Collaborative Research: Machine-enabled modeling of terminus ablation for Greenland's outlet glaciers

合作研究：格陵兰岛出口冰川终点消融的机器模型

• 批准号: 2146704
• 负责人: Denis Felikson
• 金额: $ 11.45万
• 依托单位: Universities Space Research Association
• 依托单位国家: 美国
• 项目类别: Standard Grant
• 财政年份: 2022
• 资助国家: 美国
• 起止时间: 2022-08-15 至 2023-04-30
• 项目状态: 已结题
• 来源: https://www.nsf.gov/awardsearch/showAward?AWD_ID=2146704&HistoricalAwards=false
• 关键词: Collaborative; Research; Machine; enabled; modeling

Predicting future sea level relies on improved modeling of how the climate forces the ice sheet to change. This is particularly challenging at the ice-ocean boundary where there are multiple processes occurring simultaneously. At present, no single equation adequately describes the changing ice-ocean boundary, which poses problems for coupling ice sheet models to climate models. This project will improve understanding of how the variables that influence the ice-ocean boundary may change over time and space using machine learning to search for relationships amidst available data. This technique will allow the research team to categorize glacier terminus behavior and identify the relevant parameters forcing change for a particular glacier around the Greenland Ice Sheet. Results of the machine learning exercise will be used to develop an equation to represent ice-ocean interactions in an ice sheet model which will be used to determine future changes to the ice sheet forced by the ocean into the future. Models of future ice sheet change yield reliable forecasts of sea level rise only when all the critical processes controlling ice sheet evolution are appropriately accounted for. However, many physical processes are currently poorly understood. One such process is ablation (iceberg calving and submarine melt) at the terminus of outlet glaciers, which has been shown to be the dominant control on mass change at particular glaciers. The goal of this project is to improve model forecasts of sea-level from Greenland by using machine learning analyses of glaciological observations to inform physics-based modeling of outlet glaciers, with a focus on the ice-ocean boundary. Machine learning tools will be used to determine what controls changes in terminus position over a range of time scales for all glaciers in Greenland over a period of pronounced historical change (the satellite era). Analysis of the model performance will enable the research team to determine the dominant controls on terminus position for individual and groups of glaciers and to test how well the model performs as new glaciological and environmental data become available. The machine learning model of terminus positions will be used to improve projections of outlet glacier mass change using a physically-based numerical ice flow model. The team will examine how robust model prediction is on various time-scales as more and more data become available over the course of this project. The project will result in refined projections of dynamic loss from the Greenland Ice Sheet, which is important for policy makers needing to make critical infrastructure and resource decisions globally. This goal is a central focus for research within NSF's Office of Polar Programs, NSF's Navigating the New Arctic, and other national (e.g., NASA, NOAA) and international priorities. The project integrates researchers across disciplines, genders, and career stages. Data products and methods produced through this project will be make publicly available and will be useful to the broader scientific community. This project is co-funded by a collaboration between the Directorate for Geosciences and Office of Advanced Cyberinfrastructure to support AI/ML and open science activities in the geosciences.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.

预测未来的海平面依赖于气候如何迫使冰盖变化的改进模型。这在冰-海边界尤其具有挑战性，在那里有多个过程同时发生。目前，没有一个单一的方程能够完全描述不断变化的冰-海边界，这给冰盖模型和气候模型的耦合带来了问题。该项目将提高对影响冰海边界的变量如何随时间和空间变化的理解，利用机器学习在可用数据中搜索关系。这项技术将使研究小组能够对冰川末端的行为进行分类，并确定迫使格陵兰冰盖周围特定冰川发生变化的相关参数。机器学习的结果将被用来在冰盖模型中开发一个方程式来表示冰与海洋的相互作用，该方程式将被用来确定未来被海洋强迫的冰盖的未来变化。只有在适当考虑了控制冰盖演变的所有关键过程时，未来冰盖变化的模型才能产生可靠的海平面上升预测。然而，目前人们对许多物理过程知之甚少。其中一个过程是出口冰川末端的消融(冰山崩解和海底融化)，这已被证明是对特定冰川质量变化的主要控制。该项目的目标是通过使用对冰川观测的机器学习分析来改进格陵兰的海平面模型预报，以便为以物理为基础的出口冰川建模提供信息，重点是冰-海洋边界。将使用机器学习工具来确定格陵兰所有冰川在一个显著的历史变化时期(卫星时代)在一系列时间尺度上的末端位置变化是由什么控制的。对模型性能的分析将使研究小组能够确定单个和组冰川末端位置的主要控制因素，并测试随着新的冰川学和环境数据的获得，模型的表现如何。终端位置的机器学习模型将用于通过基于物理的数值冰流模型来改进对出口冰川质量变化的预测。随着越来越多的数据在项目过程中变得可用，该团队将检查模型预测在不同时间尺度上的健壮性。该项目将对格陵兰冰盖的动态损失进行精确预测，这对需要在全球范围内做出关键基础设施和资源决策的政策制定者来说非常重要。这一目标是NSF极地计划办公室、NSF导航新北极以及其他国家(例如NASA、NOAA)和国际优先事项研究的中心重点。该项目汇集了不同学科、性别和职业阶段的研究人员。通过该项目产生的数据产品和方法将公之于众，并将对更广泛的科学界有用。该项目由地球科学局和高级网络基础设施办公室共同资助，以支持AI/ML和地球科学领域的开放科学活动。该奖项反映了NSF的法定使命，并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。