Doctoral Dissertation Research: Predicting the location of hominin cave fossil sites with a machine learning approach

在一个地区调查新的化石遗址是一项费时费力的活动。然而，在遥感图像中训练的机器学习模型可以用来预测这些站点的位置。本研究的目的是建立一个适合人类化石洞穴遗址独特环境和保存背景的洞穴遗址预测机器学习模型。本博士论文研究旨在(1)确定在这种背景下，哪些环境和地貌变量对洞穴分布影响最大；(2)通过整合遥感和机器学习技术，改变洞穴化石遗址的发现过程。这是一个跨学科项目，旨在支持STEM领域的研究生培训和研究，并寻求通过在线数据共享、会议演讲、出版物和公众参与来促进未来的合作项目。在这个项目中，除了地质数据外，研究区域的遥感高分辨率图像还记录了地表反射和海拔，作为机器学习预测模型中输入预测变量的来源。这些模型利用这些输入来识别上新世-更新世古人类洞穴遗址特征之间的复杂模式，并定位具有类似特征的地区进行化石勘探。本研究比较了不同的机器学习算法，以确定哪一种算法最适合在这种情况下预测洞穴遗址，对表现最佳的模型进行实地真实评估，并分析模型在研究区域之外的效用。本项目考察地表地下化石遗址的地貌特征，寻找新的潜在化石勘探地点。这种定量方法有助于减少阻碍遗址发现的障碍，并旨在增加古人类学的化石恢复。本项目由生物人类学计划和促进竞争研究的既定计划（EPSCoR）共同资助。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。

Surveying an area for new fossil sites is a time- and labor-intensive activity. However, machine learning models trained in remotely sensed imagery can be used to predict the location of these sites. The purpose of this research is to build a predictive machine learning model for cave sites suited to the unique environmental and preservation context hominin fossil cave sites. This doctoral dissertation research seeks to (1) determine which environmental and geomorphological variables are most influential in the distribution of caves in this context, and (2) transform the process by which cave fossil sites are discovered by integrating remote sensing and machine learning techniques. This is an interdisciplinary project that supports graduate training and research in STEM and seeks to promote future collaborative projects via online data sharing, conference talks, publications, and public engagement.In this project, remotely sensed high-resolution images of the study area that document surface reflection and elevation, in addition to geologic data, serve as the sources for input predictor variables in machine learning prediction models. The models use these inputs to identify complex patterns between characteristics of Plio-Pleistocene hominin cave sites and locate areas with a similar suite of features for fossil prospecting. This research compares different machine learning algorithms to determine which one is best suited to predict cave sites within this context, conducts in-field ground-truthing assessments of the best performing model, and analyzes model utility outside the study area. This project examines the geomorphological characteristics that indicate subterranean fossil sites at the surface level and seeks out new potential localities for fossil prospecting. This quantitative approach helps reduce the barriers impeding site discovery and aims to increase fossil recovery in paleoanthropology. This project is jointly funded by the Biological Anthropology Program and the Established Program to Stimulate Competitive Research (EPSCoR).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.