Lane Change Rates at Freeway Weaving Sites: Trends in HCM6 and from NGSIM Trajectories

Concerns have been raised about the HCM6 weaving method's lack of sensitivity to weaving segment length. This study explores the trends in HCM6 as they relate to lane change estimates and their impact on the segment speed and level of service (LOS). The study also compares HCM6 estimates of lane changes against empirical data from an NGSIM weaving site. Thus, the objectives of this study are twofold: (a) critically investigate the effect of weaving length on lane change and associated speed model estimates in HCM6, and (b) analyze trends in lane changes against congestion levels using detailed NGSIM trajectory data, comparing against HCM6 estimates. For (a) it was found that the lack of sensitivity to weave length is because of the absence of this parameter in the nonweaving lane change and speed models. For (b), a comparison of HCM6 lane change rates with NGSIM, US-101 data confirmed that the HCM6 estimates for weaving vehicles are fully consistent with those at the NGSIM site, controlling for density. In contrast, nonweaving lane change estimates in HCM6 did not deliver the expected trends, with more discretionary lane changes predicted as congestion increased. Finally, analysis of lane change patterns at the NGSIM site revealed a tendency for early merging for freeway to ramp traffic and uniform merging for ramp to freeway traffic over the length of the weave. Interestingly, a speed analysis showed that in most cases, a higher frequency of discretionary lane changes yielded lower travel times for drivers executing them.