Calculation of accident dose-risk estimates with the RADTRAN code requires input data describing the population likely to be affected by the plume of radioactive material (RAM) released in a hypothetical transportation accident. In the existing model, population densities within 1/2 mile (0.8 km) of the route centerline are tabulated in three ranges (Rural, Suburban, and Urban). These population densities may be of questionable validity since the plume in the RADTRAN analysis is assumed to extend out to 120 km from the hypothetical accident site. We present a GIS-based population model which accounts for the actual distribution of population under a potential plume, and compare accident-risk estimates based on the resulting population densities with those based on the existing model. Results for individual points along a route differ greatly, but the cumulative accident risks for a sample route of a few hundred kilometers are found to be comparable, if not identical. We conclude, therefore, that for estimation of aggregate accident risks over typical routes of several hundred kilometers, the existing, simpler RADTRAN model is sufficiently detailed and accurate.