Recent developments for GPR full-waveform inversion of experimental data

To improve the characterization of the near surface and the critical zone, high resolution geophysical methods are essential to enhance our understanding of small-scale, shallow subsurface processes. Ground Penetrating Radar (GPR) is one such method that has significant advantages as it can be applied quickly and efficiently at different scales, and at the same time provide high resolution images through electromagnetic properties: dielectric permittivity (related to wave velocity); and electrical conductivity (related to the attenuation of the wave). Furthermore, the permittivity can be linked to porosity/soil water content, and the conductivity can be related to soil texture and salinity.

In the last decade, full-waveform inversion (FWI) has been adapted for vectorial GPR waves in the megahertz range and is able to resolve sub-wavelength high resolution images. We discuss recent developments of FWI for experimental GPR data, and show that we are able to obtain higher resolution quantitative medium properties compared to using conventional ray-based approaches. For a large number of unknowns, gradient-based optimization methods are commonly used and require a good starting model to prevent them from being trapped in local minima. For a limited number of unknown parameters, a combined global and local search using the shuffled complex evolution (SCE) can be used. An overview of the methodological developments will be given, and several applications will be discussed for GPR surface and crosshole applications. Recent advances in the capabilities and performance of forward modeling has enabled the use of improved 3D GPR models within FWI approaches. Over the last few years we have been working on the combination of a 3D Finite-Difference-Time-Domain (FDTD) forward modeling tool with surface and crosshole FWI techniques. In this way, future developments include the inversion of time-lapse GPR data, 3D FWI with integrated realistic antenna models, and gradient-based on-ground GPR FWI.