Assessing Traffic Safety of Dutch Weaving Sections

A Validation of the Surrogate Safety Assessment Model combined with VISSIM

More Info


Weaving sections are frequently applied on Dutch freeways. Guidelines are present for designing those freeway weaving sections. However, the origin of these guidelines is unclear and it is not always possible to apply the guidelines. Dutch road designers and road safety experts are searching for methods to evaluate the safety of a (proposed) design more quantitatively than is currently done by expert judgement. An alternative option would be to determine safety of Dutch weaving sections using VISSIM micro-simulation models in combination with the Surrogate Safety Assessment Model (SSAM). SSAM is able to calculate the number of conflicts that occurred in a micro-simulation model using the surrogate safety measures time-to-collision (TTC) and post-encroachment-time (PET). A conflict is an observable situation in which two or more road users approach each-other such that there is a risk of collision if their movement remains unchanged. The goal of this master thesis research is to assess whether this number of conflicts observed from VISSIM microsimulation models using SSAM is representing safety (crash rate) of Dutch weaving sections. For that nine Dutch weaving sections were selected, and were ranked based on four criteria: (I) the crash rate from the BRON crash database, (II) the conflict rate (conflicts per number of vehicle kilometres) calculated from VISSIM micro-simulations using SSAM, (III) the expected number of crashes calculated using the crash prediction model developed by Iliadi et al. [29], and (IV) the judgement of road safety experts. The Spearman Rank Correlation Coefficient was calculated between these rankings, to validate the ranking based on the conflict rate. For initial VISSIM and SSAM settings, a correlation of 0.567 was found between the crash rate and conflict rate ranking, indicating a reasonable fit. However, due to the small sample size this correlation is not significant (𝑃-value is 0.112). The effects of some micro-simulation settings, conflict analysis thresholds and the calibration method are assessed in a sensitivity analysis. Reducing the TTC threshold from 1.5 seconds to 0.5 seconds resulted improved the correlation between the conflict rate and the crash rate. However, changing other settings resulted in a weaker correlation. Although different than expected, extending the calibration process by adding calibration on speeds resulted in a weaker correlation. Hence care should be taken when using conflicts calculated by VISSIM and SSAM as (only) predictor for safety of Dutch weaving sections. It is recommended to expand the selection of weaving sections to assess whether a larger sample results in a stronger, significant, correlation. Also, it can be assessed how a more extensive calibration of VISSIM parameters affects the correlation between the conflict rate and other measures. Furthermore, crash numbers might have more impact on policy makers and public perceptions than alternatives such as conflicts and surrogate safety measures.