Extracting subsurface information from GPR reflection and transmission data

For several decades single-offset GPR data sets are being recorded to image and characterize the shallow subsurface from the surface and in boreholes. Over the last two decades the GPR community has moved from scattered publications on advocating the use of multi-offset GPR data for improved imaging and subsurface characterization to routinely recording multi-offset GPR data for specific applications where detailed subsurface information is required. When we try to infer structural information from the data at scales smaller than the dominant wavelength, it becomes crucial to consider uncertainties in the data. Uncertainties in the data related to antenna position and orientation lead to uncertainties in the derived velocity distribution, which in turn lead to image features that may be interpreted as geological features such as a rough fracture surface. When we try to infer specific physical properties, such as water content, porosity, clay content, or grain size distribution, uncertainties in the required petrophysical model add on to the acquisition related uncertainties. If we use full waveform inversion we necessarily need adequate models to account for antenna coupling effects to obtain quantitative subsurface electromagnetic properties. We investigate the accuracy requirements in position and orientation information as well as the effect of uncertainties in antenna coupling models on the accuracy of the information that can be extracted from GPR data. We find that antenna positions should be known to within one tenth of the dominant wavelength to construct high quality images that can be used for feature extraction. This also holds for antenna endpoint locations, which leads to requirements for orientation information accuracy. Different much used petrophysical models for water content determination show similar trends, but have different values. These differences can be accounted for when calibrated on a specific site.