The microstructures of heterogeneous materials can strongly influence a material’s mechanical and physical properties. The nature of this influence can be primarily determined by the spatial distribution of defects within the structure. To provide model validation and explore a breadth of microstructural features, we pursue synthetic reconstructions of materials that match the spatial statistical features of experimental microstructures and account for variability in such structures. The spatial distribution of features in heterogeneous materials such as porous media or composite materials is assumed stationary in most literature, an often inappropriate assumption, and although high resolution images are available, they are often too small to be representative of the entire material. We implement dimension expansion (Bornn et al, 2011; Shand and Li, 2017), a modern nonstationary spatial modeling approach, to rigorously characterize the spatial distribution of microstructure defects and enable straightforward reconstruction. We successfully use this modeling approach to reconstruct and simulate new synthetic microstructures across multiple densities of a pressed pellet material.