How Good is Your Fracture Model? Evaluating Human Biases and Uncertainty in Geoscientific Interpretations

Applied geoscience relies on robust structural models that are appropriately scaled and detailed to address specific challenges. However, the process of developing these models is influenced by human biases shaped by personal and professional experiences, area of expertise, cognition, and values. This diversity of approaches to geoscience interpretation, such as fracture characterisation, impacts the reliability of structural models, which are critical for geoenergy, resource and infrastructure applications. Ensuring robust interpretations is vital for improving safety, enhancing decision-making, and securing project success.

This study investigates the variability in fracture interpretations made by geoscientists analysing an aerial drone image of a fractured outcrop. We compare outputs such as fracture frequency, orientation & density, and network topology across participants to assess the variability in their observations and the uncertainty this develops.

Previous studies on 3D seismic data have shown that geoscientists’ experience and approach significantly impact structural models. Our research systematically assesses similar variability using remotely sensed outcrop data and shows that while our cohorts of geoscientists agree of the “big stuff”, there is less consensus when we examine the detail.

By illustrating these uncertainties, we can begin to inform improved interpretation workflows, team arrangements, assurance processes, and geoscience education and communication. Understanding biases in fracture interpretation is a critical step towards enhancing interpretational accuracy. Coupled with a clear idea of how “good” the structural model needs to be for the problem being solved and appropriate mitigation measures, (if necessary) this ensures better project outcomes and supports the development of reliable geoscience outputs across applications.