Potential and pitfalls of GPR traveltime tomography for ancient buildings investigation: experiments on a small-scalem real and synthetic calcarenitic block

Even if its use is not widespread in the archaeological community, GPR tomography is a viable tool in the maintenance of Cultural Heritage and for the diagnosis of internal defects in masonry, originating either at the building stage or later because of normal decay or natural disasters. Two-dimensional GPR traveltime tomography aims to obtain information on the distribution of the dielectric constant on a section of the investigated medium from the picked direct arrival traveltimes between sources and receivers. This paper shows the results of a GPR tomographic experiment on a calcarenitic stone block with an empty central hole, using 1000 MHz as transmitter and 1800 MHz as receiver antennas. The original aims of this work were firstly, to assess the usefulness and limits of very basic tomographic tools, accessible also to the non academic community, in the limited case of locating voids in small-scale structures (pillars or columns) and secondly, to identify possible pitfalls due to acquisition/processing procedures or to inadequacy of the inversion algorithm. We examine some problems encountered in data acquisition and we propose a method to estimate the effective bandwidth of antenna pairs of different nominal frequencies and to estimate the zero time correction. The experiment shows that picking the first arrivals is a very delicate operation when the airwaves interfere with the transmitted ones and that using the wrong picked traveltimes in the inversion could lead to inconsistencies or to strong reconstruction artefacts. Finite-difference numerical modelling is helpful both for identifying the correct arrivals to be picked and for exploring the dependence of the tomographic inversion on cell size, geometry of transmitters and receivers and initial model. The inversion results show a strong dependence on the angular ray coverage. The general improvement observed by increasing the illumination directions confirms the opportunity of using, whenever all sides are accessible, as in the case of columns or pillars, both parallel and orthogonal antenna positioning. In the presence of strong velocity gradients, as in this case, even using the best acquisition configuration for transmitters and receivers, the straight-ray tomography based on the SIRT algorithm can only detect the anomaly but is unable to resolve adequately its geometry and dielectric constant. Although more time-demanding, curved-ray tomography or more sophisticated algorithms are therefore necessary for a better characterization of internal defects in most problems of structural assessment.