
In September, 2022, the City of Detroit Public Works Department released a report on pedestrian safety Streets for People: Detroit Comprehensive Safety Action Plan with a vision to make Detroit streets safer for not just drivers, but for pedestrians, cyclists, and public transit users as well. The plan lays out the local and national problem we have with pedestrian safety, noting that USA pedestrian deaths reached a 40 year high in 2021, doubling rates from 14 years ago. For Detroit in particular, they note that 108 people were killed in all crashes on average annually from 2017-2021.
High Injury Network – City of Detroit Streets for People

The authors created a High Injury Network (HIN) that identified the most dangerous surface streets for all modes of transportation by locating corridors with crashes that had the highest concentrations of crashes that resulted in fatalities and incapacitating injuries between 2017 and 2021. On average, 42 of those killed annually in crashes are pedestrians and cyclists, and while the Streets for People plan gives plans for addressing their safety, the HIN map does not make a distinction for areas that cyclists and pedestrians specifically are at risk for injury. With the development of the Joe Louis Greenway and the increasing number of bike lanes and complete streets over the years in Detroit, we were particularly interested to see what we could learn from available data about pedestrian and cyclist safety in particular.
For all of the maps in this post, you can click the two arrows >> in the top right corner of the map to hide or show the legend. The icon right below the legend icon with the stacked squares gives you access to turn different layers on and off.
The blue dots on this map represent cyclist-involved crashes, and the gold dots represent pedestrian-involved crashes that involve a driver from 2017-2021 for which police reports exist. There are a total of 3,107 crashes represented on this map; 2,349 of those are pedestrian-involved crashes, and 758 are cyclist-involved crashes. According to research by Jaimy Fisher at Simon Fraser University in British Columbia, Canada, only around 20% of cycling crashes are reported, indicating the number of cyclist- and pedestrian-involved crashes could be upwards of 15,500. The only certain counts we have are from the police report data, so we will work with the 3,107 crashes for the rest of this post. Included in these data represented on the map are all ranges of crashes, from those with no injuries to those with fatalities, so while we can see areas of concentrations of crashes and compare them to areas with fewer reported, this map doesn’t tell us both where crashes happened and the severity of injuries that resulted.
Comparing crash locations to the High Injury Network map, we can see the density of crashes showing up on similar roads – Gratiot going northeast out of downtown, and Grand River going northwest out of downtown; 7 mile running east-west on both the east and west sides of Detroit; McNichols, Plymouth, and Warren running east-west on the westside. There are also a few standouts not represented on the HIN map, particularly Jefferson Avenue east and west of downtown; Mack on the eastside; the cluster in downtown and midtown; Morang snaking on the far eastside; and clusters around both Vernor and Michigan Avenue in Southwest Detroit. These visible clusters on this map that don’t show up on Detroit’s HIN may indicate that they are areas uniquely dangerous for cyclists and pedestrians, and not for drivers. However, Detroit’s HIN incorporates the severity of injury sustained in the crashes and not just frequency, so we need to include the severity of resultant injury to know if these areas are dangerous beyond prevalence.
With locations of pedestrian- and cyclist-involved crashes and filtering crashes by injury severity, we can put together a high injury network specifically for cyclists, pedestrians, and the combination of both. We used the Safer Streets Priority Finder that takes crash and road data and analyzes ½-mile segments to find areas that are most likely to be dangerous. With predictive modeling, it scores 1/10 mile increments by weighing the most severe crashes higher, giving us corridors where pedestrian- and cyclist-involved crashes resulting in fatalities and incapacitating injuries are more likely.
On this map, blue represents high injury corridors for cyclists, gold represents high injury corridors for pedestrians, and the dark brownish-green represents high injury corridors for both cyclists and pedestrians. There are certainly some commonalities to the overall Detroit overall HIN, but on this network some discrepancies appear. Morang in northeast Detroit is uniquely dangerous for cyclists and pedestrians but not drivers, similar for Cass avenue in Midtown. While there are high counts of crashes on Grand River, smaller areas than the overall HIN are dangerous particularly for cyclists. The westside is well represented on both high injury networks, with overlap and corridors unique to each.
So far we’ve been able to locate where reported crashes are happening, and areas where those crashes result in severe injury, but we don’t yet know things like the volume of vehicle traffic these areas have, the volume of pedestrians and cyclists, or what their destinations might be.
There are sources for cycling volume that, while also incomplete, can be used to draw inferences. Strava is a GPS-based route tracker, and through their Strava Metroview service we were able to produce this map of anonymized, cyclist-recorded volume in 2021 by cyclists. While this map is limited to cyclists using this particular service, according to Strava’s research, when volume counts have been done their maps are usually accurate as representative of relative volume. On this map, areas with higher volume are darker blue with thicker lines, while lighter blue, thinner lines have the lowest recorded volume other than streets that have no recorded volume. For instance, a section of Belle Isle near Sunset Point has 11,715 cycling trips recorded using Strava. In addition to the volume of these recorded cycling trips, this map is overlaid with 2017-2021 cyclist-involved crash locations. Unfortunately, since the road network used for the crash locations is different from the road network that Strava uses, an analysis that compares the number of cyclist crashes with cyclist trip volume is beyond the effort we can put into this blog post. But a simple visual investigation shows that compared to downtown and midtown where crashes are happening on high-Strava-usage routes, much of the westside has many crashes showing up on routes with lower usage. Gratiot in particular has low-Strava-usage but also has cyclist-involved crashes all along its path through the city, perhaps indicating a relatively high crash rate to volume of cycling activity. Belle Isle, Palmer Park, and near the waterfront connecting to the Dequindre Cut have high volumes of cyclists but few crashes.

The pedestrian data from Strava is far more limited than the cyclist data, and while Southeast Michigan Council of Governments (SEMCOG) does take some counts of pedestrian and cyclist volume, the information is so limited as to not be useful in this analysis.
On this map we compare the locations of pedestrian- and cyclist-involved crashes with daily vehicle traffic volume. Roads with higher vehicle traffic volume would provide more opportunities for pedestrian- and cyclist-involved crashes. SEMCOG Annual Average Daily Traffic (AADT) volume is “collected and provided to SEMCOG by county road commissions, local communities in Southeast Michigan, the Michigan Department of Transportation, and by consultants specializing in traffic data collection.” The green lines on this map are thicker where there is a higher AADT, and this dataset excludes AADT for highways which otherwise have the highest AADT. Analyzing areas of high volume overlap, we found that 52% of pedestrian- and cyclist-involved crashes are happening where there is a traffic volume of 10,000 or more. Although we are able to analyze these areas of volumes for patterns, further analysis is difficult due to limitations with the data. SEMCOG data does not contain traffic counts for every street in Detroit, nor are traffic counts completed every year, so this AADT map does not reflect how traffic volume has shifted during the 2017-2021 range of the pedestrian- and cyclist-involved crash data. Additionally, some of the AADT measurements are from years prior to the pedestrian- and cyclist-involved crash data.
As the city of Detroit expands bike lanes and greenways, we can overlay the cycling infrastructure map with pedestrian- and cyclist-involved crashes to investigate whether the infrastructure increases safety.
Reminder, you can turn layers on and off using the icon in the bottom right corner
On this map, the blue lines are locations of bike lanes, green lines are greenways, gold lines are bike routes (which, according to the SEMCOG definition used by the City of Detroit, are ”bicycle friendly roadways which may not have bicycle/pedestrian facilities or official hike-bike routes; but, due to a combination of low traffic volumes and low posted speed, are none-the-less more comfortable to travel on”), and black lines are “other,” which is not clearly defined by the source. Comparing crash data to cycling infrastructure, we learn that from 2017 to 2021, 78% of the cyclist-involved crashes happened where there is currently no cycling infrastructure. With less than a quarter of all cyclist-involved crashes happening where there is cycling infrastructure, it is possible that this indicates that the infrastructure in these areas has a positive impact on cyclist safety, but the conclusions we can confidently draw from these data are limited. Since the city is continually expanding its cycling infrastructure and the data does not include implementation dates, we don’t know if all of this infrastructure was in place when these crashes occurred. Ideally we would be able to incorporate an analysis with the Strava cycling volume, but we are again limited in the scope of this blog post by the different road networks used for the data sources.
The 15-Minute City
In 2016, City of Detroit mayor Mike Duggan proposed a vision for “20 minute neighborhoods,” meaning neighborhoods where one can access whatever one needs in a 20 minute walk or bike ride. Internationally, this movement had begun with the 15 Minute City, typically in cities built before automobiles and car-oriented design, unlike Detroit which would need to develop in new patterns to facilitate 20 minute access to places of interest (POI). While the city administration has been quiet about this priority since the announcement in 2016, it has continued developing pedestrian- and cyclist-oriented infrastructure, as the Streets for People action plan demonstrates. CityAccessMap has provided global heat maps that combine pedestrian infrastructure and places of interest sourced from OpenStreetMap, with population density to show hotspots of where POI are likely within a 15 minute walk or bike ride for more people.
When we overlay pedestrian- and cyclist-involved crashes on this 15-minute city heat map created by CityAccessMap, we can see that the dots for crashes align with the 15-minute-walk accessible hotspots in yellow. Particularly, downtown and midtown, southwest, northwest, and northeast have aligned hotspots; and, more generally, the lengths of Woodward and Gratiot are hotspots for both. There are, however, crash hotspots in the near westside that don’t show up in the 15-minute POI hotspots (particularly bounded by Tireman on the south, Livernois on the east, Greenfield to the west, and McNichols to the north), and Morang on the eastside shows up again as an area with a concentration of crashes with few points of interest.
The CityAccessMap is not a map layer that can be incorporated into our ArcGIS with interactivity, so we can’t analyze the connection between the hotspots and crashes other than observations. Additionally, this map only indicates where the likely hotspots are, not whether they actually have higher volumes of pedestrians and cyclists visiting those areas, so we can’t determine whether these areas are safer or more dangerous.
In Conclusion
While each of these data sources are individually incomplete, in combination we start to build a picture of where the greatest risks are to pedestrians and cyclists, and what is the existing infrastructure at those locations. Hopefully, all the stakeholders with an interest in making Detroit a safely multi-modal city will focus their future data collection efforts on the specific usage of Detroit’s transportation infrastructure by pedestrians and cyclists, so we can more easily make data driven conclusions about what successes and failures we’ve had along the way.
So far, these data have only represented crash locations and injury severity mapped to the geography of road infrastructure that remains mostly the same. In part two, we will dig into the characteristics of the crashes, who is involved, when they happen, and what types of injuries result.
Read more in Part 2: Facts and Figures!