Sub-metric Resolution FWI of Ultra-High-Frequency Marine Reflection Seismograms. A Remote Sensing Tool for the Characterisation of Shallow Marine Geohazard

A quantitative high-resolution physical model of the top 100 meters of the sub-seabed is of key importance for a wide range of shallow geohazard scenarios: identification of potential shallow landsliding, monitoring of gas storage sites, and assessment of offshore structures stability. Cur- rently, engineering-scale sediment characterisation relies heavily on direct sampling of the seabed and in-situ measurements. Such an approach is expensive and time-consuming, as well as liable to alter the sediment properties during the coring process. As opposed to reservoir-scale seismic exploration, ultra-high-frequency (UHF, 0.2-4.0 kHz) multi-channel marine reflection seismic data are most often limited to a to semi-quantitative interpretation of the reflection amplitudes and facies geometries, leaving largely unexploited its intrinsic value as a remote characterisation tool. In this work, we develop a seismic inversion methodology to obtain a robust sub-metric resolution elastic model from limited-offset, limited-bandwidth UHF seismic reflection data, with minimal pre-processing and limited a priori information. The Full Waveform Inversion is implemented as a stochastic optimiser based upon a Genetic Algorithm, modified in order to improve the robustness against inaccurate starting model populations. Multiple independent runs are used to create a robust posterior model distribution and quantify the uncertainties on the solution. The methodology has been applied to complex synthetic examples and to real datasets acquired in areas prone to shallow landsliding. The inverted elastic models show a satisfactory match with the ground-truths and a good sensitivity to relevant variations in the sediment texture and saturation state. We apply the methodology to a range of synthetic consolidating slopes under different loading conditions and sediment properties distributions. Our work demonstrates that the seismic inversion of UHF data has the potential to become an important practical tool for marine ground model building in spatially heterogeneous areas, reducing the reliance on expensive and time-consuming coring campaigns.