Conveyal
Published in

Conveyal

Modeling Bicycle Comfort with Conveyal Analysis (Part 1)

[ Editor’s note: This guest post is authored by Juliet Eldred, who recently joined Trillium Solutions as a Project Manager. It is the first part in a series on street network modifications and other new Conveyal features. ]

Recent and upcoming Conveyal updates make it easier to account for Bicycle Level of Traffic Stress (LTS) when creating isochrones and analyzing access to opportunity. In this post, we review the LTS concept, explain Conveyal’s initial LTS labeling method, discuss options for improved labeling, and demonstrate some of these features with a case study of access to grocery stores in Washington, DC.

What is Level of Traffic Stress (LTS)?

The Level of Traffic Stress (LTS) metric, first proposed in the 2012 paper “Low-Stress Bicycling and Network Connectivity,” categorizes road segments and networks by their relative stress levels for people riding bicycles. LTS classifications are determined using the following criteria:

  1. The total number of traffic lanes
  2. The posted speed limit
  3. The presence and type of any existing bike infrastructure (such as separated bike lanes, painted bike lanes, sharrows), and if applicable, the proximity of bike infrastructure to parking

Through these criteria, roadways can be classified as one of four levels:

LTS 1: Tolerable for children. This includes low-speed, low-volume streets, as well as those with separated bicycle facilities (such as parking-protected lanes or cycle tracks).

Example of LTS 1: parking-protected Cycle Track on 15th Street NW in Dupont Circle, Washington DC (photo by author).

LTS 2: Tolerable for the mainstream adult population. This includes streets where cyclists have dedicated lanes and only have to interact with traffic at formal crossing.

LTS 3: Tolerable for “enthused and confident” cyclists. This includes streets which may involve close proximity to moderate- or high-speed vehicular traffic.

LTS 4: Tolerable for only “strong and fearless” cyclists. This includes streets where cyclists are required to mix with moderate- to high-speed vehicular traffic.

A wide urban street with traffic moving in both directions
Example of LTS 4: 16th Street NW in Dupont Circle, Washington, DC (photo by author).

Using LTS metrics to assess a larger street network can help planners understand how useful it is for a variety of types of cyclists. While confident and experienced riders who are comfortable on most urban streets are able to utilize the majority of the network to get where they need to go, higher-stress streets may restrict the route choices and destinations of newer or more cautious. A large, connected network of low-stress streets is a key way to expand accessibility for potential riders, beyond just the “strong and fearless,” and understanding the extent to which a current network meets (or fails to meet) this goal is important for planning future infrastructure that can meet a wider array of cyclists’ needs and comfort levels.

Washington, DC’s Bicycle Infrastructure

Washington, DC has one of the highest rates of bike-commuting in the country. The city received a ‘Gold’ ranking for bike-friendly cities from the League of American Bicyclists in 2018, and as of that year, approximately five percent of residents biked to work. While utilization of Capital Bikeshare, DC’s bikeshare system, dropped 56% since the COVID-19 stay-at-home orders began in March 2020, it has steadily begun to increase since then, and overall interest in bicycling has risen substantially: local bike shop owners have seen major increases in sales, as more people desire bikes for both transportation and recreation. With an influx of new riders, many of whom are likely to be on the more cautious side, LTS metrics can help planners understand access for these new riders.

Washington, DC’s Department of Transportation (DDOT) has built a total of 89 miles of bike lanes since 2001. In 2009, DDOT also began installing protected bike lanes (PBLs), also referred to as Cycle Tracks or separated bike lanes, and has since constructed 12 miles of these facilities. Protected bike lanes and cycle tracks are generally defined as LTS 1, but painted bike lanes can vary widely in terms of LTS levels, depending on where and how they are implemented.

Beginning in 2020, DDOT is embarking on a plan to build over 20 miles of protected bike lanes over the next three years, more than doubling the existing network of protected bike infrastructure. Given that protected bike infrastructure typically has an LTS 1 rating, these projects will increase the District’s low-stress bicycle network, thus making bicycling a safer and less intimidating transportation option for a larger population of potential riders.

Conveyal Estimated LTS

While LTS limits have been available as an experimental feature in Conveyal for a while, recent updates have made them easier to apply in the web-based user interface.

In 2015, the Conveyal team started working on a way to estimate LTS from commonly tagged attributes of OSM ways. Conveyal Estimated LTS uses the tags commonly available in OpenStreetMap road networks, which are uploaded when a Conveyal user creates a Network Bundle. The initial method may miss certain nuances of bike connectivity, but it does facilitate LTS analysis in the absence of specialized data. If a user has detailed LTS data available, they can be applied to an OpenStreetMap network, with the custom “lts=” tag, before uploading as a Network bundle in order to override the default Conveyal Estimated LTS labeling.

The sequential logic for assigning default Conveyal Estimated LTS values in v5.10.0 of the routing engine can be viewed in its source code and is summarized below:

Does not allow cars: LTS 1
Is a service road: Unknown LTS
Is residential or living street: LTS 1
Has 3 or fewer lanes and max speed 25 mph or less: LTS 2
Has 3 or fewer lanes and unknown max speed: LTS 2
Is tertiary or smaller road:
Has unknown lanes and max speed 25 mph or less: LTS 2
Has bike lane: LTS 2
Otherwise: LTS 3
Is larger than tertiary road
Has bike lane: LTS 3
Otherwise: LTS 4

Then, an intersection adjustment step is applied. At each unassigned network vertex, the highest LTS of any entering/exiting street is assigned to all entering/exiting streets. Although service roads (which generally include alleys) are not initially assigned an LTS, in this adjustment step, the LTS of intersecting street segments will spill over to them. As such, the update step may lead to counterintuitive results — for example, segments of an alley (highway=service) or cycle path crossing an LTS 4 street will be assigned LTS 4, even though the path on its own would be expected to have LTS 1.

It is worth checking whether LTS values are sensible after this update step, especially in areas with many alleys, such as Washington, DC.

In this case study, the low-stress network based on Conveyal Estimated LTS ends up being more disjointed than the actual DC bike network. This is because many bikeways that are initially tagged at LTS 1, such as the 15th Street NW Cycle Track, cross higher-stress streets and are accordingly re-tagged with higher LTS in the intersection adjustment step. For example, take the intersection of 15th Street NW and U Street NW in the map below. While the initial step tags the Cycle Track as LTS 1, the intersection is tagged as LTS 4, which spills over into the cycle track in the intersection adjustment step.

By contrast, sometimes OpenStreetMap users draw in crosswalks at intersections. This means that through the Conveyal tagging process, these crosswalks can “protect” the LTS rating of a PBL, such that the rating of an intersecting street won’t spill over to the PBL. At the intersection of 15th St and V St NW, just one block north of the previous intersection, an OSM contributor drew in the crosswalks, which “protect” the LTS rating of the PBL, but give the section of the PBL that goes through the intersection the same rating as V Street (LTS 2).

A map showing the LTS tags of the 15th street cycle track at the intersections of U and V streets.
15th Street Cycle Track tags at U Street and V Street intersections. Light Blue = LTS 1, Dark Blue = LTS 2, and Red = LTS 4. Note how the Cycle Track is properly tagged as LTS 1 north of V Street because its rating was “protected” by the crosswalk drawn in at 15th and V (Map made by author in QGIS, using OSM basemap).

As multimodal options continue to be refined in the R5 routing engine, the Conveyal team is outlining various approaches that would improve on this initial approach to LTS, including better labeling and logic around vertices, as well as cycling elevation profiles. In the meantime, there are three ways to adjust LTS so that it better matches on-the-ground conditions:

  1. You can use an offline script to edit your street network from OSM before uploading it to Conveyal, which can match existing cycling infrastructure and apply explicit LTS tags that override the default steps. We do not use this method in this post, but it can be done with tools like this one for LADOT.
  2. Make more extensive use of the crosswalk “protection” quirk by adding crosswalks to intersections in OpenStreetMap. We expect that crosswalks will be more extensively represented in OSM over time, especially as AI is increasingly used for OSM updates.
  3. Use Conveyal’s Modify Streets feature to make quick, ad-hoc adjustments. Draw a polygon around the street(s) whose LTS rating you’d like to adjust, and correct it accordingly. This feature can also be used to specify infrastructure such as pop-up bike lanes or DDOT’s plan to double protected bike infrastructure.
Use Conveyal’s Modify Streets function to alter the LTS tag of any street in your network

The map below shows all streets tagged as LTS 1 by Conveyal’s Estimated LTS method (excluding alleys/service roads) in green, and Washington DC’s bike network in pink. While there is some overlap, especially on streets with cycle tracks like 15th Street NW, there are many areas where there is dedicated bike infrastructure, but the LTS ratings are 2 or higher. This is because DC’s bike network includes painted lanes, which do not provide the same level of protection as PBLs, and may be adjacent to higher-speed traffic. This is important to note; areas that show up on a conventional bike-path map may not register to riders as a “low-stress network,” because many of those lanes are on streets that are higher-stress than LTS 1.

A map of Washington, DC showing the Conveyal-Tagged LTS 1 streets and DC’s bike network.
Comparing Conveyal-estimated LTS 1 streets with DC’s bike network (map by author).

LTS Analysis in Conveyal

Regardless of how LTS is labeled, when using Conveyal, users can now specify a maximum LTS tolerance in the analysis panel. The default parameter is a maximum LTS of 4, which includes all streets. Users can modify the maximum LTS level they would like the routing engine to consider by selecting different values in the drop-down menu. On any street links exceeding the selected limit, the routing engine will assume cyclists dismount and walk their bikes.

Users can also set up two analyses with different LTS maximums, by toggling off the Identical Request Settings switch. This feature allows comparing, for example, how far one can travel by bike by either solely using LTS 1, versus LTS 1–3 streets, from a defined origin point.

Setting up LTS parameters in the Conveyal UI. Select the Max Level of Traffic Stress for your first isochrone (left), then uncheck the Identical Request Settings button (right) to set a different Max LTS for your second isochrone.

From this example origin in Dupont Circle, the substantial red area extending beyond the purple overlap with LTS 1 access indicates that braver and more experienced cyclists can travel much farther than those who stick to lower-stress routes.

Isochrones showing areas accessible by bike via LTS 1 streets (blue) and LTS 1–3 streets (red) from 21st St and P St NW in Dupont Circle within 15 minutes.

For fine-grained cycling analysis, users may want higher resolution than Conveyal’s default 250 meter grid. A new experimental feature allows boosting isochrone resolution, as shown below.

Increasing the zoom level to 12 (highlighted above) increases the resolution of your isochrones.

Doing so will increase the resolution of your results, and can show particular roads and network gaps more clearly. Conveyal plans for future releases include wider support of similar higher-resolution analysis.

In order to see how access to opportunities and amenities can be affected by LTS, you can select an Opportunity Dataset. These can include datasets such as population, demographic characteristics, and amenities.

In this example, I used a Shapefile of the locations of all of DC’s Grocery Stores from Open Data DC in order to measure how LTS levels of streets affect grocery store access by bike. In the graph on the left side of the window, you can see how many grocery stores are from the Origin Point by bike within fifteen minutes, via LTS 1 streets (in blue) compared to LTS 1–3 streets (in red). From 21st St and P St NW in Dupont Circle, one can access 11 grocery stores within 15 minutes of biking on only LTS 1 streets, but can access 20 stores if using LTS 1–3 streets.

The graph on the left shows how many grocery stores are available within a 15-minute bike ride of the origin point (21st and P Streets NW) via LTS 1 (blue) versus LTS 1–3 (red) streets.

To get a snapshot of how grocery store access varies across neighborhoods, I chose eight locations — one in each of DC’s eight wards — from which to measure grocery store access within fifteen minutes of biking. The results of my findings are in the table below:

Though the example origins in each ward had at least one grocery store accessible via LTS 1 streets, grocery store access can vary widely when one is restricted to lower-stress roads. The highest concentration of grocery stores are in Wards 1 and 2, which is not too surprising, given that these Wards contain some of the densest neighborhoods in the District. However, these areas also contain the highest variability between stores accessible solely via LTS 1 streets and those accessible via LTS 1–3 streets.

Ward 6 also has a large disparity between LTS 1 and LTS 1–3 accessible stores, with 4 accessible via the former, and 12 via the latter. While access to grocery stores in Wards 7 and 8 is not affected by LTS levels, these Wards have nearly three-fourths of all of the District’s Food Deserts, which means that there is relatively low access to nearby grocery stores as a whole. Additionally, Wards 7 and 8 have very little dedicated bike infrastructure, which means that active or potential cyclists in those areas may be reluctant to take up cycling because of safety concerns.

In Summary

Conveyal’s bicycle LTS features allow users to analyze how a number of outcomes, including access to services and amenities, can vary by the stress levels of given streets.

The features described above also work for multimodal trips, in which bikes are used to access and depart from public transit — the analysis panel enables flexible combinations of access, transit, and egress modes. Multimodal trips including public transit are Conveyal’s specialty, and recent upgrades like these LTS features enable more nuanced analysis of such multimodal trips.

In the second part of this series (coming soon), we will use the more comprehensive Regional Analysis feature to evaluate grocery store access across DC’s eight wards and their various populations, and use the Add Streets tool to model how potential bike infrastructure improvements could affect access to grocery stores.

--

--

--

We enable people and institutions to make smarter decisions by applying technical expertise in open source software and open data to create improved transportation systems, more livable cities, and a better world.

Recommended from Medium

Customer Segmentation Report for Arvato Financial Services

Behind Your Searches

Learnings about handling fraud with data and machine learning

Predictions of Ames Housing Prices

Things I wish I knew before a Data Science bootcamp

Examining how Occupations affect the number of Male and Female Positions in the Workforce

Applied Data Science with Python by University of Michigan: A Course-by-Course Review

Monty Hall Paradox and Bayes’ Theorem

Get the Medium app

A button that says 'Download on the App Store', and if clicked it will lead you to the iOS App store
A button that says 'Get it on, Google Play', and if clicked it will lead you to the Google Play store
Juliet Eldred

Juliet Eldred

More from Medium

Exploring Trends in US Bee Data in GBIF

Data Abstraction and the Grivan-Newman Algorithm

Handling and Statistically Modeling the Unknowns

#2. The Bigness of Bus Data