Freight Demand Modeling and Data Improvement Handbook

Chapter 4. Conclusions

As data development and modeling continue to push forward, foundational lessons from past work should be used to inform new efforts. This chapter summarizes principles and considerations that are general and foundational, the individual issues that the projects encountered, and the specific lessons learned that can be applied to similar future efforts.

Principles and Considerations for Data and Modeling

Future projects in freight data and modeling should use the foundational principles and considerations discovered or emphasized in the 11 Implementation Assistance Program (IAP) projects. Those principles and considerations are summarized below.

Developing a Regional Freight Stakeholder Group. Nearly every project in the group of seven included the establishment or use of an existing freight stakeholder groups. These groups performed many tasks, including project oversight and providing expert level input on specific issues. The early development of a freight stakeholder groups will benefit any new freight data development project. Additionally, the stakeholders will lend weight and importance to any data development effort. They will likely be influential in obtaining assistance from agencies and companies that are requested to participate in the effort in some capacity.
Administering Surveys to Local Freight Producers and Consumers. Many of the data projects involved large data collection efforts that were dependent on surveys to gain input from freight-related businesses and others. Poorly designed or deployed surveys produce notoriously poor response rates. The experiences of the seven data projects contain lessons that can maximize the return on any surveys conducted and maximize the data collected.
Developing New Data Sources. The data projects clearly show that developing new data sources is a time-consuming and difficult task. Freight data is often fragmented and uncoordinated. Much of it was developed for purposes other than freight analysis and so contains data standards that vary widely and may make transformation of the data into usable freight information difficult. For these reasons, choose data sources wisely. Prioritize data sources that can be most readily transformed into usable information and used for the greatest number of analyses and purposes. Keep, however, a list of alternate data sources, as well. A preferred source may turn out to be more difficult to use than originally thought, and an alternate source may be needed.
Incorporating Behavior-based Modeling. Behavior-based aspects of freight decisionmaking are being incorporated into models. These improvements allow the model to more accurately portray the complex factors that are involved in decisionmaking by freight shippers and carriers.
Using an Open Format Code. Models of all the projects were developed utilizing an open source programming language that is the basis used by other States and MPO's. Use of this platform allows improvements made to the model to be available to all other modelers. They also utilized available public freight data sets along with more localized data compiled specifically for use within the model. Data sources for localized freight movement may need to be updated into the model as they become available.
Collecting Establishment Data. The collection of local establishment data can be difficult even with strong support from partners and a robust data collection outreach process. Agencies should be ready to use all available methods to collect local data because not any one method will produce enough data. Paper surveys and trip diaries, smartphone apps, vehicle monitoring data, site visits, and any other methods should all be considered for use.
Developing Localized Modeling. Most projects included the development of an integrated freight model that is able to provide supporting data and information on a more refined basis. The models consider local freight movements and the ability to identify commodity types, volumes, routes, as well as, existing infrastructure issues, (e.g., congestion, conditions, safety) and project priorities within a region.
Providing Training Materials. Many of the projects developed freight model guides and provided either on-site training sessions and/or written documentation to educate users in the function and maintenance of the models. This training helped to ensure the success of the model, as well as set the stage for future improvements to refine the model.

Issues and Lessons learned

Each IAP project faced its own issues, the understanding of which will be helpful to those agencies that would like to incorporate these practices in their own freight planning activities. Some of the issues, challenges, lessons learned, and benefits for each proof of concept are summarized in Table 12. Users can reference this table and ascertain the applicability of the data or model issues and benefits to their projects.

Table 11. Modeling pilot project needs and gaps comparison.
Agency	Issues/Challenges	Lessons Learned/Benefits
Capital District Transportation Committee—New York	Majority of the datasets are confidential, requiring extensive dialog to ensure sharing. Each agency follows its own units of measure, aggregation methods, and variables while preparing data. Some databases need more time to be cleaned and processed before being released.	Conducting outreach activities provided opportunities to build relationships with the freight community. It takes a long time to complete the data collection process. Identifying sources and beginning the process of acquiring the data early is important. Data was available to use at all levels of aggregation, including national, interstate, and zip code.
Delaware Valley Regional Planning Commission— Pennsylvania	Some data items initially identified turned out to be either too difficult to obtain or limited in coverage. Personnel changes at partner organizations slowed the acquisition of some datasets. Due to the proprietary nature of the shipping and logistics industry, commodity flow data typically represents estimates or samples of the shipments and varies in geographic resolution.	Obtain formal agreements with partner organizations to minimize the impacts of staff turnover and get advice on potential data/ data sources and benefit from their shared experience. Explore alternative data sources and/or alternative data to obtain desired data. Having an open-source product results in the creation of new data sets, users (e.g., Travel Demand Modeling Group), and uses for the data (e.g., commodity flow studies, estimating delay impacts on revenue, costs, air quality, etc.).
Florida Department of Transportation	Fuel tax data was incomplete for a large portion of records. Acquiring data from private vendors was difficult to validate due to proprietary technology. Information sharing and video data was difficult to obtain.	GPS data, driver surveys, and tax records enhanced Modeling and simulation tools. With continued improvements, video image processing and license plate recognition are promising technologies to support data collection. Work to maintain the commitment of partners and stakeholders throughout the project to avoid lost investment.
Mid-America Regional Council—Missouri	Immense data sets required sifting through hundreds of millions of records and consumed resources. Agencies hoping to use freight waybill data should have an understanding of the logistics industry before drawing conclusions directly from the data. Due to available data, congestion analysis included only congestion occurring on the NHS and did not include minor arterials and other secondary roadways.	A comprehensive scan of other teams' research with respect to data sources, methods of extracting and using the data, data limitations and proxy data sources, and outcomes can provide insight into potential issues when advancing the state of the knowledge or developing new methods of analysis. Supporting financial resources should be packaged with the research so it can be readily and easily used by practitioners. The level of effort needed to work with large datasets should not be underestimated.
South Dakota Department of Transportation	Truck counts and roadway conditions for the local transportation system were not available, preventing complete understanding of local truck movement. Local and county road use for agricultural production was not well understood (e.g., how much production stays on the farm versus transported, timing of agricultural movements, granularity of data collection and analysis).	Several publicly available and national agricultural data sources can be used as inputs into the transportation planning process. The illustrative decision support tool allows agencies to more explore "what if" scenarios using modeling estimates rather than waiting for data collection and analysis. Additional data regarding truck movement on the local transportation system would further enhance the new methodology. It is possible to combine transportation and agriculture datasets to develop a systematic approach that assists in making local transportation investment decisions.
Washington State Department of Transportation	There were a variety of independent truck owners/operators serving many farms and relying on rural and county roads. It was difficult to get interviews and survey data from local businesses. Automated technologies could not reliably collect truck counts and trip generation data.	Survey responses can be improved by establishing relationships with local businesses and organizations. Human spotters can collect truck count and trip generation metrics when automated data collection technologies are impractical. The project provided metrics on two major supply chains (wheat production and food delivery) enabling scenario planning under a variety of potential future conditions.
Winston-Salem Metropolitan Planning Organization—North Carolina	Incomplete freight node information from initial data source needed to be supplemented with additional publicly available sources and aerial imagery. Freight facility information was sometimes less accurate and contained errors or omissions. The initial survey administration plan netted very few responses so a modified/ adjusted collection process was developed to improve response rate.	Wherever possible, pre-populate survey information. Administering the surveys in-person yielded the highest response rate. Freight facility visits can be used to identify and correct errors. Surveys should only include the most valuable information. The project resulted in improved collaboration between the public and private sector in the region for future data updates.
Maricopa Association of Governments (MAG)—Arizona	Using the framework at a nationwide scale requires some aggregation in order to deal with computational issues. Performing analysis at a sub-area scale might require more detailed data. Obtaining establishment data from industry partners is often challenging. Many industry partners prefer to keep information on what is being shipped where and when proprietary to maintain their competitive edge.	The majority of the work accomplished was not region-specific and could be replicated by other agencies with comparable data access and resources. Data licenses can be expensive and data restrictions may limit the ability to use the data in new or unique ways. The new model can provide insight into supply chain decisions, including distribution channels, models, and shipment sizes.
Maryland Department of Transportation and Baltimore Metropolitan Council	Some data took longer than expected to obtain and assimilate into the model. Resolving various data issues (e.g., missing data, various granularity of data) in bringing together the various data sources took longer than expected. There was a lack of local data to use for answering specific and local questions. Establishment survey responses did not meet the sample size target.	Agencies can replicate the activities completed within this project to tailor and implement a similar behavior-based freight demand model. Smartphone apps can be useful for data collection but may have limited participation. Having all the relevant information (i.e., a documented, organized set of folders of data, files, analysis outputs, and scripts) in a single location is a useful resource to accompany the model and facilitate future model updates.
Metro—Oregon	Even with a robust data collection plan and insight from industry insiders, there was difficulty collecting freight behavior data through an establishment survey due to low participation rate from industry in the region. Smartphone application for data collection can provide high quality data, but there are challenges, such as potential for driver distraction. It was difficult to analyze non-truck freight modes.	The new model added a more accurate depiction of trucks on the network by including a component focusing on service vehicles and will allow planners to estimate the movement of these vehicles separately from typical freight vehicles. The model development approach should be adapted as early as possible to the expected types of data to prevent issues with scope and schedule. Data collection for models of this type can be a significant effort due to the many types and sources of data; identify and include strategies for integrating data from disparate sources and in varying formats during the planning stages.
Wisconsin Department of Transportation	The model took longer than expected to adapt to Wisconsin.	Even off-the-shelf model solutions require significant time to customize for the region. New models can and should be benchmarked against existing models and techniques to evaluate their effectiveness.