NSF Convergence Accelerator Track D: Artificial Intelligence and Community Driven Wildland Fire Innovation via a WIFIRE Commons Infrastructure for Data and Model Sharing

NSF融合加速器支持以使用为灵感，以团队为基础，多学科的努力，解决国家重要性的挑战，并将在不久的将来产生对社会有价值的可交付成果。“融合加速器”第一阶段项目的更广泛影响和潜在社会效益是创建WIFIRE Commons，这是一种数据驱动、人工智能（AI）支持和基于模型的科学方法，最终旨在通过使用先进技术支持火灾缓解、准备、响应和恢复，限制甚至防止野火的破坏性影响。野火数据、人工智能和火灾行为物理在WIFIRE公共资源的主要设计中的结合，推动了多学科合作，并与教育工作者、市政领导和火灾管理人员进行了接触，以确保公共资源的设计可用于翻译。数据和模型共享是这项工作的核心，战略伙伴关系以及与私营和公共部门的密切合作也是核心。项目团队包括来自西班牙裔服务机构的教育工作者，以及倡导增加女性在消防劳动力和数据科学领域的参与。此外，WIFIRE Commons的AI网关机器学习、可扩展计算和交互式地理空间分析工具将适用于任何可以从建模中受益的领域。该项目旨在对人工智能集成的野火研究和响应进行融合研究，并建立一个我们称之为WIFIRE Commons的框架，以利用人工智能对基于物理的野火模型和用于实时监测和预测野火的异构数据集的不断发展的组合进行创新优化。第一阶段的工作将通过设计思维方法实现这一目标，包括五个可交付成果流：1)社区融合研讨会，2)原型数据和模型共享框架，3)用例启发案例研究，以展示拟议的人工智能创新，4)教育、推广和公共信息活动的原型；5)第二期规划。长期愿景是创建一个可持续的、开源的人工智能驱动的数据和模型公地，以促进和利用合作，“利用人工智能创新”，支持以使用为灵感的社会和科学野火应用。在设计思维的推动下，在我们团队成员（WIFIRE、MINT、QUIC-fire）之前研究的基础上，提出的WIFIRE Commons融合研究、数据和模型共享框架将使开发新的人工智能技术和可重用模型成为可能，这些技术和模型可以在许多应用中使用。这个Commons基础设施将为人工智能驱动的火灾科学编目、管理和集成数据和模型，以云兼容的形式保持对数据的开放编程访问，可以通过网关接口集成到人工智能过程中，并确保数据和模型的来源。这种基于人工智能的智能数据/模型集成将改变基于科学的野火决策的敏捷性，允许快速吸收新类型的模型和数据，并允许不断扩大的用户基础来理解不确定性水平。该奖项反映了美国国家科学基金会的法定使命，并通过使用基金会的知识价值和更广泛的影响审查标准进行评估，被认为值得支持。

The NSF Convergence Accelerator supports use-inspired, team-based, multidisciplinary efforts that address challenges of national importance and will produce deliverables of value to society in the near future. The broader impact and potential societal benefit of this Convergence Accelerator Phase I project is to create the WIFIRE Commons, a data-driven, artificial intelligence (AI) enabled and model-based scientific approach that ultimately aims to limit and even prevent the devastating effects of wildfires by using advanced technologies to support fire mitigation, preparedness, response, and recovery. The combination of wildfire data, AI and the physics of fire behavior in the main design of WIFIRE Commons drives multidisciplinary collaboration and engagement with educators, municipal leaders, and fire managers to ensure the Commons is designed for translational use. Data and model sharing are core to the effort, as is strategic partnerships and close collaboration with the private and public sectors. The project team includes educators from Hispanic-serving institutions and advocates for increasing participation of women in the fire workforce and data science fields. In addition, WIFIRE Commons’ AI Gateway machine learning, scalable computing and interactive geospatial analysis tools will be applicable to any area that can benefit from modeling.This project seeks to undertake convergence research on AI integrated wildland fire research and response, and to build a framework we call the WIFIRE Commons for using AI to enable innovative optimization of the evolving combinations of physics-based wildfire models and heterogeneous data sets used to monitor and predict wildfires in real-time. The Phase I effort will contribute toward this goal through a design-thinking approach with five streams of deliverables: 1) community convergence workshops, 2) a prototype data and model commons framework, 3) use-inspired case studies to demonstrate the proposed AI innovations, 4) prototyping of educational, outreach, and public information activities; and 5) Phase II planning. The long-term vision is to create a sustainable and open source AI-driven data and model commons to facilitate and leverage collaborations to “harness AI innovations” supporting use-inspired societal and scientific wildland fire applications. Driven by design-thinking and building upon prior research by our team members (WIFIRE, MINT, QUIC-fire), the proposed WIFIRE Commons convergence research and data and model sharing framework will enable development of novel artificial intelligence techniques and reusable models that can be utilized in many applications. This Commons infrastructure will catalog, curate and integrate data and models for AI-driven fire science, maintain open programmatic access to data in a cloud-compatible form that can be integrated into the AI process through a gateway interface, and ensure provenance of data and models over time. This AI-enabled smart data/model integration will transform the agility of science based wildland fire decision making, allowing for new kinds of models and data to be assimilated rapidly and allowing an expanding base of users to understand levels of uncertainty.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.