InterPore2025

From Fracture Roughness to Fault Conductivity: A Probabilistic Framework for Multi-Scale Uncertainty in CO2 Leakage Risk Assessment

19 May 2025, 09:55

1h 30m

Poster Presentation (MS14) Uncertainty Quantification in Porous Media Poster

Speaker

Dr Sarah Perez (Heriot-Watt University)

Description

Quantifying uncertainties associated with fault-related leakage is a significant challenge to ensure safe and efficient storage of $C{O}_2$ in the subsurface. The complexity of this problem arises from structural uncertainties in fault damage zones across various scales, which remain poorly resolved. These uncertainties range from seismically sub-resolved fractures to fine-scale structural and material features, such as fracture wall roughness, that are typically neglected in large-scale modelling and can lead to significant model misspecifications. To reliably quantify potential $C{O}_2$ leakage rates, it is thus crucial to understand how uncertainties on both the data and models propagate across the scales and influence larger-scale predictions of fracture flow and fault conductivity.

This study directly addresses these challenges by proposing a robust uncertainty quantification (UQ) framework for estimating fracture hydraulic conductivity. The approach is grounded on a probabilistic model that employs Bayesian inference, combining data-driven approaches with physics-based correction of the previous model misspecifications. The framework dynamically adjusts to different sources and magnitudes of uncertainty through adaptive, task-specific weighting, enabling to mitigate modelling and data uncertainties, and improve their predictive capabilities.

A critical insight of our work is that traditional reliance on mechanical aperture measurements alone is insufficient for accurately characterizing flow behaviour in rough fractures. Fracture wall roughness introduces geometric effects that significantly alter hydraulic conductivity and lead to over-estimations of the permeability either through the empirical Cubic Law or Darcy flow-based upscaling. To address this, we propose an automatic geometrical correction to infer latent hydraulic aperture fields that account for these roughness effects. This correction enhances the robustness of the modelled hydraulic properties while simultaneously quantifying their associated uncertainties.

Our approach produces detailed hydraulic aperture and permeability maps, with their associated uncertainties, which are crucial for robustly characterizing fracture conductivity derived from the mechanical aperture observations. Moreover, these mappings are designed to be upscaled to larger and more complex fracture networks. This multi-scale capability ensures that the influence of smaller-scale uncertainties, such as those stemming from roughness and sub-resolution features, remains integrated into larger-scale models, preserving essential information for risk assessment and decision-making.

Overall, this work highlights the importance of addressing uncertainty propagation across scales in fault-related leakage problems. By integrating advanced UQ methodologies, it provides a pathway for improving predictions of $C{O}_2$ leakage risks, thereby supporting safer and more efficient carbon storage solutions.