Fusion of Electrical Resistivity Tomography (ERT) and Resistivity Cone Penetrometry (RCPT) Data for Improved Hydrogeophysical Characterisation

Electrical resistance tomography (ERT) is potentailly an appropriate subsurface imaging tool for hydrogeophysical characterisation, due to strong correlations between resistivity and hydrological parameters such as clay content and permeability. However, the ability of ERT to locate accurately sharp boundaries, such as permeability contrasts and wetting fronts, is rather poor because of the relatively large measurement support volume of the technique. This poor ability to locate interfaces also creates problems for ERT applications in geotechnical engineering, where practitioners need accurate determination of lithological interface depths. In contrast, the technique of resistivity cone-penetrometry (RCPT) provides resistivity data with a high vertical resolution, allowing interfaces to be located accurately. However, RCPT bores are usually widely spaced, so horizontal resolution is poor. Hence, combination of RCPT and ERT has significant advantages. Here, we investigate fusion of ERT and RCPT data for hydrogeophysical characterisation. The ability of RCPT data to guide ERT inversion towards improved solutions is investigated using both synthetic models and field data. A series of synthetic RCPT and Wenner ERT data for a sand body within a clay background was generated. The ERT data were contaminated with 2% and 5% Gaussian noise. A series of reference models for ERT inversion was generated from the synthetic RCPT data. The ERT data were inverted with and without these RCPT-derived reference models. The models produced by inversion (the final models) with the best fit to the original synthetic model were identified. The final models were then ranked using a weighted sum of (i) the least-squares misfit of the original synthetic data to the data produced via forward modelling from the final model, (ii) the least- squares misfit between the final model and the reference model, and (iii) the final model smoothness. This ranking technique identified the same best final models as identified using the best fit to the synthetic model. This result indicates that the reference model approach described here can safely be used with field data, i.e. for an unknown 'true' resistivity distribution. Using RCPT data as a constraint significantly improved interfacial depth accuracy and horizontal boundary location in the final geoelectrical model. RCPT and ERT data were collected from a coastal site near Withernsea, East Yorkshire, UK, where fluvioglacial sand lenses exist within clay tills. The ERT data were inverted with and without RCPT-derived reference models to produce range of geoelectrical models. These were quantitatively assessed using the procedures identified during the modelling stage. The models were then compared to a logged cliff section. We discuss the implications for the design of combined ERT and RCPT investigations.