Trip Modes¶
Quick-Reference Summary¶
| Mode | Modeled? | Key Details |
|---|---|---|
| Private Auto | Yes | Includes rental cars; can appear for zero-vehicle households |
| Auto Passenger / Carpool | Yes | Children/teens, Census commute carpoolers |
| Taxi / TNC | Yes | Regression model (land use, POI activity, walkability, transit scores) |
| Public Transit | Yes | All GTFS-sourced modes: bus, rail, light rail, subway, ferry, gondola |
| Ferry | Yes | Modeled via GTFS; cross-check with operator ridership counts recommended |
| Biking | Yes | Purposeful trips only; no e-bike separation |
| Walking | Yes | Purposeful trips only; no recreational/looping trips |
| Freight (Medium Commercial) | Yes | FHWA classes 4-6 (14,000-26,000 lbs) |
| Freight (Heavy Commercial) | Yes | FHWA classes 7-8 (>26,000 lbs) |
| Other | Partially | Likely flights near airports; not recommended for analysis |
| Motorcycle | No | Not modeled |
| School Bus | No | Trips fall under auto passenger, private auto, or walking |
| Paratransit / Private Shuttle | No | Not modeled |
| Golf Cart / ATV / Scooter | No | Not modeled |
| E-Bike | No | Not separated from regular biking |
| Airline / Air Travel | No | Some may appear as "Other" near airports |
Primary Mode Hierarchy¶
When a trip involves multiple modes, Replica assigns a single primary mode using this ranking (highest priority first):
- Public Transit
- Driving (Private Auto) / Auto Passenger / Taxi / TNC
- Biking
- Walking
A walk-to-bus-to-rail trip is a single trip with primary mode "Public Transit." A drive-to-bike trip has primary mode "Driving." When filtering by mode in Hex or BigQuery, remember that the primary mode may mask secondary legs of multi-modal trips.
Multi-Modal Trips¶
Replica faithfully models multi-modal journeys. A single trip can contain multiple segments (e.g., a 2-mile bus ride connecting to a 10-mile rail trip). Both segments are modeled, but the trip as a whole receives one primary mode according to the hierarchy above.
Private Auto¶
Private auto trips include trips taken in rental cars. Zero-vehicle households can still generate private auto trips because ACS data shows these households reporting private auto as their commute mode -- Replica accounts for borrowed, rented, or employer-provided vehicles.
Replica does not currently provide vehicle occupancy data.
Auto Passenger / Carpool¶
Auto passenger mode is assigned based on the synthetic population's attributes and behaviors. It is applied to:
- Children and teenagers below driving age
- People who report as carpoolers or passengers in Census commute data
After mode choices are applied, the model calibrates against ground truth data to achieve correct mode splits.
Taxi / TNC (Ride-Hailing)¶
Taxi and TNC trips cannot be reliably identified from mobile location data alone (they look similar to carpools). Instead, Replica uses a regression (direct demand) model that considers:
- Land use characteristics
- Day-of-week and time-of-day POI activity
- Walkability and transit accessibility scores at origins and destinations
Public Transit¶
Replica models all transit modes available in a region where GTFS data exists. Transit sub-modes include:
- Bus
- Rail
- Light rail
- Subway
- Ferry
- Gondola
Linked vs. Unlinked Trips¶
Replica trip counts generally reflect linked trips -- a single trip from origin to destination, regardless of how many transit vehicles are boarded. A transit-to-transit transfer is part of the same trip.
Unlinked trips (individual boardings) are visible when clicking on an individual transit station or summing trips by transit sub-mode in the Summaries panel.
Transfers and Access Modes¶
A single transit trip can include multiple transit legs (e.g., transferring between services). Currently, all access to transit is modeled via walking or driving only -- not by passenger drop-off, biking, or for-hire vehicle. The entire trip still receives a primary mode of "Public Transit" and counts as one trip.
Wait Time¶
The first transit leg's start time is optimized so there is no modeled wait time before boarding. For subsequent transfers, wait time can be derived from BigQuery by comparing end and start times of consecutive trip legs.
Ferry¶
Ferry travel is modeled within Replica's multimodal network using route, schedule, and ridership data from ferry providers where available. However, ferry trips are harder to model than other transit modes due to lower sample sizes and rider preferences that do not follow typical utility maximization (e.g., choosing a ferry for the experience rather than efficiency). Cross-check Replica ferry volumes with observed ridership counts from operators wherever possible.
Walking and Cycling¶
Replica models purposeful walking and biking trips -- trips with a clear origin and destination (commuting, errands, school trips). It does not model:
- Recreational or looping trips (jogging, dog walking)
- Trips without a defined destination
Key details:
- Trails are included when they serve as the most direct route or link to a place of interest. Trail volumes may be lower than observed because recreational trips are excluded.
- E-bikes are not separated from regular biking trips.
- Pedestrians take whichever side of the network is faster based on start/end points.
- Cyclists follow the flow of vehicular traffic.
- The mode choice model considers origin/destination, travel time, congestion, cost, income, mode availability, and ACS Journey to Work data. In dense urban areas, machine learning assesses patterns over time to assign walking/biking where infrastructure and trip characteristics support it.
Freight / Commercial Vehicles¶
Freight is modeled as a distinct trip mode, separate from person trips. Freight trips have no person demographic data (age, home location, etc.) -- trips with missing person data are either freight or visitor trips.
Vehicle Classifications¶
| FHWA Class | Category | Weight |
|---|---|---|
| 4-6 | Medium commercial | 14,000 - 26,000 lbs |
| 7-8 | Heavy commercial | > 26,000 lbs |
Data Sources and Calibration¶
- Raw data comes from GPS transponders installed on vehicles for fleet management.
- Transponders record time and speed as vehicles enter and exit portions of the route network.
- Replica builds a tour-based model: tours are identified from truck GPS data and scaled using ground truth counts. Scaling coefficients vary by geography and time.
What Freight Data Does NOT Include¶
- Dwell/idle times at stops
- Goods data (what is being carried)
- Vehicle/company identification (providers do not specify which retailer owns a truck, though POI data along trips can highlight stops at major retailers, ports, refineries, airports, and factories)
"Other" Mode¶
Trips with an "Other" primary mode near airports most likely represent people arriving on flights. Replica does not model airline trips directly and intentionally removes most suspected flight movements. Some still appear as "Other" in model outputs. Do not analyze "Other" trips as a meaningful category.
Modes NOT Modeled¶
The following are not explicitly modeled by Replica. Trips by these modes may be absorbed into other categories:
- Motorcycle
- School bus (categorized as auto passenger, private auto, or walking)
- Paratransit
- Private shuttles
- Golf carts, ATVs, scooters
- E-bikes (included in biking, not separated)