Knowledge of particle morphology is vital to understand the behaviour of natural geomaterials including heat transfer. The effects of particle shape on heat transfer have been mostly quantified with two-dimensional (2D) particle descriptors or at most with a single three-dimensional (3D) descriptor. However, these particle shape descriptors may fail to capture the shape of all irregular particles. To redress this issue, we developed a method to reconstruct particles from micro-computed tomographic (pa) images and to extract 3D sphericity and roundness of individual particles in the assembly. Sphericity and roundness of five real sand packings are calculated using the new proposed method. Furthermore, the effective thermal conductivity (ETC) of each sample is estimated using finite element modelling. Our results show that packings with higher sphericity or roundness tend to render higher ETC. A further examination of the microstructure in the assemblies indicates that sphericity or roundness corresponds to inter-particle contacts. (C) 2019 Elsevier B.V. All rights reserved.