Identifying Dangerous Curves

3234
10
01-20-2017 09:59 AM
DorotheaKnigge
New Contributor III

Hello,       

We would like to identify dangerous tight curves in our road database.

To start I calculated a field called 'Sinuosity' with the following expression:

"!Shape.Length! / (math.sqrt(math.pow((float( !Shape.FirstPoint.X!) - float( !Shape.LastPoint.X!)), 2) + math.pow((float( !Shape.FirstPoint.Y!) - float( !Shape.LastPoint.Y!)), 2)))"

 

I then downloaded a geoprocessing script called 'Identify True Curves' which creates two output feature classes:

OutputCurvePoints and OutputWithCurves

It also creates two temp files: tmpSourceVertices and tmpDensifiedVertices

I have successfully run the model but it is really not what I am looking to accomplish.

Does anyone have a way to find these dangerous hairpin turns?

I am using ArcInfo 10.1.

Thank you,

Dorothea

Tags (1)
0 Kudos
10 Replies
JoeBorgione
MVP Emeritus

Look for accident data...  I don't think a true curve is what you want; it's more or less a drawing trick.  Not sure your math will find them either: if I understand them they are just looking at the beginning and end.  The sinuosity of any line is defined along the way, not just end points, right?

That should just about do it....
0 Kudos
NeilAyres
MVP Alum

It is common in river research to define sinuosity as a ratio of channel length over straight line distance.

See here :

sinuosity

But in terms of roads, I agree with jborgion

Rather look for better evidence of what constitutes "dangerous".

0 Kudos
DorotheaKnigge
New Contributor III

Good morning,

Thank you Neil and Joe for your comments.  Joe you are correct that the sinuosity of any line is defined along the way, not just the end points.  I have not been happy with the "math" results.

We do maintain our own accident data so I will take a look at where the 'hot spots' are occurring.

Thank you for your re-direction and I wish you both a great day!

Dorothea

0 Kudos
AnthonyPankala
New Contributor II

Looking at only 'hot spots' won't be very useful if your area you are looking in is fairly large as the hot spots will be mostly at intersections.  If the traffic accidents are tracked with other details such as type of crash you could just look at 'lane departure' types of crashes.  Or possibly remove all crash data that is within so many feet of an intersection, say 300 or 500 feet.  But all of these potential solutions require access to good crash data.  Which really is going to help you identify what is truly a dangerous curve.  A tight curve with no crash histories I would argue is not a 'dangerous' curve.  Then again 1 crash every three years isn't bad at all, and definitely not a 'hot spot'.  But if every one of them is a fatal crash since the vehicle falls over a giant cliff I think everyone would agree to be a very dangerous curve.  So sorting by severity of injury would help as well....assuming your crash data tracks that as well.

0 Kudos
DorotheaKnigge
New Contributor III

Anthony,

Thank you for your comments, we do track "type" and "severity" among many other attributes. Using linear referencing we are also tracking by quarter mile segments which shows a different aspect on what's happening on the roads.

I'm looking at all possible ways to identify dangerous curves that may need signage.  That 'magic' script is not yet out there!

Dorothea

0 Kudos
SteveCole
Frequent Contributor

If you also have speed limit information, perhaps you can look for road segments where the speed limit changes "significantly" on either side of a given road segment. If you have a road sign inventory, you could also look for signs that indicate hairpins or sharp turns as well.

DorotheaKnigge
New Contributor III

Hello Steve,

I appreciated your take on finding hairpin/sharp turns using speed limit information and road sign inventory. We have incorporated those aspects into our search.

Thanking you for your input,

Dorothea

0 Kudos
DaleHarris2
New Contributor II

Hi Dorothea,

Interesting thread. I’ve had some experience in using GIS to identify high risk curves in New Zealand. Our approach is based on the idea is that a tight curve is not necessarily a dangerous curve – it all depends on the context. For example, drivers are much more likely to lose control and have a serious crash on a tight curve at the end of a long straight, compared to a tight curve along a winding section where driving speeds are already very low. Relying on crash data can give you an idea of problem areas, but it's not very helpful for low volume roads where crashes tend to be more sporadic and "random".

In our GIS models, we first identify curves from a road centreline. Then we model vehicle speeds along the road centreline (acceleration on straights, deceleration on curves) to determine the approach speed to each curve. The difference between the approach speed and the appropriate speed for the curve then determines whether it is “dangerous”. Our models are relatively complex in that we also factor in the overall terrain (eg vehicle speeds are generally lower on mountainous roads) and driver behaviour (eg drivers tend to brake within 70 metres of a curve and don’t accelerate on straights less than 200 metres).

Our methodology uses Model Builder and some Python scripts - no special software. We do quite a bit of editing to get the centreline into a suitable format for extracting curves and running our models.

If you are interested in our methodology, I wrote it up an article for the New Zealand Surveying+Spatial magazine last year (see page 19 onwards here: https://issuu.com/nzis/docs/s_s_march_2016).

Hope this is helpful? I'm happy to elaborate further if you have any questions.

Dale

DorotheaKnigge
New Contributor III

Hi Dale,

I find your process very intriguing.  It seems the whole process hinges on having a very accurate centerline.  Do you incorporate 'true curves' when editing?  What is the suitable format for extracting curves?

Thank you for any additional information.

Dorothea

0 Kudos