Fig 3 - uploaded by Tong Xu
Content may be subject to copyright.
Land-use patterns in Xiamen City in 1990 (a), 2000 (b), and 2010 (c). These maps were used to compute enrichment factors and verify large neighborhood effects.
Source publication
Urban cellular automata (CA) models are broadly used in quantitative analyses and predictions of urban land-use dynamics. However, most urban CA developed with neighborhood rules consider only a small neighborhood scope under a specific spatial resolution. Here, we quantify neighborhood effects in a relatively large cellular space and analyze their...
Context in source publication
Context 1
... P ij is the total development probability of cell ( i , j ) at moment t in the new urban CA model; ( P l ) ij is the local development suitability of cell ( i , j ), which is calculated with the method proposed in Section 2.1 and has the same inputs and parameters as the Logistic-CA model proposed in Section 2.2; ( P U ) ij is the impact of the 3 Â 3 kernel neighborhood at moment t À 1; con () is the constraints on urban growth at moment t À 1; ( P ln ) ij represents large neighborhood effects at moment t À 1; and P r is the stochastic perturbation term. The model presented here is called Logistic- LNCA, which considers extended neighborhood effects. The Logistic-CA model proposed in Section 2.2 does not have such considerations. The components/procedures of the urban CA considering large neighborhood effects are illustrated in Fig. 2. The simulation experiment was designed to include a model calibration in 1990 e 2000 (Sections 3.4, 3.5, and 3.6) and an independent validation in 2000 e 2010 (Section 3.7). Referring to pre- vious studies on urban CA modeling and testing (Feng and Liu, 2013; Feng et al., 2011; Li et al., 2014; Van Vliet et al., 2013), the generated transitional rules were applied to urban expansion simulations during the same two periods. During the validation stage, observed land-use changes were treated as an independent data set and the transition rules calibrated during the fi rst period were used to predict urban expansion dynamics at the end of second period (Van Vliet et al., 2013). According to Eqs. (2) and (3), urban CA models are mainly characterized by model inputs and parameter settings. From this point of view, the Logistic-CA model is de fi ned by a series of spatial variables and corresponding parameter values used for determining local development suitability. In contrast, the Logistic-LNCA model also includes inputs and parameters for the extended neighborhood rule, except those for the local development suitability component. During the independent validation period (2000 e 2010), the parameter con fi gurations related to the afore- mentioned transition rules in the two urban CA models were all obtained during the calibration period (1990 e 2000), and the model inputs remained unchanged. Xiamen City in southeast China, with an area of 1574 km , was selected as the study area for this research. In 2010, Xiamen City had a total population of 1,802,060 (Tang et al., 2013; XCSB, 2011). Its urban landscape consisted of two urban centers (Siming District and Huli District on Xiamen Island) and four sub-centers outside the island (Jimei, Haicang, Tongan, and Xiangan). Recently, Haicang, Jimei, and Xiangan have gradually become the new industrial centers. Xiamen City is one of the fi rst four special economic zones in China. Its urbanization has accelerated over the past three decades, and its built-up area increased from 34.01 km 2 in 1989 to 197 km 2 in 2008 (Chen and Xu, 2005; XCSB, 2008). Such rapid urban expansion poses serious challenges for the region in terms of sustainable development, environmental loads, and ecosystem ser- vices (Shaker, 2015). Land-use maps of the study area in 1990, 2000, and 2010 were obtained from the Institute of Remote Sensing and Digital Earth of the Chinese Academy of Sciences. Each land-use map has a 30 Â 30 m spatial resolution and includes farmland, woodland, grassland, urban residential land, industrial land, water, and rural areas (Fig. 3). First, the land-use datasets from 1990 to 2000 were used to compute the extended enrichment factors of different land-use types in the neighborhood of residential land during the periods of 1990 e 2000 and 2000 e 2010 and to verify the existence of neighborhood effects. Subsequently, the land-use maps from 1990, 2000, and 2010 were regrouped into three categories (built-up land, non built-up land, and water bodies). Industrial land and residential land were merged into built-up land, and the spatial resolution remained unchanged during the aggregation process. The reclassi fi ed data for 1990 and 2000 were used to calculate the extended enrichment factor and con fi gure the initial states of the Logistic-LNCA model during the simulation periods. When measuring neighborhood effects with the enrichment factor formula, the enrichment factors of all land-use types can be computed for all locations of a speci fi c type or only for locations that have changed their land-use states/types (Van Vliet et al., 2013). We used the second approach to calculate the enrichment factors for each land-use type for locations that changed to industrial or urban residential land during the simulation period. We also determined the average enrichment factor through logistic trans- formation for comparison and curve drawing in the speci fi c neighborhood scope. The extended enrichment factor of urban or industrial land near new residential land was greater than that of agriculture, woodland, grassland, water, or rural land, but its value decreased with distance (Fig. 4), suggesting that new urban and industrial land are more likely to emerge in the neighborhood of urban land use than other land-use types. The depicted enrichment factor curves indicated that the in fl uences of neighboring land-use types decrease with increasing neighborhood radius. However, the measured extended enrichment factors for urban or industrial land in certain large neighborhood scopes (1 e 3 km, equal to 30 e 100 cells in the neighborhood radius) of new urban land still maintained relatively high values. The extended enrichment factor curves presented an undulating pattern. These neighborhood characteristics demonstrated that there was a considerably large neighborhood effect in cellular space. Compared with traditional small neighborhood windows that generally cover several cells (for instance, 1 e 4 cells, up to a 9 Â 9 Moore neighborhood) in the neighborhood radius, it is more appropriate to construct the transition rules of the urban CA model using a large neighborhood. To further estimate the in fl uence of a particular large neighborhood, the values of D r , R min , and R max in Eq. (7) were set to 10, 11, and 81, respectively. Thus, the large neighborhood of the central cell was divided into seven annular bands, and the radius of each annular band was 10 cells. It is worth noting that the window size used here was visually obtained from Fig. 4, and its in fl uence on the model results is still unknown. The large neighborhood module dynamically calculated the extended enrichment factors for the central cell at different annular bands. Similarly, whether the cell was to become built-up land was treated as the dependent variable, and 20% of the sample points were extracted through proportional random-strati fi ed sampling. The obtained sample data were used to perform a binary logistic regression in SPSS. Eq. (9) was used to calibrate the model, and parameter values of the large neighborhood module were obtained. Before executing the logistic regression calibration, the enrichment factor for the large neighborhood was normalized and the standardization was realized ...
Similar publications
The investigation and modeling of land use dynamics can be conducted at different scales based on the objective of the study. However, few studies have looked at comparing various scale aspects, such as spatial resolution and the related neighborhood effect, for practical case study applications. In this chapter, we contribute to this under-explore...
Citations
... Transition suitability reflects the relationship between land cover changes and driving factors. It can be derived from historical data using statistical or machine learning methods such as logistic regression, multicriteria evaluation, and support vector machines (Fu, Wang, & Yang, 2018;Liao et al., 2016;Yang, Li, & Shi, 2008). Among them, the random forest algorithm can effectively mine the relationship between multiple nonlinear variables and quantify the importance of each variable. ...
Urbanization-induced land cover changes significantly impact ecological environments and socioeconomic growth. Vector-based cellular automata (VCA) models are an advanced cellular automata (CA) method that use irregular cells and perform well in simulating land use changes within urban areas. However, the applicability and parameter setting of VCA models for land cover change simulation are still challenging for researchers. To address this issue, this study applied a VCA model and two raster-based models, i.e., a pixel-based CA model and a patch-based CA model, to simulate and compare their performance in simulating land cover changes. The results show that VCA and patch-based CA were superior, with VCA's FoM being 39.74% higher than pixel-based CA and 11.00% over patch-based CA. VCA effectively tracks construction land expansion in rapidly developing areas, while patch-based CA excels in central urban and suburban shifts, fitting broader study scopes. Additionally , a spatial scale sensitivity analysis of the VCA model revealed that a smaller VCA cell size improves accuracy but introduces a risk of spatial pattern errors. Notably, the scope of study impacts VCA accuracy more than cell size. These findings bolster land cover change modeling theory and offer insights for precise future land cover change simulations and decision-making.
... This quantification provides a simple abstraction of local interactions, meaning that all neighboring cells have the same impact on the focal cell. Therefore, scholars have designed several neighborhoods with inhomogeneous rules to improve the simulation performance of urban CA, such as distance-decay, orientation-weighted, and trend-adjusted neighborhoods (Feng & Tong, 2019;Liao et al., 2016;Pan et al., 2021;Roodposhti et al., 2020). ...
... This strategy provides a novel approach for the delineation of the spatial scope of cells' self-organization but is weak in specifically defining the interaction rules between the focal and neighboring cells. This deficiency also exists in most current urban CA; that is, only the clustering process of urban cells is considered when characterizing the self-organized process of urban growth (Liao et al., 2016). Although this clustering process is widely recognized as a direct manifestation of self-organized urban growth, human activities are the essential driving force behind this kind of self-organization (Batty, 2005;Mustafa et al., 2017;Parker et al., 2003). ...
... This study investigated the effects of stakeholders' interactions within the self-organized process of urban growth on urban CA modeling. Although many earlier studies have explored the sensitivity of neighborhood type and size to urban growth simulations and proposed diverse neighborhood definitions, they did not take into account human behaviors, which are an important driving force of self-organization in urban growth (Dahal & Chow, 2015;Liao et al., 2016;Menard & Marceau, 2005;Moreno et al., 2009;Wu et al., 2019). Through integrating stakeholders' interactions with the size-adaptive strategy, the ability of CA to model the self-organized process of urban growth has been significantly improved, with an increase in FoM of about 11.5 % compared to the commonly used CA. ...
Urban modelers have long interpreted urban growth as a coupled evolution of two processes, namely spontaneous and self-organized processes. However, most scholars have always paid attention to the exploration of the driving mechanisms of the spontaneous process. While for the self-organized process, most simulations based on cellular automata (CA) simply abstract it as the clustering process between land use types, ignoring the important role of human behaviors in urban growth. This study presents a framework that integrates a size-adaptive strategy with the multi-agent system (MAS) to establish a neighborhood with stakeholders' interactions (NSI). We employed the CA with NSI (Nsi-CA) to simulate the urban growth of Wuhan from 2000 to 2020 and analyzed its performance. The findings indicate that Nsi-CA performs better than the commonly used CA models at both local and global scales, with an increase in global Figure of Merit (FoM) of about 11.5 %. Moreover, the effectiveness of parameters controlling stakeholders' interactions in the NSI was revealed through a sensitivity analysis. By setting different scenarios with these parameters, the future urban land patterns of Wuhan in 2035 were predicted and used to provide planning suggestions for its future development.
... The neighborhood effect refers to the local interaction between the central cell and surrounding cells, characterizing the local agglomeration effect of the urban expansion process. (Clarke et al. 1997, Liao et al. 2016, Zeng et al. 2023a). Calibration of the neighborhood effect can be divided into the definition of the neighborhood configuration and its quantification Engelen 2000, Zhang andWang 2021). ...
... Many studies have found that increasing the neighborhood size appropriately can allow the central cell to more fully accept the peripheral information and improve the simulation accuracy. (Liao et al. 2016, 2023, Zeng et al. 2023a). However, the impervious surface data exhibit a different pattern. ...
... The agglomeration effect serves as a fundamental driver leading to the formation and growth of cities (Han et al. 2019). Current neighborhoods in CA models mostly focus on characterizing the spatial agglomeration effect, which assumes that urban expansion is more likely to occur near urban land (Liao et al. 2016. However, recent advancements in remote sensing have revealed new insights. ...
... Specifically, the neighborhood rule defines how the state of the focal cell is influenced by its neighboring cells, and the neighborhood scope delineates the geographical range of these local interactions among cells, both of which may vary spatially depending on location-specific characteristics (Phipps, 1989;Tobler, 1979;White & Engelen, 1993;White et al., 1997). In CA modeling research, the neighborhood rule is often simplified as the tendency of urban cells to cluster (Liao et al., 2016;Yeh & Li, 2006). While some studies have introduced distance-weighted and orientation-weighted neighborhood rules to elucidate the varying influences of neighboring cells on the focal cell, their definitions of neighborhood rules remain consistent across all cells' neighborhoods (Roodposhti et al., 2020;Wu et al., 2019). ...
Contemporary investigations into cellular automata (CA) modeling often neglect the considerable influence of spatial heterogeneity on both spontaneous and self-organized processes of urban growth. In this research, we combined a partitioned quantity control strategy, the geographically weighted artificial neural network (GWANN), and a neighborhood size-adaptive approach to formulate a CA framework for simulating spatially heterogeneous spontaneous and self-organized urban growth (SHSS-CA). We examined the simulation performance of SHSS-CA in Beijing, Shanghai, and the Pearl River Delta during 2000-2010 for calibration and 2010-2020 for validation. The findings demonstrate that incorporating spatial heterogeneity enhances CA performance in approximately 90% of regions. Introducing spatially heterogenous spontaneous or self-organized urban growth rules alone can improve the simulation performance of CA. Furthermore, their integration leverages the strengths of both rules and makes SHSS-CA the optimal choice, resulting in a notable improvement in the figure of merit (FoM)-approximately 9% during calibration and around 5% during validation-when compared with ANN-CA. This study contributes new methodologies for developing urban growth simulation rules and has the capacity to effectively help urban planners understand and analyze the complicated urban growth processes.
... Additionally, in conventional CA-based urban simulation models, neighborhood interaction is often represented in the form of development density (i.e., counting the number of land units that are developed within a given neighborhood), and the neighboring land units are either equally weighted or weighted by the distance from the neighborhood center (or its variants) (Wang et al. 2021, Roodposhti et al. 2020, Liao et al. 2016, Zhang et al. 2023a. It is questionable to use the traditional one-fits-all density calculation to represent the neighborhood interaction in different spatial contexts. ...
Cellular automata (CA) have been prevalently used for the simulation of urban land change. However, how to effectively learn the spatial-temporal dynamics of urban development from time-series data remain an important challenge for CA-based models. To address this issue, we propose a new model for the simulation of urban development based on convolutional long short-term memory (ConvLSTM) neural networks. The core of the proposed model is a sequence of vanilla ConvLSTM cells integrated with the modules of channel attention and contextual embedding. Compared with conventional CA-based models, the proposed ConvLSTM model is more advanced in that it can better leverage the open access annual urban land maps to capture simultaneously the spatial structure and the temporal dependency of historical urban development, and further predict multiple maps of annual development for subsequent years (i.e., Maps-to-Maps). The performance of the ConvLSTM model is evaluated through the case studies in China’s three mega-urban regions, and ConvLSTM outperforms other state-of-the-art deep learning architectures at both the pixel level and the coarser grid level. The results also suggest the satisfactory transferability of ConvLSTM in that the model trained in one mega-urban region can be successfully re-used in others without fine tuning.
... According to the transition trends of historical land-use data, the baseline scenario in this study constrained existing urban land from remaining unchanged. Considering the stochasticity and uncertainty in land-use change, the stochastic disturbance term RA was introduced in the CA models as a stochastic factor (Feng and Liu 2013;Liao et al. 2016;White and Engelen 1993;Xie et al. 2020;Zhai et al. 2020). Thus, the transition probability can be expressed as ...
Understanding the spatio-temporal evolution of urban expansion is essential for urban planning and sustainable development. Recently, cellular automata (CA)-based models have emerged as highly effective and widely utilized approaches for simulating urban expansion. However, they suffered from complex structural information inherent in neighborhood effects, including spatio-temporal dimension disjunction and neighborhood sensitivity. To address these issues, herein, we propose a spatial hierarchical learning module based cellular automata model (SH-CA). Specifically, to tackle the spatio-temporal dimension disjunction, we take spatial dependence and historical expansion trends into consideration. We redefine the neighborhood structure and introduce lightweight convolutional neural networks to capture the complex spatio-temporal interaction in neighborhood effects. For the neighborhood sensitivity, we develop a gate filter to aggregate multiscale neighborhood effects for ensuring the synthesis of diverse neighborhood effects disparities. The proposed SH-CA model was implemented to simulate urban expansion in three distinct main urban areas of Beijing, Guangzhou, and Chengdu in China during 2010–2015. The results showed that the proposed SH-CA greatly improves the figure of merit and simulates the most real land-use patterns compared with other four sophisticated CA models. Moreover, the hierarchical learning module effectively modeled spatio-temporal interaction in neighborhood effects, mitigated neighborhood sensitivity, and showed a strong scalability to existing popular CA-based models.
... Some studies have considered multiple attributes of land, such as its natural, socio-economic, and ecological aspects, when constructing a comprehensive evaluation index system. However, methods such as the weighted index summation [67], artificial neural network (ANN) [27], and cellular automaton (CA) [68] that have primarily been used focus on the vertical process of land evaluation, neglecting the horizontal process. However, the MCR model is more effective in reflecting the horizontal process of land expansion. ...
Comprehensive land consolidation (CLC) has become an effective tool for promoting the coordinated development of production, living, and ecological spaces (PLES) in rural China. Given the remarkable territorial differentiation, planning strategies that are geared towards local conditions are indispensable for implementing CLC projects. This study employs the minimum cumulative resistance (MCR) model to simulate the horizontal competition among PLES in Chongqing. The suitability evaluation index system for PLES was developed using natural ecological data, socio-economic data, and land use data from Chongqing Municipality. The results show that: (1) Based on the principles of productivity, livability, and sustainability, the suitability of PLES in Chongqing is classified into highly suitable, moderately suitable, generally suitable, unsuitable, and extremely unsuitable areas. The spatial distribution of suitability across different levels in Chongqing exhibits certain degrees of overlap, intersection, and clustering. (2) Based on the different resistance relationships, 1031 townships in Chongqing were divided into seven types of CLC areas. The northeastern and southeastern regions of Chongqing Municipality exhibit distinct ecological and functional advantages, whereas the northern and western parts of the city are characterized by greater multifunctionality. (3) Tailored CLC measures are suggested for various suitability scenarios, aligning with local conditions and planned developments. The MCR model and PLES theory integrated zoning methods for CLC are practicable and effective, providing a scientific foundation for the construction of land consolidation plans in Chongqing and important references for regional sustainable development.
... This assumption obviously violates the spatial heterogeneity in reality, even with a well analysis of size sensitivity in related studies [21,22]. Recently, there have been some studies taking the distance-decay, multi-layer, and orientation weighted into account to investigate the influence of heterogeneous neighborhoods on individual cells [23,24]. These studies on heterogeneous neighborhoods have significantly contributed to the expansion of our knowledge regarding spatially varying interactions among adjacent cells. ...
In cellular automata (CA) modeling, spatial heterogeneity can be delineated by geographical area partitioning. The dual constrained space clustering method is a prevalent approach for providing an objective and effective representation of differences within urban regions. However, previous studies faced issues by ignoring spatial heterogeneity, which could lead to an over- or under-estimation of the simulation results. Accordingly, this study attempts to incorporate spatially heterogeneous area partitioning into vector-based cellular automata (VCA), producing more accurate and reliable simulations of urban land-use change. First, an area partition strategy with DSC algorithm was employed to generate multiple relatively homogeneous sub-regions, which can effectively capture the spatial heterogeneity in the distribution of land-use change factors. Second, UrbanVCA, a brand-new VCA-based framework, was utilized for simulating land-use changes in distinct urban partitions. Finally, the constructed partitioned VCA model was applied to simulate rapid urban development in Jiangyin city from 2012 to 2017. The results indicated that the combination of DSC clustering and UrbanVCA model could obtain satisfying results as the average FoM values for the partitions and the entire study area exceeded 0.22. Furthermore, a comparative analysis of results from traditional area-partitioned CA models revealed that the proposed area partitioning approach had the potential to yield more accurate simulation outcomes as the FoM values were higher and SHDI and LSI metrics were closer to real-world observations, indicating its good performance in simulating fragmented urban landscapes.
... The mechanism of cellular automatons (CAs) makes them appropriate for UUS growth simulation because they can describe the spatial interactions in neighboring or layered areas. A growth simulation of UUS, similar to urban space growth simulation, focuses on the self-organization and spatial evolution of UUS or urban space development, where CA models have been widely used in urban space growth simulation [34][35][36] to explore the urban expansion mechanisms of various cities [37][38][39]. CA models can also effectively aggregate various socioeconomic and geographic factors, including demographics, traffic, and location, to simulate spatial utilization dynamics by defining reasonable transition rules based on spatial interactions between cell states and regions [40,41]. ...
... CA models are effective because of their flexible structure and support for bottom-up simulations [42,59,60]. CA models [34][35][36][37][38][39][40][41] have been widely used in urban growth simulation, land-use change (LUC), and analysis over the past two decades because they have the advantage of representing the drivers of urban development through transition rules and reflecting the spatial externalities of urban sprawl through neighborhood rules, as illustrated in Table 1. ...
... The urban CA, in general, can be described using the following equation [36,40,55,56,84], and the meta-CA for UUS growth simulation also obeys such a definition: ...
The growth simulation of urban underground space (UUS) under the consideration of ecological constraints can effectively reveal the characteristics and trends of UUS changes, and provide a basis for planning the construction of sustainable and livable ecological cities. Therefore, this study considers urban ecological space as a constraint mechanism for UUS development and conducts a simulation study of the dynamic and complex UUS growth process, with a view toward guiding UUS planning under a long-term overall vision. In this study, a patch-based cellular automaton (CA) model is constructed to simulate the dynamic and complex growth process of UUS, subject to the ecological constraints generated by the agent-based land allocation optimization model. The spatial drivers of UUS growth simulation are determined based on the Random Forest (RF) algorithm. The results of the research case in Tianfu New District, Chengdu City, demonstrate that UUS expansion with ecological constraints exhibits sustainable characteristics. However, the growth rate of the UUS development scale is significantly lower when ecological constraints are present compared to when they are not. This study’s results contribute to urban management by finding a balance between UUS development and ecological space conservation, and providing theoretical support for rational UUS planning and decision making in the construction of low-carbon cities.
... More significantly, it is this component that is subject to microeconomic analysis. Different spatial configurations yield different economic returns (Bramley et al., 2009;Guth et al., 2009;Griffith and Wong, 2007;Kok et al., 2011;Boussauw, 2011;Liao et al., 2016), ...
Urban spatial structure remains the center of quest and modeling. The decentralization concept is among the leading literature discourses that guide spatial analyses. In line with the decentralization discourse, the application of land-use optimization as a modeling method has grown significantly. Despite decentralization dominating the contemporary spatial analysis literature, no study so far explicitly declares an end to the centers (and subcenters). While centers (and subcenters) are alive, the land-use optimization has never taken this macro-morphological structure into consideration. This case study frames land-use optimization within the agglomeration and decentralization concepts based on the view that no single conceptual framework addresses spatial analysis sufficiently. On a theoretical level, the link is between coarse morphological assumption (basis of economic geography) and decentralization (basis of sustainable built environment). The paper blends these dual theories, one governing urban macro-morphological structures and the other governing decentralization literature. On a methodological modeling level, it blends centers and other discretized uses. Optimizing four objectives across the complete centralization through multiple centers of gravity to complete the decentralization of urban spatial structures applying Non-dominated Sorting Genetic Algorithm II, the case study findings justify the importance of explicit modeling of the macro-morphological element. It has been observed that multicenter urban forms perform well above both the single center and the dispersed scenarios. It is, therefore, argued that an appropriate approach to land-use optimization is modeling both the macro-spatial element and fine spatial elements. The result further indicates that local land-use planning regulations place the structure of city in a suboptimal state.
