Detecting Looting Activity through Earth Observation Multi-Temporal Analysis over the Archaeological Site of Apamea (Syria) during 2011–2012

Abstract

This study aims to investigate the potentials of open-access, freely distributed Earth Observation images for detecting large-scale looted areas. The analysis was conducted using medium-resolution Landsat 7 ETM+ images over the archaeological site of Apamea, at Syria. The site has been systematically and extensively looted in the recent past, attracting the interest of scholars. We propose a multi-temporal analysis of cloud-free multispectral Landsat 7 ETM+ images throughout the period between January 2011 to April 2012, just at the beginning of the Syrian war. The analysis was completed through the interpretation of pseudo-color temporal composites, investigation of the multi-temporal spectral profiles, correlations between the spectral bands, and the application of principal component analysis (PCA). The overall analysis was limited within the spectral range of 450–1750 nm. This wavelength range corresponds to the first five spectral bands of the Landsat images. Furthermore, we explored the big data cloud platform Google Earth Engine to detected looted areas. A supervised classification strategy was designed and performed on this cloud platform employing the Random Forest classifier. Finally, a time-stamp change detection approach was implemented. The overall findings were compared with available images from Google Earth Digital Globe and published articles and reports related to the Apamea archaeological site. It was found that the high revisit temporal resolution of the Landsat sensor was able to detect and map the looting activity in the area as a result of the spectral change in the archaeological landscape, despite its 30 m spatial resolution. At the same time, however, the analysis has provided other false-true detections in other areas in the landscape.