A study of inter-analyst variability in the classification of stone-walled structures in southern Gauteng, South Africa

Abstract

The study of stone-walled structures within the last 1000 years in southern Africa can help archaeologists to understand how the landscape was peopled. Google Earth and GIS can make data capture easier, but when diferent analysts are involved, there can be significant variability in their results. By comparing data classified by three researchers in the same study area, it was possible to quantify inter-analyst variability and to query where and why it occured. The quality of Google Earth imagery had much influence in introducing inter-analyst variability. Subjective decisions on classification also introduced high amounts of variability. With more intensive training on classification and better imagery, inter-analyst variability can be reduced.