Urs Boeniger, Jens Tronicke

• Decomposition of geophysical signals (e.g., seismic and ground-penetrating radar data) into the time-frequency domain can provide valuable information for advanced interpretation (e.g., tuning effects) and processing (e.g., inverse Q-filtering). The quality of these subsequent processing steps is strongly related to the resolution of the selected time-frequency representation (TFR). In this study, we introduce a high-resolution spectral decomposition approach representing an extension of the recently proposed Tree-Based Pursuit (TBP) method. TBP significantly reduces the computational cost compared to the well known Matching Pursuit (MP) technique by introducing a tree structure prior to the actual matching procedure. Following the original implementation of TBP, we additionally incorporate waveforms commonly used in geophysical data processing and present an alternative approach to take phase shifts into account. Application of the proposed method to synthetic data and comparison of the results with other typically used decompositionDecomposition of geophysical signals (e.g., seismic and ground-penetrating radar data) into the time-frequency domain can provide valuable information for advanced interpretation (e.g., tuning effects) and processing (e.g., inverse Q-filtering). The quality of these subsequent processing steps is strongly related to the resolution of the selected time-frequency representation (TFR). In this study, we introduce a high-resolution spectral decomposition approach representing an extension of the recently proposed Tree-Based Pursuit (TBP) method. TBP significantly reduces the computational cost compared to the well known Matching Pursuit (MP) technique by introducing a tree structure prior to the actual matching procedure. Following the original implementation of TBP, we additionally incorporate waveforms commonly used in geophysical data processing and present an alternative approach to take phase shifts into account. Application of the proposed method to synthetic data and comparison of the results with other typically used decomposition approaches, illustrate the ability of our approach to provide decomposition results highly localized in both time and frequency. Applying our procedure to field GPR data illustrates its applicability to real data and provides examples for potential applications such as analyzing thin-bed responses and modulating the data frequency content.…