Adequate representation of underground structures is necessary for modeling and simulation (M&S) applications to provide quality training for both military and civilian use cases. Urban and asymmetric warfare is becoming more predominant, and training applications require technological advances in order to replicate the complexities of these urban environments. Underground roadways, basements, subways, sewer systems, and even subsurface multilevel buildings or garages are examples of the underground elements that play a role in urban and asymmetric warfare. Underground structures are important outside of urban environments as well. Naturally occurring underground features, such as cave systems, present special case challenges to existing synthetic environment algorithms that are tailored around open environments or structured building interiors. Given the prominence of aerial bombing to prepare for a ground invasion, complex military infrastructure will increasingly be found underground in bunkers or cave systems. In urban environments, infrastructure such as power, water, internet, and sewage, are already primarily underground, and require modeling to realistically replicate the urban battlespace. Simulations that include underground environments need to consider environmental effects, infrastructure interruptions, propagation of smoke and sound, fire propagation, and ventilation. This paper discusses an approach for representing and dynamically manipulating complex underground environments, both natural and man-made. The approach uses voxelization techniques found in modern game engines, movies, and medical imaging systems. The voxelized data for soil and structures is subdivided into chunks and organized into variations of B+ trees for efficient access and storage. Smooth-mesh algorithms analyze the voxel data and generate meshes for rendering and navigation mesh generation. We describe and propose an architecture for providing underground capabilities to existing and future simulation systems in a distributed fashion. Within this architecture, we also define functional components that will be modularized in order to take advantage of cloud computing technologies.