The Exploratory Advanced Research Program

National Multimodal Freight Analysis Framework Research Workshop

Workshop Summary Report - December 11, 2013

Part Two: Workshop Conclusions

Research Opportunities

During the workshop, participants discussed a variety of opportunities for improving the data, analysis, and modeling of freight travel at the national level. Two potential research directions that surfaced in this workshop were as follows:

1. Behavioral-based (or agent-based) national freight-demand modeling. Incorporating agent-based modeling could represent a significant step forward for the FAF and could enhance the model's predictive capabilities. Seasonal fluxes, the impact of business decisions' and other variables that have been previously unaddressed by the FAF, yet influence the decisionmaking of shipping entities, could be incorporated in a meaningful way.

2. Freight data development and enhance-ment to support national freight transportation analysis, modeling, and forecasting practices. To create a predictive model, the FAF requires more detailed, higher quality data. New methods of data collection and integration for the FAF could represent a significant leap in terms of FAF potential capabilities.

Freight Analysis Framework: Future Direction

Workshop participants offered potential research areas to enhance the FAF in the categories of data and modeling. As the FAF dataset is based on a national-scale compilation of different surveys and field databases, challenges with the FAF data include:

• Applying the data for reasoning, "what if?" scenario analyses, and trend or pattern study.
• Provisional and future year estimation.
• Inadequate cost and temporal factors.
• Calibration and validation issues of the FAF data that are due to a lack of reference data.
• Insufficient geographic scale.
• Data deficiencies of coverage, aggregation, sparseness, consistency, and accuracy.

Workshop participants made the following conclusions in the data category:

• Key data use is validation.
• Rather than tracking international shipments, the intent is to characterize international shipments (e.g., port of origin or final destination and mode to or from a U.S. port) and to track domestically from the port to the final destination or in reverse from the point of shipment to the port for export.
• Temporal data may be an important enhancement from data mining.
• Cost and economic data may be an important enhancement from data mining.
• It is important to reduce the reporting lag time of local data to keep the FAF current but not necessarily to the extent where near-real-time data are being collected and reported at the national level.

Workshop participants made the following conclusions in the modeling category:

• Considering both the Federal and non-Federal uses of the next FAF and future FAFs is important to consider before making any substantial modifications to the FAF.

• In considering different user groups, there exists the potential for providing an application-programming interface.

• An important modeling component would be the accurate capture of transfers among modes.

Addressing Challenges

To address these challenges, participants noted that new innovations in freight data development and its management are needed. The research outcome in the form of nationwide, disaggregated freight-flow data will feed a broad range of further studies and applications. One of the major beneficiaries is on the freight-travel demand model improvement side. The future FAF will likely be developed in the form of a national supply-chain-based, comprehensive multimodal freight-travel demand model. FAF could extend its capacity by supporting national and regional freight policy making, strategic scenario analyses, and future freight and economic impact estimations in a timely manner.

Improved FAF modeling could aid in economic impact studies, road maintenance plans, cost–benefit analyses, air quality, and toll or pricing studies. Major economic sectors and industries, including general public domain, could also benefit from the geographically detailed, cost-sensitive, and temporal nature of datasets on their policy and decisionmaking.

Suggested approaches discussed during the workshop include, but are not limited to:

• Designing and testing novel, cost-effective, transparent, and accurate data collection methodologies to feed national and regional freight-travel demand models and analyses.

• Maximizing the usage of current FAF data sources by improving methods and techniques in data collection, format, and processing.

• Developing better validation and calibration methods for freight models and analyses in both the national and regional levels through the use of improved data.

• Developing novel data collection and modeling strategies for estimating policy-sensitive variables (e.g., cost) for more robust "what if?" analysis.

• Developing novel data disaggregation methods from national and yearly levels to regional, metropolitan, and seasonal levels.

• Developing methods for using new data sources and formats to increase connectivity and integration of locally collected data in different levels of data density, seasonality, accuracy, and geography (e.g., "crowd sourcing," bottom-up data-sharing, and "grass-root" data build-up) to support FAF and similar products.

Freight Analysis Framework Issues

Although FAF has many strengths and benefits, there are also issues to be addressed. These issues include:

• FAF objective. The FAF objective should be reviewed and examined on a continual basis. Areas such as Federal, State, and local roles and needs, private (rail) business and public-private (highway-truck) codependency, as related to multimodal exploration, should be thought through.

• Static data sources. On the basis of mostly census data, surveys, and field data collections, FAF is a static snapshot of national goods movement. This conglomeration of nationwide freight data is helpful to understand the big picture of America's economic and freight activity today; however, FAF is not currently designed as a systematic analytical tool with modeling power in its structure. As a result, it lacks the power to explain the causes and effects of dynamic freight activities and their ripple effects, both now and in the future. The current architecture of FAF limits what a national freight-demand modeling tool can pro-vide (e.g., national "what if?" scenario analyses and forecasting functions).

• Incomplete and disconnected datasets. Many datasets from different sources and levels in different formats are combined to create the FAF. Gaps and level of detail constrains FAF's suitability to mainly national- and regional-scale analysis. To fill the gaps, advanced statistics and fitting methods are applied along with experts' judgment and adjustment. As a result, FAF achieves reasonableness and pattern of goods movement in both national and regional levels; however, there are accuracy and reliability limitations in adopting the FAF at the regional and local level. There are also consistency and reproducibility issues related to the data-cleaning and integration steps.

• Timeliness. The FAF has a significant time lag in processing and providing availability for its interim outputs and further applications due to the dependency of its data sources on the most recent census and surveys. In several cases, it might reduce its utility (e.g., time-sensitive alternative studies).

• Detail versus scale. The background need for the creation of the FAF was to capture and understand the national freight movement of the time. FAF has been focusing FHWA's internal use in drawing the big picture in an aggregated format with less interest on local disaggregated detail; however, with increasing roles inside and growing popularity outside FHWA, there are growing demands to support a deeper and wider scope of tasks. With numerous resource limitations, proprietary and confidentiality concerns, and technical issues, it is important to challenge and balance those limits to maximize the benefits of the freight-demand model.

• Information integration and sharing. FAF's study area covers all 50 States and the District of Columbia. Without local data feedback, it is almost impossible to keep the FAF data and its modeling accurate and up to date. Most of the State Departments of Transportation and Metropolitan Planning Organizations have their own limitations on data and modeling development and their updates. The State Departments of Transportation, Metropolitan Planning Organizations, and academia tend to resort to FAF as a base for their freight studies. There are growing needs from both ends for a breakthrough, new approach in information integration and sharing. Systemized, standardized, and technically sound methods will likely help in the steps of collecting, integrating, and sharing information-not only of data but also in modeling.

For more information on FAF, visit http://ops.fhwa.dot.gov/freight/freight_analysis/faf/index.htm.

Appendix

Appendix A: Workshop Participants

Attended in person

Attended via Webinar and call-in