We are grateful to the German Research Foundation – Project Number 2901391021 – SFB 1266 for funding the presented research.
1Magnetic prospection of archaeological sites is widely applied. After basic processing, like median filters, essential spatial information about the site can be derived from the magnetic anomaly map. The interpretation yields the layout of the site and anomalies are classified as different building types, pits, ditches, etc. Apart from the differentiation of ‘weak’ and ‘strong’ anomalies, the magnetic properties of the subsurface structures are not taken into account in the image interpretation. Hence, there is unused information available in the magnetic map that can be used to quantify subsurface structures. Here, we present an interpretation scheme for magnetic prospection data with which we first calculate the magnetization, second quantify the building remains in terms of their magnetization per area and third statistically evaluate the data for about 2300 buildings to deduce internal patterns of the site.
2The Chalcolithic site Maidanetske is located in the Southern Bug-Dnieper interfluve (Ukraine) and was inhabited between approximately 3950 and 3650 BCE. Maidanetske is one of the ‘Tripolye megasites’ with a distinct concentric spatial layout, central free spaces, public buildings, and thousands of mostly burned dwellings (Fig. 1). The magnetic map covers 175 ha with anomalies of 2300 burnt clay builds. For the dwellings different types of architecture are known with specific constructive characteristics and a high degree of standardization (Chernovol, 2012). We aim to investigate the architecture by calculating the magnetization as an indicator of the used masses of clay. Furthermore, this data can reveal intra-site patterns that might be related to social structure.
Figure 1. Magnetic map of the site Maidanetske (data provided by Römisch-Germanische Kommission, Deutsches Archäologisches Institut and Institute of Pre- and Protohistoric Archaeology, Kiel University) and the location of the site in Ukraine. The red rectangle marks the area we investigate in Figure 3 as test area.
3The interpretation scheme is based on three conditions: (1) Excavations have shown that the burnt clay is located in a distinct depth range. (2) By modeling, we have shown that the burnt clay is the main source of the magnetic anomalies and therefore hosts the magnetization (Pickartz et al., 2019). (3) Based on three excavated buildings, we have shown that the mass distribution of the burnt clay correlates with the magnetization distribution.
4The interpretation scheme is as follows.
5The magnetization distribution is calculated via inverse filtering (Tsokas & Papazachos, 1992; Tassis et al., 2008). In this approach the magnetic anomaly field T is understood as a convolution of the magnetization intensity (amplitude function D) and the Green’s function (shape function R) of a basic body. Consequently, the magnetization can be calculated by convolving the magnetic anomaly field with the inverse (in terms of convolution) of the shape function. The basic body has to be chosen so that it can be used to reconstruct the target’s geometry and has unit magnetization within the same direction as the targets. We use a cube in the depth range of the layer of burnt clay (0.35 – 0.6 m) and the size of a grid cell (0.25 m × 0.25 m). The magnetization is in the direction of the earth’s magnetic field (inclination 65.9°, declination 6.7°) at the time of the surveys that has been derived from the International Geomagnetic Reference Field. The shape function is modeled with ‘Fatiando a terra’ (Uieda et al., 2013) and its inverse is derived by minimization of E2=( R*R−1−I )2, where R−1 is the truncated inverse version of R. For this approach, the constant and known depth range of the magnetized layer and the direction of the magnetization are essential.
6This concept is illustrated in Figure 2. Figure 2a shows a magnetization model that resembles the circumstances at the site. This magnetization model is used to calculate the magnetic anomaly map with ‘Fatiando a Terra’ (Fig. 2c). Figure 2b shows the shape function of the basic body as described above and Figure 2d shows the inverse filter. The result of the inverse filtering is shown in Figure 2e, whereas Figure 2f illustrates the deviations between original model and the inversion result.
Figure 2. (a) Magnetization model for testing the inverse filtering. (b) Shape function R of the basic body. (c) Magnetic anomaly of the magnetization model calculated with ‘Fatiando a Terra’. (d) Inverse of the shape function, hence the inverse filter R−1. (e) Result of inverse filtering, hence the convolution of c with d. (f) Difference between (a) and (e) showing the fit.
7These deviations are small showing this concept is suitable to determine the magnetization distribution.
8The magnetization distribution is determined for the complete site. To compare the buildings, the sum of their magnetization is divided by the floor area. The total magnetization per area has shown to be an indicator of different building types (Pickartz et al., 2019).
9The magnetization per floor area is the base for a statistical evaluation of the settlement. We aim to investigate questions like: Is in a specific area of the settlement a specific building type dominant or absent? Does the mass have a correlation with time, e.g. have the oldest buildings less mass than the younger ones? Are there any other intra-site patterns visible within the distribution of the masses of building remains when taking the location within the site and temporal settlement phase into account? What does this tell us about the life within the settlement or the society of the inhabitants?
10Figure 3 illustrates the results of first tests. Figure 3a shows the magnetic map of the test area. The houses are arranged in concentric ellipses in the western part and in radial rows in the eastern part. There is a corridor of free space between two concentric rows. Figure 3b shows the magnetization distribution that has been calculated with the inverse filter technique and the resulting magnetization per floor area (Fig. 3c). So far, these first tests have shown that the dwellings along a concentric corridor of free space have a higher total magnetization per floor area than those located far from the corridor (Fig. 3d). Moreover, the public buildings, termed ‘megastructures’ are located within this corridor. This first test clearly shows that we can identify internal site patterns, which gives us the possibility to pose and answer new questions with the magnetic prospection data.
Figure 3. (a) Magnetic map of the test area. (b) Magnetization distribution in the test area. (c) Magnetization per floor area for the buildings inside the test area. The white area shows the free corridor, the blue and red strips indicate the buildings that are directly located along the corridor. The gray area marks the buildings far from the corridor. (d) Boxplots of the magnetization per area for the houses with in the three groups.