Quantifying the uncertainties in a fault stability analysis of the Val d’Agri oilfield

Induced seismicity as an effect of injection of fluids in a geological basin is a widely observed phenomenon which is nowadays under public scrutiny. In most cases, the numerical simulation and interpretation of these phenomena requires models with a high degree of realism. In turn, this requires accounting for complex interactions between the fluid and solid components, as well as a detailed description of geometry and a high degree of heterogeneity in the material properties. Also, issues related to the uncertainty of parameters, to an incorrect modeling of sub-scale or multiscale effects and to a limited knowledge of initial conditions are unavoidable and can be detrimental to the reliability of the results. In this work, a statistical analysis including uncertainty quantification and sensitivity analysis on variances has been applied as a post-processing to data coming from a set of numerical simulations of a real world setting (the Val D’Agri oilfield), with the aim of studying the stability of a fault that is known to have experienced a good amount of microseismicity during the modeled period. The uncertainty quantification targeted the effects of fault surface local orientation and pore pressure fluctuations, as well as variability in the friction coefficient of the fault. The analysis of variances focused on the effects of varying the permeability of the fault damage zones (the area enveloping the faults) and the geometrical orientation of the fault as well. The model shows that the zone where the microseismicity has been measured is included in a wider region of moderate instability, which is higher the lower the permeability of the fault damage zone. From the model results the fault seems to be far from a critical state, but the analysis offers, nevertheless, some useful information on the relationship of slip tendency with geometrical and flow quantities in the system, and suggests some improvements in the dynamical model assumptions and settings.