Application of Subset Full-Waveform Inversion to Multioffset Surface Ground-Penetrating Radar Data

Tan Qin,Thomas Bohlen,Niklas Allroggen,Anja Klotzsche,Yudi Pan

DOI: https://doi.org/10.1109/tgrs.2024.3388480

IF: 8.2

2024-04-30

IEEE Transactions on Geoscience and Remote Sensing

Abstract:Ground-penetrating radar (GPR) full-waveform inversion (FWI) can determine high-resolution electromagnetic (EM) properties of the subsurface and has gained increasing attention in near-surface geophysics. However, the application of GPR-FWI to field surface data is limited due to the high computational costs. Here we apply for one of the first times a 2-D time-domain gradient-based FWI to synthetic and field multioffset surface GPR dataset. We use subset FWI (SFWI) to reduce computational costs by implementing the simulation on a model subset rather than the whole model. Thus we obtain theoretical speedup and memory saving factor equal to the size ratio of the model to its subset. The properties of the model subset depend on the illumination of the chosen acquisition geometry, based on which we provide rules of thumb for selecting the model subset. Our study reveals that SFWI can be further improved by parallelizations, where source parallelization allows for higher efficiency than model domain parallelization. Both 2-D synthetic data and field data validate that SFWI provides results comparable to FWI but without the redundant computations present in FWI. In a field example with surface geometry, SFWI provides a speedup of more than six times for a 45-m survey line. In another field example with cross-hole geometry, SFWI achieves more than a three times acceleration for a 21-m multiborehole plane. Our study proves that SFWI has the potential for high-performance computation to solve large-scale GPR survey problems.

engineering, electrical & electronic,imaging science & photographic technology,remote sensing,geochemistry & geophysics