CAREER: CAS- Climate: Climate Adaptation Pathways in Eco-Hydrologic Systems with Physics-Informed Machine Learning

职业：CAS-气候：基于物理的机器学习在生态水文系统中的气候适应途径

• 批准号: 2144332
• 负责人: Scott Steinschneider
• 金额: $ 50.66万
• 依托单位: Cornell University
• 依托单位国家: 美国
• 项目类别: Continuing Grant
• 财政年份: 2022
• 资助国家: 美国
• 起止时间: 2022-09-01 至 2027-08-31
• 项目状态: 未结题
• 来源: https://www.nsf.gov/awardsearch/showAward?AWD_ID=2144332&HistoricalAwards=false
• 关键词: CAREER; CAS; Climate; Adaptation; Pathways

Sustainable water resources planning and management in the 21st century requires adaptation strategies that are robust to deep uncertainties in future climate and environmental change. To manage this uncertainty, recent decision-making frameworks have promoted flexible adaptation pathways that respond dynamically to the trajectory of climate as it unfolds. This project will explore the hypothesis that the optimal design of flexible adaptation pathways – including the sequence, timing, and permanence of adaptation actions - depends on the mechanisms of dynamic and thermodynamic climate change influencing the water system, and the degree of natural climate variability and predictability across time scales. To test this hypothesis, this work will develop innovations in physics-informed machine learning that will enable process-guided climate simulation, hydrologic prediction, and sub-seasonal-to-seasonal forecasting that support risk-based adaptation planning. These approaches will be applied in the Lake Ontario eco-hydrologic system to examine three fundamental questions: 1) What are the primary patterns of dynamic and thermodynamic climate change over the Great Lakes, and how can they be integrated into risk-based simulation frameworks? 2) How do these climate mechanisms influence the hydrologic and ecological response of the Lake Ontario system and the diverse interests of stakeholders, and how predictable are these impacts at sub-seasonal to decadal timescales? and 3) How should adaptation pathways for lake level management and coastal resilience be designed to cope with these mechanisms of climate change? These questions will be addressed alongside a co-production model of community engagement and knowledge sharing with Great Lakes communities, students, and other stakeholders.This work will impact hydroclimatic modeling for eco-hydrologic systems by developing physics-informed machine learning techniques for feature identification, spatiotemporal modeling, emulation, functional dependence, and synthetic data generation, with the ability to propagate physically meaningful features through sequentially linked systems while accounting for uncertainty. The goal is to develop a computationally efficient and probabilistic modelling framework for risk-based simulation and forecasting needed to develop robust climate adaptation pathways. Expected outcomes include six major scientific advancements: 1) a diagnostic understanding of historical and projected future thermodynamic and dynamic climate processes relevant to eco-hydrologic systems; 2) the development of stochastic models that can reveal how those physical processes shape future climate risk to water infrastructure; 3) enhanced predictability of eco-hydrologic response to climate across sub-seasonal to decadal time scales; 4) credible emulation of system objectives to support uncertainty propagation in adaptation planning; 5) endogenous learning strategies to detect mechanisms of climate change from projections and noisy observations; and 6) the identification of general principles for how to develop adaptation pathways in water systems exposed to multi-scale climate variability and different mechanisms of climate change. This project will integrate research, teaching, and service missions through a pedagogic and scholarly model of community engagement that promotes knowledge co-production and translation between students, academics, extension and education specialists, Great Lakes communities, and an international board of water managers. Undergraduate and Graduate Education: Models and partnerships developed through this work will enhance community-engaged project experiences for undergraduate and graduate students in courses on hydrologic engineering, climate change, and machine learning. Active learning modules will embed data science literacy directly into these educational experiences. Student Training: This project will provide an interdisciplinary training experience for 1 PhD student, and undergraduates will also be recruited to participate through course-based research. Public Forums and K-12 Education: Through collaborations with the Sciencenter of Ithaca NY and New York Sea Grant, this work will develop public forums to educate and learn from Lake Ontario communities, particularly those in rural, low-income areas, about water level variability, management, and impacts on community resilience. These collaborations will also support middle school curriculum development, disseminated widely across the Great Lakes shoreline. Real-World Decision-Making: By collaborating with the International Joint Commission, this work will enhance the adaptive management plan of one of the largest managed, freshwater lakes in the world that is undertaking one of the largest wetland restoration efforts in North America.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.

21 世纪的可持续水资源规划和管理需要能够应对未来气候和环境变化的深刻不确定性的适应战略。为了管理这种不确定性，最近的决策框架提倡灵活的适应途径，可以动态地响应气候变化的轨迹。该项目将探讨这样一个假设：灵活适应路径的优化设计（包括适应行动的顺序、时间安排和持久性）取决于影响水系统的动态和热力学气候变化机制，以及跨时间尺度的自然气候变化程度和可预测性。为了检验这一假设，这项工作将在基于物理的机器学习方面进行创新，从而实现过程引导的气候模拟、水文预测和次季节到季节的预测，从而支持基于风险的适应规划。这些方法将应用于安大略湖生态水文系统，以研究三个基本问题：1）五大湖动态和热力学气候变化的主要模式是什么，以及如何将它们整合到基于风险的模拟框架中？ 2）这些气候机制如何影响安大略湖系统的水文和生态响应以及利益相关者的不同利益，以及这些影响在次季节到十年时间尺度上的可预测性如何？ 3）应如何设计湖泊水位管理和沿海恢复力的适应路径来应对这些气候变化机制？这些问题将与五大湖社区、学生和其他利益相关者的社区参与和知识共享的联合生产模型一起得到解决。这项工作将通过开发用于特征识别、时空建模、仿真、功能依赖和合成数据生成的物理信息机器学习技术，影响生态水文系统的水文气候建模，并能够通过顺序链接的系统传播物理上有意义的特征。 同时考虑不确定性。目标是开发一个计算效率高的概率建模框架，用于基于风险的模拟和预测，以开发稳健的气候适应路径。预期成果包括六项重大科学进展：1）对与生态水文系统相关的历史和预测的未来热力学和动态气候过程的诊断性理解； 2) 开发随机模型，揭示这些物理过程如何影响水基础设施的未来气候风险； 3) 增强次季节到十年时间尺度上生态水文对气候响应的可预测性； 4）系统目标的可信仿真，以支持适应规划中的不确定性传播； 5）内生学习策略，从预测和噪声观测中检测气候变化机制； 6) 确定如何在暴露于多尺度气候变率和不同气候变化机制的水系统中制定适应途径的一般原则。该项目将通过社区参与的教学和学术模式整合研究、教学和服务任务，促进学生、学者、推广和教育专家、五大湖社区和国际水管理委员会之间的知识共同生产和翻译。本科生和研究生教育：通过这项工作开发的模型和合作伙伴关系将增强本科生和研究生在水文工程、气候变化和机器学习课程中的社区参与项目体验。主动学习模块将数据科学素养直接嵌入到这些教育体验中。学生培养：该项目将为1名博士生提供跨学科的培养体验，同时也将招募本科生通过课程研究的方式参与。公共论坛和 K-12 教育：通过与纽约州伊萨卡科学中心和纽约海洋补助金合作，这项工作将开发公共论坛，向安大略湖社区，特别是农村、低收入地区的社区进行教育和学习，了解水位变化、管理以及对社区复原力的影响。这些合作还将支持中学课程的开发，并在五大湖沿岸广泛传播。现实世界决策：通过与国际联合委员会合作，这项工作将加强世界上最大的管理淡水湖之一的适应性管理计划，该湖正在进行北美最大的湿地恢复工作之一。该奖项反映了 NSF 的法定使命，并通过使用基金会的智力价值和更广泛的影响审查标准进行评估，被认为值得支持。