Full-waveform inversion of GPR data acquired between boreholes in Rustrel carbonates

Abstract

Abstract. Full waveform inversion (FWI) of seismic or Ground PenetratingRadar data provides high-resolution quantitative images of the constitutive parameters of the rock/soil which control seismic/GPR wave propagation. We developed a 2D inversion tool in the frequency domain adapted to the multi-parameter physics controlling GPR propagation in isotropic non dispersive media, i.e. dielectric permittivity and electrical conductivity. This inversion engine was previously tested using synthetic 2D data to mitigate the trade-off between the two parameter classes. In this paper, we present the required processing techniques and first inversion results obtained on a real GPR dataset acquired in carbonates with a cross-hole configuration. The presence of the 2 m diameter underground gallery at depth constitutes a nice target to test the robustness, efficiency and resolution of the inversion in such high-contrasts context. Starting from a time tomographic image for the dielectric permittivity and from a homogeneous conductivity, we show that FWI is efficient to retrieve high resolution images of dielectric permittivity but struggles with electrical conductivity. As a quality control, we compare real and synthetic radargrams computed from the tomography and final images, showing the efficiency of the process to reconstruct some events but also underlying some issues, particularly on large incidence angles amplitude traces.

Matrix‐free crosshole elliptical‐anisotropy tomography: parametrization analysis and ground‐penetrating radar applications in carbonates

“…Based on several tests, we found that values of d a = 28 m and d b = 33 m were efficient, while in our case, the maximum sourcereceiver offset is slightly larger than 40 m. Figure 14(b) shows the isotropic solution without smoothing, while Fig. 14(c) displays an isotropic velocity model obtained from fullwaveform inversion (FWI) inversion using the same dataset by Pinard et al (2016). While tomography results are smoother, Figure 12 Distribution of the 6741 picked first-arrival traveltimes as a function of source-receiver elevations.…”

“…Our synthetic model has been constructed considering isotropic full‐waveform‐inversion results obtained from the same real crosshole ground‐penetrating radar (GPR) data as those used hereafter (Pinard et al ., 2016). We consequently designed a target model for vertical velocity (Fig.…”

“…14(c) displays an isotropic velocity model obtained from full‐waveform inversion (FWI) inversion using the same dataset by Pinard et al . (2016). While tomography results are smoother, the introduction of regularization and of prior model information reduces artefacts around the gallery and increases the quality of the gallery reconstruction.…”

“…(c) Isotropic velocity model derived from full‐waveform inversion, courtesy of Pinard et al . (2016).…”

“…Smoothed (a) and raw (b) velocity models derived after 10 iterations of isotropic inversion using an homogeneous initial velocity of 80 m/µs. (c) Isotropic velocity model derived from full-waveform inversion, courtesy ofPinard et al (2016).…”