Som, S.M., Hagadorn, J.W., Thelen, W.A., Gillespie, A.R., Catling, D.C. & Buick, R.
Computers & Geosciences, 54, pp 231-238.
doi:10.1016/j.cageo.2012.11.019, 2013.
Brief summary
In this methods paper, we present a new way to measure geological objects (such as air pockets, mineral-filled air pockets, crystals, and fractures) trapped within geological materials (such as volcanic rocks, other volcanic debris, and construction cement). Such objects are typically of different densities compared to the material they are trapped inside of. This allows the use of X-rays to identify them (akin to dentists using X-ray to get a closer look at a patient’s teeth). Coupled with X-ray detection, we present new software-based techniques that we have developed to extract these objects from the raw X-ray images in order to obtain quantitative information about them, such as their volume distribution and associated statistics. This type of information is important because they can tell a story about what happened to those rocks.
Extended summary
The size and size distributions of geological objects trapped in geological materials can be useful to scientists and civil engineers because they can reveal the history of the rocks, provide constraints on the environmental conditions that prevailed while they were formed, and reveal physical processes that have acted on them since formation.
Measuring such trapped objects is challenging because physically accessing them can cause their destruction as rocks are very brittle. While different techniques do exist, X-ray tomography has proved to be the most effective method, particularly when there is a strong contrast between the material and the object (for example: a fracture in cement, or an air bubble in a volcanic rock). The use of X-rays to measure objects with variable densities (such as voids filled in with several minerals, each with a different density) has not been significantly studied.
The result of X-ray scans are gray scale images, where “black” represents very low density and “white” represents very high density. Voids and fractures appear black, while rock and cement are gray. When working with voids, it is easy to tell a software to identify “black” as the object of interest while ignoring the rest. This method becomes impossible to apply when the objects of interest are also gray (but of varying different shades of gray from the parent material). To enable such studies, we have developed new software-based techniques to isolate such variable density objects from the parent material. The method relies on the grayscale gradient, rather than its absolute value. To illustrate the method, we have implemented it as a proof-of-concept on 2.7 billion year old lava flows, where the air bubbles have been filled-in by subsequent mineralization.
We also present mathematical techniques to obtain the mean size of those objects in a way that allows statements to be made about the statistics of the entire rock unit, rather than the sample alone. This involves bootstrap mathematics and application of the Central Limit Theorem.


