A Model for Estimating Free-Flow Speed on Brazilian Expressways

Abstract

Free flow speed (FFS) is defined in the HCM as the “average speed of vehicles on a given segment, measured under low-volume conditions, when drivers are free to travel at their desired speed and are not constrained by the presence of other vehicles or downstream traffic control devices”. FFS is a very important parameter for the estimation of LOS and capacity for uninterrupted highways facilities. Ideally, FFS is determined using field measurements, but the HCM 2010 provides models for its estimation, when field data is not available. Such models must be recalibrated for local conditions when the HCM 2010 is adapted for use in countries outside North America. In this paper, Multivariate Analysis (MVA) techniques are used to investigate which infrastructure characteristics are related to FFS on Brazilian expressway facility segments, in order to provide a framework for the development of models to replace those used in HCM 2010 to estimate FFS. This study used a database with a large number of speed-flow observations obtained at 36 sites on Brazilian expressways (freeways and divided multilane highways), along with information on seven variables describing segment characteristics such as: expressway type; roadside environment; number of lanes; rise and fall; bendiness; access point density; and posted speed limit for passenger cars. Principal Component Analysis (PCA) was used to select the variables to be included in the proposed model. The results of the PCA indicated that three components are responsible for most of the variance on observed FFS: (1) posted speed; (2) the combined effects of road class, number of lanes and density of access points; and (3) the interaction of geometric design and roadside environment. Stepwise linear regression was used to fit a FFS model from these three components. This model was mathematically manipulated to provide a FFS that is structurally similar to the one used by HCM 2010 but in fact accounts for the interactions among variables. The proposed model can replace the HCM 2010 model and does not require the user to choose an initial value for the base FFS.