城镇化进程中基于蚁群行为规则的滇池流域不透水表面扩张智能体建模与模拟

The continuous aggravating of urban Non-point source pollution is due to the rapid increasing of impervious surface(IS) area for Dian-chi Lake watershed， and impervious surface coverage(ISC) is a double assessment indicator of urbanlization level and risk pre-warning of Non-point source pollution. Thereby, it is necessary to study the scenario simulation model of IS expansion in order to promote the harmonious development between urbanization and water environment conservation in the watershed by simulating, analyzing and forecasting the rules and tend of urban spatial pattern evolution, as well as explaining drvien mechanism of IS expansion in urbanization. Presently, it is a major research trend how to simulate IS expasion used Agent-based modeling(ABM) in Geosimulation research area, and behavior modeling and it's optimization based on ABM also is one of the key challenges in this area. Therefore, in this research proposal, taking the optimization of ant behavior rules used in ABM as a goal, we study the coupling mechanism between ant colony optimization and agent behavior modeling, as well as their concept model, and design the heuristic rules of agent behavior and decision-making on both microcosmic and macroscopic point, as well as the methodologies of model validation and verification for parameters sensitivity. And finally we will develop a simulation model of IS dynamic change. Through those researches above all, it is hopeful to provide the scientific theoretical supports and information-based decision-making tools for benign land use mode in urbanization of watershed, as well as to enrich the theories and methods system of optimized Agent-based modeling.

不透水表面（IS）的快速增加是导致滇池流域城市非点源污染不断加剧的主要原因，而不透水表面覆盖率（ISC）是城镇化发展水平与非点源污染风险预警的双重评价指标，有必要研究流域不透水表面扩张的情景模拟模型，模拟、分析和预测流域城镇化空间格局演变的规律，进而解释流域不透水表面扩张的动力作用机制，从而促进流域城镇化与水环境保护相协调的可持续性发展。当前，基于智能体建模（ABM）的不透水表面变化模拟是地理模拟领域的研究趋势，而智能体行为建模及优化是该领域面临的关键性挑战之一。为此，本研究以蚁群行为规则在智能体建模中的优化为目标，研究蚁群算法与智能体行为建模的耦合机制及概念模型，设计微观与宏观相结合的智能体启发式行为与决策规则、参数敏感性检验与模型验证等方法，最终构建滇池流域不透水表面扩张的动态模拟模型，为流域城镇化发展的良性用地模式提供科学理论依据和信息化决策工具，有望丰富城镇化用地扩张的智能体建模优化理论与方法

不透水表面及其覆盖率既是可调控的城镇化水平评价指标，也是可量化的城市水环境监测与预警指标。为此，本研究以蚁群行为规则及其人工智能算法在智能体建模中的优化为理论依据，以滇池流域的城镇化用地变化为研究对象，以不透水表面覆盖率为分析和评价指标，构建了滇池流域城镇化进程中不透水表面扩张的动态变化模型，模拟并预测了滇池流域城镇空间格局演变的时空过程，为滇池流域城镇化进程的良性用地模式提供了科学依据和信息化决策工具。具体而言，课题组综合运用了数学建模、空间分析、蚁群智能优化和GIS可视化等多学科交叉方法实现上述研究目标，这包括：构建了滇池流域不透水表面扩张的多因子驱动模型；设计了蚁群行为规则约束下的智能体建模框架；构造了蚁群智能体行为规则和用地变化的元胞转换规则；集成GIS、REPAST和Matlab组件研发了滇池流域城镇化空间扩张情景模拟与时空分析系统。试验结果表明：模型模拟精度分别达到77.6%（平均点对点统计精度）、75.1%（Kappa系数）和53.6%（平均形状指数），模型通过相关验证。由此可知，在蚁群行为规则及其优化算法的约束下，所设计和构建的不透水表面扩张模型不仅整体模拟效果良好，而且还显著提升了人地交互激烈的用地类型的模拟精度。综合上述成果，课题在如下2个方面做了一些开拓性工作：（1）基于空间和属性双约束聚类方法，构造了分区异步元胞自动机模型，较好地解决了元胞转换规则的时空异质性问题；（2）耦合蚁群优化算法与人地交互作用过程，构建了元胞自动机与智能体混合的用地变化模拟模型，较好地提升了此类模型的并行计算效率和模拟精度。最后，依托本项目，课题组发表论文23篇，其中SCI期刊源收录6篇，EI期刊源收录3篇，中文核心收录9篇；出版专著2部；获得软件著作权6项；培养博士研究生1名，硕士研究生4名，本科生8名。项目投入经费49万元，支出30.25万元，剩余经费18.75万元。各项支出基本与预算相符，结余经费计划用于本项目后续研究的支出。