Archaeoinformatics - Data Science

BA/MA: Thesis in Data Science - Magnetic Pattern Classification

Contact: Steffen Strohm, M.Sc.

Magnetic prospection data depicts the remains of archaeological sites. We investigate the sites of the Cucuteni-Tripolye Culture via the magnetic anomalies of burned houses to infer the social strucutre. From the magnetic anomaly we derive the magnetization distribution which correlates with the mass of burnt clay. The magnetization patterns of the building remains are directly related to the architecture, which in turn can reflect social structures.

For the settlements of the Cucuteni-Tripolye Culture is characteristic: large sites with up to several thousands of buildings, concentric settlement layout and a high degree of standardization concerning the floor plan of the houses. The Figure shows a section of the magnetic anomaly map (a), the thereof derived magnetization distribution (b) and three examples of house anomalies in comparison to a standardized floor plan (c).

Example of a magnetic anomaly map, derived magnetization distribution and examples of house anomalies. 

(Click on picture for details in new tab)

There are two main questions related to the magnetic data of these sites, which can be tackled with supervised or unsupervised learning or data mining techniques:

(1) How can the building remains be automatically identified?
(2) Are there different clusters of buildings in terms of their magnetization patterns? If so, how are these clusters related to the ideas of standardized floor plans?


This work is a collaborative project with scientists from subproject G2 (Geophysics) and subproject D1 (Archaeology) within the CRC 1266 "Scales of Transformation" at Kiel University. No previous knowledge of Geophysics or Archaeology is required.