Archaeological AI Development

Over the course of my career, I have been active in the advancement of computer automation methods (like machine learning and deep learning models) within the realm of archaeology and cultural heritage. I have many different projects that aim to develop and refine existing techniques for automatically identifying archaeological sites from remote sensing datasets. I am also, currently, a member of a recently funded project through the European Cooperation in Science and Technology (eCOST) that aims to develop training opportunities and best practices for the development of AI approaches within archaeological research.

Inverse Depression Analysis (IDA)

I have developed a method for identifying mound architecture using existing hydrological algorithms originally designed for sinkhole detection. This method does not require training data (which is helpful in contexts where data is limited) and has been successfully deployed in a variety of environments and case studies. In South Carolina, the method (IDA) allowed my colleagues and I to identify new Native American mound sites. In Sweden, IDA allowed us to identify enough charcoal production sites to train a deep learning convolutional neural network algorithm. The technique has also been successful in identifying underwater archaeological sites off the coast of Florida and shipwrecks along the coast of the United States. IDA successfully identified late Pleistocene/Early Holocene archaeological sites

Machine Learning and Deep Learning

Automated Detection of Samoan Terracing Agriculture

In collaboration with Dr. Seth Quintus, I am helping to develop a deep learning model to assist archaeologists working on Samoa in detecting and digitizing agricultural terracing features across all of the different Samoan islands. The results of this automated detection and digitization procedure are creating a comprehensive and systematically acquired dataset of terracing agriculture that will allow for in depth studies of niche construction, food production, and population distributions on Samoa over time.

Mapping Native American Mound Architecture in the Southeastern United States

Since 2018, this project has evolved to incorporate remotely sensed archaeological features to re-evaluate settlement patterns in this region. Using LiDAR data and different computer automation methods, my colleagues and I are using and spatial modeling methods to investigate shell mound and shell ring construction practices. We began using object-based image analysis (OBIA) methods, including image segmentation, template matching, and inverse depression analysis (described above), which helped identify many new ring-like features, several of which have been confirmed as archaeological shell rings, to date. In continuing work, we are continuing to improve our original algorithms using convolutional neural networks (or CNN), a form of deep learning.


Using CNNs, we were able to evaluate the spatial extent of shell ring building practices between Georgia and South Carolina. The results of this work were published in Journal of Archaeological Science. The future of this project will continue to examine Archaic settlement patterns along the entire Eastern coastline of the United States and investigate their relationship with climate events and environmental conditions.

Niche Construction on Madagascar

On Madagascar, my colleagues and I have used machine learning algorithms to identify areas of extended human land-use. By using time-averaged satellite datasets, we were able to identify subtle differences between archaeological sites and their surroundings based on vegetation and soil characteristics. This allowed us to train a random forest algorithm to identify places that were modified by past human activities and accelerate archaeological discovery. The results were published in Frontiers in Ecology and Evolution.
In ongoing work, we are now trying to improve our use of machine learning to identify different kinds of past land use from archaeological signatures. This will help clarify the role of different traditional land use practices (like foraging, pastoralism, and agriculture) on ecosystems in this area over time.