Automated reflection picking and inversion: Application to ground and airborne GPR surveys

We apply an automated picking and inversion algorithm to ground-based and airborne glaciological GPR surveys, in order to recover the internal stratigraphy, density distribution, and water content of alpine glaciers. Current glacier monitoring techniques encompass topographic mapping, direct measurements, and GPR surveys. However, the resulting models strongly depend on the assumptions made about the glacier's internal EM velocity and density distributions, which are usually set either constant or slow-varying, with the only constraints given by locally sampled values. Our inversion procedure uses amplitudes and timespace positions of the recorded reflections to recover the EM velocity and thickness of each layer by reconstructing the travel path of each reflected wavelet. The internal density distribution of glaciers is then recovered using well-known empirical formulas. The input reflections are automatically picked using an algorithm designed to detect and track any recorded event characterized by lateral phase continuity. Such a procedure is mostly independent of the interpreter and only requires a few input parameters and thresholds. High data densities lead to accurate and statistically sound models, while 4-D GPR surveys allow monitoring of the temporal variations of a glacier and the estimation of its mass balance.