The dolmens erected in the province of Drenthe between 3350-2700 BC are the most ancient monuments of the Netherlands. They consist of a long assemblage of rocks capped with boulders which served as burial chambers for the Funnel Beaker population. Despite their robust look, dolmens are vulnerable. For example, one of the caprocks of dolmen D14 fell off in 2019, once again. The method currently used to repair megaliths is not optimal: a crane is mobilized and the rocks are repositioned using a trial and error strategy. The reconstruction scenario should be selected and fine-tuned digitally beforehand. Moreover, the structural stability of the new assemblage of rocks should also be checked numerically. This necessitates a digital model of all the rocks that have to be rearranged. Not only the visible part of these rocks has to be digitized. The buried parts of the support rocks have to be modelled too as these rocks might have to be displaced and tilted to obtain a more compact and interlocked structure. Since dolmens occupy a prominent position in the Dutch heritage, only non-destructive see-through techniques have to be used for imaging the hidden contours of the bearing rocks. In addition, the bearing stones have complex geometries and are not isolated, which increases the complexity of the problem.

Two suitable geophysical methods are the Ground Penetrating Radar (GPR) and seismic techniques with a focus on reflection measurements. For the GPR specifically, we choose a Common Offset Survey, which can map reflections from the subsurface. For the seismic techniques, we choose a line array measurement, among others. We use the GPR to estimate the buried rock contour of the keystone Sl2 of megalith D14, which is a bearing stone formerly supporting capstone D9. We perform several reflection tests on various rocks unrelated to D14 using different seismic sources and receivers to estimate the reflection depths. We follow a proposed approach for both methods.

To evaluate the GPR data from the field, we assume a simplified GPR with zero-dimensional antennas (GPR point model). Subsequently, we develop two mathematical models (GPR point-to-GEO and GEO-to-GPR point model), based on this conceptual model in order to I) calculate the (buried) rock surfaces from field data and II) model field data from estimated buried rock contours.

We first perform the Common Offset Survey on a non-buried boulder on the campus of the TU Delft to evaluate the accuracy of the developed GPR point-to-GEO model and to optimise the second survey on keystone Sl2. We first perform the seismic reflection measurements on several rock samples to determine the best seismic source. Finally, we perform a line array measurement on a cylindrical basalt column using 300 kHz transducers.

We calculate rock contour coordinates from the GPR data and these show a reasonable fit with the contour of the TU Delft boulder, with an accuracy of 5-10 cm. For the keystone Sl2, the maximum burial depth is determined to be 80 cm at the southern side. The bottom of the keystone is sloping downward starting from ground level at the northern side. The southern, eastern and western rock faces are steep, almost vertical, which is confirmed by historic photographs. However, the calculated (buried) rock surface coordinates consist of an incoherent set of coordinates with locally a lack of data or blind-spots. Estimating a coherent buried rock contour, therefore, requires shortcuts and a decrease of accuracy is to be expected especially for rock surfaces near blind-spots in the GPR data. Furthermore, the identification of relevant reflection surfaces is rather subjective and combined with blind spots in the acquired GPR data, this can lead to wrongful interpretations of the buried rock contour.

The seismic reflection measurements we perform give clear reflections for the 300 kHz transducers on rocks of limited size with simple geometries. However, the transducers should first be applied on rocks with increasingly more complex geometries before being applied in the field. The accuracy in the order of 1 cm can be considered promising, but its applicability for complex geometries and reflection depths larger than 0.5 m remains unknown.