(Center Identification Number: 79050-02-A)
A significant need exists for creating a model to estimate demand for intercity bus services, especially in rural areas. Many states and rural operators are unsure about the potential demand for rural intercity bus service, and many of the existing models are unreliable due to poor data (Fravel et al. 2011). To address this need, a TCRP project by Fravel et al. (2011) developed a sketch-planning guide that could be used by state transportation department program managers and both public and private rural intercity bus service providers to forecast demand for rural intercity bus services. The route-level modeling techniques used in this TCRP report provide a useful tool for estimating ridership on rural intercity routes, but it has some limitations. It does not account for through passengers that are using the service simply because it connects to others routes and it is not sensitive to changes in fares or frequency. The model proposed for this study will attempt to address these issues through the use of a regional network model. The intent of the Fravel et al. (2011) study was to develop a tool to help determine ridership on proposed feeder routes, rather than to construct a network model. A network model, however, would be a useful tool that would account for through passengers while estimating route-level demand and would show impacts of service changes and demand changes on the entire network.
Create an intercity bus network model for the Upper Midwest that would be used to estimate boardings at each stop and ridership on each link in the network, and use the model to estimate the impacts of possible service changes.
This study will employ a network model that would follow the basic urban transportation modeling approach but apply it to the intercity bus network. The model would be a single-mode bus model that would include trip generation and trip distribution steps but not mode split. The model will be applied to the Upper Midwest states of North Dakota, South Dakota, Minnesota, Iowa, and Wisconsin.
The network model would describe the current intercity bus network in terms of existing routes, frequencies, and travel times between stops.
The region to be studied is North Dakota, South Dakota, Minnesota, Iowa, and Wisconsin. The intercity bus operators in this region include Jefferson Lines, Rimrock Trailways, Standing Rock Public Transit, River Cities Public Transit, and Burlington Trailways. Jefferson Lines, in particular, is the major intercity bus carrier in the region. Service data within the region for all these carriers will be incorporated into a GIS model.
Many of the trips within this five-state network originate and terminate within the network, but some origins or destinations are outside the network. To account for these internal-external, or external-internal trips, a few major stops outside the region will be included in the model, such as Billings, Omaha, Kansas City, and Chicago.
The trip generation step would estimate the number of intercity bus trips generated at each stop in the network. This would be accomplished using a version of the point demand model proposed by Fravel et al. (2011). The point demand model estimates the number of persons boarding at a particular stop. Such a model would require data from intercity bus providers on the number of persons boarding at individual stops, which would serve as the dependent variable in a regression model.
Boardings would be estimated as a function of the size of population served, demographic characteristics of the population served, and service characteristics. The size of the population served would be measured as the number of people living within a certain number of miles of the stop. Demographic characteristics included in the model are the size of the population that is low-income, elderly, or minority, as these population groups may be more likely to use intercity bus.
Service characteristics in the model that could influence demand include service frequency, fares, and connections to high-demand destinations. Regarding connections to high-demand destinations, the model would indicate if the service connects to an airport or some other special generator. Because the demand for boardings could also be affected by the places served by the bus service, another explanatory variable would be the combined population size of destinations that could be reached within a certain time period. It is hypothesized that, everything else equal, boardings will increase if potential riders can reach higher population areas within a shorter period of time.
If sufficient time series data are available, the impacts of gas prices on ridership will also be included in the model.
The results of the trip generation step provides estimates on the number of boardings at each stop, but it does not estimate the number of riders on each route, since many riders may have boarded at previous stops and are traveling through. To accomplish this requires distributing and assigning the trips across the network.
The trip distribution step would pair trip origins with trip destinations. A gravity model would be used to distribute the trips. Average intercity bus trip length data from the NHTS or from bus carrier ridership statistics could be used to calibrate the friction factor in the gravity model. The model will be run so that the number of trips that begin at each stop equals the number of trips that end at that stop.
The trip assignment step would place the trips onto the links in the network, so that overall ridership on each link could be estimated. For each origin-destination pair, trips will be assigned to the route with the shortest travel time. In many cases, there may be only one available route option.
The model could be validated using actual route-level ridership data from the bus carriers.
Uses for Model
The developed model could be used to estimate ridership on new routes, the number of boardings at new stops, and the impacts of changing routes, adding new links, or eliminating routes or links. The impacts of population changes or demographic shifts on demand could also be estimated, and if sufficient time series data are available, the impacts of changing gasoline prices on bus ridership could also be explored.
Data from intercity bus operators on the average number of boardings at each stop is required to estimate the point demand model, and route-level ridership data are needed to calibrate and validate the model. As Fravel et al. (2011) wrote, obtaining ridership data from private carriers is difficult. However, they were able to obtain stop-by-stop and route-level ridership data from Jefferson Lines, the main intercity bus carrier in the region being studied. Such data from Jefferson Lines would be required for this study, and similar data from the other intercity bus providers would improve the model. If these data cannot be obtained but data from Greyhound are available, then the study may have to focus on a different region that is served by Greyhound.
Average intercity bus trip length data from NHTS could also possibly be used to calibrate the gravity model.
Service data such as service frequency and fare levels at each stop would also be collected. The route information and travel time data needed to create the network model can easily be obtained from the bus carriers, much of which is available online.
Population data and demographic data will be obtained from the U.S. Census.
Task 1. Review literature
Task 2. Collect data
Task 3. Develop network model in GIS
Task 4. Estimate point demand model
Task 5. Run gravity model for trip distribution
Task 6. Assign trips on the network
Task 7. Validate the model
Task 8. Evaluate different scenarios with service changes or demographic shifts
Task 9. Prepare final report