Active Transportation Trips

Overview

The active transportation table contains information about fully disaggregate, individual walking and biking trips for a typical weekday and weekend day in each season. Each row of data corresponds to one trip. This dataset can be utilized to generate a number of metrics including origin destination tables and network volumes. Data is released twice a year and is currently available for 2019, 2021, 2022 and 2023. It is available for download in CSV format.

Demographic, socioeconomic and employment information about the trip taker for each individual trip can be added by joining this table with the Demographics & Employment table using the person_id and household_id fields.

Replica does not model scooter trips and does not separate out e-bike trips.

Methodology

Replica’s active transportation data relies on a number of public and private sources including census and census derivative products (e.g. Public Use Microdata Samples (PUMS), American Census Survey (ACS)), built environment data (e.g. the identification of existing physical infrastructure like bike lanes, sidewalks, and paths as well as granular land use data), and observed trip and behavior data sources like Location Based Services (LBS), Point-of-Interest (POI), and consumer transaction data.

Replica uses these sources of data to generate a synthetic population, representative of everyone in the built environment without compromising any specific person's or group's privacy. This approach allows us to account for everyone and estimate the universe of travel, while maintaining some longitudinal consistency despite changes in any individual data source, like penetration fluctuations. Our location data sets are used to understand travel sequences and routes, allowing us to build tens of millions of behavior models that are then assigned to every synthesized person. They are not used for mode determination, a practice that we have found to be inaccurate.

Replica then uses three models to assign movements to the individuals in the synthetic population, including the activity sequence model, the location choice model, and finally the mode choice model. The latter determines how each trip will be made based on the state of the transportation network, accounting for available transit options, multiple driving routes, and availability of bike lanes and pedestrian-accessible facilities. For each trip that we determine is made, for each individual (based on everyone in the country) we model several options in different modes and ascribe a probability score for each. This probability is informed by each person's primary commute mode, as indicated by the census, the access the person has to different modes, patterns from location data sources on the local activity there (i.e. including up-to-date WFH preferences), as well as the state of the built environment (e.g. how long trip alternatives will take), factoring in time-specific variables like congestion. Where available, the data is calibrated and validated against ground truth count data, sourced from both private and publicly available data sources as well as customers. At this time, Replica only includes purposeful non-motorized travel in its model. Recreational trips (such as jogging for a workout, or a walk around a neighborhood or park that starts and ends at home) are not included.

Use the link below to see Replica's full seasonal mobility model methodology:
Seasonal Mobility Model Methodology Summary (Places)
Seasonal Mobility model Methodology Extended (Places)

Schema