Crash prone 'modern roundabouts'

[roadfro]

Here is a list of 40 multi-lane roundabouts that I think could be meaningful to this conversation.  This was queried from a database of over 5,100 modern roundabouts.  This was the criteria used to query out the list:

-roundabouts constructed within the last 10 years    Reasonable
-all approaches have 2 entry lanes    Why exclude roundabouts that may have one or two legs with a single lane entry?
-roundabouts have 4-legs    Exactly 4 or >=4? Why exclude 3-leg 2x2 roundabouts?
-main & side streets are major state or county routes (ie. likely high AADT roundabouts)    Not an accurate assumption...while likely, there still are a number that are solely local arterial roads, and not numbered/lettered routes. Ex: Thompson/Lein/Zeier, Madison, Thompson/Main, Sun Prairie
-interchange roundabouts (and frontage road roundabouts) not included    Why? Just because interchange roundabouts don't have 4 complete approaches?

As I said above, you're being too limiting in your search. Some might say you're cherry-picking data. The lack of complete crash data skews the whole process. I could list many multilane roundabouts that you haven't mentioned, but they would/should get thrown-out because the intersection did not exist before the roundabout, thus there is no before period to compare to.

This is going in nothing but circles (pun not intended). I tried to find common ground in your reasonable query on roundabout size, but you are blinded into focusing solely on size. Is size a factor? Yes. Does it deserve more scientifically-acceptable research? Most certainly. Is it the only factor where crashes are a problem? No. Roundabout design isn't that simplistic. You started a discussion on "crash-prone 'modern roundabouts'", when a more accurate subject would be 'why modern roundabouts are too small'.

DaBigE, I think some of your assumptions about tradephoric's data and filtering criteria are a little unreasonable.

In order to effectively analyze roundabouts from various locations to draw some general conclusions, you have to limit the data in some way in order to have the analysis make sense and reduce effects from other variables. One way to do that is looking for substantially similar conditions at various sites. Limiting the sample to 4-leg roundabouts with double lane approaches on all sides makes valid sense in this data analysis–otherwise, you're not making an "apples to apples" comparison and can not make the same generalizations because other design factors can influence the data. (Although I would agree that the limitation to include just state/county routes isn't necessarily a good one, but I understand the rationale in trying to find roundabouts with similar AADTs. Finding that data in relation to crash data for specific sites might not be an easy endeavor for armchair/hobbyist analysis purposes.)

When you have raw data like this in a tabular format, with similar site characteristics, any good data analysis would attempt to find patterns in said data. Tradephoric found one possible trend between a possible variable (inscribed circle diameter) and accident rates, and has commented on this specifically. I don't think we could say with any reasonable certainty that there is a direct correlation without more analysis, but it is a plausible theory worth further examination (possibly worthy of a civil/traffic engineering journal article). And by no means is ICD the only possible factor influencing the accident rates at these sites–but to examine multiple possible factors independent of others would require much more research and site visits.; however, it is one that is easy to examine without making a detailed site visit.

[Rothman]

I'm going to have call baloney on this repeated assertion as well regarding the New Scotland roundabout.  You're saying that the average in the first three years was 9.6 per year and then the average in the three years afterwards was 38.3. per year.

That's total baloney.  In essence, you're saying that there was at least one incident in the roundabout every 10 days for three years after it was built. I highly suspect that you have warped the data into something for which it was not intended.

I don't understand what is so hard to believe.  Does 38 crashes a year sound too high to you?  To put it into perspective, there are over 50 intersections in SE Michigan that have averaged at least 38 crashes a year.

http://semcog.org/Data-and-Maps/High-Frequency-Crash-Locations

Yep.  38 crashes a year at that roundabout in the three years since it was built does sound far too high.

[DaBigE]

DaBigE, I think some of your assumptions about tradephoric's data and filtering criteria are a little unreasonable.

And I respectfully disagree. The initial discussion was based on multilane roundabouts in general, not specifically 2x2 roundabouts. In the end, we've mainly been discussing entering and exiting crashes, which don't tend to discriminate very much between a 2x2 or a 1x2 roundabout; as long as there is some multilane component, similar vehicular conflicts can and will exist.

I don't think there is anyone here that disagrees that ICD may play a role or that there is a pattern at many multilane roundabouts; what many of us disagree with is making a broad generalization on very limited data. That said, it is definitely worth looking into further, but from a more detailed approach. I think the size discussion has gone about as far as it can on an enthusiast level. The discussion may even be completely moot, as the recommended range of multilane roundabout size is already greater than the 100-ft threshold of this discussion (Exhibit 6-9, NCHRP Report 672). Many of the roundabouts being discussed likely made smaller because of design concessions/site constraints (e.g., costly right-of-way acquisition, some unmovable feature, etc.).

For similar reasons, I also disagree with the IIHS study being promoted any further because, as tradephoric also pointed out, it was done on fewer roundabouts and over 15 years ago. A lot has changed within the past 5 years, let alone 15. Similarly, as I've said earlier, many of us (professional colleagues included), have fundamental issues with the studies WisDOT has carried out as well.

Analyzing roundabouts is not an easy task, whether it's crash data or capacity. There is a lot of disagreement in the design world. This is one reason why I have been harping on the danger of some of the comparisons of older roundabouts that have not been touched since they were first constructed a decade ago. Design philosophies are still evolving, from a forecasting perspective as well. From that point alone, it will be very hard to ever get an apples-to-apples comparison. Throw in different driving characteristics in the various areas of the country and the similarities disappear even faster.

Looking into the future of multilane roundabouts, some new experiments are gaining traction in a few locations in Texas and a couple in Minnesota: non-striped or limited-striping circulatory roadways. This will blur the analysis picture even more.

[jakeroot]

Looking into the future of multilane roundabouts, some new experiments are gaining traction in a few locations in Texas and a couple in Minnesota: non-striped or limited-striping circulatory roadways. This will blur the analysis picture even more.

I quite like this idea. I've always thought that we go overboard with signage and pavement markings. Do you have any links to these roundabouts?

[DaBigE]

Looking into the future of multilane roundabouts, some new experiments are gaining traction in a few locations in Texas and a couple in Minnesota: non-striped or limited-striping circulatory roadways. This will blur the analysis picture even more.

I quite like this idea. I've always thought that we go overboard with signage and pavement markings. Do you have any links to these roundabouts?

Not yet. Most are either still in final planning phases or are just going to construction.

[lordsutch]

Non-striped circulatory roadways are standard in the UK for unsignalized, multilane roundabouts. For what it's worth...

[jeffandnicole]

Looking into the future of multilane roundabouts, some new experiments are gaining traction in a few locations in Texas and a couple in Minnesota: non-striped or limited-striping circulatory roadways. This will blur the analysis picture even more.

I quite like this idea. I've always thought that we go overboard with signage and pavement markings. Do you have any links to these roundabouts?

Look up nearly any traditional NJ Traffic Circle, and this is what they are.

Yes, people that should stay left drift to the right, and those that should stay further to the right drift left.  Occasionally, someone from the left exits the circle.  Horns honk.  Words are uttered (mostly behind closed windows).  But overall, very few accidents.

[tradephoric]

Looking into the future of multilane roundabouts, some new experiments are gaining traction in a few locations in Texas and a couple in Minnesota: non-striped or limited-striping circulatory roadways. This will blur the analysis picture even more.

I quite like this idea. I've always thought that we go overboard with signage and pavement markings. Do you have any links to these roundabouts?

Look up nearly any traditional NJ Traffic Circle, and this is what they are.

Yes, people that should stay left drift to the right, and those that should stay further to the right drift left.  Occasionally, someone from the left exits the circle.  Horns honk.  Words are uttered (mostly behind closed windows).  But overall, very few accidents.

When a non-striped traffic circle in Augusta, Maine was converted to a modern roundabout, it saw a 60% drop in crashes (the size of the circle remained the same).  I don't think there is any consensus that non-striped roundabouts would be more effective then striped roundabouts.

http://www.pressherald.com/2013/04/17/terms-for-traffic-circles-cause-confusion_2013-04-18/

[tradephoric]

The initial discussion was based on multilane roundabouts in general, not specifically 2x2 roundabouts.

The initial discussion was regarding crash prone modern roundabouts.  DaBigE, you questioned why I didn't include interchange roundabouts.  Based on the WisDOT studies it became clear that interchange roundabouts perform better than non-interchange roundabouts.  Nearly every interchange roundabout in the WisDOT studies saw a drop in crash rate.  You've made it known that you have fundamental issues with the WisDOT studies, but I doubt the studies were based on fabricated crash data.  The best performing multi-lane roundabouts in SE Michigan also happen to be at interchange roundabouts.  Logically, it makes sense why interchange roundabouts may have a lower crash rate then other multi-lane roundabouts.  Some legs only service entry or exiting traffic and it could be argued the turning movements are more predictable (certain turning movements won't be that heavy).  In general, interchange roundabouts just don't appear to be that accident prone.

[DaBigE]

The initial discussion was based on multilane roundabouts in general, not specifically 2x2 roundabouts.

The initial discussion was regarding crash prone modern roundabouts.  DaBigE, you questioned why I didn't include interchange roundabouts.  Based on the WisDOT studies it became clear that interchange roundabouts perform better than non-interchange roundabouts.  Nearly every interchange roundabout in the WisDOT studies saw a drop in crash rate.  You've made it known that you have fundamental issues with the WisDOT studies, but I doubt the studies were based on fabricated crash data.  The best performing multi-lane roundabouts in SE Michigan also happen to be at interchange roundabouts.  Logically, it makes sense why interchange roundabouts may have a lower crash rate then other multi-lane roundabouts.  Some legs only service entry or exiting traffic and it could be argued the turning movements are more predictable (certain turning movements won't be that heavy).  In general, interchange roundabouts just don't appear to be that accident prone.

First, no one is claiming that the studies are coming from "fabricated crash data". The studies' flaws had more to do with site selection. Secondly, if you're trying to prove whether size has a significant contributing factor, why throw out roundabouts with lower crash rates? In order to make a generalization such as size being a "key factor", you need to look at all the data, not just the ones that have the best chance of proving your point. Maybe the ones you chose not to include also help prove a larger size theory.

[DaBigE]

Non-striped circulatory roadways are standard in the UK for unsignalized, multilane roundabouts. For what it's worth...

That is one of the reasons why it is being tried in the US, in addition to less striping maintenance. There is a design theory that stripes should be applied to increase capacity rather than to keep people in-lane. Lack of circulatory striping also removes the confusion when a large vehicle is moving through, as there are no lane lines to have to straddle.

[tradephoric]

First, no one is claiming that the studies are coming from "fabricated crash data". The studies' flaws had more to do with site selection.

I was curious what you took issue with regarding the WisDOT studies.  Some roundabout proponents may attempt to discredit any study that shows roundabouts in a negative light.  It feels like there's a little bit of "let's shoot the messenger"  with these WisDOT studies.

Secondly, if you're trying to prove whether size has a significant contributing factor, why throw out roundabouts with lower crash rates? In order to make a generalization such as size being a "key factor", you need to look at all the data, not just the ones that have the best chance of proving your point. Maybe the ones you chose not to include also help prove a larger size theory.

More data is not always useful.  Here are some low volume 2x2 roundabouts that services subdivisions, commercial office complexes, schools, etc.  Would crash data at these roundabouts (which might average 0.6 crashes/year) be useful in this discussion?  It would become an apples to bananas comparison.   
 
https://www.google.com/maps/@39.22989,-77.2528,18z/data=!3m1!1e3
https://www.google.com/maps/@35.84866,-86.35948,365m/data=!3m1!1e3
https://www.google.com/maps/@30.10538,-81.4157,19z/data=!3m1!1e3
https://www.google.com/maps/@38.9684442,-76.9233194,117m/data=!3m1!1e3
https://www.google.com/maps/@38.9383773,-104.734292,98m/data=!3m1!1e3

[johndoe]

... increase the central island diameter of the roundabout from 95 feet to 150 feet.

Sorry if this has been covered...but is there a reason you don't do your analysis based upon the inscribed diameter?  I ask since that's usually how they're usually laid out (in my experience).

[DaBigE]

First, no one is claiming that the studies are coming from "fabricated crash data". The studies' flaws had more to do with site selection.

I was curious what you took issue with regarding the WisDOT studies.  Some roundabout proponents may attempt to discredit any study that shows roundabouts in a negative light.  It feels like there's a little bit of “let’s shoot the messenger” with these WisDOT studies.

And I'm not sensing that same vibe. Sure, there's a few hell-bent on 'roundabouts everywhere' just like there's similar chest-beaters for traffic signals or DDIs. Most in the design community are taking these studies and trying to find why are these crashes happening, not why are these studies wrong. However, we want to work with the most accurate and complete data we can. Most of us don't get paid based on the number of 'xyz' intersection type are constructed.

In the interest of preventing future false assumptions, I am not one of those roundabout proponents, rather I am a proponent of using the proper tool for the proper location. Roundabouts won't work everywhere, nor will traffic signals or cloverleaf interchanges. I'm just as interested in finding out why crashes are higher in some places and not others. That's not to say there aren't earmarked projects, like many HSIP-funded intersection projects.

More data is not always useful.

The only time more data is not useful is if the data is flawed, incorrect, or objectively erroneous data. In any case, bad data should be objectively removed before attempting any formal conclusions. You don't establish reliable trend lines based on only a little data. That is one of the reasons why LOS analysis equations for roundabout capacity prediction are continually being adjusted.

Here are some low volume 2x2 roundabouts that services subdivisions, commercial office complexes, schools, etc.  Would crash data at these roundabouts (which might average 0.6 crashes/year) be useful in this discussion?  It would become an apples to bananas comparison.   
 
https://www.google.com/maps/@39.22989,-77.2528,18z/data=!3m1!1e3
https://www.google.com/maps/@35.84866,-86.35948,365m/data=!3m1!1e3
https://www.google.com/maps/@30.10538,-81.4157,19z/data=!3m1!1e3
https://www.google.com/maps/@38.9684442,-76.9233194,117m/data=!3m1!1e3
https://www.google.com/maps/@38.9383773,-104.734292,98m/data=!3m1!1e3

You answered your own question by saying they are low-volume. Only if an adjustment is made to correct for volume disparities would this data be useful for anything but analyzing low-volume roundabouts. Your original list of qualifications was looking for multilane roundabouts of a certain AADT. All I am saying is that intersections of similar AADTs shouldn't be omitted in this case just because they have a low crash rate (referring to the afore-mentioned interchange or frontage road roundabouts). Maybe their size helps prove a size/crash relationship.

On a somewhat unrelated topic, I do see some poor yield line placement in a few of your latest examples.

[DaBigE]

... increase the central island diameter of the roundabout from 95 feet to 150 feet.

Sorry if this has been covered...but is there a reason you don't do your analysis based upon the inscribed diameter?  I ask since that's usually how they're usually laid out (in my experience).

That is correct. Central island size is generally a byproduct from the ICD and circulatory lane width(s), in addition to possibly allowing for future staged capacity enhancements.

[lordsutch]

All I am saying is that intersections of similar AADTs shouldn't be omitted in this case just because they have a low crash rate (referring to the afore-mentioned interchange or frontage road roundabouts). Maybe their size helps prove a size/crash relationship.

Indeed. This is what is referred to as "selecting on the dependent variable." You can't (validly) conclude that roundabouts of a certain type with a certain AADT or above have a high crash rate if you omit all the roundabouts of the same type and AADT that have zero or low crash rates.

What you could validly do is compare the characteristics of roundabouts that have high crash rates with those with low crash rates (but similar AADT and lane counts) to identify if there are meaningful patterns that distinguish the high-crash ones from low-crash ones (for example, entry angle, striping design, signage, use of traffic calming treatments like rumble stripes, signed speed limits, advisory speed plaques, lighting, landscaping...).

[ScottRAB]

Look up nearly any traditional NJ Traffic Circle, and this is what they are.

not sure if I got the right attribution:
Modern roundabouts are traffic circles, but traffic circles are not modern roundabouts.
   Many people confuse other and older styles of circular intersections with modern roundabouts.  East coast rotaries, large multi-lane traffic circles (Arc D'Triomphe, Dupont Circle), and small neighborhood traffic circles are not modern roundabouts.   If you want to see the difference between a traffic circle, a rotary (UK roundabout) and a modern roundabout (UK continental roundabout), go to http://tinyurl.com/kstate-RAB  to see pictures.   And here's another site that shows the difference between an older rotary and a modern roundabout: http://tinyurl.com/bzf7qmg

[tradephoric]

Larger roundabouts may be safer assuming the entry angles of the roundabout are designed properly and circulating speeds remain low.  Consider the following scenario.  CAR A enters the roundabout mistakenly believing CAR B is about to exit.  Would u want CAR B to take 1.7 seconds or 3.4 seconds to travel the blue path?  More reaction time is a good thing!

95 FT central island diameter:


160 FT central island diameter:

You make the assumption that vehicle B will be traveling the same speed in both scenarios. The larger the circle, the faster vehicle B can comfortably travel. The faster vehicle B is traveling, the harder it is for vehicle A to judge the gap.

Your general assumption that drivers feel more comfortable driving faster in larger circles is wrong.  The red min-van may feel comfortable traveling through this circle at 90 mph (in a straight line). 



Relating specifically to modern-roundabouts, the same type of straight line fastest path can occur.  Here's an example of a small diameter modern-roundabout in Oconomowoc, Wisconsin where the fastest path through the roundabout is approaching a straight line.  I know it sounds counter-intuitive, but larger diameter roundabout can actually slow down the speeds of circulating traffic.



The 160 ft diameter roundabout had a 48% increase in crashes whereas the 95 ft diameter roundabout had a 335% increase in crashes.  Is it possible that drivers at the 160 ft diameter roundabout can judge gaps in traffic better than the 95 diameter one and that is reflected in the crash data?

[iBallasticwolf2]

Larger roundabouts may be safer assuming the entry angles of the roundabout are designed properly and circulating speeds remain low.  Consider the following scenario.  CAR A enters the roundabout mistakenly believing CAR B is about to exit.  Would u want CAR B to take 1.7 seconds or 3.4 seconds to travel the blue path?  More reaction time is a good thing!

95 FT central island diameter:


160 FT central island diameter:

You make the assumption that vehicle B will be traveling the same speed in both scenarios. The larger the circle, the faster vehicle B can comfortably travel. The faster vehicle B is traveling, the harder it is for vehicle A to judge the gap.

Your general assumption that drivers feel more comfortable driving faster in larger circles is wrong.  The red min-van may feel comfortable traveling through this circle at 90 mph (in a straight line). 



Relating specifically to modern-roundabouts, the same type of straight line fastest path can occur.  Here's an example of a small diameter modern-roundabout in Oconomowoc, Wisconsin where the fastest path through the roundabout is approaching a straight line.  I know it sounds counter-intuitive, but larger diameter roundabout can actually slow down the speeds of circulating traffic.



The 160 ft diameter roundabout had a 48% increase in crashes whereas the 95 ft diameter roundabout had a 335% increase in crashes.  Is it possible that drivers at the 160 ft diameter roundabout can judge gaps in traffic better than the 95 diameter one and that is reflected in the crash data?

Assuming the vehicle decides to go the same speed in both roundabouts then there would be more time to judge a gap. I would assume the speed would be 15 mph.

[DaBigE]

Larger roundabouts may be safer assuming the entry angles of the roundabout are designed properly and circulating speeds remain low.  Consider the following scenario.  CAR A enters the roundabout mistakenly believing CAR B is about to exit.  Would u want CAR B to take 1.7 seconds or 3.4 seconds to travel the blue path?  More reaction time is a good thing!

95 FT central island diameter:


160 FT central island diameter:

You make the assumption that vehicle B will be traveling the same speed in both scenarios. The larger the circle, the faster vehicle B can comfortably travel. The faster vehicle B is traveling, the harder it is for vehicle A to judge the gap.

Your general assumption that drivers feel more comfortable driving faster in larger circles is wrong.  The red min-van may feel comfortable traveling through this circle at 90 mph (in a straight line). 

There was no assumption. Can doesn't mean they necessarily will, but the probability increases.

Relating specifically to modern-roundabouts, the same type of straight line fastest path can occur.  Here’s an example of a small diameter modern-roundabout in Oconomowoc, Wisconsin where the fastest path through the roundabout is approaching a straight line.  I know it sounds counter-intuitive, but larger diameter roundabout can actually slow down the speeds of circulating traffic.

Thanks for helping prove my point by attempting to look at the fast path. Circle size is only part of the equation. The Oconomowoc roundabout also has some path overlap which typically leads to entering crashes as well. This example also shows what tight urban constraints can place on the desirable roundabout size. IIRC, they had to buy out the property in the NW corner. Any larger of a circle, and one or both of the gas stations would have to go as well.

[tradephoric]

Thanks for helping prove my point by attempting to look at the fast path. Circle size is only part of the equation. The Oconomowoc roundabout also has some path overlap which typically leads to entering crashes as well. This example also shows what tight urban constraints can place on the desirable roundabout size. IIRC, they had to buy out the property in the NW corner. Any larger of a circle, and one or both of the gas stations would have to go as well.

If ROW constraints prevent the desired roundabout size from being constructed, maybe it shouldn't be constructed to begin with.  A traffic signal at Ellsworth & State in Ann Arbor, Michigan that averages 16 crashes a year is a better alternative than a roundabout that experiences 170 crashes in a year.  A similar argument can be made for the roundabout at Auburn & Main in Rockford, Illinois (crashes increased 14X and had nearly 100 crashes in one year). 

Do you have a compelling reason why there has been a significant crash rate at these two roundabouts?  Can you cite a large diameter modern-roundabout that has had 100 crashes in one year?  The fact is the worst performing modern roundabouts have small central island diameters.  Nothing you say will change that fact.

[DaBigE]

If ROW constraints prevent the desired roundabout size from being constructed, maybe it shouldn’t be constructed to begin with.  A traffic signal at Ellsworth & State in Ann Arbor, Michigan that averages 16 crashes a year is a better alternative than a roundabout that experiences 170 crashes in a year.  A similar argument can be made for the roundabout at Auburn & Main in Rockford, Illinois (crashes increased 14X and had nearly 100 crashes in one year).

And no signalized intersection, no interchange ramp design has ever had design compromises? Ever driven a curve that has an advisory speed lower than the speed limit? Design compromises were made in those cases, building a curve smaller than what the laws of physics dictate are safe. Years ago, this ramp saw several overturns per year (IIRC, there were three that happened in the period of just a few months in one year). Should this interchange not have been constructed because the curves are tighter than desirable? ROW was a big consideration in that case as well. FWIW, the other similar ramps at the same interchange have had near-zero similar overturn incidents.   

Do you have a compelling reason why there has been a significant crash rate at these two roundabouts?  Can you cite a large diameter modern-roundabout that has had 100 crashes in one year?  The fact is the worst performing modern roundabouts have small central island diameters.  Nothing you say will change that fact.

And nothing I have said has disputed that fact. Please stop distorting my comments to fit your agenda. Have you investigated anything other than the size of the central island? What about lane width? As it's been pointed out, ICD is the proper way of classifying a roundabout's size, since the central island is a byproduct of the width of circulatory roadway and the number of circulating lanes. The Oconomowoc example you provided has many other deficiencies besides size (path overlap, fast path, entry deflection, using a mandatory right-turn lane within the regular approach when it should be a partial bypass at the very least...). Looking solely at size is lazy at best, foolish at worst. How many ways do I need to say it before it sinks in (the last time I'm going to say it)...
A roundabout's size MAY indeed be a factor in crashes, but needs additional in-depth study*, in addition to investigating other possible deficiencies.

* More detailed and scientific than can be achieved on this forum and are suitable for publishing in a venue such as a TRB or NCHRP paper.

[tradephoric]

Have you investigated anything other than the size of the central island?

Fastest path:  The fastest path at small central island diameters approaches a straight line.  The Oconomowoc roundabout has a small central island diameter that exemplifies this concept.

Entry deflection angle: Small central island diameters have less pronounced entry deflection angles than larger central island diameter roundabouts.  Just imagine a small neighborhood traffic circle; the entry deflection angle is nearly non-existent. 

Painted spiral markings:  The New York roundabouts cited in this thread have painted spiral markings as opposed to the spiral being designed into the truck apron.  Interestingly, the larger 130' diameter spiral roundabout in Bethlehem outperformed the smaller 103' diameter spiral roundabout in Malta (299% vs. 445% increase in crashes respectively). 

Non-interchange vs. interchange roundabouts:  Interchange roundabouts appear to be less accident prone than non-interchange roundabouts.  Some legs at an interchange roundabout only service entry or exiting traffic and it could be argued the turning movements are more predictable (certain turning movements won't be as heavy).  A pretty strong trend can be seen when looking through the Wisconsin Phase 1 & Phase 2 studies. 

Reducing circulating lanes:  The following crash prone roundabouts have reduced the number of circulating lanes in an attempt to improve safety. 
-14th Street & Superior in Lincoln, Nebraska (3x2 reduced to 2x2)
-Maple & Drake in Farmington Hills, Michigan (3x2 reduced to 2x2)
-Maple & Farmington in Farmington Hills, Michigan (3x2 reduced to 2x2)
-Homer Watson Boulevard and Block Line Road in Kitchener, Ontario (3x2 reduced to 2x2)
-Bailey & Radio in Woodbury, Minnesota (2x2 reduced to 2x1)
-Broadway Avenue & Lake Street in Forrest Lake, Minnesota (2x2 reduced to 1x1)
-Main Ave & Broadway in De Pere, Wisconsin (2x2 reduced to 2x1)

Pavement markings and signage: The study done on the Richfield, Minnesota roundabout at 66th &  Portland has been discussed on this thread multiple times.  The study suggests that standard arrows are less confusing to drivers than fishhooks (the most recent repaving at this intersection includes fishhook arrows though :? ).   

Reaction time: Larger roundabouts may be safer assuming the entry angles of the roundabout are designed properly and circulating speeds remain low.  Consider the following scenario.  CAR A enters the roundabout mistakenly believing CAR B is about to exit.  Would u want CAR B to take 1.7 seconds or 3.4 seconds to travel the blue path?  More reaction time is a good thing!

95 FT central island diameter:


160 FT central island diameter:


YES.

[DaBigE]

Have you investigated anything other than the size of the central island?

...
YES.

No, you have not. What you have reiterated are more generalities for a small sample of intersections. When I say investigate, I mean take all of the crash-prone roundabouts and investigate each roundabouts features: fast path speeds, deflection, path overlap, lane configuration, signing, pavement marking, PHI-angles, in addition to size. See how each of those characteristics compare to roundabouts with low/"acceptable" crash rates as well as what the current desirable metrics are for each. Size alone may or may not be the presiding factor in these cases. Making the circle arbitrarily larger will do you no good if you have poor speed control on the approach and/or bad path overlap.

Until then, or some new information comes to light that has not been beaten to death in the last 8 pages, I am done. We're going in circles. As I and roadfro have said, we've taken this size discussion as far as one can in an enthusiast setting.

[iBallasticwolf2]

We're going in circles.

OH the irony!