Machine Learning Applications with Lidar to Predict Locations of Natural and Cultural Features in the Maya Lowlands

Summary

This is an abstract from the "SAA 2021: General Sessions" session, at the 86th annual meeting of the Society for American Archaeology.

This project entails creating machine learning models to predict the locations of caves and archaeological features using airborne Lidar (laser scanning) data. The goal of this work is to bridge the gap between machine learning pursued by computer scientists and the types of on-the-ground projects of interest to scientists who seek to improve management and conservation practices. This project began in 2018 with the goal of creating a targeted method of finding cave entrances in the dense tropical forests of Guatemala and Belize. In 2019, we used a random forest classifier and a training dataset of known caves to successfully identify several previously undocumented caves. This included a 200-foot-deep collapsed cave complex. Building on this work, modeling has been expanded to include other types of hidden and obscured features that colleagues are interested in studying, including ancient Maya archaeological features in Guatemala and Mexico, as well as shipwrecks. These models are based on existing convolutional neural network architectures. The first completed iteration of the models has an F1 score of 0.92. The models can be used to create more accurate maps of natural and archaeological features to aid management objectives, study patterns across the landscape, and find new features.

This Resource is Part of the Following Collections

• Society for American Archaeology 86th Annual Meeting, Online (2021)   •
• SAA 2021: General Sessions

Keywords

Spatial Coverage

Record Identifiers

Abstract Id(s): 33090