Cost estimation is probably the most decisive factor in the process of computer-aided, preliminary planning for low-volume road networks. However, the cost of construction is normally assumed to be route-independent for a specific project area, resulting in sub-optimal layouts. This is especially true for mountainous terrain and in areas with unstable subsoil. Here, we present a model for more accurately estimating spatial variability in road life-cycle costs, based on terrain surface properties as well as geological properties of the subsoil. This parametric model incorporates four structural components: embankment, retaining structures, pavement, and drainage and stream-crossing structures. It is linked to a geo-database that allows users to derive location-specific parameter values as input. In applying this model, we have demonstrated that variability in costs ranges widely for mountainous areas, with the most expensive construction being approximately five times greater there than on more favorable sites. This variability strongly affects the optimal layout of a road network. First, when location-specific slope gradients are considered, costs are reduced by about 17% from those calculated via currently available engineering practices; when both slope gradient and geotechnical formations are included, those costs are decreased by about 20%. Second, the length of the road network is increased by about 4% and 10% respectively, compared with current practices.