Hu used two colocation measures—global and local—to analyze the spatial relationship between pedestrian- and bicyclist-automobile crashes and intersections.

• Global colocation. The global measure considers the overall relationship between any two types of objects or events—in this case some form of recorded crash and some iteration of an intersection. To search for a pattern, the analysis starts with the observed number of times that a specific intersection type has a specific crash type in close proximity to it. That number is then compared with those expected to occur by chance through multiple simulation runs. If the resulted metric is greater than one, it indicates a colocation pattern (i.e., fatalities are attracted to traffic light-controlled intersections). If the result is less than one, it denotes an isolation pattern (i.e., fatalities are separated from traffic light-controlled intersections).

• Local colocation. The global analysis looks at the general pattern across the entire city but cannot reveal the pattern at specific locations. The local colocation method examines the localized pattern between specific points of one variable and the overall set of points from the other. The local analysis shows how a specific traffic light-controlled intersection either attracts or is isolated from all fatality incidents nearby. This approach identifies the locations of those crashes that were spatially tied to certain types of intersections.

On major roadways, nonsignalized intersections are also problematic, Hu said. Wide roadway—such as Bellaire Boulevard, Westheimer Road and Richmond Avenue in Houston—have signalized intersections that are far apart. Hu said this contributes to crash incidents at intersections without signals or signs when pedestrians jaywalk. Hu said that more pedestrian crosswalks with signals at closer intervals may be needed on some streets.

In addition to the presence or absence of traffic signals, Hu found that intersection structure and proximity to bus and rail stops also matter. Complicated intersections with four or more roads leading into it and those that have bus and rail stops nearby are 2.4 times more likely to see a fatal accident, whereas intersections with fewer than four entry points are only 0.33 times more likely to have a fatality. Actions that simplify traffic flow at complicated intersections and considerations of how to design transit stops in safer ways, such as midblock boarding bulbs could be considered, Hu said.

The key finding of this research—that bicycle- and pedestrian-automobile crashes are collectively colocated with signalized intersections—suggests that these intersections require attention to make our streets safer for all. While traffic light intersections are the most controlled in terms of signage, it is clear that they are not the most safe. Both the global and local findings provide useful policy implications for planners and decision-makers.

Physical improvements that alert users to proceed with caution around other road users or in complicated intersections—such as pedestrian visibility via crosswalks, traffic lights or protected medians—could make a big impact. Likewise, attempts to induce slower speeds in or near intersections would likely benefit safety. This might mean reducing speed limits as they approach high-risk intersections and adding additional notification and warning features like rumble strips or painting. Finally, attempts to shorten the distance pedestrians and people on bikes have to cross via protected medians or narrower streets would likely reduce risks as well.

—“Dangerous Crossings”

Resources

• Dangerous Crossings: The Relationship Between Intersections and Crashes in Houston