UNSUPERVISED CLASSIFICATION TO IDENTIFY SETTLEMENT AREAS IN AMERICAN SAMOA

As coverage of aerial lidar has been extended, it has become an increasingly valuable tool for locating and evaluating historic and ancient settlements around the world. While semi-automated methods are used in some cases for finding settlements, few unsupervised classifications exist, and typically settlements are located using a fully manual procedure. Aerial Lidar was flown for American Samoa in 2012, and has facilitated a much better understanding of interior settlements on the smaller islands in the archipelago, Ofu, Olosega, and Ta‘u. This work has been done through a combination of manual and semi-automated classification, yet these techniques are difficult to extend to the much larger island of Tutuila. This issue is addressed by the development of an unsupervised classification method based on slope, and hypsometric index: a measure of topographic roughness. The classification method was tested on Ofu and Olosega, as these islands have well documented known settlements, and then used to locate likely settlement areas on Tutuila. The findings from this work suggest that settlement patterns are strongly dependant on topography, and that settlements in Tutuila are likely smaller than those found on Ofu and Olosega.