Collaborative Research: Applying 3D Deep Learning to Site Detection in Tropical Regions

Researchers at several universities will develop Artificial Intelligence (AI) methods to investigate long-term human impacts on tropical ecosystems. Archaeologists can provide new insights to AI because they study both spatial and temporal components critical for understanding cultural processes that shape past and present ecosystems. The remote sensing technique of airborne light detection and ranging (lidar) captures 3D data that permits researchers to identify previously unknown archaeological features beneath forest canopy and in inaccessible places, generating new data and fundamentally changing the capacity for understanding the spatial aspects of anthropogenic landscapes.

However, in tropical regions, researchers face a challenge because they must manually examine the 2D images of lidar data to identify archaeological features, which is time-consuming, expensive, and produces results that commonly exclude small archaeological features, such as households. This project overcomes these issues by developing new methods that directly analyze 3D lidar data that can be used in addition to the 2D images.

The research team will develop transformative methods applicable to industry, academia, and beyond providing insights into current issues of the interconnections of landuse, land transformation, and the importance of the tropics in human-environment dynamics for resilience and sustainability. The study has broad implications for the local, national, and global challenges we face on multiple fronts related to climate change, urbanization, and population growth that coincides with increasing social inequality and environmental consequences.

An interdisciplinary team will use AI to develop machine learning methods that allow researchers to automatically detect archaeological features of varying sizes as well as anthropogenic landscape modifications in lidar data in relation to topography and vegetation. These methods will enhance understanding of human impacts on tropical ecosystems because they (a) produce more comprehensive documentation of the built environment, allowing for more accurate demographic reconstructions and total household counts, (b) fill in gaps in measurements of the smallest structures that constituted the majority of ancient Maya households allowing for more accurate reconstructions of household and neighborhood inequality and social networks, and (c) create more accurate maps of human-environment relationships.

Beyond archaeology, these methods will benefit biology, geology, geography, civil engineering, architecture, and urban studies, which rely on accurate reconstructions of small spatial features. The collaborative focus of the project will also create and enhance educational and training opportunities for students in geospatial techniques and computer science, and strengthen connections between the US-based institutions, international agencies, and indigenous communities.

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.