The presented study compares three alternative straight-line approximations of the longitudinal road-profile power spectral density (PSD). The one straight line and two and three split straight-line approximations were analyzed. The minimizing of the root-meansquared error (RMSE) was used as criteria of goodness of fit for the approximation models. The statistics of the PSD approximation model parameters, i.e., unevenness index, waviness, break angular spatial frequency, and the standard error of the regression for 145 test sections from the long-term pavement performance program (SPS-6) were calculated. The approximations were compared with connected versus unconnected lines. The approximation with two connected lines seems to be a reasonable substitution of a raw spectrum. A mean improvement of approximately 35% compared to the one straight-line approximation was identified.