Multiscale anthropogenic feature detection in the Argentinian Andes: satellite machine learning, UAV, and pedestrian survey

The Cusi Cusi micro region in Argentina’s Andes presents a unique challenge for cultural heritage research. This rugged area lies within the Puna of Jujuy at elevations from 3800 to 4200 metres above sea level. Archaeological, historical, and features indicate long term human presence, including livestock enclosures, agricultural terraces, and settlement structures. The region is situated within the ‘Lithium Triangle’ and faces challenges in resource ownership, Indigenous rights, and landscape change, highlighting the importance of documenting cultural and environmental heritage. We present an integration of satellite and unpiloted aerial vehicle (UAV) imagery with pedestrian survey datasets, machine learning (ML), and physical ground truthing to explore anthropogenic activities in the landscape. Satellite-based ML detects features while UAV data improves resolution and assessment of present activities. Targeted pedestrian survey validates results and provides potential dating. This combined approach identifies loci of activities and offers a framework for investigating other rugged regions.

Paper

Full text

PDF

Multiscale anthropogenic feature detection in the Argentinian Andes: satellite machine learning, UAV, and pedestrian survey

Semantic Scholar · Environmental Science · 2026

Abstract

The Cusi Cusi micro region in Argentina’s Andes presents a unique challenge for cultural heritage research. This rugged area lies within the Puna of Jujuy at elevations from 3800 to 4200 metres above sea level. Archaeological, historical, and features indicate long term human presence, including livestock enclosures, agricultural terraces, and settlement structures. The region is situated within the ‘Lithium Triangle’ and faces challenges in resource ownership, Indigenous rights, and landscape change, highlighting the importance of documenting cultural and environmental heritage. We present an integration of satellite and unpiloted aerial vehicle (UAV) imagery with pedestrian survey datasets, machine learning (ML), and physical ground truthing to explore anthropogenic activities in the landscape. Satellite-based ML detects features while UAV data improves resolution and assessment of present activities. Targeted pedestrian survey validates results and provides potential dating. This combined approach identifies loci of activities and offers a framework for investigating other rugged regions.

Similar papers

© 2026 NYSGPT2525 LLC