An Impact Assessment of Intersection Design and Operational Elements on Red-light Running Canadian Associate of Road Safety Professionals Conference May 2019 Bob Henderson, CET, LEL Pedram Izadpanah, Ph.D., P.Eng
Background Angle collisions account for: ~10% of all collisions at traffic signals in the Region of Waterloo More severe and fatal injuries compared to other collision types
Study Objective Identify geometric and operational factors that impact Chargeable Incident An event whereby a red-light camera determines that a red-light violation occurred Average Intrusion Time The time an incident occurs after the onset of a red-light indication The higher the number, the greater likelihood of a more severe angle collision occurring May be a key indicator that drivers are not detecting the presence of an intersection
Study Design 16 signalized intersections with red-light cameras Red light camera data from 2015 - 2019 Categorize intersections by geometrical elements Presence of a median No Approach Median: 7 Intersections Approach median: 9 Intersections Crossing width Ranging from 14 m to 47 m Approach Lanes 2, 3, and 4 lanes Number of legs 4-legged and 3-legged
Study Design Categorize intersections by operational elements Posted speed limit Ranging from 50 km/h to 70 km/h Fixed vs. Semi-actuated signal operation One-way versus two-way operation Number of signal heads 2, 3, or 4 signal heads Lateral Signal head spacing Ranging from 6 m to 18 m
Study Design Categorize intersections by operational elements All-red Phase Ranging from 2.0 Sec to 2.6 Sec Amber Phase Ranging from 3.7 Sec to 4.2 Sec Average Annual Daily Traffic (AADT) Ranging from 1,855 veh/day to 19,152 veh/day Important Note: All intersections assessed were operating with pedestrian countdown signals (PCS) during the study period A previous Region of Waterloo study found that PCS reduced red-light running by 40%
Study Methodology Descriptive Analysis Statistical Modelling Chargeable Incident Average Intrusion Time
Descriptive Analysis
Descriptive Analysis
Descriptive Analysis
Chargeable Incident Model CI=∝ × AADT β1×W β2×e β3 × Leg×e β4×Median Where: CI Chargeable Incidents per Month AADT W Crossing Width (m) Leg 1 if the intersection is 4-legged and 0 otherwise Median 1 if the approach with RLC has a median and 0 otherwise Chargeable Incidents Increase as AADT increases (exposure) Increase as crossing width increases Increase at 3-legged versus 4-legged intersections (less conspicuous?) Decrease with the presence of medians (visual cue?) Ln(∝) 𝛽 1 𝛽 2 𝛽 3 𝛽 4 Coefficient Estimates -8.49370 1.41737 0.22492 -1.90825 -0.50611 p-value < 2e-16 0.00122 Generalized Linear Model (GLM) with a Negative Binomial (NB) error structure Over dispersion parameter: 4.8052 AIC: 6061.4
Average Intrusion Time Model AvgInt =𝑎+ 𝑏 1 ×𝐴𝐴𝐷𝑇+ 𝑏 2 ×𝐿𝑒𝑔+ 𝑏 3 ×𝑂𝑛𝑒−𝑊𝑎𝑦 Where: AvgInt Average Intrusion Time per Month (Sec) AADT Leg 1 if the intersection is 4-legged and 0 otherwise One-Way 1 if the approach with RLC is a one-way street and 0 otherwise a 𝑏 1 𝑏 2 𝑏 3 Coefficient Estimates 6.843 -0.000524 2.4738 3.236 p-value 8.61e-14 < 2e-16 0.000166 Adjusted R-squared: 0.2261
Conclusions Locations with higher volume experience larger frequency of red light running but fewer instances of hazardous situation with large intrusion time 3-legged intersections experience more chargeable incidents compared to 4-legged intersections but the average intrusion time is higher at 4-legged intersections
Conclusions Presence of an approach median reduces the frequency of red light running Presence of an approach median reduces intrusion time of red light running although not statistically significant at a 95% confidence interval
Conclusions One-way operation increases the intrusion time One-way operation increases the frequency of red light running although not statistically significant at a 95% confidence interval
Conclusions The results of this study can be used to: Improve intersections with higher potential for red light running especially those with larger intrusion time; and Select intersections that would benefit most from the deployment of red-light cameras
Limitations Only 16 sites evaluated Potential to expand study to include red-light camera sites from across Ontario.
Acknowledgements Pedram Izadpanah, Ph.D., P.Eng Modelling, Statistical Analysis Jeffery Catlin, City of Toronto Data production
Thank you!