A Template-matching Approach Combining Morphometric Variables for Automated Mapping of Charcoal Kiln Sites

• Published in: Archaeological prospection 2015-01, Vol.22 (1), p.45-62
• Main Authors: Schneider, Anna, Takla, Melanie, Nicolay, Alexander, Raab, Alexandra, Raab, Thomas
• Format: Article
• Language: English
• Subjects: airborne laser scanning; charcoal kilns; Digital elevation model; geoarchaeology; small landforms; template matching
• ISSN: 1075-2196; 1099-0763
• DOI: 10.1002/arp.1497

Analysing the spatial distribution of anthropogenic relief structures can contribute to the understanding of past land‐use systems. With automated mapping routines, small relief forms can be detected efficiently from high‐resolution digital terrain models (DTMs). In this study, we describe an approach for the automated mapping of charcoal kiln sites from an airborne laser‐scanning DTM. The study site is located north of Cottbus, Germany, where an exceptionally large historic charcoal production field has been documented in previous archaeological surveys. The goal of this study was to implement, evaluate and improve an automated GIS‐based routine for mapping these features based on the template‐matching principle. In addition to the DTM, different morphometric variables were evaluated for their suitability to detect kiln sites. The mapping results were validated against a comprehensive database of kiln sites recorded from archaeological excavations and via manual digitization. The effects of irregular kiln‐site geometry and DTM noise were evaluated using synthetic DTMs. The results of the synthetic DTM mapping show that the template‐matching results differed depending on the morphometric variable used for the mapping process. In accordance with this observation, a validation of the mapping procedure for the field site suggests that feature mapping can be improved. In particular, the number of false detections can be reduced using a combination of morphometric variables. For the validation area, the kiln sites with diameters of at least 10 m were mapped using the automated routine, with detection rates that were close to those of manual digitization. Therefore, the described method can considerably facilitate the mapping and distribution analysis of kiln sites or similar small relief forms that are prominent in a specific landscape. Copyright © 2014 John Wiley & Sons, Ltd.