The application of extreme value theory (EVT) to road safety can produce quick and reliable safety evaluations. In view of a future widespread practical application, this paper investigates for the first time the issues of model transferability and seasonality in EVT. A case study dealing with motorway rear-end collision risk is presented. Vehicle-by-vehicle traffic data were collected in 14 motorway sections with similar characteristics for a year, and Time-To-Collision (TTC) was computed for each couple of consecutive vehicles. Two sets of Generalized Pareto distributions were fitted with the Peak-Over-Threshold approach: in Set#l TTC values were aggregated across all sections, for each month; in Set#2 for each section and each month, only TTC collected in all the other sections in the same month were considered. Model performance was evaluated comparing predicted and observed rear-end collisions. According to the findings of this work, it is possible to aggregate data from several road sections, provided that they share similar geometric, traffic and weather characteristics, to estimate EVT models; moreover, such models are transferable with very good results to other similar road sections. In addition to this, results show that it is important to take into account seasonality effect, as predictions made considering data collected for only a short continuous observation period may be largely overestimated or underestimated.

Transferability and seasonality in extreme value theory applications to road safety: A case study in an Italian motorway

Orsini F.;Gecchele G.;Gastaldi M.
;
Rossi R.
2020

Abstract

The application of extreme value theory (EVT) to road safety can produce quick and reliable safety evaluations. In view of a future widespread practical application, this paper investigates for the first time the issues of model transferability and seasonality in EVT. A case study dealing with motorway rear-end collision risk is presented. Vehicle-by-vehicle traffic data were collected in 14 motorway sections with similar characteristics for a year, and Time-To-Collision (TTC) was computed for each couple of consecutive vehicles. Two sets of Generalized Pareto distributions were fitted with the Peak-Over-Threshold approach: in Set#l TTC values were aggregated across all sections, for each month; in Set#2 for each section and each month, only TTC collected in all the other sections in the same month were considered. Model performance was evaluated comparing predicted and observed rear-end collisions. According to the findings of this work, it is possible to aggregate data from several road sections, provided that they share similar geometric, traffic and weather characteristics, to estimate EVT models; moreover, such models are transferable with very good results to other similar road sections. In addition to this, results show that it is important to take into account seasonality effect, as predictions made considering data collected for only a short continuous observation period may be largely overestimated or underestimated.
File in questo prodotto:
Non ci sono file associati a questo prodotto.
Pubblicazioni consigliate

Caricamento pubblicazioni consigliate

I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.

Utilizza questo identificativo per citare o creare un link a questo documento: http://hdl.handle.net/11577/3351289
Citazioni
  • ???jsp.display-item.citation.pmc??? ND
  • Scopus 4
  • ???jsp.display-item.citation.isi??? ND
social impact