This paper proposes a methodology to deduce the spatial variation of the elastic characteristicsof a two-dimensional earth model from seismic data, given the media’s response to interrogating SH waves. Areduced dimension, self regularized treatment of the inverse problem using partition modeling is introduced,where the SH wave velocity field is discretized by Voronoi tessellations, and the number and geometry of thesetessellations dynamically alter during the inversion to adapt the form of geologic units and geomorphologicalfeatures (e.g. transitions between soil layers, faults, concentration of materials, etc.). The subsurface materialcharacteristics is treated as a random field where the measure of uncertainty associated with the deduced subsurfaceimage is casted into a form of a probability density function at each point in space. This allows for a)propagating full probabilistic descriptions of material properties obtained from geophysical data to full probabilisticdescriptions of geotechnical properties; b) borrowing a stratigraphic earth random model to populate anyother mechanical property, after the probabilistic identification of the location of boundaries between materialsis obtained (i.e. spatial probabilistic definition of the geomorphological features); and c) making available theprobabilistic identification of geomorphological features to be merged with other geomorphological features,retrieved from probabilistic inversions related to different geophysical technologies. Consequently, it is anticipatedthat the resulting probabilistic descriptions of the earth model will contribute to improve the confidencein the identification of subsurface hazards such as over-pressured water or gas hydrate occurrences, to improvethe decision-making on the definition of the scope of the geotechnical surveying, and to improve the reliabilityassessment of geotechnical structures, among others, including the improvement on the risk assessment ofoffshore drilling and field structural developments.
展开▼
