ABSTRACT The operating speed-profile is an important tool for the safety evaluations of new and existing roads. In particular, the operating speed-profile is useful to evaluate the consistency of an alignment identifying the excessive speed differential and to calculate the correct sight distance necessary along the road. Several models have been developed to predict the operating speed on curves and on tangents of two-lane rural roads. However, to trace an operating speed-profile close to the real speed-profile, it is necessary to join these speeds on successive elements using deceleration and acceleration rates that represent the real rates experienced by drivers. Therefore, to develop a deceleration model that could be effectively used in the operating speed-profile models, driver speed reduction behaviour and its dependence on the geometric characteristics of the road were investigated on 18 horizontal curves of two-lane rural roads. The results show that a single constant deceleration rate does not effectively describe the observed speed reduction behaviour. Therefore, its use in a speed-profile model results in a wrong estimation of the real distance used to decelerate before a curve. In fact, the deceleration rate observed varies significantly between the various curves. The data analysis revealed that the deceleration rate is correlated to the radius, the curve length, the approaching speed and the speed reduction. Therefore, two deceleration models were developed. The first one estimates the deceleration rate as a function of the radius of the impending curve and the second one as a function of the approaching speed and of the speed reduction. These models can be effectively used for the construction of the operating speed-profile because they predict deceleration rates consistent with the speed reduction behaviour observed and, consequently, deceleration lengths closer to the real distance travelled during the speed reduction.
