Accessibility analysis for more informed transport planning
Some local findings
Whichever mode you prefer, the primary goal of a transportation system is to let people get from Point A to Point B. So what if we took this approach to measuring our transportation system and evaluating different projects? In Bend, Oregon, we used accessibility indicators to help us understand how well the current transport network enables people to drive, take transit, bike, and walk. We are using this to evaluate different future scenarios as part of a long-term planning process — here’s how it’s already helped inform our process.
Transportation planning needs a holistic approach
Building a transportation plan usually involves comparing different future scenarios: we start by considering the “big picture” needs across the whole city, develop scenarios that include different projects to address the needs, and then test the scenarios to see how well they perform against a set of criteria.
No matter where you are, this process is often very car-centric. We do need to focus on cars — it’s how most people get around. But without tools to consider other modes, we’re missing part of of the picture. In Oregon, recommended evaluation criteria include mobility indicators (e.g. congestion, travel time reliability, trip delay) and travel behaviour indicators (vehicle miles traveled, mode split, transit ridership). Most of these criteria are modeled using a travel demand model (TDM), which predicts what travel patterns will look like in the future. TDM can be really useful, especially for making comparisons. There’s a lot we can glean from this modeling and a lot of work has gone into fine-tuning the process. But there are two big limitations with this approach:
- TDM predicts what people will do, which is especially uncertain when considering 20-year timeframes. Travel patterns and modes are changing because of technology; with new mobility solutions and autonomous vehicles on the horizon, it’s especially difficult to predict what mobility will look like many years from now and account for this change in a model.
- TDM focuses on cars and congestion — it is not (usually) designed to forecast pedestrian or bicycle trips (see comment from E-LO at the bottom of this post). A TDM can forecast a few metrics for mode shift, transit ridership, etc, but these are especially uncertain (there just aren’t great data for predicting shifts to cycling or walking; it’s a hard thing to forecast).
Without more robust tools, non-car modes play a much more limited role in regional scenario planning, and they are often been measured with basic metrics that give a very limited view of the system.
For example, the bike network is sometimes measured by miles of bike lanes. This is a start, but it doesn’t consider where those lanes are, whether they form a connected network, or whether they enable people to reach their destinations.
Transit quality is often assessed by the number of people living within a quarter-mile of bus routes. Again, this tells part of the story but it doesn’t consider whether the bus system is a viable mode — Where do buses go? How often do they run? How long would it take to get from A to B? These questions have traditionally been difficult to answer, but they are key to understanding whether transit serves people’s needs. Without this information, it’s difficult to understand current conditions and assess the impacts of potential changes. And without being able to quantify the benefits of a project, it’s also more difficult to make informed decisions regarding funding.
Accessibility analysis
To develop a more comprehensive understanding of their transportation systems, some cities have begun to consider accessibility as an additional lens. Accessibility measures what people can reach in a given timeframe, such as jobs, schools, parks, or other destinations. This is based on the built environment (roads, paths, etc.) and transit schedules. Rather than trying to predict what people will do, accessibility measures what they could do. For example, this answers questions like, “How many jobs can the average resident reach in 45 minutes? 60 minutes?”. This helps to understand the current transportation systems, and can be used to compare the impacts of different potential projects and changes.
Outputs can be summarized into an overall metric for the area and are often visualized as maps showing colour bands (“isochrones”) of what people can reach in a given timeframe. Jarrett Walker calls this “a map of your freedom”. Here’s an example for New York City area:
Having visualizations and metrics for access gives an additional perspective for considering connectivity, especially for biking, walking, and transit. This can help to understand the current state of a transportation system, and to quantify what sort of improvement you could achieve with a particular investment.
This can be used in combination with travel demand modeling — the two approaches can complement one another and strengthen overall findings.
How we’re using it in Bend
As part of its long-term transportation planning process, the City of Bend in Oregon used employment accessibility metrics as part of its regional scenario planning (which also included TDM). I evaluated a series of tools that could be used and determined that Conveyal’s Analysis tool would best fit our particular needs. Some of our scenarios weren’t straightforward to model with this tool because they involved new infrastructure, so I developed and documented a workflow to prep our data.
I modeled four future scenarios for 2040, which included detailed population and employment forecasts. Scenario maps are here, and can be summarized as:
- Baseline scenario: Includes current transportation network, plus some new roadways that are under construction or expected to be built in the near-term
- Scenario A: Builds new corridors (to increase connectivity)
- Scenario B: Widens and enhances existing corridors (to ease congestion)
- Scenario C: Maximizes the existing transportation system (includes enhanced transit, does not include any new or wider roads)
For each of the scenarios, I modeled access to jobs within 30 minutes (for the 50th percentile resident) — Bend is a small city so this seemed like a more realistic time limit than 45 or 60 minutes. I did this for cars, transit, pedestrians, and bicyclists, and also included a few variations. Here’s a summary of the results:
This tells us a couple of things right away:
- Access for cars is very good. Regardless of how the road network may change, the average person could reach all jobs in Bend in 30 minutes or less. (Note that this doesn’t account for potential congestion — that’s something that’s part of travel demand modeling and doesn’t factor into accessibility analysis.)
- Access for pedestrians is fairly low and doesn’t change between scenarios. This is because Bend isn’t very dense and people can only walk so far in half an hour, so there’s a geographic limit to what they would be able to reach. This isn’t something that would change based on the transportation network — it’s a function of land use and density.
The more interesting modes were transit and bicycle.
What we learned about transit
Results suggest that it’s possible to commute by bus to many jobs, but it takes significantly longer than other modes.
Transit allows the average future resident to reach only 5% jobs in 30 minutes; this increases when we model higher-frequency routes and local mobility hubs (Scenario C), but it’s still very low. This is largely because 30 mins doesn’t allow much time to walk to a bus stop, catch the bus, and walk to a final destination. As with pedestrian access, this is likely a function of land use; Bend is not very dense.
With this in mind, I also modeled access to jobs in 60 minutes, in order to understand how the local bus network works as a whole. This was particularly important because cross-city bus travel often requires transferring through a central station, and I wanted to understand how routes operated in relation to one another. In an hour, the average future resident could reach 40% of jobs, or 67% with transit enhancements (Scenario C).
This suggests that transit would probably not be a choice mode, even with the potential new enhancements. No matter how many people live near a bus route, they just can’t access destinations as easily by this mode. If residents could drive to most jobs in as little as 15 minutes, but it takes up to 60 mins (and probably a transfer) to get there by bus, then it’s easy to guess what most people will do if they have the ability to choose. Understanding which groups of people are more likely to use transit, and why, can help inform how to plan this service to suit their needs.
What we learned about bikes
Traffic stress really matters
Not all roads are created equal for bicyclists; riding in mixed traffic on a busy arterial is very different from riding on a quiet side street or a protected bike facility. These differences can be quantified in a standard approach using the Level of Traffic Stress, which categorizes roads from Level 1 (low stress) to Level 4 (high stress) based on their physical infrastructure characteristics and posted speed limits. If a road feels too stressful for bicyclists, data suggests that they won’t use that road; people on bikes are more sensitive to safety and comfort than users of any other mode.
Since our modeling focused on understanding the transportation system for the average person, I modeled the bike network so that it only included LTS 1 and LTS 2 roadways. The majority of people (75–85% of adults and most children over 10 years) would be comfortable on these facilities.
Bend contains many high-stress corridors that can make it difficult to cross the city, which can cause neighbourhoods to function as “islands” that can’t connect to one another. This makes it really important to be able to measure citywide connectivity and access.
To understand how high-stress roads limit connectivity, I also modeled what access to jobs would look like if the whole road network (except major highways) were usable for cyclists. Here’s a snapshot of just the bike results:
These results indicate that citywide bike connectivity isn’t very good. High-stress corridors do act as barriers to people on bikes. In every scenario I modeled, the average future resident could access about twice as many jobs if the whole road network felt safe and comfortable. High-stress roadways are more than just orange lines on a map; they place significant limits on how people can get around the city.
Modeling both types of network enabled us to better understand what was happening in our future scenarios, and why. When I modeled only low-stress facilities, the results seemed to indicate that widening roads would significantly increase access to jobs (from 29% to 41%) , but that building new connections would not have a big impact (from 29% to only 31%). This is because Scenario B includes major road widening projects that are assumed to include low-stress bike facilities, so they make significant improvements over the current (high-stress) conditions. This is correlation, not causation. When I model the full network and see how the different scenarios perform, we see that widening a roadway has no inherent impact on what bikes can access (results go from 64% in the baseline to 64% for Scenario B). We also see that, if the whole network were already low-stress, then the new connections really would have an impact; access to jobs increases from 64% to 77%.
We also learned that the new connections (in Scenario A) were often needed, but were underperforming because of the surrounding connectivity issues. When I modeled only low-stress roadways, I noticed that big infrastructure projects in Scenario A didn’t seem to have a big impact on access (results went from 29% in the baseline to 31% in Scenario A). This was counter-intuitive since those projects were intended to increase bike and pedestrian connectivity. This prompted us to take closer look at the projects to see why they were performing this way. The answer: the projects didn’t address the broader network. Here’s an example of two projects that try to increase connectivity to get bikes across two state highways and into a shopping district:
As a result, the staff team examined each bike/ped project in context to identify where modifications would be necessary in order for the projects to achieve their intent. We identified additional segments of roads that would need to be made low-stress as part of certain projects, and which projects should be done as a group in order to maximize their benefits. We took these results to our public committee and a regional advisory committee to get their approval to consider the changes.
Bike connectivity is a regional need
Analyzing access changed our understanding of the bike network as a regional need. Bike connectivity and barriers to bicycling are often part of a later, neighborhood needs component of a transportation plan. Being able to quantify how high-stress routes were limiting connectivity citywide meant that bike connectivity could be considered as a regional, “big picture” need.
City staff and consultants had already been developing the concept of a “low-stress network” of future trunk routes for bicyclists. This would have been discussed at a later stage of the planning process. However, quantifying the citywide need for bike connectivity led us to consider these needs sooner.
City staff pored over the future low-stress network and the scenario projects to understand where the key east-west and north-south routes were, where there were gaps, and which gaps would be the highest priority needs for developing citywide connectivity.
Through this process, we identified four additional bike connectivity projects that appear regionally-significant, but which were not initially included in our scenarios. We took these results to our public committee and a regional advisory committee to get their approval to consider the new projects. This enables the new projects to be considered for inclusion as part of the backbone of the city’s network.
Takeaways and next steps
Overall, accessibility indicators gave us a tool to better understand transit and bicycling in a much more comprehensive way. By factoring in land use, road infrastructure, traffic stress, and transit schedules, we were able to understand how systems worked as a whole and how they may enable residents to reach potential destinations. This enriched our understanding of the transit system and of the barriers to connectivity for bikes. As a result, we worked with our committees to modify certain potential projects so that they would perform better, and to include regionally significant bike connectivity projects that would otherwise have emerged as “big picture” needs.
Moving forward, we’ve started to consider how we can use an accessibility approach to understand access to schools across the city, based on the presence of sidewalks and low-stress bike routes. This could help us to identify barriers, quantify their impacts, consider how this relates to equity, and identify high-priority needs to help ensure that kids can safely get to school.
On the technical side, next steps involve modifying our workflow so that data inputs can be produced systematically, with less manual manipulation.
Another thing I’d love to explore: optimization. Let’s say we have a given amount of funding and we want to improve access as efficiently as possible. How can we get the highest results for that investment? Being able to run permutations of potential projects and compare results to costs would help inform decision-making.
