Collaborative Research: Community tools for automated paleoenvironmental interpretation from sedimentary field data

地质野外数据对于重建地球和火星的历史条件、发现和开发自然资源以及管理自然灾害是必不可少的。沉积地质学依赖于岩石露头中的一组模式，这些模式提供了关于过去存在的自然环境(如海洋、河流、沙漠或湖泊)的信息。该项目将开发从沉积露头数据自动解释过去环境的工具。地球科学家和计算机科学家对露头数据集的合作评估将为如何完成地质露头解释提供新的视角，并提供对野外数据自动解释所需信息的更深层次的理解。自动露头解释将扩大研究人员快速挖掘现有和新的数字数据集的能力，以获得有助于对地球和火星表面条件的历史做出新见解的信息。作为该项目的一部分开发的工具将得到广泛传播，并将与现有的外地数据数字平台相结合。这些工具将有助于促进广泛的地球科学家之间的合作，包括那些不能轻易在偏远地区进行野外工作的人。该项目将通过交叉培训具有高级计算机科学技能的沉积地质学家来加强劳动力发展，并将为计算机科学专业的学生提供机器学习方法的实际应用范例。为了实现这些目标，该项目将利用现有的数字露头数据集并收集新的目标数据集，以探索从露头图像和地表地形数据中提取沉积特征的自动化方法。该项目将重点提取对古环境重建至关重要的沉积特征，包括交错层理类型，并调查露头质量、规模和方向如何影响沉积特征的可恢复性。此外，该项目将探索从自动化露头观测中可以推断出沉积特征的三维方向的程度。该项目开发的工具将用构造地质学和露头地貌学数据集进行测试，以评估作为这项工作的一部分开发的工作流程如何有助于更广泛的地球科学学科。结果将被打包成一个可访问的在线工具，可与数字外地数据储存库结合使用。该工具的广泛传播和社区的广泛参与将随着时间的推移加强算法并改进社区资源。将开发适合本科生沉积地质学课程的教材，并将有助于扩大本科生对计算机科学的接触。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。

Geological field data is essential for reconstructing historical conditions on Earth and Mars, finding and developing natural resources, and managing natural hazards. Sedimentary geology relies on a set of patterns in rock outcrops that provide information on where natural environments (like oceans, rivers, deserts, or lakes) existed in the past. This project will develop tools for automated interpretation of past environments from sedimentary outcrop data. Collaborative evaluation of outcrop datasets by geoscientists and computer scientists will yield new perspectives on how geological outcrop interpretation can be accomplished and provide a deeper understanding of the information required for automated interpretation of field data. Automating outcrop interpretation will broaden researchers’ ability to rapidly mine existing and new digital datasets for information that will contribute new insights into the history of surface conditions on Earth and Mars. The tools developed as part of this project will be broadly disseminated and will integrate with existing digital platforms for field data. These tools will help facilitate collaborations among a broad range of geoscientists, including those who can’t easily conduct fieldwork in remote locations. This project will strengthen workforce development by cross-training sedimentary geologists with advanced computer science skills and will provide examples of practical applications of machine-learning approaches for computer science students.To accomplish these goals, this project will leverage existing digital outcrop datasets and collect a new targeted dataset to explore automated approaches for extracting sedimentary features from outcrop image and surface-topography data. The project will focus on extracting sedimentary features that are critical for paleoenvironmental reconstruction, including types of cross bedding, and investigate how outcrop quality, scale, and orientation influence the recoverability of sedimentary features. Additionally, this project will explore the degree to which the three-dimensional orientation of sedimentary features can be extrapolated from automated outcrop observations. The tools developed in this project will be tested with structural geology and geomorphology outcrop datasets to evaluate how workflows developed as part of this work could aid a broader range of geoscience disciplines. Results will be packaged as an accessible online tool that can be used in combination with digital field data repositories. Widespread dissemination of this tool and broad community participation will strengthen the algorithm and improve the community resource over time. Educational materials suitable for undergraduate sedimentary geology courses will be developed and will help expand undergraduate geosciences students’ exposure to computer science.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.