We evaluated the effectiveness of a distance sampling from roads program for estimating population sizes of white-tailed deer (Odocoileus virginianus) from 2001 to 2015 in parks of the National Capital Region (NCR), National Parks Service. Distance sampling is a method for estimating the density of organisms using a distribution of distances to observed individuals. Re-analysis of survey data for 9 of 11 NCR parks found that although the original park analyses likely estimated deer densities correctly, the uncertainties (coefficients of variation or CV) of the original estimates were likely underestimated. Power analyses based on the current analysis methods showed that survey effort at some parks was likely insufficient to reach the NCR target of a 20% CV. We simulated 7 different types of deer populations and 3 survey designs to assess how violations of the assumptions of distance sampling might have impacted population estimates. A significant interaction between survey type and population type explained most of the variation in population estimates across simulations. Simulation results suggested that (1) non-road surveys were more robust to bias in seven deer population distributions than were road surveys, (2) effectiveness of each of 3 survey types was dependent on the way deer were distributed across the landscape, and (3) non-road surveys produced unbiased estimates of populations affected by roads, whereas, road surveys did not. Based on this study, we recommend revisions of the NCR distance sampling program, including additional sampling effort for some parks and suggest alternative survey strategies to ameliorate potential assumption violations of distance sampling.