论文部分内容阅读
Received: July 25, 2011 / Accepted: November 7, 2011 / Published: January 20, 2012.
Abstract: Cellular automata (CA) models (deterministic, stochastic or hybrid) have garnered tremendous popularity as spatial simulation techniques. As powerful modelling tools, they have been successfully employed in conjunction with land-use allocation and statistical simulation models to portray the dynamics and patterns of growth in such simulation contexts as population dynamics, polycentricity, urban land-use evolution, gentrification and urban sprawl. In this study, the authors present a methodological approach based on a synergistic integration of a deterministic CA model, a Markov Chains transition model and a multi-criteria land evaluation model to determine and portray the patterns and dynamics of land-use change in a rural setting. The site chosen for this study is the Lake Chad Basin, an endorheic basin supporting a population of over twenty million people living in four African countries (Chad, Cameroon, Niger and Nigeria) where the people are among Africa’s most chronically vulnerable to food insecurity due to the drastic impact of natural and anthropogenic agents of ecological transformation in the basin. Using remotely sensed images captured at three different dates (1975, 1987 and 1999) in conjunction with other supporting attribute data, simulation runs were executed to predict and construct two different future scenarios (2011 and 2023) of land-use/land-cover states. Based on the numerical and graphical results of the simulation runs, Water, Wetland, Openland and Forest land-use classes were predicted to register net losses while the Road, Settlement and Farm land-use classes were predicted to register net gains in both 2011 and 2023 simulated land-use scenarios. These predictive modelling results suggest that the Lake Chad ecosystem may inexorably undergo extensive land-use/land-cover transformation in the absence of mitigating intervention measures. The study demonstrates that the proposed methodological approach of integrated rural land-use scenario building and analysis relying on the CA-based land-use simulation model is a powerful and compelling tool with significant potentials to support planning and policy for sustainable rural land development.
Key words: Cellular automata, grid cell, state transition, land-use/land-cover, transition potential model, Markov Chains model, transition rule.
Issues relating to land management and land-use change in both rural and urban settings dominate the development agenda of all countries of the world and, over the years, have remained highly political and contentious. This is expected in the context in which optimal land-use planning is perceived as an indispensable factor for ensuring food security, environmental sustainability and economic development. One of the recent studies on the impact of increased pressure on land and the effects of land-use and land management practices on the dynamic character of rural ecosystems shows that a strong correlation exists between balanced, sustainable land development and, human, food and environmental security [1]. Other recent studies on land management indicate that, in many parts of the world, the rural
landscape is experiencing rapid land-use/land-cover(LULC) changes [2]. It is not surprising therefore that the issue of rural land use change occupies the front-burner of the development initiatives of responsible governments all over the world. The desire of planning authorities and municipal governments is to articulate policies and programmes capable of maintaining a balanced ecosystem or to mitigate or prevent, on a sustainable basis, the devastating consequences of severe land-use change when they occur. Achieving this goal requires that such authorities must adopt responsible, holistic and sustainable development strategies which must axiomatically be carried out using spatial decision support tools and methodologies.
The spatio-temporal dynamics of rural ecosystems is an invariably complex phenomenon that involves a complex nexus of interacting forces of causal agents. It has been demonstrated that, in order to disentangle the complex suite of socio-economic and bio-physical forces that influence the rate, spatial pattern and distribution of land use change and to estimate the impacts of such changes, spatially-explicit land use models are indispensable [3]. As reproducible tools, such spatial simulation models have the potential to expand the planner’s knowledge domain and to support the exploration of future land use changes under different scenario conditions, thus supplementing his existing mental capabilities to analyze land use change and to make more informed decisions. Such tools can help to predict ecological responses to changing landscape heterogeneity and to gain insights into the variability of landscape patterns and processes over time.
The research on dynamic landscape modelling using spatial simulation models is quite extensive and varied[4]. These models have been successfully applied in such domains as land-use allocation and land-use planning. The most recent of these spatial simulation models are based on simulated annealing, genetic algorithms, cellular automata (CA), or agent based models. Until this recent development, spatial simulation models were exclusively based on techniques such as differential equations, partial differential equations and empirical equations. CA models (deterministic, stochastic or hybrid) have recently garnered tremendous popularity as a spatial simulation technique in a wide range of urban and rural land-use simulation and modelling domains and, as such, the vistas of research in this direction are rapidly expanding. Over the past few years, CA models have found application in spatial simulation involving a plethora of themes including, among others, population dynamics, polycentricity, urban land-use evolution, gentrification and urban sprawl. Compared to conventional mathematical tools of spatial simulation such as differential equations, partial differential equations and empirical equations, CA models are relatively simple yet produce results that are quite meaningful and useful to support decision making in a planning context. Although geographical information systems (GIS) are powerful tools to collect, store, manage and analyze spatial data, current GISs have shown considerable weakness and limitation in spatial decision making which are due, in great part, to their lack of sophisticated analytical and spatial modeling tools [5, 6]. Many studies have shown that the integration of geographic information systems (GIS), cellular automata (CA) models, land use allocation models (such as multi-criteria evaluation) and statistical simulation models (such as Markov chains) provides a powerful environment to simulate and predict dynamic phenomena such as urban and rural spatial growth [7].
In this study, the authors present a methodological framework that leverages the suitability-based cellular automata land-use simulation model. The proposed approach involves loose-coupling geographical information systems with Markov chains, multi-criteria evaluation and cellular automata models to model and simulate the rural land-use dynamics using spatial and thematic data sets covering a section of the Lake Chad
aspect map, slope map, distance map, population surface map, etc.), spatial constraints (e.g. a binary map of the exclusion zones such as maps of underground treatment plants, water canal, landfill sites, etc.) and non-spatial constraints. Some MCE modelling methods require that the factors be standardized prior to the computation of the final transition suitability (potential) map [12].
1.3 Cellular Automata
Cellular Automata (CA) are dynamic models that can be employed to simulate the evolution or dynamics of a wide variety of natural and human systems. They are processing algorithms that were originally conceived by Ulam and Von Neumann in the 1940s to study the behavior of complex systems [13]. CA models present a powerful simulation environment represented by a grid of space (raster), in which a set of transition rules determine the site attribute of each given cell taking into account the attributes of cells in its vicinities. These models have been very successful in view of their operationality, simplicity and ability to embody both logic and mathematics-based transition rules, thus enabling complex global patterns to emerge directly from the application of simple local rules. A cellular automaton system consists of a regular grid of cells, each of which can be in one of a finite number of k possible states, updated synchronously in discrete time steps according to a local, identical interaction(transition) rule. The state of a cell is determined by the previous states of a surrounding neighborhood of cells. The types of spatial problems that can be approached using CA models include spatially complex systems(e.g., landscape processes), discrete entity modeling in space and time (e.g., ecological systems, population dynamics) and emergent phenomena (e.g., evolution, earthquakes). From the application perspective, CA are dynamic models that inherently integrates spatial and temporal dimension.
CA is composed of a quadruple of elements as defined in Eq. (3) [14].
ILWIS-based script. In designing the CA transition rule, the Moore neighbourhood kernel (see Section 1.3) was adopted with a kernel threshold of 3. A cell would therefore undergo state transition to the state of the predominant cell in its 8-cell neighbourhood if its transition suitability value is greater than zero and if it is not “Settlement” or “Road” and if its transition suitability value is greater than the suitability threshold value and if the count of the predominant cell is greater than or equal to the neighbourhood kernel threshold value of 3. It is to be noted that several other constraints can be integrated into the transition rule to achieve some set configuration. Algorithm 1 shows a simple basic-like CA-based transition algorithm developed for the simulation. The script was designed to reference the five raster-structured maps generated in the previous sub-section as input maps to compute final(simulated) raster maps corresponding to the 1999-2011 and 1999-2023 scenarios based on neighbourhood map calculation. Fig. 5 shows the resulting simulated maps for the two proposed scenarios.
Abstract: Cellular automata (CA) models (deterministic, stochastic or hybrid) have garnered tremendous popularity as spatial simulation techniques. As powerful modelling tools, they have been successfully employed in conjunction with land-use allocation and statistical simulation models to portray the dynamics and patterns of growth in such simulation contexts as population dynamics, polycentricity, urban land-use evolution, gentrification and urban sprawl. In this study, the authors present a methodological approach based on a synergistic integration of a deterministic CA model, a Markov Chains transition model and a multi-criteria land evaluation model to determine and portray the patterns and dynamics of land-use change in a rural setting. The site chosen for this study is the Lake Chad Basin, an endorheic basin supporting a population of over twenty million people living in four African countries (Chad, Cameroon, Niger and Nigeria) where the people are among Africa’s most chronically vulnerable to food insecurity due to the drastic impact of natural and anthropogenic agents of ecological transformation in the basin. Using remotely sensed images captured at three different dates (1975, 1987 and 1999) in conjunction with other supporting attribute data, simulation runs were executed to predict and construct two different future scenarios (2011 and 2023) of land-use/land-cover states. Based on the numerical and graphical results of the simulation runs, Water, Wetland, Openland and Forest land-use classes were predicted to register net losses while the Road, Settlement and Farm land-use classes were predicted to register net gains in both 2011 and 2023 simulated land-use scenarios. These predictive modelling results suggest that the Lake Chad ecosystem may inexorably undergo extensive land-use/land-cover transformation in the absence of mitigating intervention measures. The study demonstrates that the proposed methodological approach of integrated rural land-use scenario building and analysis relying on the CA-based land-use simulation model is a powerful and compelling tool with significant potentials to support planning and policy for sustainable rural land development.
Key words: Cellular automata, grid cell, state transition, land-use/land-cover, transition potential model, Markov Chains model, transition rule.
Issues relating to land management and land-use change in both rural and urban settings dominate the development agenda of all countries of the world and, over the years, have remained highly political and contentious. This is expected in the context in which optimal land-use planning is perceived as an indispensable factor for ensuring food security, environmental sustainability and economic development. One of the recent studies on the impact of increased pressure on land and the effects of land-use and land management practices on the dynamic character of rural ecosystems shows that a strong correlation exists between balanced, sustainable land development and, human, food and environmental security [1]. Other recent studies on land management indicate that, in many parts of the world, the rural
landscape is experiencing rapid land-use/land-cover(LULC) changes [2]. It is not surprising therefore that the issue of rural land use change occupies the front-burner of the development initiatives of responsible governments all over the world. The desire of planning authorities and municipal governments is to articulate policies and programmes capable of maintaining a balanced ecosystem or to mitigate or prevent, on a sustainable basis, the devastating consequences of severe land-use change when they occur. Achieving this goal requires that such authorities must adopt responsible, holistic and sustainable development strategies which must axiomatically be carried out using spatial decision support tools and methodologies.
The spatio-temporal dynamics of rural ecosystems is an invariably complex phenomenon that involves a complex nexus of interacting forces of causal agents. It has been demonstrated that, in order to disentangle the complex suite of socio-economic and bio-physical forces that influence the rate, spatial pattern and distribution of land use change and to estimate the impacts of such changes, spatially-explicit land use models are indispensable [3]. As reproducible tools, such spatial simulation models have the potential to expand the planner’s knowledge domain and to support the exploration of future land use changes under different scenario conditions, thus supplementing his existing mental capabilities to analyze land use change and to make more informed decisions. Such tools can help to predict ecological responses to changing landscape heterogeneity and to gain insights into the variability of landscape patterns and processes over time.
The research on dynamic landscape modelling using spatial simulation models is quite extensive and varied[4]. These models have been successfully applied in such domains as land-use allocation and land-use planning. The most recent of these spatial simulation models are based on simulated annealing, genetic algorithms, cellular automata (CA), or agent based models. Until this recent development, spatial simulation models were exclusively based on techniques such as differential equations, partial differential equations and empirical equations. CA models (deterministic, stochastic or hybrid) have recently garnered tremendous popularity as a spatial simulation technique in a wide range of urban and rural land-use simulation and modelling domains and, as such, the vistas of research in this direction are rapidly expanding. Over the past few years, CA models have found application in spatial simulation involving a plethora of themes including, among others, population dynamics, polycentricity, urban land-use evolution, gentrification and urban sprawl. Compared to conventional mathematical tools of spatial simulation such as differential equations, partial differential equations and empirical equations, CA models are relatively simple yet produce results that are quite meaningful and useful to support decision making in a planning context. Although geographical information systems (GIS) are powerful tools to collect, store, manage and analyze spatial data, current GISs have shown considerable weakness and limitation in spatial decision making which are due, in great part, to their lack of sophisticated analytical and spatial modeling tools [5, 6]. Many studies have shown that the integration of geographic information systems (GIS), cellular automata (CA) models, land use allocation models (such as multi-criteria evaluation) and statistical simulation models (such as Markov chains) provides a powerful environment to simulate and predict dynamic phenomena such as urban and rural spatial growth [7].
In this study, the authors present a methodological framework that leverages the suitability-based cellular automata land-use simulation model. The proposed approach involves loose-coupling geographical information systems with Markov chains, multi-criteria evaluation and cellular automata models to model and simulate the rural land-use dynamics using spatial and thematic data sets covering a section of the Lake Chad
aspect map, slope map, distance map, population surface map, etc.), spatial constraints (e.g. a binary map of the exclusion zones such as maps of underground treatment plants, water canal, landfill sites, etc.) and non-spatial constraints. Some MCE modelling methods require that the factors be standardized prior to the computation of the final transition suitability (potential) map [12].
1.3 Cellular Automata
Cellular Automata (CA) are dynamic models that can be employed to simulate the evolution or dynamics of a wide variety of natural and human systems. They are processing algorithms that were originally conceived by Ulam and Von Neumann in the 1940s to study the behavior of complex systems [13]. CA models present a powerful simulation environment represented by a grid of space (raster), in which a set of transition rules determine the site attribute of each given cell taking into account the attributes of cells in its vicinities. These models have been very successful in view of their operationality, simplicity and ability to embody both logic and mathematics-based transition rules, thus enabling complex global patterns to emerge directly from the application of simple local rules. A cellular automaton system consists of a regular grid of cells, each of which can be in one of a finite number of k possible states, updated synchronously in discrete time steps according to a local, identical interaction(transition) rule. The state of a cell is determined by the previous states of a surrounding neighborhood of cells. The types of spatial problems that can be approached using CA models include spatially complex systems(e.g., landscape processes), discrete entity modeling in space and time (e.g., ecological systems, population dynamics) and emergent phenomena (e.g., evolution, earthquakes). From the application perspective, CA are dynamic models that inherently integrates spatial and temporal dimension.
CA is composed of a quadruple of elements as defined in Eq. (3) [14].
ILWIS-based script. In designing the CA transition rule, the Moore neighbourhood kernel (see Section 1.3) was adopted with a kernel threshold of 3. A cell would therefore undergo state transition to the state of the predominant cell in its 8-cell neighbourhood if its transition suitability value is greater than zero and if it is not “Settlement” or “Road” and if its transition suitability value is greater than the suitability threshold value and if the count of the predominant cell is greater than or equal to the neighbourhood kernel threshold value of 3. It is to be noted that several other constraints can be integrated into the transition rule to achieve some set configuration. Algorithm 1 shows a simple basic-like CA-based transition algorithm developed for the simulation. The script was designed to reference the five raster-structured maps generated in the previous sub-section as input maps to compute final(simulated) raster maps corresponding to the 1999-2011 and 1999-2023 scenarios based on neighbourhood map calculation. Fig. 5 shows the resulting simulated maps for the two proposed scenarios.