Extracting the subsurface stratigraphy by means of a novel automated reflection strength tracking procedure: Theory and application to ground-penetrating radar data

We developed an algorithm designed to automatically detect and accurately track reflections within geophysical data sets, whether seismic or georadar, based on their reflection strength. Each signal trace is divided into a series of energy packets containing either reflected, diffracted, interfering, or noise-related signals. Acceptable pairs across nearby traces are then connected using the locally-estimated reflection dips, thus constructing a network of energy packets covering the entire data set. Starting from automatically determined initial seeds, the algorithm moves along the network, selecting at each node the path that most likely follows the reflection that is supposedly being tracked. After constructing all possible horizons, the procedure disregards the expected noise-related false positives, which randomly connect unrelated energy packets in areas lacking any recognizable structure. Among its advantages, the proposed algorithm can be applied with minimal signal preprocessing and it requires only limited input from the interpreter, thus leading to more objective results. Moreover, the procedure is able to track a reflection even when its lateral continuity is locally disrupted by either low signal-to-noise ratios, interfering events, or splitting horizons, which typically hinder auto-picking algorithms that are based on waveform-related attributes. The presented algorithm is tested on a georadar profile acquired on the Boulder Clay Glacier, located in Victoria Land, Antarctica, showing an overall good performance across the data set, which includes a well-stratified internal section with wide-ranging reflection dips, as well as more complex structures along the bedrock and the moraine. • The algorithm automatically detects and accurately tracks the recorded reflections. • Reflection dips are locally estimated from the gradient of the reflection strength. • The local dips create a web of interconnected energy packets covering the data set. • The web allows to tracks reflections whose lateral continuity is locally disrupted. • The procedure needs only limited input from the user and minimal signal processing.