A proper understanding of subsurface gas migration patterns resulting from belowground natural gas (NG) infrastructure is essential for safe and efficient first responder operations and to improve leak repair actions. Subsurface migration of leaked gas occurs predominately by pressure-driven advection close to the leak. Far migration occurs by concentration-driven diffusion and eventually defines the plume’s edge. Therefore, proper formulation of the diffusion behavior in mathematical models of gas transport is critical to safety and response. This study evaluated six diffusivity parametric functions (DPFs) for an accurate representation of soil gas diffusivity through soils by comparing predicted and measured values for mineral, clay, and sand soils. DPFs were further tested by implementing them in a multiphase flow model previously developed for belowground NG transport. DPFs were also tested by comparing with field scale NG leakage experiments conducted in varying soil conditions (mineral soil, sand, and sand/clay) at controlled NG leak rates ranging from 5 to 21 slpm. Results showed that soil-type and structure-dependent DPFs performed better than soil-type and structure-independent DPFs in simulating the diffusive transport of NG. Further, results showed that DPF selection does not significantly influence NG migration simulations under dry soil conditions during the leak period. However, when the leak is terminated, simulation results followed the same trend of over- and under-predictions as of Dp/Do parameterization. Further, wet soil simulations show more than a 10% CH4 composition deviation and 2.5-day time differences among DPFs in reaching explosive limits. Unique to this work, the selection of a DPF can result in large differences in flux estimates in transport models, masking the effects of the gas behavior, especially in cases where gas diffusion to outer boundaries is of interest. The findings, therefore, support the idea that careful selection of DPFs based on soil-type, structure and moisture condition is essential to predict NG plume behavior during a leak event and after repair.