2D full-waveform inversion of multi-offset surface ground-penetrating radar data with 3D-to-2D transformation and borehole verification

  • Venue:

    Bldg. 06.42 - Room 001 (seminar room) / Online

  • Date:

    June 11, 2024


  • Speaker:

    Vu, Tuan Dung

    Master thesis presentation

  • Time:

    9:30 am


Ground-penetrating radar (GPR) full-waveform inversion (FWI) is an effective technique for high-resolution multi-parameter subsurface imaging. It can reconstruct dielectric permittivity and electrical conductivity models by iteratively minimizing the misfit between observed and modelled waveforms. Due to the 2D modelling cannot account for the spherical spreading of the electromagnetic waves in the 3D. Therefore, in order to implement 2D GPR-FWI for multi-offset surface GPR data, which is obtained in the field by point source, one needs a procedure to transform the observed data generated by a point source in the field to the one generated by an equivalent line source. This transformation, called the 3D-to-2D transformation, has been widely applied to seismic data, but it is quite new for multi-offset surface GPR data.

This study investigates the performance of the 3D-to-2D transformation of different waves for multi-offset surface GPR data, and its efficiency for 2D GPR-FWI. First, the effectiveness of the transformation is demonstrated by using synthetic data. After that, it is applied to multi-offset GPR field data acquired at the Rheinstetten test site in Germany. The results of 2D GPR-FWI for the field data show that the reconstructed permittivity and conductivity models are efficient in delineating the main geological feature, which is a refilled trench from the 18th century. This confirms the effectiveness of the 3D-to-2D transformation for multi-offset surface GPR data.

To evaluate the GPR-FWI results, the water content and grain size of the borehole soil samples were measured in the laboratory. I then calculated the relative dielectric permittivity from the water content using petrophysical relationships identified by the grain size. I found that the GPR-FWI result and the borehole data have a similar permittivity trend with respect to depth. This indicates that the estimated models from FWI are reliable.