Modeling geospatial processes: Integrating artificial intelligence with complexity, networks and geographic information sciences

迅速的城市化和与此相关的对自然资源的需求对宝贵的农业用地和林地造成了越来越大的压力，并造成了很高的损失风险。研究耦合的人类和自然系统（CHANS）之间的接口，相互作用和反馈是一项多学科的工作，需要扩展思维，结合社会和自然科学中常见的传统研究方法。土地利用/土地覆被变化过程是一个典型的CHANS过程，是一个复杂的系统，其根源在于地方层面的人类与环境的相互作用，并在多个空间尺度上产生影响。复杂性科学理论与地理信息科学的结合需要创新的地理空间建模和地理模拟方法。这些新的方法将提供有效的方法：预测可能的情景;评估不同管理和政策战略的影响;并帮助减轻LULC变化的后果。该研究计划的具体目标是：（1）开发新一代人工智能地球模拟方法，用于代表过程中各种行为体的行为，并预测LULC变化的空间动态和模式：（2）用网络科学概念增强所开发的模型，使其能够在更大的空间范围内使用;（3）设计模型测试、校准和验证程序，以评估和比较所开发的模型对各种LULC问题的有效性。地理自动机，特别是基于地理空间代理的建模，将与人工智能的前沿技术以及新兴的网络科学学科相结合，所有这些都在地理信息系统（GIS）和科学框架内。空间建模方法将使用地理信息系统和遥感数据集，并将在空间决策和城市大都市地区土地使用规划以及森林覆盖范围内实施，主要是在加拿大、不列颠哥伦比亚省和大都会温哥华地区。拟议的研究计划将提高理解和评估LULC变化过程的潜在结果的能力。它还将创建地理空间建模方法和工具，以改善地方和国家两级的城市和自然资源规划、决策和政策制定。所产生的知识可以转移到其他方面。

Rapid urbanization and the associated need for natural resources create increasing pressures on and impose a high risk of loss for valuable agricultural and forested lands. Examining the interfaces, mutual interactions and feedbacks between coupled human and natural system (CHANS) is a multidisciplinary effort and require an extension in thinking that combines the traditional research methods common in the social and natural sciences. The land-use and land-cover (LULC) change process is a typical CHANS, characterized as a complex system, and rooted in local-level human interactions with the environment that have consequences at multiple spatial scales. Innovative geospatial modeling and geosimulation approaches are needed to integrate the theory of complexity science with geographic information science. These new approaches will provide effective ways to: forecast possible scenarios; evaluate the effects of different management and policy strategies; and aid in mitigating the consequences of LULC change. The proposed research program has the specific objectives to: (1) develop an advanced generation of artificial intelligence geosimulation approaches for representing the behavior of various actors in the process and forecasting the spatial dynamics and pattern of LULC changes; (2) enhance the developed models with network science concepts to enable their use in larger spatial extents; (3) design model testing, calibration and validation procedures to evaluate and compare the effectiveness of the developed models for various LULC problems. Geographic automata, particularly geospatial agent-based modeling, will be integrated with techniques at the forefront of artificial intelligence as well as with the emerging discipline of network science all within geographic information systems (GIS) and science frameworks. The spatial modeling approaches will use GIS and remote sensing datasets and implemented in the context of spatial decision-making and land-use planning for the urban metropolitan areas and forest covers primarily in Canada, British Columbia and the Metro Vancouver Region. The proposed research program will enhance the capability to understand and assess the potential outcomes of the LULC change process. It will also create geospatial modeling methodologies and tools to improve urban and natural resources planning, decision-making and policy-building at local and national levels. The knowledge generated would be transferable to other contexts.