[0019] Preferably, in step 1), the conversion probability value of the non-urban cell i is calculated as follows:
[0020]
[0021] in:
[0022] S i is the urban development suitability of cell i predicted by the geographically weighted regression model;
[0023] is the urban development density of cell i in its neighborhood at time t;
[0024] Z i is the planning and ecological constraints imposed on cell i, representing the limitations of planning and ecological protection;
[0025] It is a dynamic random number that evolves with the iteration time, representing the influence of unknown factors in the process of urban expansion.
[0026] Preferably, the S i ∈[0,1], the S i The expression is:
[0027]
[0028] in:
[0029] x ik is the value of the independent variable at cell unit i, and the independent variables include the elevation at cell unit i, the slope at cell unit i, the distance between cell unit i and the main river channel, the distance between cell unit i and the core city node in the Bay Area, the distance between cell unit i and the important city node in the Bay Area, and the distance between cell unit i and the ordinary city node;
[0030] (u i ,vi ) is the coordinate of cell i;
[0031] β 0 (u i ,v i ) is a constant term at cell i;
[0032] β k (u i ,v i )(k=1,2,...,m) is the regression coefficient of the kth independent variable in the regression analysis, and m is the number of independent variables;
[0033] ε i is the error term.
[0034] Preferably, the The expression is:
[0035]
[0036] in:
[0037] N urban is the number of cells of type town in the neighborhood W;
[0038] r is the radius of the window of the neighborhood range W, where r = 1;
[0039] The neighborhood range W is the 8 adjacent cells around the cell i.
[0040] Preferably, the Z i The calculation formula is:
[0041]
[0042] in:
[0043] Δx W It is a key area defined in urban planning and is a meta-logical variable;
[0044] represents the integration operator;
[0045] Represents the state of the W neighborhood of cell unit i;
[0046] Ψ i It is an ecological retention control condition.
[0047] Preferably, in step 2), the calculation formula for town size is:
[0048]
[0049] in:
[0050] and are the town sizes of city j at time t and t+1 respectively;
[0051] W a and W b They represent the weights of the two processes of balancing the intrinsic thrust and the network radiation force, respectively. a +W b =1;
[0052] is the intrinsic thrust of city j at time t+1;
[0053] ) is the network radiation power of city j at time t+1;
[0054] is the proportion of urban development in city j at time t;
[0055] ) is the urban development ratio of city j at time t-1;
[0056] is the upper limit of the town size of city j;
[0057]
[0058] Preferably, in the method for calculating the town size, The expression is:
[0059]
[0060] in:
[0061] Curve(.) is the fitted urban agglomeration Northam curve function;
[0062] is the proportion of urban development of city j at time t;
[0063]
[0064] in:
[0065] a, b parameters representing urbanization level;
[0066] μ and ρ are parameters that control the shape of the Sigmoid function curve;
[0067] t is time;
[0068] ) is the urban development ratio of city j at time t-1;
[0069] In the calculation method of town size, the DW j The expression is:
[0070] DW j =Σ k DW jk
[0071] in:
[0072] DW jk is the strength of the spatial connection between city j and other cities k, j≠k;
[0073] The DW jk The expression is:
[0074]
[0075] in:
[0076] M j and M k are the urbanization quality of cities j and k respectively;
[0077] D jk is the Euclidean distance between cities j and k;
[0078] The M j The expression is:
[0079] M j =GDP j ×POP j ×Tran j ;
[0080] GDP j is the gross regional product of city j;
[0081] POP j is the total population of city j;
[0082] Tran j is the number of public transportation services in city j.
[0083] Preferably, the expression of the state of cell i in the city j during iteration is:
[0084]
[0085] in:
[0086] and are the states of cell i in city j at iterations t and t+1 respectively;
[0087] f j (.) is the urban expansion simulation conversion rule of city j;
[0088] is the town size of city j at the tth iteration;
[0089] is the transition probability value of non-urban cell i at the tth iteration.
[0090] The advantages of the present invention are:
[0091] The present invention provides a method for simulating urban expansion based on gravity field and geographically weighted regression cellular automata, which fully considers the phased laws of urban agglomeration development, spatial heterogeneity, interaction between cities, historical development trajectory, macro land demand, and ecological protection planning constraints. The gravity field model and Northam curve are used to construct a simulation model of urban expansion scale, and the geographically weighted regression model and cellular automata are used to construct a micro-layout model of urban agglomeration expansion, which fully considers the mutual connection and difference between cities in the urban agglomeration, spatial heterogeneity and other characteristics. The method of the present invention can more likely simulate the urban expansion process of urban agglomerations in my country. BRIEF DESCRIPTION OF THE DRAWINGS
[0092] Figure 1 It’s the Northam curve;
[0093] Figure 2 It is the urban built-up area and ecological protection area in 2000, 2010 and 2020;
[0094] Figure 3 It is the inverse S curve fitting of the urban land development ratio;
[0095] Figure 4 is the interaction between cities and the size of node towns;
[0096] Figure 5 It is the driving factor and suitability of the GWR model;
[0097] Figure 6 It is the forecast of urban land layout in Guangdong-Hong Kong-Macao Greater Bay Area in 2050;
[0098] Figure 7 It is the error detection of urban network scale under different scenarios;
[0099] Figure 8 This is the actual distribution map of urban land in the Guangdong-Hong Kong-Macao Greater Bay Area in 2020;
[0100] Fig. 9 This is a simulation map of urban land in the Guangdong-Hong Kong-Macao Greater Bay Area in 2020;
[0101] Fig.10 It is planning case verification;
[0102] Fig.11 It is the land development intensity from 2020 to 2050. DETAILED DESCRIPTION
[0103] The advantages of the technical solution of the present invention will be further described below in conjunction with the accompanying drawings.
[0104] A method for simulating urban expansion based on gravity field and geographically weighted regression cellular automaton, comprising the following steps:
[0105] 1) Simulation of urban scale network evolution
[0106] The simulation of the evolution of urban scale comprehensively adopts the Markov chain model to quantify the historical inertia of the evolution of urban scale in each city, characterizes the internal driving force of the evolution of each city development stage by fitting the Northam curve of the urban agglomeration, establishes a gravity field model to measure the radiation power of the urban agglomeration network, and relies on multi-source information fusion technology to quantify the ecological constraints of urban expansion. The quantification of various driving forces is as follows:
[0107] 1-1 Historical inertia force:
[0108] The traditional Markov chain describes a state evolution sequence. In the Markov process, the future state depends only on the known present and has nothing to do with the past. It can be described in mathematical language as follows: in, and are the town sizes of city j at time t and time t+1, M j is the transition probability of the town size of city j to the Markov process, It is to obtain the increment of each city's town size from the historical process. According to the Markov chain process of each city, the historical inertia force of its town size evolution is quantified.
[0109] Due to ecological constraints on urban expansion, the urban size of each city is Q j There is an upper limit Therefore, the evolution of urban agglomeration town size always follows rules, namely ecological constraints, and calculate Z i Similarly, the maximum urban size of each city is calculated based on the ecological protection and various planning restrictions of each city. Ecological constraints as a simulation of urban agglomeration expansion.
[0110] 1-2 Intrinsic thrust:
[0111] An urban agglomeration is an integrated and homogenized urban group formed by a number of closely connected cities. Therefore, the urbanization development history and trends of each city are similar, while their urbanization levels are different due to the different development stages they are in. The present invention intends to statistically analyze the historical changes in the scale of urban land use of each city in the urban agglomeration, using the Urban Development Proportion (UDP) of each city at different periods as sample data, and fitting the urbanization Northam curve of the urban agglomeration to depict the evolution of the development stages of the cities in the urban agglomeration and quantify the inherent driving force of the general law of urbanization on the evolution of urban scale. UDP is the urban scale Q of the city at time t. t Occupancy of its developable land Q max The ratio of UDP t =Q t / Q max .
[0112] According to existing research, the evolution curve of urbanization level, that is, urban development ratio UDP, is an inverse S curve, that is, Northam curve, which is very close to the shape of Sigmoid function curve. The evolution curve of urbanization speed is close to the normal distribution curve, such as Figure 1 Since the Sigmoid function is the constructor of the Logistic model, the urban development ratio of each city in the urban agglomeration at different times is used as a sample to fit the curve and obtain the parameters of the Logistic equation:
[0113]
[0114] In the formula, a and b represent the parameters of urbanization level; μ and ρ are parameters that control the shape of Sigmoid function curve; by fitting and training with the urban development ratio of each city in different periods as samples, the parameters of the Logistic equation with time t as the independent variable and urban development ratio UDP as the dependent variable can be obtained.
[0115] The inner thrust of city j The calculation formula is: Where Curve(.) is the fitted urban agglomeration Northam curve function. Intrinsic thrust The actual meaning of is the ratio of the predicted urban development proportion increment to the historical urban development proportion increment. If the urbanization of city j is in the acceleration stage, then otherwise
[0116] 1-3 Network radiation:
[0117] Integrate the flow of population, economy, transportation and other factors between cities, build a multidimensional urban flow model that characterizes the interaction of urban networks, and quantify the network radiation of urbanization development based on the urban agglomeration network structure generated by distance relationships. Calculate the interaction process between cities based on the gravity field model, take cities as nodes, and generate the network connection structure between cities in the urban agglomeration based on Euclidean distance. The strength of the spatial connection between city j and other cities k (j≠k) is:
[0118]
[0119] In the formula, M' j , M' k is the urbanization quality of cities j and k, D jk is the Euclidean distance between cities j and k, where M' j =GDP j ×POP j ×Tran j , by GDP j , Total Population POP j 、Number of public transportation services j The network radiation power of city j is DW j =Σ k DW jk The gross regional product represents its economic strength. The spatiotemporal big data service of Baidu Map Smart Eye is used to obtain the population in-migration and out-migration data of each city in the urban agglomeration during the selected period, and the population flow intensity between cities is calculated. The average monthly number of public transportation such as high-speed rail, motor train, ordinary express train and long-distance bus between cities is collected from ticketing websites to calculate the traffic flow intensity between cities.
[0120] 1-4 Ecological constraints:
[0121] According to the relevant planning and ecological protection documents of the urban agglomeration, the ecological protection and various planning restriction boundaries of each city are obtained by georeferencing and vectorization, and the maximum urban scale of each city is calculated based on the above range restrictions. And use it as an ecological constraint for urban agglomeration expansion simulation.
[0122] 1-5 Urban scale network evolution model integrating multi-dimensional driving forces:
[0123] The town scale network evolution model integrates the above multi-dimensional driving forces, and its scale evolution is calculated as follows:
[0124] Where W a +W b =1,
[0125] Among them, W aand W b It represents the weight of balancing the intrinsic thrust and network radiation force. The sum of these two processes is 1. The default value is set to 0.5. It can be debugged and optimized according to actual needs.
[0126] 2) Conversion rules for town expansion simulation:
[0127] 2-1 Calculate the urban growth suitability of each grid cell in the urban space based on the geographically weighted regression model:
[0128] Suitability is generally used to represent the probability of urban growth at the grid scale (the probability can also be interpreted as the suitability of urban growth). Suitability is the basis for conducting spatiotemporal evolution simulation of urban agglomerations. Usually, the evaluation or estimation of grid probability needs to consider a series of natural and human environmental variables, such as topography, transportation network, population density, etc. However, in terms of medium- and long-term simulation of urban agglomeration scale, factors such as transportation network and population density are unstable spatial driving variables. For example, the construction of a new high-speed railway or expressway can completely change the driving pattern, and the relaxation of household registration control may lead to large-scale population growth in a region. These unstable factors have a great impact on China's strongly controlled urban development. In summary, relatively stable spatial factors are selected as environmental variables driving the evolution of urban agglomerations. These independent variables mainly include: a) elevation, b) slope, c) distance from the main river (Pearl River system), d) distance from the core city nodes of the Bay Area (Guangzhou, Shenzhen, Hong Kong, and Macau), e) distance from important city nodes in the Bay Area, and f) distance from ordinary city nodes. The dependent variable can be determined by the changes in urban growth in historical data, which is a binary logical variable. When the state of a cell unit changes from non-urban land to urban land, the value is 1, otherwise it is 0.
[0129] Each city at the urban agglomeration scale may have its own evolutionary characteristics, so local autocorrelation will be strengthened. In order to better consider the geographic spatial differentiation at the urban agglomeration scale, the geographically weighted regression method is used to estimate the initial probability of urban growth at each grid scale. The calculation formula is as follows: In the formula, x ik is the value of the independent variable at cell i, mainly including a) elevation, b) slope, c) distance from the main river (Pearl River system), d) distance from the core city nodes in the Bay Area (Guangzhou, Shenzhen, Hong Kong, and Macau), e) distance from important city nodes in the Bay Area, and f) distance from ordinary city nodes. (u i ,v i ) is the coordinate of the cell i; β 0 (u i ,v i ) is the intercept term; β k (u i ,vi )(k=1,2,...,m) is the regression coefficient of the kth independent variable in the regression analysis, m is the number of independent variables; ε i is the error term.
[0130] 2-2 Calculate the neighborhood attributes of each cell unit in the urban space:
[0131] Neighborhood Development Density It is the urban development density function of cell unit i within the neighborhood range W at time t. The neighborhood range W generally refers to the 8 adjacent cells around cell i. The calculation formula of neighborhood development density is: Among them, N urban is the number of town cells in the neighborhood, r is the radius of the neighborhood window, where r = 1.
[0132] 2-3 Calculate the spatial constraints of each cell in the urban space:
[0133] Spatial constraints mainly consider indicators such as planning and ecological protection, and the calculation formula is: Where Δx W It is a key area designated by urban planning and is a meta-logical variable. represents the integration operator, represents the state of the W neighborhood of cell i, that is, the cell within the neighborhood that is planned as a key development area is marked as 1 during calculation, so as to strengthen the spatial guidance effect of planning; and Ψ i is the ecological preservation control condition, Ψ ij =0 indicates an ecological reserve unit, where urban development is not allowed.
[0134] 2-4 Calculate the random disturbance of each cell unit in the urban space:
[0135] It is a dynamic random number that evolves with iteration time, representing the influence of unknown factors in the process of urban expansion.
[0136] 2-5 Construction of conversion rules
[0137] Transition rules of cellular automata models Calculate the comprehensive growth probability of cell unit i at time t, when Then the cell unit is converted into a town; otherwise, the cell unit is not converted, and λ is set to 0.3 here; By suitability S i , Neighborhood Development Density Spatial Constraint Z i and random perturbations Coupling obtained.
[0138] 3) Scenario setting for urban expansion simulation:
[0139] In order to test the performance of the model and verify its advantages, four simulation scenarios are set up, according to the historical inertia advancement (ErS1), the scenario driven by the anti-S curve law (ErS2), the scenario after the network radiation is pulled up (ErS3), and the scenario ErS4 after comprehensive consideration of the three driving forces. The model comparison uses the root mean square error to evaluate the training and prediction accuracy of urban development potential, and uses Kappa, Figure of Merit (FoM) and other accuracy indicators to evaluate the accuracy of urban expansion simulation, and conducts planning case demonstration to achieve model performance testing and optimal parameter adjustment.
[0140] The application of the method of the present invention will be described below with reference to specific embodiments.
[0141] (1) Data preparation
[0142] A large amount of historical data is needed to spatialize and spatially register the existing land use data from 2000 to 2010, socioeconomic statistics, and the results of manual interactive interpretation of historical remote sensing data, to form a land use sequence map from 2000 to 2010. The land use data is collected and entered into the new urbanization spatial layout simulation and prediction system for system analysis.
[0143] The total land area of the Guangdong-Hong Kong-Macao Greater Bay Area is approximately 5.6×104km 2 , consisting of Guangzhou City and Shenzhen City in Guangdong Province, and the two special administrative regions of Hong Kong and Macao, it is the region with the most complex administrative system in the world. From the perspective of secondary administrative units, Guangdong Province consists of 50 county-level units, and Hong Kong and Macao have 18 and 5 secondary administrative regions respectively. These regions not only differ in size, but also in population size and socio-economic development status. In order to better carry out simulation and analysis, this study refers to the idea of dividing standard regional units related to spatial planning, that is, standard regional statistical units, to solve the problem of incomparability of different administrative units, such as NUTS of the European Union's spatial planning and Hong Kong Planning Unit (Tertiary Planning Unit, TPU). The study divides the Greater Bay Area into county-level administrative units, combines the land area and the current urbanization level, and reorganizes 36 standard regional units, coded as follows: Figure 1 shown.
[0144] First, the study used the historical changes from 2000 to 2010 to explore the spatiotemporal evolution of urban agglomerations, and used the observation data from 2020 to test the simulation accuracy. Due to the large scale of the Greater Bay Area, Landsat satellite images were used to extract the construction land in the Greater Bay Area, and Google Earth high-resolution images were used for sampling detection to evaluate the classification accuracy. The construction land pattern in the Greater Bay Area in 2000, 2010 and 2020, the average classification accuracy of construction land is about 90%, which fully meets the research application needs.
[0145] Secondly, based on the geographical conditions of the Greater Bay Area, this study further included the main rivers and surrounding wetlands in the Pearl River Basin, as well as mountains with a relative height difference of more than 50m into the ecological conservation area, prohibiting urban development. In China, although permanent basic farmland is prohibited from development, permanent basic farmland is not static, but can be adjusted appropriately according to development needs, but the approval level is very high. As a national strategy, the Greater Bay Area will inevitably occupy part of the cultivated land for urbanization development. However, once the ecological space is severely damaged, the regional ecological environment cannot be guaranteed. In summary, only the completely rigid constraints of the basic ecological control line are considered, and the impact of basic farmland is not considered for the time being. Basic ecological pattern controlled by ecological water and ecological forests Figure 3 shown.
[0146] Thirdly, based on multiple sources such as the China City Statistical Yearbook, Baidu Maps, and web crawlers, we constructed a basic database based on the GDP of the Greater Bay Area, data on population inflow and outflow, and the average monthly number of public transportation services such as high-speed trains, EMUs, ordinary trains, and long-distance buses.
[0147] Finally, the following historical data is sorted out for analysis:
[0148] Figure 1 - Location map of Guangdong-Hong Kong-Macao Greater Bay Area;
[0149] Urban built-up areas and ecological protection areas in the Guangdong-Hong Kong-Macao Greater Bay Area in 2000, 2010 and 2020;
[0150] (2) Data preprocessing
[0151] TM remote sensing images have 7 bands. Here, false color composite images of bands 4, 3, and 2 are selected with an image resolution of 30 meters. Erdas Imagine remote sensing image processing software was used to supervise and classify TM remote sensing images from 2000 to 2020, which were divided into two categories: urban construction land and non-construction land. The classification results require that the classification template passes the error test with an accuracy greater than or equal to 85%. In addition, the classification result map uses the Kappa coefficient to evaluate the classification accuracy. For Kappa coefficients lower than 0.6, the classification results should be corrected by visual interpretation based on computer interpretation.
[0152] Since the projection coordinates and geographical scope of remote sensing images in different years are different, the classification map should be combined with GIS tools for image registration and image cropping, so as to facilitate the comparative analysis of data over the years. At the same time, the main roads, national roads, provincial roads and other important routes in the traffic map are digitized and stored. At the same time, important landforms are selected according to the maps of previous years and digitally tracked and stored.
[0153] (3) Simulation framework
[0154] The urban agglomeration is regarded as a network, each city node is used as a network node, each network node controls the scale of urban development, the Markov process is used to simulate the self-transformation of network nodes, and the group interaction is used to correct the node self-transformation process. All nodes interact with each other, and the network automaton is constructed in turn. The law of urbanization process (inverse S curve) and ecological red line constraints (control of maximum development scale) are used. The evolution scale of each node is simulated by the network automaton. Then, the evolution of the spatial pattern is simulated under the control of each city scale through the partitioned cellular automaton.
[0155] (4) Simulation of urban scale network evolution
[0156] First, the urban agglomeration is divided into several sub-regions, each of which serves as an important node in the urban agglomeration and constitutes an urban network. For example, Guangdong Province can mainly use county-level administrative units and merge some overly small urban units. Cities such as Shenzhen, Dongguan, and Zhongshan, which are only county-level in size, can be directly treated as one unit, while Hong Kong and Macao can each become an independent unit due to their particularity. Secondly, the initial transfer probability is obtained by using the urban growth of each node from t-1 to t, and the Markov chain process is run under the constraints of the law of urbanization process to predict its new scale from t to t+1. Thirdly, the gravity field radiation model is used to calculate the comprehensive radiation potential energy of each urban node in the entire network at time t, and the intensity is used as the weight coefficient to make the urban network oscillate and rebalance, that is, the new scale from t to t+1 is redistributed to each node according to this weight (the number of network oscillations can be based on needs. In theory, the more oscillations, the more balanced the nodes will be). Fourthly, the ecological security control range is analyzed according to the needs of the ecological security pattern, and the maximum buildable scale of each city is estimated (the new reclamation is highly regulated by policies, and this paper does not consider the changes in the sea area), that is, each node must not exceed the maximum development scale.
[0157] The inverse S curve that fits the Greater Bay Area is fitted. The study integrates the development ratios of each node in the Greater Bay Area in 2000 and 2010, and extends it to 1 according to the trend forward method. The law of the Greater Bay Area's inverse S curve is as follows: Figure 2The specific application strategy of the urban development ratio function is that if the development ratio of the current node n is close to the time t (such as 0.1792), then driven by the law of urbanization, its next moment is expected to jump to the time t+1 (i.e. 0.3320). Through the promotion of the S development law, the thrust coefficient of the land urbanization process can be obtained to avoid linear uniform growth.
[0158] Regarding the network radiation pull, we first used the Generate Spatial Weights Matrix tool of ArcGIS to identify the inverse distance weights between urban network nodes and calculate the network radiation force. After four iterations, we studied the external radiation potential network of each urban node in the Greater Bay Area in 2050. Figure 3 As shown: 1) The yellow circle in the figure represents the size of the city in 2010, and the red circle represents the relative size of the town in 2050; after proportional processing, the larger the circle, the larger the scale, and the greater the difference between the red circle and the yellow circle, the greater the increase in the unit in the later period; on the contrary, the small difference between the red circle and the yellow circle indicates that the unit has faced the growth limit or has no growth potential. 2) The blue line in the figure represents the intensity of interaction between the urban network. The thicker the line, the stronger the driving effect of the node on the development of the surrounding cities, and the stronger the synchronization between the two.
[0159] (5) Conversion rules for urban expansion simulation
[0160] Urban growth simulation modeling usually uses data mining methods to explain the relationship between independent variables and dependent variables. The most classic and widely used method is logistic regression. Logistic regression treats the region as an overall fit, and each spatial driving variable shares a regression coefficient. However, at the scale of urban agglomerations, each city may have its own evolutionary characteristics, so local autocorrelation will be strengthened. In order to better consider the geographic spatial differentiation at the scale of urban agglomerations, the GWR method is used to estimate the initial probability of urban growth at each grid scale. In order to better use the GWR method, the study resampled the 30m resolution logistic changes from 1995 to 2015 (0 means no change, 1 means newly added city) according to the hectare grid (100m*100m) to a density of 0-1 continuous distribution, and the spatial distance variable was standardized by 0-1 before performing geographically weighted regression. The result of GWR estimation is only a probability distribution, which needs to be briefly revised to become the probability of urban growth suitability. The revision process mainly includes: 1) re-marking the probability of grids in ecological protection areas to 0, 2) re-marking the grids that are already urban units to 1, and 3) marking the areas that the government strategic planning has clearly identified as requiring key development, and integrating them into the landscape operator for guidance. Figure 4 shown.
[0161] (6) Scenario simulation of urban expansion in urban agglomerations
[0162] The study predicts the scale change scenarios of each unit from 2020 to 2050, and the results are shown in Table 1. According to the urban network automaton evolution prediction, the construction scale of the Greater Bay Area will be about 15277.14km by 2050. 2 , accounting for 56,000 km2 of the Greater Bay Area 2 The rate of infection in the entire Bay Area remained below the 30% warning red line.
[0163] Table 1 Prediction of urban growth scale in the Guangdong-Hong Kong-Macao Greater Bay Area from 2020 to 2050 (Unit: km 2 )
[0164] From the perspective of the distribution pattern of urban scale, by 2050, the ultimate development model of urbanization in the Greater Bay Area will present a more balanced development model. In particular, under the implementation of the new national land space development strategy, the western bay area will be effectively improved, and the "J"-shaped circle structure of the bay area will be more prominent. Figure 5 Shown are the simulated and predicted urban land sizes in the future.
[0165] (7) Verification and evaluation of urban expansion models
[0166] 7-1 Number of Town Sizes
[0167] In order to test the prediction accuracy of the network automaton, the study compared four prediction modes, namely, the scenario completely following the inertial advancement of 2000-2010 (ErS1), the scenario after being driven by the anti-S law (ErS2), the scenario after the network radiation was pulled up (ErS3), and the scenario ErS4 after comprehensive consideration of the three driving forces. The errors of the 2020 scale simulation prediction values of 36 units were compared with the true values of remote sensing observations in 2020. Overall, the average errors of ErS1, ErS2, ErS3 and ErS4 were 17.39%, 15.82%, 12.86% and 9.83%, respectively, and the unit errors simulated by the network automaton were relatively stable overall (the ideal error curve value was 0 and was a straight line). It can be seen that the adopted network automaton model framework is more reliable than the traditional simulation of Markov processes. The simulation model reduced the detection error to less than 10%, and it can be fully applied to simulate 2050, such as Figure 6 shown.
[0168] 7-2 Spatial layout of urban land
[0169] In terms of spatial pattern simulation accuracy, the study compared the simulated new value added from 2010 to 2020 with the new value added from remote sensing observations. The Kappa and FOM indices detected were 0.4416 and 0.2926, respectively, and the accuracy met the requirements. Figure 7 As shown, for comparison, Figure 8 This is the simulated urban land spatial distribution of the Guangdong-Hong Kong-Macao Greater Bay Area in 2020.
[0170] 7-3 Planning case verification of simulation results
[0171] Five cases were selected from the Greater Bay Area, namely a) Zhaoqing Development Zone, b) Guangzhou Knowledge City, c) Huizhou Central City, d) Guangzhou Nansha Development Zone, and e) Zhuhai Doumen. The base map in the figure is the latest high-resolution satellite image in 2020. From the comparison chart, it can be seen that the images basically have road networks or fill, etc., indicating that urbanization is underway in the area. Through typical case examples, the simulation model constructed in this article has well inverted the future urbanization development needs of the Greater Bay Area. Since the urban network reconstructs the development model of urban agglomerations, and the effect of this reconstruction may not be immediately observable (such as changes from 2010 to 2020), combining it with the government development plan can better illustrate the reliability of the simulation warning. The results of the planning case verification are as follows. Fig. 9 shown.
[0172] 7-4 Convergence verification of development intensity stage of simulation results
[0173] China has generally entered the final stage of urbanization and is unlikely to grow on a large scale, although the demand for bay area development still exists. Fig.10 As shown, the various units in the Greater Bay Area show a clear stratified convergence trend. If the development intensity is divided in equal proportion, there are 10 units in the first echelon with a development intensity of 0.85-1; 8 units are in the second echelon with a development intensity of 0.5-0.7; 11 units are in the third echelon with an intensity of 0.25-0.45; and 7 units are in the fourth echelon with an intensity of 0.12-0.24.
[0174] Overall size and historical stage management. Although some cities in the Greater Bay Area have completed urbanization, the Greater Bay Area as a whole is still in the acceleration stage. In particular, the implementation of the Guangdong-Hong Kong-Macao Greater Bay Area Development Plan has made the integrated development of cities in the Bay Area a new driving force. The government plan predicts that it will enter a relatively stable period between 2035 and 2050. Fig.11 As shown in the figure, as the Chinese government vigorously promotes the system of intensive and economical land use, stratified convergence is an inevitable trend.
Claims
1. A method for simulating urban expansion based on gravity field and geographically weighted regression cellular automata, characterized in that: The urban expansion simulation method based on gravity field and geographically weighted regression cellular automaton comprises the following steps: 1) Determine the conversion probability value of the non-urban cell i, wherein the conversion probability value of the non-urban cell i is calculated as follows: in: S i is the urban development suitability of cell i predicted by the geographically weighted regression model; is the urban development density of cell i in its neighborhood at time t; Z i is the planning and ecological constraints imposed on cell i, representing the limitations of planning and ecological protection; It is a dynamic random number that evolves with the iteration time, representing the influence of unknown factors in the process of urban expansion; The S i ∈[0,1], the S i The expression is: in: x ik is the value of the independent variable at cell unit i, and the independent variables include the elevation at cell unit i, the slope at cell unit i, the distance between cell unit i and the main river channel, the distance between cell unit i and the core city node in the Bay Area, the distance between cell unit i and the important city node in the Bay Area, and the distance between cell unit i and the ordinary city node; (u i ,v i ) is the coordinate of cell i; β0(u i ,v i ) is the constant term at cell i; β k (u i ,v i )(k=1,2,...,m) is the regression coefficient of the kth independent variable in the regression analysis, and m is the number of independent variables; ε i is the error term; Said The expression is: in: N urban is the number of cells of type town in the neighborhood W; r is the radius of the window of the neighborhood range W, where r = 1; The neighborhood range W is the 8 adjacent cells around the cell i; The Z i The calculation formula is: in: Δx W It is a key area defined in urban planning and is a meta-logical variable; represents the integration operator; Represents the state of the W neighborhood of cell unit i; Ψ i Ecological conservation control conditions 2) Determine the town size of city j; 3) According to the transition probability value of the non-urban cell i obtained in step 1) and the urban scale of city j determined in step 2), the state of cell i in city j during iteration is constructed, and the urban expansion is simulated according to the state of cell i in city j during iteration.
2. The urban expansion simulation method based on gravity field and geographically weighted regression cellular automaton according to claim 1, characterized in that: In step 2), the calculation formula for town size is: in: and are the town sizes of city j at time t and t+1 respectively; W a and W b They represent the weights of the two processes of balancing the intrinsic thrust and the network radiation force, respectively. a +W b =1; is the intrinsic thrust of city j at time t+1; ) is the network radiation power of city j at time t+1; is the proportion of urban development in city j at time t; ) is the urban development ratio of city j at time t-1; is the upper limit of the town size of city j; 3. The urban expansion simulation method based on gravity field and geographically weighted regression cellular automaton according to claim 2, characterized in that: In the calculation method of town size, the The expression is: in: Curve(.) is the fitted urban agglomeration Northam curve function; is the proportion of urban development in city j at time t; in: a, b parameters representing urbanization level; μ and ρ are parameters that control the shape of the Sigmoid function curve; t is time; ) is the urban development ratio of city j at time t-1; In the calculation method of town size, the DW j The expression of = is: DW j =∑ k DW jk in: DW jk is the strength of the spatial connection between city j and other cities k, j≠k; The DW jk The expression is: in: M' j and M' k are the urbanization quality of cities j and k respectively; D jk is the Euclidean distance between cities j and k; The M' j The expression is: M' j =GDP j ×POP j ×Tran j ; GDP j is the gross regional product of city j; POP j is the total population of city j; Tran j is the number of public transportation services in city j.
4. The urban expansion simulation method based on gravity field and geographically weighted regression cellular automaton according to claim 3 is characterized in that: The expression of the state of cell i in the city j during iteration is: in: and are the states of cell i in city j at time t and t+1 respectively; f j (.) is the urban expansion simulation conversion rule of city j; is the town size of city j at the tth iteration; is the transition probability value of non-urban cell i at the tth iteration.
