Sampling the uncertainty associated with segmented normal fault interpretation using a stochastic downscaling method

• A method to subdivide a 3D fault into possible en-echelon segments is proposed.
• The method stochastically produces overlapping segments and connecting faults.
• Application to an 11 km long fault interpretation is performed.
• We analyze the emerging statistics about segment number, size, overlap and spacing.

A large-scale normal fault may be composed of several overlapping fault segments separated by relay zones at a finer scale. Fault segmentation may be critical to the understanding and the forecast of physical phenomena at the reservoir or basin scale (e.g. fluid flow, seismic rupture propagation). In this paper, we propose an automatic and stochastic method to sub-divide (downscale) below the data resolution a segmented normal fault into en-echelon segments that may be linked by connecting faults, based on variations in the orientation of the fault. The downscaling algorithm is composed of three main steps. (1) The first step involves detecting the segments using geometrical criteria. (2) Then the overlapping segments are modeled using isolated fault descriptions and statistics. (3) Lastly, the maturity of the simulated relay zones is evaluated based on relay geometry. If a given maturity threshold is reached, the relay ramp is breached. These three downscaling steps depend on seven parameters that can be defined as constant or randomly chosen from probability distributions to sample uncertainties. The method was applied to a large normal fault that laterally limits a hydrocarbon reservoir and is poorly imaged from seismic data. This resulted in one hundred possible 3D fault array models. The analysis of the emerging parameters shows a strong variability of the number and length of segments, and of the overlap and spacing between segments.