1A ground penetrating radar prospection campaign completed in 2019 at the eastern end of the Roman small town of Bridiers in La Souterraine (Creuse, France), in a context of granite soil, revealed the presence of a previously unknown residential area covering roughly 0.8 ha (Baret, 2019) (Fig. 1). The prospection was carried out and interpreted by AnalyseGC (G. Caraire and T. Jubeau). The choice of a 200 MHz GPR StreamX, from IDS, for the survey was determined by its “multi-channel” configuration, which allows the simultaneous recording of seven radargrams covering nearly one metre in width (one every 12 cm). The GPR data were positioned using an RTK (Real Time Kinematic) RS+ GPS from Emlid. The RTK base was readjusted using RTKLIB software and data from the IGN permanent GNSS network.
Figure 1. Global results of the GPR survey.
2The ancient Roman urban district is planned around a main east–west street and two possible perpendicular streets dividing the district into three densely built blocks north of the main axis. About ten masonry complexes can be individualised in the GPR images. In 2020, archaeological excavation at the western end of the parcel opened up an area of 180 m² located on the westernmost building revealed in the GPR results (Baret, 2020) (Fig. 2). One of the objectives of this first excavation was to confirm the GPR results in this part of the area in order to validate the proposed interpretations and correct the reading of the other parts.
Figure 2. Western building detail (interpretation by AnalyseGC).
3The overall agreement between the two sets of data was satisfactory, but a more detailed analysis revealed some differences that are interesting to point out (Fig. 3). For one thing, the excavated depth of the features appeared greater than the estimation proposed by the geophysicists (AnalysisGC company) from their calculation model (the speed of propagation of the waves into the ground and the round trip time of the wave). Moreover, large granite blocks (doorsills, pipe sections) failed to appear in the GPR images (slice or B-scan: 77 images spaced 5 cm deep and 14 “summed” images every 20 cm) (Fig. 4). In the case of the doorsills, they are missing from the images for the depth of 0.45 m, which is the level at which they were found in the excavation (the tops of walls were excavated at a depth of 0.3 m), thus appearing as a break in the continuous anomaly corresponding to the wall; the masonry below and extending into the wall foundation reappears on the slices of levels deeper than the thresholds (starting at 0.60 m depth). Moreover, a well with a casing of granite rubble, possibly unmortared, uncovered during the excavation, also did not appear in the GPR images.
Figure 3. Comparison of GPR and excavation results.
Figure 4. Successive slices at the location of the doorsills showing the no detection of the granite blocks.
4A number of hypotheses can be put forward to explain these phenomena:
-
detection of masonry mortar by contrast of the material (rich in lime) with the granite soil environment, which would leave the building elements, which are themselves of granite, undetected;
-
demonstrations by mathematical processes of anomalies corresponding to multi-reflectors (masonry walls with lime mortar and granite rubble stones) at the expense of large-sized blocks, which are homogeneous and therefore have only one reflection interface;
-
a very strong response from the masonry which during signal processing induces in horizontal images a weakening of anomalies with less intense signal, which subsequently “disappear” during mathematical processing.
5In conclusion, the poster aims to attract attention to an observation made possible by the succession of two different archaeological operations (GPR prospecting followed by excavation) on the same building, one year apart, allowing a precise comparison of the results obtained by the two methods. It is interesting that in the event of positive results of GPR surveys with very marked anomalies (an association of heterogeneous and homogeneous sets in the constructions), homogeneous elements of large dimensions tend to disappear in the data processing, which appears to highlight multi-reflective anomalies.
6The objective is not to oppose GPR to excavation but to highlight the differences when GPR results appear less rich (Löcker et al., 2015). Indeed, the absence of visible anomalies does not necessarily mean the absence of remains and a dialogue between GPR and excavation is a must in such cases.