基于Agent的城市扩张占用耕地动态模型及模拟

与传统的在景观尺度上对城市扩张占用耕地时空特征及其演变规律的研究不同，论文从微观层面提出基于Agent 的城市扩张占用耕地动态模型（Urban Encroachment on CroplandDynamic Model，UECDM），多角度深入探索耕地向城市用地转变的驱动机理及其时空动态特征，并提出协调二者平衡的可持续发展对策。在UECDM模型中引入视域分析，通过模拟地形和观察点将视觉景观的量化评价结果应用到居民居住区位决策过程，从而分别构建包含可视域的能够更全面表征环境要素及其内部关系的元胞自动机层和不同类型Agent 相互作用的多智能体层，兼顾人类决策行为和城市空间生长自组织性。应用该模型对天津城市扩张占用耕地进行动态模拟研究，并从数量与空间格局两方面展开精度验证，Kappa 系数达到0.795，依据4 个景观格局指数计算得到模拟结果的平均相对误差为3.86%，表明模型具有较高可信度。对不同时期模拟结果的分析表明，天津城市扩张占用耕地空间主要表现为“摊大饼”式外延扩展和跳跃式“卫星城”，研究结果将为未来城市发展规划和耕地保护提供决策支持。

An Agent-based Approach for Urban Encroachment on Cropland Dynamic Model and Simulation

Unlike traditional studies about the spatio-temporal characteristics and its evolution law of urban encroachment on cropland on landscape scale, this paper proposes agent-based Urban Encroachment on Cropland Dynamic Model (UECDM) from the micro-level, exploring in depth the mechanism and dynamics of urban encroachment on cropland from diverse perspectives, so as to bring forward sustainable development countermeasures for reconciling both balance. UECDM introduces viewshed analysis, which applies quantitative assessment results of the visual landscape to residential decision-making processes by simulating terrain and observation point. The model consists of cell automata layer of viewshed to fully characterize the various environmental elements along with their internal relations and agent layer containing the interactions among different kinds of agents, which takes account of human decision-making and self-organization of urban growth. It is applied to simulate and analyze the spatio-temporal dynamics of Tianjin's urban encroachment on cropland and validation follows. The Kappa coefficient is up to 0.795 and based on four landscape indices, the average relative error of the simulation result is about 3.86%, both indicating a high credibility of this model. Analysis of the simulation and prediction results of different periods indicates that the spatial pattern of Tianjin's urban encroachment on cropland is mainly outward expansion of the "pie" type as well as "satellite towns" of the "jump" type, which makes it possible to provide decision support for future urban development planning and cropland protection.