On June 15, the National Association of Transportation Officials and BBSP kicked off the third annual BBSP Bike Share and Cities for Cycling Roundtable. This year, the roundtable welcomes the host cities of Baltimore, MD; Alexandria and Arlington, VA; and Washington DC
Due to COVID-19, things will look a little different this year — the roundtable has gone virtual! While it will be a shift from previous years, 2020’s two-week virtual event will still create plenty of space for city practitioners to “connect for a peer learning opportunity to discuss the changing landscape of shared micromobility, and explore how increased ridership might impact bike lane design and bike infrastructure,” writes NACTO.
Between the panels, interactive workshops and design exercises, participants will walk away with “new tools and implementation techniques to build better bike networks, as well as brainstorm process changes and challenges for managing and regulating shared micromobility.”
It’s important to note that the roundtable is only open to those working directly for cities, with the exception of two public events that welcomes everyone. But to keep you in the loop, BBSP will be regularly sharing mini-blogs about the daily events and what they entailed — we’ve covered day one and day two, so let’s continue with day 3!
Day 3, June 17 – Using Analytics for Mobility Planning
- Michael Fichman, lecturer, Master of Urban Spatial Analytics Program (MUSA) – University of Pennsylvania
- Ken Stief, Ph.D., Director of MUSA Program – University of Pennsylvania
- Eugene Chong, Dual MUSA/Planning student – University of Pennsylvania
- Ophelia Liu, Recent graduate – University of Pennsylvania
Workshop Objective: Interested in how open analytics can be used to guide mobility planning? Instructors and recent graduates of the University of Pennsylvania’s Master of Urban Spatial Analytics Program lead an interactive workshop showcasing some innovative data science approaches to mobility planning and allowed participants to design their very own dream mobility planning dashboard.
“We just feel like, as data science and technology become more requisite for governments, there are more opportunities for communities to come together to develop these projects and make them sustainable on their own.” – Ken Stief
The big question for day 3 was how could planners better collect, analyze and use data to make planning decisions? The MUSA program at the University of Pennsylvania teaches the intersection of data science and public policy; this year, its students participated in a semester-long MUSA Practicum, which is an applied data science project where the students worked with actual governments to analyze scooter equity and demand in several American cities.
Explained by Eugene Chong and Ophelia Liu, the students’ Scooter Planning Toolkit presentation consisted of an introduction to scooter share system in the United States as well as the wireframe app they created for this project, their prediction of equity analysis, and finally, a deep dive into the rebalancing data and compliance within their selected cities.
- Presently, there are 340 scooter-share programs operating in 242 municipal areas and campuses across 40 states and Washington DC.
- While some cities, like Louiseville, have instituted requirements for equitable distribution, not all cities have done so — and many companies aren’t complying with these distribution requirements anyway.
- The students conceptualized their project as a municipal scooter planning toolkit to help cities that are interested in launching scooter share systems, or in adding some additional oversight, learn from other municipalities that have already put those systems in place.
Data analysis snippet:
- The students aggregated the total number of scooter trips taken between July and September of 2019 in each census tract.
- Their maps showed trip outflows at the census tract level, confirming that scooter trips tend to concentrate in the center of inner-city areas.
- Their training data came from six cities that had good experiences with scooters: Austin, Chicago, Washington DC, Kansas City, Louisville and Minneapolis.
- To build their machine learning model, they tried four different model frameworks, including linear regression, generalized linear model, extreme boost and random forest. Additionally, they conducted cross-validation and model tuning to find out which model performed the best — the random forest model was the winner.
- Next, they selected 10 pilot cities with no scooter experiences and applied their model to predict scooter demand in those cities: Asheville, Jacksonville, Hartford, Houston, Jersey City, Madison, Omaha, San Antonia, Syracuse and Philadelphia.
Equity of access analysis snippet:
- Actual and predicted ridership were concentrated in downtowns and near universities, which reflects company decisions and deployment strategies
- People living in disadvantaged areas have or will have insufficient access to scooters
- They created an equity score — or a metric to measure the equity of accessibility — by comparing income, percentage of white population and age across areas with high and low ridership
- They observed which cities tend to have lower equity scores: Louisville, Washington DC, Austin, Minneapolis, Kansas City and Chicago
Louisville Rebalancing Compliance snippet:
- The students’ work shows how to convert vehicle flow into data audits for equitable distribution
- Compliance was low — during most audits, not enough of the vehicle fleets were within the redistribution zones
- The caveat, however, was that the students weren’t certain when the distribution requirements took effect for each company
- In the end, they learned more monitoring and enforcement are needed
“The purpose of the workshop is to get participants to collaborate and design their ideal micro-mobility planning dashboard with particular emphasis on across-agency use cases/data,” said Stief. During this time, the roundtable was led into three breakout sessions, where participants had discussions about the kinds of operational and planning problems related to micromobility that could be addressed and solved with data.
Here are examples of some use cases that would require cross-agency data:
- Predicting where there are broken vehicles in scooter fleets
- Demand predictions for coming up with site selection criteria for streetscape improvements and corrals for scooters based on demands of the width of sidewalks in a given area
- Identifying opportunities for mode shift to bike share
- Dashboard: Mode Shift Utility Predictor, which allows you to zoom into an area to see the predictions. You can see the characteristics of the area as well to make sure the populations who need to be served are being served.
Question for you: If you could think of a metric to generate in order to prioritize locations in the probability of mode shift, what would that look like?
The Better Bike Share Partnership is a JPB Foundation-funded collaboration between the City of Philadelphia, the Bicycle Coalition of Greater Philadelphia, the National Association of City Transportation Officials (NACTO) and the PeopleForBikes Foundation to build equitable and replicable bike share systems. Follow us on Facebook, Twitter and Instagram or sign up for our weekly newsletter. Story tip? Write firstname.lastname@example.org.