This paper analyzes the impact of road grade variability on the identifiability of vehicle chassis parameters. The paper is motivated by the need for accurate parameter estimation in both active safety and chassis model validation applications. The literature addresses this need through several algorithms capable of both estimating vehicle chassis parameters from experimental data and assessing the accuracy of these estimates. However, the dependence of these algorithms' estimation accuracy on road grade variability remains relatively unexplored. We address this research problem in the specific context of utilizing nonlinear least squares methods for estimating the parameters of a longitudinal vehicle dynamics model. Specifically, we use Fisher information analysis to obtain Cramér-Rao bounds on the accuracy of the least squares parameter estimates. We then examine the impact of mean square road grade variability on these error bounds, both in simulation and experimentally. This examination is performed for a heavy-duty commercial truck, which is instrumented and driven on a variety of rural and interstate roads for the purpose of gathering experimental data. The impact of terrain variability on parameter identifiability is shown to be quite significant, both in simulation studies and experiments.