Contribution of Satellite imagery and DEMs to the Detection of Neolithic Settlements in Thessaly , Greece
Thessaly is a region of low relief in Greece where hundreds of Neolithic settlements/tells called magoules were established from Early Neolithic period until Bronze Age. Multi sensor remote sensing was applied to the study area in order to evaluate its potential to detect Neolithic settlements. Moreover, different kinds of digital elevation models were used such as SRTM, DEM constructed by interpolation of contours from topographic maps, DEM constructed by aerial photos and DEM constructed by Aster images, where tells can be identified as small contrasting spots within the elevation pattern of the natural variation of the land surface. A range of image processing techniques such as colour composite, principal components analysis, decorrelation stretch, followed by visual interpretation, were originally applied to the hyperspectral imagery in order to detect the settlements and validate the results of GPS surveying. The next step was to collect spectral signatures of these tell sites, to correlate them within the same spectral range of the different sensor systems and finally to proceed with their statistic analysis. Various filters were applied to all images to explore the high spectral and spatial variability of the settlement patterns, such as Sobel 3*3 right diagonal and Laplace filter. Classification of all the images using different hard and soft classifiers and application of vegetation index NDVI was followed. To cope with the difficulties of pixel based methods, object–oriented classification techniques were also applied to Ikonos imagery to classify tells according to their shape and geometry. In addition, sophisticated filters were applied to each DEM in order to detect the settlements. After validating the results with real altitude data, we concluded which of them are more reliable either for general topographic studies of the area or more specifically for the detection of the settlements. The final step was the application of fuzzy algorithms for the classification of the possibility of settlement existence. Although there are specific difficulties encountered in the classification of archaeological features composed by a similar parent material with the surrounding landscape, the results of the research suggested a different response of each sensor to the detection of the Neolithic settlements, according to its spectral and spatial resolution. Moreover, the integrated use of remote sensing imagery and the digital elevation models produced an important enhancement to the design of a predictive model of the Neolithic settlements of Thessaly by combining the spectral, spatial and topographic attributes of the tells.
