Geophysical clues about oceanic plateau formation from magnetic anomalies and machine learning

基本的科学问题是了解海洋高原的起源。海洋高原是巨大的水下火山，由大量的火山爆发造成。它们约占海洋地壳的5%，可以轻易地覆盖像阿拉斯加这样大的地区，厚度达到数十公里。 大洋高原的形成与古气候变化和生物大灭绝相吻合。因此，研究海洋高原是一项重要的科学工作，它将促进我们对地球动力系统的了解，并可能促进对过去气候变化的了解。然而，人们对海洋高原的形成和内部结构知之甚少，因为它们位于水下数千公里处，距离陆地数千公里。该项目将侧重于利用海洋磁性数据和机器学习来推进我们对海洋高原形成和结构的理解。主要的假设是，海洋高原形成于扩张脊附近，并以线性磁异常为主，因此，线性火山活动。为了检验这一假设，该小组将把来自三个选定的具有良好磁测数据覆盖率的大洋高原的海洋磁测异常的空间模式与来自两个大洋中脊（即东太平洋海隆和大西洋中脊）的海洋磁测异常的空间模式进行比较。该团队将采用三种经过验证的机器学习方法，而不是依赖于对磁异常的目视检查，以帮助定量比较五个地区磁异常的空间模式。该项目将有助于加快基于磁性数据的新发现，并将有助于留住和吸引年轻人才进入磁性领域，这是关键矿物和行星探索的一种行之有效的方法。此外，项目活动将为来自代表性不足群体的学生创造教育和培训机会，并帮助发展美国的人工智能劳动力。海洋高原的流行观点是，它们是由大规模火山作用在已经存在的海洋地壳上形成的集中式盾状火山。然而，最近的研究表明，塔穆和奥里地块是由线性火山作用形成的，这一过程类似于海底扩张火山作用。该项目将提供新的证据，以了解这两种理论中的任何一种是否有效，或者海洋高原是否以混合方式形成（即，两者在一定程度上都是正确的）。除了调查磁异常是否是线性的之外，这项研究还将为人们所知甚少的海洋高原地质结构提供新的线索。这项研究的一个重要方面是将机器学习应用于海洋高原海洋磁异常的分析，尽管机器学习在包括地球科学在内的许多其他领域取得了巨大成功，但还没有尝试过。因此，拟议的研究还将促进我们对机器学习的理解，这是地球科学界一个新的但迅速增长的兴趣，当与海洋磁性数据结合并应用于海洋高原时，它将产生新的知识。该研究还将有助于了解机器学习方法在应用于海洋磁力数据时的优势和局限性。该奖项反映了NSF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 fundamental science problem is understanding the origin of oceanic plateaus. Oceanic plateaus are massive underwater volcanoes resulting from voluminous volcanic eruptions. They represent about 5% of oceanic crust and can easily cover areas as large as Alaska and reach thickness of tens of kilometers. The formation of oceanic plateaus coincides with significant paleoclimate change and mass extinction. Therefore, studying oceanic plateaus is an important scientific pursuit which will advance our understanding of Earth’s dynamic systems and, possibly, past climate changes. However, the formation and internal structures of the oceanic plateaus are poorly understood because they are located kilometers underwater and thousands of kilometers away from land. The project will focus on advancing our understanding of the formation and structure of oceanic plateaus using marine magnetic data, and machine learning. The main hypothesis is that oceanic plateaus formed near spreading ridges and are dominated by linear magnetic anomalies, and therefore, linear volcanism. To test this hypothesis, the team will compare the spatial patterns of marine magnetic anomalies from three selected oceanic plateaus with good magnetic data coverage against those from two mid-ocean ridges, namely, the East Pacific Rise and the Mid-Atlantic Ridge. Instead of relying on visual inspection of magnetic anomalies, the team will implement three proven machine learning methods to help quantitatively compare the spatial patterns of the magnetic anomalies from the five regions. The project will help to accelerate new discoveries based on magnetic data and will help to retain and attract young talent to the area of magnetics, a proven method in critical mineral and planetary exploration. In addition, the project activities will create educational and training opportunities for students from underrepresented groups and help to develop America’s artificial intelligence workforce.The prevailing view of oceanic plateaus is that they are centralized shield volcanoes formed by massive volcanism onto already existing oceanic crust. However, recent research proposed that Tamu and Ori Massifs formed by linear volcanism, a process similar to seafloor spreading volcanism. This project will provide new evidence to understand whether either of these two theories is valid or whether oceanic plateaus formed in a hybrid manner (i.e., both are true to an extent). In addition to investigating whether magnetic anomalies are linear, the research will give new clues about the geologic structure of oceanic plateaus, which is poorly known. An important aspect of the study is the application of machine learning to the analysis of marine magnetic anomalies over oceanic plateaus, which has not been attempted even though machine learning has achieved great success in many other areas including geoscience. Therefore, the proposed research will also advance our understanding of how machine learning, a new yet rapidly growing interest in the geoscience community, generates new knowledge when combined with marine magnetic data and applied to oceanic plateaus. The research will also help understand the advantages and limitations of the machine learning methods when applied to marine magnetic data.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.