What settlements leave behind — pXRF compositional data analysis of archaeological layers from Tell el-Fara'in (Buto, Egypt) using machine learning
| Autor: | Martin Seeliger, R. Hartmann, Andreas Ginau, Jürgen Wunderlich, U. Hartung, Robert Schiestl, D. Steiniger, Marina Altmeyer |
|---|---|
| Rok vydání: | 2020 |
| Předmět: |
010506 paleontology
Context (archaeology) Geoarchaeology Settlement (structural) business.industry Paleontology Sampling (statistics) Excavation 010502 geochemistry & geophysics Oceanography Machine learning computer.software_genre 01 natural sciences Archaeology Human settlement Artificial intelligence business Compositional data computer Ecology Evolution Behavior and Systematics Geology 0105 earth and related environmental sciences Earth-Surface Processes Chronology |
| Zdroj: | Palaeogeography, Palaeoclimatology, Palaeoecology. 546:109666 |
| ISSN: | 0031-0182 |
| DOI: | 10.1016/j.palaeo.2020.109666 |
| Popis: | Modern portable and handheld XRF devices (pXRF) allow quick measurement of large geochemical datasets without the necessity for laboratory facilities. Such facilities are rare in Egypt and modern dating techniques which are indispensable in Geoarchaeology to establish a robust chronology are not available, as sample transport is restricted and Egypt does not provide OSL or AMS 14C laboratories. With these preconditions, we evaluate the usability of pXRF geochemical data for the dating of archaeological sediments with machine learning techniques. The sample material was collected via sampling of archaeological sections and profile walls from archaeological excavations in the northwestern Nile delta at the settlement site Buto (Tell el-Fara'in) and Kom el-Gir. Additionally, samples were taken from sediments and cultural layers uncovered from their surroundings using vibracore corings with open steel auger heads. In this methodological approach, we examine the applicability of pXRF methods and test the sample data for distinct geochemical differences between the main settlement phases with multivariate statistical methods. The dating is based on the training of artificial neural networks with pXRF data from archaeological material of well-dated context to date test data of cultural layers within the vibracores. This allows us to link fundamental changes in the landscape with the settlement history of Buto and neighboring tells. |
| Databáze: | OpenAIRE |
| Externí odkaz: |
|
Full Text from ScienceDirect