Youssef Dehghani

An investigation of prediction errors in work trip mode choice models

Abstract This paper reports some analyses of predictive accuracy of disaggregate work trip mode choice models. The prediction error is separated into three components: model specification error, aggregation error, and transfer error. The main results are (1) the total forecasting errors can be very large, especially if the model transfers poorly; (2) poor transfer-ability was found between cities where the transit or shared ride markets are much different; (3) prediction from models having statistically different coefficients need not be much different and (4) the type of level-of-service data, i.e. manually coded or network based, used in model estimation has little influence on forecasting accuracy. These and other results are provisional for reasons discussed in the main text.

Models for transportation level of service

The paper describes the development of models for predicting travel times of door to door trips for both transit and automobile trips. The models for access times have a distribution associated with them and permit a truly disaggregate assignment of travel time components. The equations for linehaul travel time of the highway using modes (bus, auto) are volume dependent and can thus be used in equilibrating travel demand and level of service. All the models are related directly to transportation policy options--changing bus line spacings, bus headways, number of (priority) lanes, etc.--and translate the effects of such policies into specific values of the level-of-service attributes without the need to code networks and run paths. The use of the models in a practical application is also discussed.

Comprehensive Planning Model for Ferry Ridership Forecasting Analysis in Puget Sound Region

The methods used to develop a travel forecasting model for the Washington State Ferries (WSF) are described. The model is being used to develop ridership forecasts for the WSF Long-Range System Plan and other project planning efforts. The WSF model covers a 12-county ferry system service area, including the four counties within the central Puget Sound region. The model was designed to be sensitive to policy variables and to produce detailed ridership forecasts on each ferry route by (a) boarding methods, (b) mode of access and egress combinations for walk-on passengers, and (c) vehicle occupancy for those persons boarding by vehicle. The model is fully incremental and uses observed travel patterns from a comprehensive on-board origin-destination survey of ferry riders conducted in 1993 for the afternoon peak period. Choice of terminal is an intermediate destination choice for walk-on ferry riders and is a prime example of trip chaining between an origin and destination. A new method involving matrix algebra has been used to formulate choice of terminal in terms of matrix products. This method estimates demand matrices corresponding to individual legs of the trip chain for each access and egress mode combination. The modeling system was validated for replication of ridership on each route by boarding methods and by mode of access and egress combinations.

WASHINGTON STATE FERRIES' TRAVEL DEMAND MODEL DEVELOPMENT AND FORECASTING ANALYSIS PROGRAMS: A 10-YEAR PERSPECTIVE

Recent Washington State Ferries (WSF) origin–destination travel survey data collection and model development activities are described. WSF performed similar activities in 1993. The lessons learned from the initial project helped to develop a better survey design and sampling plan. The overall survey response rate was 37%, about 28% higher than the response rate for the 1993 travel survey. The improved survey design and sampling plan also provided a richer data platform from which to establish more representative observed baseline ferry travel patterns, by boarding method and access and egress mode combinations. Compared with 1993, the 1999 travel survey results reflect a significant increase in the systemwide use of transit for access to and egress from ferry terminals. This increase was primarily attributed to improved transit service, timed ferry-bus connections, and ferry-bus passes implemented by Kitsap Transit and other transit agencies. In addition to the 1999 travel survey, WSF conducted a stated preference survey. The combined data from these two surveys provided information on ferry users’ revealed and stated preferences, which allowed a more rigorous statistical mode and route choice estimation analysis to be performed. This analysis helped develop a more robust and internally consistent model, as reflected in model performance. For walk-on and automobile-boarding passengers, the updated WSF model has produced more accurate route-level validation results than the previous model. The results from the model estimation and validation analyses are presented. Specific improvements made in the updated model are highlighted.