Multidimensional GPR array processing using Kirchhoff migration

Abstract

We compare the ability of several practical ground-penetrating radar (GPR) array processing methods to improve signal-to-noise ratio (SNR), increase depth of signal penetration, and suppress out-of-plane arrivals for data with SNR of roughly 1. The methods include two-dimensional (2-D) monostatic, three-dimensional (3-D) monostatic, and 3-D bistatic Kirchhoff migration. The migration algorithm is modified to include the radiation pattern for interfacial dipoles. Results are discussed for synthetic and field data. The synthetic data model includes spatially coherent noise sources that yield nonstationary signal statistics like those observed in high noise GPR settings. Array results from the model data clearly indicate that resolution and noise suppression performance increases as array dimensionality increases. Using 50-MHz array data collected on a temperate glacier (Gulkana Glacier, AK), we compare 2-D and 3-D monostatic migration results. The data have low SNR and contain reflections from a complex, steeply dipping bed. We demonstrate that the glacier bed can only be accurately localized with the 3-D array. In addition, we show that the 3-D array increases SNR (relative to a 2-D array) by a factor of three.