Effect of antennas on velocity estimates obtained from crosshole GPR data

To obtain tomographic images with the highest possible resolution from crosshole ground-penetrating radar (GPR) data, raypaths covering a wide range of angles between the boreholes are required. In practice, however, the inclusion of high-angle ray data in crosshole GPR inversions often leads to tomograms so dominated by inversion artifacts that they contain little reliable subsurface information. Here, we investigate the problems that arise from the standard assumption that all first-arriving energy travels directly between the centers of the antennas. Through numerical modeling, we show that this assumption is often incorrect at high transmitter-receiver angles and can lead to significant errors in tomographic velocity estimates when the antenna length is a significant fraction of the borehole spacing.