Rate per HMVM

[teeman]

I am reviewing a report that looks at a 1.2 mile stretch of urban interstate freeway. The report breaks the freeway into segments of 300 ft, 4100ft, 300ft and 1500ft and calculates accident rates per Hundred Million Vehicle Miles for each segment.The 300 foot segments are associated with interchange ramps and the longer segments are between the interchanges.

My gut reaction is that utilizing very short roadway segments (300ft) in the analysis skews the results in a way that makes the interchanges appear to be more accident prone than they actually are?

Is this a valid application of accident analysis procedures?

Any thoughts on what would be the best way to look at accidents along a short stretch of urban freeway?

 

[francesca]

For the same number of accidents per veh, there are 13.67 times more accidents per veh-mi on the 300 ft segment. However, seeing as the section is 13.67 times shorter, this seems fair.

I might argue that, by not distinguishing the "weaving" sections (before/after the intersection) from the "free flow" sections, bias is introduced, making the areas immediately before and after an intersection appear safer than they actually are.

 

[ACtrafficengr]

I'd take another look at your methodology to make sure you don't have an apples and oranges situation.

I would look at the interchange area as a whole, and compare it to other interchanges with similar AADT's, and compare mainline segment with other mainline segments with similar AADT's. That is because if you graph cr/mvm vs volume, you'll see it isn't linear.

You will probably find that the interchange crash rates are usually higher than the segments between them. That is to be expected, because of the merging, diverging and weaving that goes on.

Just looking at high crash rate sections isn't always the best way to improve safety. Look for the "low hanging fruit," too. If you look at mainline segments separately from interchanges, you may find some mainline segments with safety problems that would otherwise be missed because of interchange crashes. You may spend all you effort on intractable problems at interchanges when a few chevrons on a curve may solve a run-off-road crash problem.

Of course, there comes a time when you have to say, which do you like better, apples or oranges?


"...students of traffic are beginning to realize the false economy of mechanically controlled traffic, and hand work by trained officers will again prevail." - Wm. Phelps Eno, ca. 1928

"I'm searching for the questions, so my answers will make sense." - Stephen Brust

 

[teeman]

Thank you for your thoughts. I guess my ultimate concern is I am not convinced the data provided from this calculation is meaningful in this circunstance. The whole purpose of the calculation appears to be to allow comparison to a critical accident rate from an undocumented source.

Most of the example calculations I have seen for using HMVM apply the calculation to roadway segments that are many miles long. Since segment lengths of less than a mile act as a multiplier, significant weight is given to segment length instead of traffic volume and the number of accidents.

9 accidents per year at an interchange with an ADT of 90000 when analyzed as a 300ft segment gets a significantly higher accident rate(482)than a 1500 ft tangent segment between interchenges with 8 accidents per year and an ADT of 60000 (128). Are those 8 accidents actually spread out along that 1500 foot segment or are they concentrated in a smaller area? The project documentation summarizes the accidents by segment but does not actually plot them along the corridor.

Due to the short length of the project area, I am inclined to think the whole (1.2) miles should be analyzed in similar segment lengths, (say 500 ft) to minimize the multiplier effect of segment length. It might also provide for better identification of accident sites.

Thanks Ted