ABSTRACT
Ground-penetrating radar (GPR) is an established geophysical method to explore near-surface sedimentary environments. Interpreting GPR images is largely based on manual procedures following concepts known as GPR facies analysis. We have developed a novel strategy to distinguish GPR facies in a largely automated and more objective manner. First, we calculate 13 textural attributes to quantify GPR reflection characteristics. Then, this database is reduced using principal component analysis. Finally, we image the dominating principal components using composite imaging and classify them using standard clustering methods. The potential of this workflow is evaluated using a 2D GPR field example collected across stratified glaciofluvial deposits. Our results demonstrate that the derived facies images are well correlated with the composition and the porosity of the sediments as known from independent borehole logs. Our analysis strategy eases and improves the interpretability of GPR data and will help in a variety of geologic and hydrological problems.