Study sites 50 unsignalized Introduction pedestrian crossings in - PowerPoint PPT Presentation

Study sites – 50 unsignalized Introduction pedestrian crossings in Warsaw • High pedestrian fatality rate in Poland: 23 person/year/mln pop. – 8436 pedestrian accidents in 2016 Modelling accident frequency at – 868 pedestrians killed • Three weeks of (28.6% of all traffic fatalities) filming at 2 sites: unsignalized pedestrian crossings • Pedestrians in Warsaw: – Radzyminska in Warsaw – Constitute 60% of fatalities – KEN – Slow improvement • All crossings located Piotr Olszewski, Beata Osi ń ska, Pawe ł W ł odarek on 2x2 lane roads Warsaw University of Technology • Objective: to model accident frequency at pedestrian crossings • Roadways separated and to identify factors that affect pedestrian safety by a median or a 30th ICTCT Workshop • Part of project InDeV : “In-Depth understanding of accident refuge island Olomouc, Czech Republic, 26-27 October 2017 causation for Vulnerable road users” 30th ICTCT Workshop, Olomouc, 26-27.10.2017 No. 2 P. Olszewski et al. 30th ICTCT Workshop, Olomouc, 26-27.10.2017 No. 3 P. Olszewski et al. KEN filming site Radzymińska filming site Accident statistics (2009-2016) • 4 + 2 pedestrian accidents in 7 years • 8 + 4 pedestrian accidents in 7 years • Distribution of sites by the number of accidents • Police accident records: 59 accidents during 7 years – 1 fatal, 14 serious – mean = 1.18 acc./site – variance = 2.51 --> over-dispersion • Problems typical for road accident dataset: – over-dispersion (greater variability than Poisson distribution) – frequent zero observations – 23 sites (46%) had zero accidents P. Olszewski et al. 30th ICTCT Workshop, Olomouc, 26-27.10.2017 No. 4 P. Olszewski et al. 30th ICTCT Workshop, Olomouc, 26-27.10.2017 No. 5 P. Olszewski et al. 30th ICTCT Workshop, Olomouc, 26-27.10.2017 No. 6

Accident prediction model Daily traffic volume estimation Variable description & statistics • Volumes of pedestrian and eq (1) vehicle traffic were counted – Six sites: 24 h counts – Remaining sites: 3x1 hour counts: 7-8, 12-13, 16-17 • Pedestrian volume model DPV = 5.65 P 7 + 9.62 P 12 + 2.1 P 16 (R 2 = 0.999) • Motor traffic volume model DTV = 3.87 T 7 + 7.53 T 12 + 4.17 T 16 (R 2 = 0.999) P. Olszewski et al. 30th ICTCT Workshop, Olomouc, 26-27.10.2017 No. 7 P. Olszewski et al. 30th ICTCT Workshop, Olomouc, 26-27.10.2017 No. 8 P. Olszewski et al. 30th ICTCT Workshop, Olomouc, 26-27.10.2017 No. 9 Results of model estimation - 1 Poisson or Negative Binomial? CURE plots • Eq (1) can be estimated assuming different statistical models: • Negative Binomial distribution is usually preferable – here it has: • Cumulative residuals can Poisson, Negative Binomial, Zero-inflated - Poisson or NB – lower (=better) AIC value but be plotted, sorted by the – lower (=worse) pseudo R 2 and lower significance of explanatory variables sum of traffic volumes • Test for over-dispersion: CT (Cameron-Trivedi) is not conclusive – it shows that Poisson distribution cannot be rejected • Poisson 1 model • Conclusion: • Negative Binomial model – Poisson acc./year • Both models perform – NB acc./year well – cumulative residuals are within 2 std. error bands P. Olszewski et al. 30th ICTCT Workshop, Olomouc, 26-27.10.2017 No. 10 P. Olszewski et al. 30th ICTCT Workshop, Olomouc, 26-27.10.2017 No. 11 P. Olszewski et al. 30th ICTCT Workshop, Olomouc, 26-27.10.2017 No. 12

Results of model estimation - 2 Extended Poisson model Conclusions • More variables were included in the Poisson model • Model Poisson 4 has the • The recommended accident prediction model is based on best fit parameters and Negative Binomial distribution: correct plot of cumulative residuals • Significant variables: • Risk factors affecting pedestrian safety at unsignalized DPV, DTV, HGV, HOUSE, PPEAK crossings: proportion of heavy vehicles, location in a housing • None of the other variables area, less peaked pedestrian volume profile were significant • Models developed will help to establish a relationship • Final equation for the number of accidents per year: between accident frequency and conflict frequency • Possible improvements: accounting for seasonal variability (DTV -> AADT) and daily variability (peak ratio) P. Olszewski et al. 30th ICTCT Workshop, Olomouc, 26-27.10.2017 No. 13 P. Olszewski et al. 30th ICTCT Workshop, Olomouc, 26-27.10.2017 No. 14 P. Olszewski et al. 30th ICTCT Workshop, Olomouc, 26-27.10.2017 No. 15 30th ICTCT Workshop, Olomouc, 26-27.10.2017 Thank you very much for your attention! This project has received funding from the European Union’s Horizon 2020 research and innovation programme under grant agreement No 635895