I'm looking for definitive references on road infrastructure, traffic and operational crash risk factors for urban local roads. By local, I mean non-arterials (access streets, collectors) and low-order arterials (some call them distributors). Generally, the roads are in the 50 - 20,000 vpd AADT range, although there would be rare exceptions.
Hi Chris, in terms of crash prediction models, most authors model rural conditions. Urban are way more complicated :-) There are traffic lights, pedestrian flows and lots of other things which are rarely registered in network-wide databases - therefore it is up to the analyst to collect his own data :-(
Nevertheless you can check several papers, which focused on urban conditions:
https://www.researchgate.net/publication/10971821_Accident_prediction_models_for_urban_roads
https://www.researchgate.net/publication/236187571_The_Influence_of_the_Infrastructure_Characteristics_in_Urban_road_Accidents_Occurrence
or Thomas Jonsson's thesis http://www.dissertations.se/dissertation/d4c7a8eb3e/
Article Accident prediction models for urban roads
Article The Influence of the Infrastructure Characteristics in Urban...
Great stuff Jiri. This is really useful. I am meeting with other experts on the subject today, so these references are very useful. I have not seen them before.
Oddly, it seems not many researchers tackle local road safety from the engineering angle, rural roads excluded (eg prof Cafiso). There are 40% of fatalities on local roads. It should be a higher priority.
We had limited experience using the ANRAM model in urban arterial roads. ANRAM uses the same risk factors as on rural roads (not ideal) but backed by urban-specific SPFs and tempered with lower speeds of course. We noted that a whole different set of factors controlled risk scores and crash predictions than on rural roads. We ought to publish this stuff one day.
Thank you Kolita. Interesting articles on road safety in Sri Lanka, although probably not the subject area I am looking for.
Hi Chris, yes, rural roads are more often in research focus - around 30-40% of fatalities happen there. But of course the rest is urban roads. I think that they are overlooked due to complexity = difficulties with data collection. We are now involved in a project looking at effect of lighting (illumination) on urban roads, so we will also start collecting data on risk factors. So let's stay in touch and share experience!
We had an internal meeting and thought to try first to develop a very simple model with variables we already have: AADT, FCD mean speeds, intersection presence and land use. The last variable, land use, should be a proxy for everything from road environment (rural, semi-urban, dense urban) to pedestrian volumes, parking, etc. This is proof of concept only. We will see what happens.
Yes, it may be a good proxy mainly for pedestrian AADT. I keep my fingers crossed - let us know if this worked.
The behavior and urgency of road user is also one of the risk factor
I am not sure of there is one set of crash risk factors locations in different countries. I suppose each country has its own factors.
For sure the literature from other countries will help and form as a guide.
Many thanks guys. I am still interested in engineering variables, mainly, with proxies for behavioural ones. I am going to ask a follow up question, related to this - effectively a new thread:
'How to overcome poor traffic accident model (SPF) correlation with historical crash data, where crashes are rare and highly dispersed?"
So, keep an eye out for that if you can inform it.
Hi Chris, interesting question. It comes down to empirical Bayes method, which combines crash history and crash prediction (from SPF). It is combined through a weight, which expresses "model quality". So when SPF is poor, as you mention, the final EB estimate is closer to crash history. And vice versa, when SPF is strong, the result is closer to SPF prediction. Does it help?
For an explanation from Ezra himself, check this link: https://www.researchgate.net/publication/252330169_Estimating_Safety_by_the_Empirical_Bayes_Method_A_Tutorial
Article Estimating Safety by the Empirical Bayes Method: A Tutorial
Hi Kolita, thank you for your suggestions. I'm doing my research for Australian conditions. It is a mature motorised road traffic system (~100 years) with a very high level of road rule compliance and overall good road user discipline. Your suggestions are 100% correct and highlight the gap in knowledge between mature and fast-growing traffic systems, e.g. in Asia. The focus on road user behaviour and compliance needs to be much greater there to achieve the gains. In my view, this could be a generational project (I hope I'm wrong!). There are a lot good things going on in this space. I have seen institutional transformation examples in Philippines, Nepal and in Kerala, e.g. in school children traffic safety education, Police enforcement, etc.
Please refer to the following article from Accident Analysis & Prevention Journal by Omranian et al.:
Exploring rainfall impacts on the crash risk on Texas roadways: a crash-based matched-pairs analysis approach
- Badges
- Science topic
- Similar topics
- Geoscience
- Asia
- India