2-D ray-based tomography for velocity, layer shape, and attenuation from GPR data

Tomographic inversion for near-surface structure from multioffset ground-penetrating radar (GPR) reflection data is implemented. The model is parameterized as layers with constant velocity and constant attenuation. The solution is divided into two parts. First, the velocity and shape of each layer are estimated from the observed reflection traveltimes; then using this geometry and the corresponding ray paths, the attenuation of each layer is estimated from the observed relative amplitudes. Solution of both tomographic problems is performed by singular value decomposition of the sensitivity matrix (to compute information density, resolution, and covariance matrices as well as the solution). Tests on synthetic data, for which the solution is known, show that the algorithm converges efficiently to the neighborhood of the correct solution. Inversion of multichannel field data in a GPR line from a fluvial/eolian environment produces a model that is consistent with structural images and velocity estimates obtained independently by traditional common-midpoint processing.