Rethinking Representation in Discrete Spatial Modeling: Theoretical Developments and a Computational Study of Hurricane Disaster Relief

地理表示是指物理和人类现象如何以符号形式构成并以数字方式存储，这是利用地理信息系统进行空间建模和决策的一个重要问题。先前的研究表明，在各种情况下，与表征相关的误差对规划和建模情景有很大影响。例如，当使用离散的网络流空间模型来有效地分配货物、路线运输和分配服务时，聚集是一个关键的表示问题，它对模型输出产生不利影响，从而影响战略规划和分析。不幸的是，聚集是许多现有离散空间模型结构的核心，这表明它们需要重新评估，并应开发新的技术。这项研究将设计新的空间模型，以减少关键决策情况下与表示相关的错误的突出来源。一般的方法是探讨是否可以重新考虑现有的空间模型，以利用地理现象的更详细和非聚合的空间表示。测试和开发这些模型的背景是飓风救灾规划。在这项研究中，将制定并执行一系列基于计算地理信息系统的模拟实验，将救援物资和服务分发到佛罗里达州受影响的地点。模拟将与新的空间模型开发结合起来设计。模拟将在考虑飓风破坏程度、政府/救援机构资源的可获得性以及其他情况特征的情况下模拟救援物资和服务的分配。新空间模型的性能将通过将其输出与使用现有方法获得的结果进行比较来评估。这项研究将增进地理信息科学中关于空间决策问题如何理论化、构建和解决的基本知识。重点将放在改进地理现象在空间模型中的表现方式上，吸取的经验教训将适用于许多其他决策问题。救灾计划具有重要的社会意义，美国东南部最近遭受的大规模飓风破坏就是明证。该项目将与飓风专家形成部门内合作，并促进与州和地方机构的信息交换，这将有助于建立一个研究基础设施，从而提高我们计划和准备应对极端天气事件的能力。最后，这项研究将有助于培养研究生，并通过为新的研究生课程开发课程，促进初级调查员的研究和教学议程的直接整合。

Geographical representation insofar as how physical and human phenomena are constituted symbolically and stored digitally is an important concern for spatial modeling and decision making using geographic information systems (GIS). Prior research has shown that representation-related error substantially impacts planning and modeling scenarios in a variety of contexts. For example, when using discrete spatial models of network flow to efficiently distribute goods, route shipments and allocate services, aggregation is a key representation issue that adversely affects model outputs and therefore compromises strategic planning and analysis. Unfortunately, aggregation is central to many existing discrete spatial model structures, which suggests that they need to be reevaluated and new techniques should be developed. The research will design new spatial models that reduce prominent sources of representation-related error in critical decision making situations. The general approach is to explore whether existing spatial models can be rethought to take advantage of more detailed and disaggregate spatial representations of geographical phenomena. The context in which these models are tested and developed is hurricane disaster relief planning. In this research, a series of computational GIS-based simulation experiments will be formulated and executed where relief goods and services are to be distributed to affected locations in Florida. Simulations will be designed in conjunction with new spatial model development. The simulations will model the distribution of relief goods and services in scenarios that take into account the extent of hurricane damage, the availability of governmental/relief agency resources, and other situational characteristics. Performance of new spatial models will be assessed by comparing their outputs to results obtained using existing approaches.The research will advance basic knowledge in geographic information science regarding how spatial decision problems are theorized, constructed and resolved. Emphasis will be placed on improving the ways geographical phenomena are represented in spatial models and lessons learned will be applicable to many other decision making problems. Planning for disaster relief is of great societal importance as evidenced by the massive hurricane damage experienced recently in the Southeastern U.S. The project will form an intradepartmental collaboration with a hurricane expert and facilitate information exchange with state and local agencies which will help to establish a research infrastructure leading to improvements in our ability to plan and prepare for extreme weather events. Lastly, the research will help to train a graduate student as well as foster the direct integration of the primary investigator's research and teaching agendas through curriculum development for a new graduate course.