A Machine-Learning-Enabled Survey for Large Residential Ancestral Pueblo Sites Using Publicly Accessible Lidar Data

Author(s): Sean Field

Year: 2025

Summary

This is an abstract from the "Lidar Research in the US Southwest" session, at the 90th annual meeting of the Society for American Archaeology.

<html>

Within the past five years, the US Geological Survey’s 3D Elevation Program (3DEP) has released large swaths of lidar data across the continental United States, giving archaeologists unprecedented access to high-resolution, landscape-scale elevation data that can be used to locate, map, and visualize medium- to large-scale archaeological sites and features. Here, we build an image classification algorithm using the Pytorch pipeline to detect and locate large, ancestral Pueblo sites in static images of hillshaded, lidar-derived digital surface models. The model is trained using site data from Mesa Verde National Park and deployed over a 15,000 km<sup>2 </sup>area in southwestern Colorado and southeastern Utah. Results demonstrate the promise of machine learning for large-scale landscape survey and caution the need for the critical application of machine learning toolkits for the creation—but not interpretation—of archaeological data.

</html>

Cite this Record

A Machine-Learning-Enabled Survey for Large Residential Ancestral Pueblo Sites Using Publicly Accessible Lidar Data. Sean Field. Presented at The 90th Annual Meeting of the Society for American Archaeology. 2025 ( tDAR id: 509991)

Record Identifiers

Abstract Id(s): 51194