The US Transportation Research Board has recently conducted a major study into the state-of-practice of travel demand forecasting theory and application in that country.
TRB Special Report 288, Metropolitan Travel Forecasting: Current Practice and Future Direction, examines metropolitan travel forecasting models that provide public officials with information to inform decisions on major transportation system investments and policies.
Part 1: The main findings
The main recommendations from the study are highlighted in part 2 of the report. The findings are based in part on the results of a survey of planning agencies reported in Part 3.
Main study findings: Most planning agencies continue to use the conventional four-step modelling approach.
The shortcomings of this approach, which has not changed in almost 50 years, have to do with model formulation (the number of variables of interest may not be represented by the models), as well as with input data gaps.
Deficiencies in the conventional modelling approach lead to an inability to model:
- Choice of start trip times
- Choice of non-motorised modes
- Values of time and values of reliability for different travel markets
- Freight movement modelling.
Advanced models, based on trip tours rather than single trips, have been developed and implemented by some of the larger urban areas. Some of these advanced models seem to perform well in the few applications to date. (TRANSIMS, the new modelling platform receives about $US2 million annually in development funding).
There is no strong evidence that advanced models can be implemented for a reasonable cost and will provide significant improvements over current practice.
There is no single ‘correct’ approach to travel forecasting for all applications. Travel forecasting tools should be appropriate for the nature of the questions being asked. The demands on forecasting models have grown significantly as a result of new policy concerns and existing models are inadequate to address many of these new concerns.
The main barriers to fundamental change include resource limitations, practitioners’ doubts on whether new practices will be better than those they replace, and lack of adequate investment in the R & D area.
There is poor technical capability in the use of models. This relates to:
- Inadequate data
- Optimism bias with a number of studies showing that forecasts for toll road and new transit projects are substantially higher than actual start-up patronage.
- Quality control issues. The best practice is to have a rigorous quality control process, with independent assurance during each step. Many agencies in the US do have such a process in place.
Part 2: Main study recommendations
The Transportation research Board in the US has recently conducted a major study into the state-of-practice of travel demand forecasting theory and application in that country. The findings are based in part on the results of a survey of planning agencies reported in Part 3.
Main study recommendations:
- It was found that current models are not adequate for many of the present applications. It is recommended that new modelling approaches be developed that are better suited for multi-modal projects, environmental evaluations and a range of policy alternatives.
- The study calls for increased funding for the implementation of activity-based modelling and other advanced practices.
- The performance of conventional and advanced models should be compared through rigorous studies using backcasting and sensitivity analysis.
- Planning agencies should conduct formal independent peer reviews of their modelling practice.
- Agencies should also conduct reasonableness checks of demand and cost forecasts for major projects.
- Individual agencies and universities could form partnerships to foster research on travel demand modelling and the implementation of advanced modelling practice.
- A national travel forecasting handbook should be developed and kept current. The handbook would describe alternative best practices for addressing different travel markets and metropolitan needs, recognising that differing approaches are needed for different applications. It should also include information on various ways to conduct quality control and model validation.
Part 3: Survey of Practitioners
A web based survey of more than 200 urban planning agencies in the US has found a general reluctance to adopt new methods for demand forecasting until they are proven to work better than existing methodologies.
The survey, which was complemented by a number of in-depth interviews with practitioners, found that very few agencies were using backcasting to quantify the extent of model and input errors.
The level of accuracy attached to forecasts was not quantified in any way by the survey.
A few agencies of the larger urban areas (greater than 1 million population) are developing new methods to address a range of issues, including congestion, peak spreading, and freight movement.
Although the vast majority of agencies are using conventional four-step travel demand modelling procedures, some are using activity or tour-based methods.
Some of the larger areas actively considering development of “advanced” models that are activity based or that include destination choice, or alternative assignment methods.
Most agencies seem to be after methods which can be easily implemented within their existing software capability, as well within the skill sets available to them.
Summary
Modelling step: Survey finding
Trip generation: Cross-classification and linear regression for trip production and attraction models respectively.
Trip distribution: Gravity model with an impedance function which combines auto and transit travel times.
Model choice: Most large areas use the multinomial or nested logit model approach.
Trip assignment: Standard conventional equilibrium assignment mostly used.
Environmental evaluation: Mobile source emissions estimated using model outputs for 97% of large urban areas.
Feedback loops: Feedback of auto and transit times to model components reported:
- to land use: 41% of large MPOs
- to auto ownership: 42% of large MPOs
- to trip generation: 33% of large MPOs
- to trip distribution: 88% of large MPOs
- to mode choice: 85% of large MPOs
(MPO: Metropolitan Planning Organisations – regional planning and funding body)
Ref: Metropolitan Travel Forecasting: Current Practice and Future Direction. Special Report 288, Transportation Research Board, 2007. Washington, DC. Download 147 page PDF from http://onlinepubs.trb.org/onlinepubs/sr/sr288.pdf