Coding for Head-On or Run-Off Crashes

Does iRAP have any guidance on the best approach to coding for Head-On crashes or Run-Off crashes?The scenario is, you have a physical median of varying width (2m to 10m) with isolated poles over a length of 1km. The SRS appears to discount Head-On LOC once you have any hazard (except a downward slope) within the median. The issue is that the Head-On crash SRS are often higher than the Run-Off crash SRS depending on AADT, median width, hazard type and the offset to the hazard.

Hi Paul, here’s my some thoughts from my colleague, James:

In terms of divided roads, the model has to determine if a driver-side object is within the median itself, or on the other side of the opposing carriageway.

It does this with median traversability, this is done through a combined attribute that compares the DS roadside object and its associated distance with the combined width of the physical median and DS shoulder.

If a DS roadside object is closer that the median is wide, the model considers the median not traversable – that is the DS roadside object is the first thing you will hit.

Whereas if a DS roadside object is further away than the combined width of the physical median and DS shoulder, then the model considers the median as traversable – and therefore the opposing flow will be the first thing you hit.

At a simple level, for divided roads:
When the median is not traversable – there is DS run-off risk
When the median is traversable – there is Head-on LOC risk

Because the way in which we code 100m lengths of road as a unit, the model cannot distinguish if an attribute covers the whole 100m or just part of it.

For example, when a roadside object is coded the highest risk category is recorded, this means the model does not understand the difference between one tree vs a row of trees.

So, in terms of median traversability, one tree in the median is treated the same as a row of trees.
Whilst this is an oversimplification, it is worth considering how the angle of departure would play into any crash.

For the example provided in the question, for there to only be DS run-off risk, it would mean that a DS roadside object was coded in the median for every 100m.

Overall, when there is only 1 post in median in a 100m section, then the model is probably overestimating the ROR risk, while the head off risk is underestimated. As the number of posts in the median increases, the risk of a head on crash will decrease and ROR risk will increase.

Hope this is helpful


Thank you for the extensive response Greg and it all makes senses to us. We appreciate that the iRAP model is at times a simplification of what actually occurs on the road network (and needs to be so), with the median traversablity being one such case. We are trying to respond to our clients concerns that on a high speed, high volume road (90km/h, +50,000) with sections of narrow medians (<1m) and tight horizontal geometry (200m - 500m radii) that Head-On crashes are a big concern. They are then concerned that introducing a couple of hazards (i.e, streetlight poles) will remove the Head-On crashes from the scoring.

I’m glad that you brought up the angle of departure concept as this is one of the factors we intend to use determine whether to code a median hazard or not. Austraods has a good discussion on this and how it determines “impact length” in AP-R628-20 “Background to the Development of the 2020 AGRD Part 6”. The concept of “impact length” is also a simplification of a more complex idea and uses average departure angles and does not take into consideration road attributes that can affect this angles such as distance to the hazard and/or horizontal road geometry, but we believe it is enough to guide us on the right path with regard to our clients concerns.

The result is that once we calculate the impact length of all hazards within the median, we can determine the likelihood that an errant vehicle will miss the hazards and then exposure the vehicle to head-on crashes. If the likelihood of a head-on crash is high enough (>40%), then we can determine which crash types results in the highest SRS and then code that section of road using the higher SRS.

The only reason we are not doing this at the moment is that we are concerned that it is not following iRAP policy, which could lead to inconsistences with other users and other Star Rating Assessment done on the road network in Australia. Would the above approach breach any iRAP policy/guidelines?