ATD: Spatio-Temporal Modeling for Identifying Changes in Land Use

土地开发动态建模是城市和区域规划中的一个关键问题。土地利用变化会对当地社区和较大规模区域的环境、生活质量、公共财政和经济发展轨迹产生影响。此外，有必要通过对未来土地开发的多种设想来评估和量化构成的威胁。土地利用模型既考虑了人类行为的主要驱动因素，也考虑了细微尺度的空间和时间相关性，是各种利益攸关方完成这项任务的宝贵工具。该项目旨在开发用于对土地利用变化的威胁进行建模、预测和评估的方法和开源软件工具。它将为各种利益攸关方、社区组织、区域规划者、政策制定者、企业以及不同的科学领域提供新的能力，以深入了解土地利用变化的关键驱动因素，并评估短期和长期发展和威胁对环境、经济和社会的影响。此外，该项目还为研究生提供了研究培训机会。为了实现所述目标，该项目利用了一个建模框架，该框架能够将结构经济地理和相关模型与精细时空数据驱动的模型相结合。此外，它还严格涉及以下技术问题：(1)为拟议的模型开发快速估计和统计推断方法；(2)开发执行无监督学习任务的技术，包括确定模型参数的制度变化，并根据动态规划算法将具有类似土地利用开发的区域进行分类；(3)开发一个框架，能够纳入勾勒未来威胁的情景的预测路径，并预测相应的土地利用结果，以及评估其影响。该方法将通过一个高度分散的时空数据集进行测试和说明，该数据集包含佛罗里达州每个地块的详细信息，由县税务审计师数据库收集和管理。该奖项反映了NSF的法定使命，并已通过使用基金会的智力优势和更广泛的影响审查标准进行评估，被认为值得支持。

Modeling land development dynamics represents a key problem in urban and regional planning. Land use changes have impact on the environment, the quality of life, public finances and economic development trajectories of local communities and larger scale regions. Further, there is a need to assess and quantify threats posed through multiple scenarios about future land developments. Land use models that account both for key drivers of human behavior, as well as fine scale spatial and temporal dependencies are valuable tools to various stakeholders for this task. The project aims to develop methods and open source software tools for modeling, predicting and assessing threats in land use change. It will provide various stakeholders, community organizations, regional planners, policymakers, businesses as well as diverse scientific fields with new capabilities to gain insights into key drivers of land use changes and also assess the environmental, economic and social impact of both short and longer term developments and threats. In addition, the project provides research training opportunities for graduate students. To achieve the stated goals, the project leverages a modeling framework that enables integration of structural economic geography and related models, with fine scale spatiotemporal data driven models. In addition it rigorously addresses the following technical issues: (i) development of fast estimation and statistical inference methods for the proposed models, (ii) development of techniques to perform unsupervised learning tasks including identifying regime changes in the parameters of the models and clustering regions with similar land use developments based on dynamic programming algorithms, and (iii) development of a framework that can incorporate projected paths from scenarios outlining future threats, and predict the corresponding land use outcomes, as well as assess their impact. The methodology will be tested and illustrated through a highly dis-aggregated spatiotemporal data set that contains detailed information for each land parcel in the state of Florida, assembled and curated from county tax auditor databases.This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.