Overview
Mapbox Movement is the world’s most comprehensive privacy-forward location dataset, collected anonymously from millions of mobile devices moving in vehicles on roads throughout the world. By applying machine learning classification and map-matching Movement data to our daily-updated global map, we produce a vehicle activity index for every drivable road segment. This traffic volumes dataset is a proxy for vehicle counts, and may be used for tasks such as transportation analysis, EV charger site selection, roadside advertising, and city planning.
In this blog post, we explore road-segment-based vehicle activity index data to analyze traffic volumes in the Detroit metro area. The granularity of our data allows us to determine, for example, that between 2019 and 2020 vehicle activity on highways around Detroit has decreased, while activity on residential roads went up. We also look at some example drill-downs to compare traffic patterns between different years on specific roads by time of day.
To confirm the validity of our analysis, we compare our results to the official Michigan Department of Transportation (MDOT) vehicle counts data for the same area. Mapbox Vehicle activity indices in more than 90% of road segments lie within 10% of the equivalent MDOT-produced statistics. Furthermore, Mapbox Movement data provides more fine-grained time-of-day insights and coverage on additional road classes.
Evaluation Area
We evaluated the 3-month period from mid-March to mid-June (Spring season) for the nine Z10 quadkey tiles shown below. This period was selected to capture the effects of COVID in 2020 and compare that with the equivalent timeframe in both 2019 and 2021. This encompasses the urban core of Detroit and its western suburbs (roads within Canada were excluded).
Analysis
Vehicle activity indexes for each road segment are calculated and visualized as line geometries on the road network (orange/red-highlighted roads have higher traffic volumes).
Traffic volumes are evaluated year-over-year to provide insight into road utilization patterns - not only how traffic trends are changing in a specific area but also more holistically. For example we examine regional patterns of road use and then further analyze by road type (motorway, primary, secondary, etc.)
Overall traffic volumes (across all classes of roads) dropped in 2020 as a result of COVID safety precautions. However, earlier this year we measured vehicle activity levels surging back to pre-COVID levels seen during 2019.
Freeway traffic dropped significantly in 2020 from 2019. Despite Spring 2021 showing a 50% increase in traffic volume year-over-year on Detroit’s freeways, freeway traffic volumes have not yet returned to their pre-COVID levels. Traffic on local streets and non-highway arterials accounted for the majority of the increased traffic volumes. This could be indicative of increased local travel behavior in conjunction with a continued pattern of suppressed freeway volumes relative to 2019.
Traffic on local streets and non-highway arterials accounted for the majority of the increased traffic volumes. This could be indicative of increased local travel behavior in conjunction with a continued pattern of suppressed freeway volumes relative to 2019.
When comparing one year to another, we drill down to road segments and explore the distribution of percentage change, year over year, between segments. Examining these results as a histogram, we see the leftward shift of most road segments, representing a decrease in vehicle activity across all road segments in Spring 2020 compared to 2019.
Comparing 2020 and 2021 in a similar fashion reveals a wider spread of relative increases in traffic. This highlights the slower resurgence of freeway traffic compared to the very strong growth in local roads’ traffic volumes.
Comparing 2019 to 2021, we observe the bifurcation of the distribution of road segments: a subset of local streets exhibit increasing traffic while the majority of freeway road segments show negative growth. We map those patterns of relative increase and decrease back onto our map of road segments. This helps highlight where traffic patterns are shifting over time. In the map below, we observe that traffic has surged back primarily on local roads in the inner ring suburbs wrapping the downtown core.
Time of Day
We can drill down to evaluate specific corridors and understand how time-of-day utilization has changed. For an example, let's analyze the Woodward Ave corridor, between Highland Park and Downtown.
Woodward Ave. is a major state trunkline connecting numerous neighborhoods between the downtown core and adjacent suburbs, as well as the key road dividing east and west sides of the city. This corridor receives significant traffic associated with both commercial retail and office activity. Using Mapbox segment volumes data, we analyze traffic volumes for a specific week period and compare traffic volume patterns by hour over the past three years.
Woodward Avenue in Detroit exhibited increased overall traffic volumes between Highland Park and Downtown, but the AM and PM peaks that were characteristic of the corridor in 2019 have not yet returned. Instead, we see increased overall traffic throughout the day, with a late afternoon peak beginning to take form (but nothing close to the volume of peak traffic observed in 2019). One explanation for this pattern is that retail activities in the afternoon and evening have largely returned, while office commute-related peak hours have not.
MDOT Comparison
Michigan Department of Transportation (MDOT) provides traffic counts data available to the public. In fact, MDOT tracks and publishes the most recent segment count data they have available for most major roads in the state. Segments are updated whenever new figures are created for a given segment, so not all segments are up to date. Nonetheless, it is a valuable resource with which to compare and validate Mapbox Movement data.
In comparing annual 2019 Mapbox segment volumes data with MDOT’s road utilization counts data within downtown and adjacent suburb areas (shown below), we found that, for 90% of road segments, the indexed vehicle activity calculated by Mapbox fell within 10% of indexed total vehicle activity counts published by MDOT.
In addition to a high correlation with MDOT data, Mapbox Movement data provides greater coverage, granularity, and recency to enable analysis on traffic volumes by time of day and day of week.
Download sample data
Mapbox Movement Data is available globally. Download sample data, view the data spec, and reach out to our team at the Mapbox Movement Data page.
