Scaling of traffic safety indicators using microscopic traffic simulation

A validation and scaling study on a use case of the ring road of Tilburg

More Info


Microsimulation can substantiate traffic safety analysis. The questions is how to use the information from microsimulation to say something about the general safety at network level. The objective of this research is to find out how microsimulation can be used to quantify traffic safety at a network level. For this, the surrogate safety assessment module was used to identify the conflicts that occur on a large scale network consisting of a distributor road with 20 intersections. These conflicts are indicated by the Time to collision (TTC) and the post encroachment time (PET). Five years of historical crash data of this same network was compared to the conflicts from the SSAM both visually and statistically. The statistical comparison used the spearman rank correlation to check the correlation for different threshold values of bot the TTC and the PET. An AM-peak and PM-peak simulation scenario were run and compared to the crashes within the same time interval during the day. The AM-peak conflicts showed a reasonable fit with the AM-peak crashes with a spearman coefficient of 0.58 and a corresponding P-value of 0.008 for TTC and PET values up to 1.0. Which is significant within a 95% confidence interval. The PM-peak conflicts showed a bad correlation to the crashes on the network with a spearman rank coefficient of 0.275 and a corresponding P-value of 0.240. To see if there was a connection with other surrogate safety methods, a safety performance function (SPF) from the federal highway administration was used to predict the crashes on the network using the traffic volume. The use of this SPF did not lead to reliable results. The last stage of this research was the application of a newly proposed framework that consists of a linear model. Within this linear model the number of conflicts for each of the conflict indicators of the SSAM was compared to the number of crashes for each intersection at the network. By solving the linear model with a least square estimate the influence of each individual conflict indicator was estimated for different number of intersections. From these estimations, it could be seen that the TTC has the best predictive value out of the conflict indicators present in the SSAM program.