Using unsupervised classification techniques and the hypsometric index to identify anthropogenic landscapes throughout American Samoa

Abstract

Aerial LiDAR data offers a valuable tool in locating ancient anthropogenic landscapes around the world. This technology is particularly ideal in places where thick vegetation obscures the ground surface, reducing the utility of satellite imagery. On the islands of American Samoa, many interior anthropogenic landscapes remain unsurveyed, largely because the terrain makes it difficult and there is only general knowledge of where the anthropogenic modification may have existed. Aerial LiDAR flown in 2012 is proving to be a valuable tool in locating these prehistoric anthropogenic areas, yet improvements can be made on the methodology. This paper provides an unsupervised classification method to identify anthropogenic landscapes based on slope and hypsometric index: a topographic measure of roughness. Areas of American Samoa with known anthropogenic modifications were used to develop the classification techniques, which were then extended to areas where anthropogenic landscapes are undocumented and unexplored. The findings presented here suggest that interior anthropogenic patterns may be strongly dependent on island topography.