Reconnecting Kendall Square(s)

MIT • 11.324 - Modeling Pedestrian Activity in Cities
In collaboration with Erin Chen, Noel Loo, Takumi Kitamura, Deena Darby, and Anna Savino

Background

Our assignment was to “improve pedestrian connectivity and legibility between the disparate parts of the greater Kendall Square area” of Cambridge, Massachusetts.

Planning/Design Proposal

Our proposal consisted of two types of pedestrian route interventions:

Commercial Corridors: vibrant corridors that accommodate multiple modes of mobility and foster active urban life through diverse flows of people and activities. We included roads in our plan based on their composition of ground-level retail. Our envisioned interventions for Commercial Corridors are uniformly-implemented (1) ground floor activation, (2) bike lanes, and (3) sidewalk amenities such as seating, lighting, and greenery.

Green Spines: pedestrian-first, quiet routes that connect existing open spaces, offering a safe and calm route into the heart of the city. The goal was to create a network of parks and paths, integrating existing green spaces that are currently disconnected from each other.

Our envisioned interventions for Green Spines differs based on the nearby urban typology: residential, school, or workplace. In residential areas, a thoughtfully-designed “back-alley” route provides a safe alternative for pedestrians to navigate their neighborhood; along existing streets, this means converting to a Woonerf-style shared street. In school areas, child-friendly activities and educational opportunities are incorporated to create spaces for kids to enjoy and parents to relax around school hours. In workplace areas, a strategic use of street furniture within office districts provides a midday respite for workers. Across all three typologies, safe crossings and clear wayfinding would be carefully designed.

These two types of interventions, Commercial Corridors and Green Spines, combine to form our final plan.

Impact Analyses Using Pedestrian Trip Modeling

Pedestrian flows were modeled using Madina, a Python tool developed by the City Form Lab at MIT.

Analysis 1: Commercial Corridors align with road segments with higher modeled pedestrian usage

Analysis 2: Complete Green Spine implementation would double the number of amenities accessible by the network

Pushing the Boundries of Pedestrian Modeling

We conducted two additional analyses for this assignment, but after looking back on the results, I am reluctant to share them publically. Nonetheless, pedestrian modeling is interesting and underutilized, so it’s worth sharing what I learned.

Both analyses attemped to predict how pedestrian trip generation would change if our plan was implemented. This analysis is built upon the belief that if walking conditions are improved along key corridors and destinations are added, trips will be both induced and re-routed. The methodology, within Madina, is to shorten the “percieved length” of each inteventions’ sidewalk segments (to account for walkability enhancements), re-run the betweenness simulation from Analysis 1, and visualize the difference.

However, there are multiple data quality concerns that cast doubt upon the accuracy of any numerical result achieved by this type of comparative modeling (i.e. drawing conclusions on anything other than the macro-scale, as done in Analysis 1):

• First, you have to trust the accuracy of the baseline model, especially the pedestrian counts that act as training data. The baseline model provided to us, with an R-squared of 0.59, was only trained on 2-hour counts conducted in 2018 at 62 locations around Kendall Square. Multiple data sources can used as the “features” - in our case, it was simulated pedestrian flows between common origins and destination types, though it could also be a travel survey - but accuracy is most dependent on how robust the training data is.
• Second, you have to trust the methodology used to adjust perceived lengths based on changes to attibutes such as sidewalk width, traffic speeds, availability of street commerce, shade, and so on. We tried to base our methodology on research on the impact of street attributes on willingness-to-walk, but the numbers (i.e. x ft. of sidewalk width decreases percieved length by y%) vary widely between studies that are conducted across very different urban contexts.
• Third, you have to embed assumptions when setting the parameters for elastic origin weights - for example, how far your simulated pedestrians are willing to walk.
• Finally, one must strike a balance between simplicity and complexity when trying to isolate the impact of proposed changes. You can choose to only adjust the percieved weights for a single corridor/project (which risks overplaying the impact of induced trips if other streetscape improvements are, in fact, implemented) or for the entire pedestrian network (which risks producting results that are difficult to interpret).

These sources of error compound in a way that makes it difficult to produce error bounds, which should be included with any good forecast.