I assisted Nabil Nazha in gathering data and developing a methodology for the geographic analysis of bicycle crashes at intersections needed to complete his master’s thesis from the University of Illinois at Chicago’s College of Urban Planning and Public Affairs (CUPPA). I graduated from CUPPA in 2010. We submitted his paper to the Transport Chicago conference and it was accepted; see Session 1 – A Safety Dance. He was out of the country at the time of the conference so I gave this presentation alone.
From 2007 to 2010 there were 6,705 bicyclists involved in 6,664 crashes involving at least one bicycle and one automobile. A majority of bicyclists received injuries and 20 bicyclists died . Bicycle crashes at intersections are the topic of the paper, Safe Cycling in Chicago.
[slide 2 – bicycling at an intersection – seen at the top]
Nabil Nazha and I both received our masters degrees in urban planning and policy at the University of Illinois at Chicago. Myself in 2010 and he last month. We collaborated on this paper, his final one for the degree, because I had the most knowledge about a certain bicycle crash data set and Nabil wanted to conduct a geographic information systems (GIS) analysis. Nabil is currently in Italy until he returns to study for a PhD at UIC in the fall.
I currently write for a blog called GridChicago.com, about sustainable transportation. I have previously worked for the Chicago Department of Transportation and as a consultant to Active Transportation Alliance. Nabil received a bachelors degree in urban planning at Politecnico di Milano in Milan, Italy.
Before I describe our analysis presented in the paper, I’d like to detail the data set on which our analysis is based.
[slide 3 – Steven’s crash report, page 1]
Every day, Chicagoans get into 221 collisions with “stuff”. This average is based on 2010 data. They hit other automobiles (parked and moving), pedestrians, bicyclists, and other users of or objects on or near the road. After the collision, responding police officers fill out form SR-1050 (seen on the screen). This is the Illinois Traffic Crash Report. If a police officer doesn’t arrive, each motorist is required to submit this form to the Illinois Department of Transportation (IDOT) within a certain time period, if there was an injury or $1,500 of property damage [formerly $500 of property damage]. The data set only includes crashes involving at least one automobile.
Crashes that don’t meet that threshold are not included in the IDOT data, but they would still be on record with the responding police department. When IDOT receives the crash report, it then attempts to correct any mistaken information.
[slide 4 – Steven’s crash report, page 2]
For example, in my own crash report (still on the screen), it was labeled as a crash type of “left-turn” when instead it should have been labeled as a crash type of “pedalcyclist” as I was the first entity to be hit. The first entity to be collided with determines the crash type; this is a curious distinction as it means one must dig deeper into the data to ensure that crashes in which an automobile hit a pedestrian or bicyclist second or third are also included in the analysis.
In the finalized crash data, which IDOT usually releases in August of the following calendar year, the report for my crash was modified to show the correct crash type.
The data set is divided into three tables: information about the crash, information about the vehicles involved in the crash, and information about the people involved in the crash. Illinois state law defines a bicycle as a device, not a vehicle, so you will not find information about bicycles in the vehicle table. But you will find details about that device’s operator in the persons table.
Each table has over 50 attributes that can be quickly searched and filtered and IDOT provides a good data dictionary to decipher them all.
Now that you know more about the data set, I can explain our analysis and some of the results.
The paper’s analysis looks at the frequency of pedestrian AND bicyclist crashes between 2007 and 2010, inclusive. I will only discuss the bicyclist crash frequencies but it’s useful to look at them together as the two figures can tell interesting stories about that intersection. For example, at Cicero Avenue and Madison Street near Garfield Park [number two in the combined ranking], there were 32 pedestrian crashes and a single bicycle crash. This disparity was not seen at any of the other top crash frequent intersections.
[slide 5 – intersection screenshot]
This screenshot is from the Chicago Crash Browser web application I’ve been developing alone for several months.
Tiny feet show pedestrian crashes. This is for additional years than the paper’s period [2005-2010].
We created our own intersections data set to be used in a GIS application using street center line data from the City of Chicago’s open data portal, a great resource for many transportation studies. The streets are classified, but a data dictionary was missing and we had to create our own which may have been imperfect. We created two intersection data sets. The first created intersection nodes where all non-highway roads intersected all other non-highway roads. It contains 52,606 intersections. The second data set comprises intersection nodes where all non-highway roads intersect highway ramps. It contains 712 intersections.
Highway ramps in Chicago are a very tricky part of any bicycle route as the roads here widen, the curbs become taller, traffic speeds increase, the number and kind of conflicts with automobiles changes. Additionally, at underpasses, the street is dark, reducing visibility. It was important for us, as we both cycle in Chicago, to capture these intersections and determine their crash frequency.
The on-ramp to southbound Kennedy is on the back side of the pictured building. Will the driver on the left change lanes, affecting people cycling in the centered bike lane?
We used Quantum GIS, also known as QGIS, an open source application, to run all of our analysis. The application is available for free for Mac, Windows, and Linux.
Once the intersections are created and loaded in the map, we added a 200 feet radius buffer to the nodes. They represented, more or less, the geometric center of the intersection, where two or more road center lines meet. We settled on 200 feet after measuring a variety of intersections. Many intersections are only 50-100 feet wide, but in most cases bike lanes, where installed, end at least 100 feet away from the intersection’s stop bars. Based on existing research, we knew that intersections are where a majority of reported crashes occur and where the most dangerous crashes occur, we assumed that the effects of an intersection extend to where a bike lane ends and cyclists would then be sharing lanes with automobile drivers.
[slide 6 – intersections screenshot from QGIS]
The buffer function must be run twice, once for each intersection data set. Afterwards, the number of crashes falling within each buffer are counted and outputted into a spreadsheet.
When some people saw an analysis of the most crash frequent intersections, they saw “danger signs” and that perhaps they should avoid cycling through that intersection. But this analysis is not complete enough to determine a level of danger for any particular intersection. One shortcoming in that kind of analysis is the lack of good exposure and ridership data. The distance, frequency, and route of Chicago cyclists is unknown. In some cases, however, the city has collected 24-hour traffic counts of cyclists, including at locations that made it in the most crash frequent intersections ranking.
Another issue to assigning a danger level to any intersection is that the sample of crashes in a multi-year period is extremely low. In 4 years there were 17 bicycle crashes at Milwaukee, Fullerton, and Elston (the top intersection), and 14 at Milwaukee, North, and Damen (the second top intersection, and perhaps the busiest intersection for cycling in the city). Many conclusions based on this data may be unfounded.
One of the goals of the paper was to describe the conditions and geometry at each of the top crash frequent intersections. The paper succeeds in describing the geometry and the presence of bike lanes but We were pressed for time and were unable to visit each intersection to gather observational data. At the end of the paper Nabil describes several design interventions that can help decrease the number of crashes. However, intersections are just one way to tell the story about bicycle crashes and injuries in a city with a growing bicycle ridership – most of one’s time cycling is not spent in intersections, but on the segments between them.
All of this analysis is part of my personal objective to expose the crash information to the public. I’ve been working on various efforts to do that since February 2011 when I published the first interactive, online map of bicycle crashes in Chicago. I later created a website, the Chicago Crash Portal, to list all of the various projects based on the IDOT crash data, and to link to other analyses and data sets around the world. Our next step is to find ways to “storify” the data. A map shows thousands of points overlaid on the Chicago street network, but it hardly describes what this can mean to people currently cycling, people who may want to cycle, politicians, policy makers, and urban planners.
[slide 7 – crash portal screenshot]
If you’d like to know more about crash data and the environment I’ve fostered around it, I invite you to visit the Chicago Crash Portal at ChicagoCrashes.org. Additionally, I frequently refer to this crash data and write “miniature” analyses for GridChicago.com [see those here].