AutoPaCA:耦合过程-模式的城镇空间增长模拟模型

理解城镇增长过程、模式、机理并预判未来发展趋势,对优化城镇空间布局,促进城市可持续发展有重要意义。元胞自动机模型（Cellular Automata, CA）是研究城镇增长过程、模式的有效技术手段,但传统CA模型对城镇增长过程和空间模式的协同考虑不足。本研究构建了基于斑块（Patch）的城镇CA模型AutoPaCA,实现对城镇增长过程及空间模式的协同表征和精细控制,将城镇增长过程分为边缘增长和跳跃增长两种形式,并实现了对不同形式下城镇斑块位置、形状和面积大小的精细控制,同时考虑城镇空间模式的集聚性、整体形态及空间约束。此外,还提出使用景观格局分析法分析历史时期城镇增长过程和模式,并结合遗传算法实现对模型参数的地域化自动校正,降低人为主观因素对模拟结果的影响。将模型应用于长沙市1995—2035年的城镇增长模拟与多情景分析。结果表明,构建的AutoPaCA模型可以取得较好的模型精度,200次模拟结果的互异邻域相似性指数均值达到0.486;在生态保护情景下,长沙市城市内部生态结构保存完好,且表现出更为明显的长株潭一体化趋势,说明本研究提出的AutoPaCA模型及参数校正方法是有效的。

AutoPaCA: Coupling Process and Spatial Pattern to Simulate Urban Growth

Unplanned urban sprawl has largely altered the territorial space of the planet. Fragmented natural habitats, shifting biomes, and altered nutrient cycling are a few examples of the repercussions of the uncontrolled global urban growth. Therefore, understanding the feature, mechanism, and future pathways of urban dynamics is of great importance for optimizing urban structure and morphology, and thus promoting sustainable urban development. Cellular Automata (CA) model has been recognized as an effective tool in the planner's packet to facilitate this understanding, and to help make informed decisions on urban development. Given the insufficient consideration on the process and spatial pattern of urban growth in traditional CA models, this study proposes a patch-based CA model for urban growth simulating, which is entitled AutoPa CA. The proposed AutoPaCA model couples the spatiotemporal process and pattern to simulate urban growth. It divides the urban development process into two different processes: the edge diffusion which represents organic urban growth and the leapfrog development which represents spontaneous urban growth. The AutoPaCA model introduces a parameterized self-organizing approach to fine control the location, shape, and size of newly generated urban patches. To be specific, both the organic and spontaneous patch-generation function yield new urban patches using two consecutive steps: Seeding and self-growing. The seeding procedure locates the pivot cell of new patch with a pruning and random selection operation based on the urban development suitability surface which is estimated using a random forest model. The self-growing procedure creates an urban patch with given size using a neighborhood-scanning operation in which a parameter is introduced to control the shape of the patch. The size of urban patches is assumed to follow a lognormal distribution. The shape controlling parameter is within the range of [0, 2], with values larger than 1 generating circle-like shape and values smaller than 1 generating elongated shape. In addition, the proposed model employs analysis methods in discipline of landscape ecology to scrutinize the characteristics of urban growth in historical periods, based on which a cutting-edge genetic algorithm is applied to achieve localized calibration of key parameters in the model. At the same time, the rigid limitation and elastic guiding effect of ecological protection on urban development are also considered in the simulation. The application of the AutoPaCA model in Changsha, China presented high simulation accuracy. The average value of the Reciprocal Similarity Comparison Index of 200 simulations in the model validation period reached 0.486, which attests the feasibility and applicability of the model.