Innovations in Freight Demand Modeling and Data Improvement (2014)

Below is the uncorrected machine-read text of this chapter, intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text of each book. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.

Examining Carrier Transportation Characteristics Along the Supply Chain Anne Goodchild, Maura Rowell, and Andrea Gagliano University of Washington Presentation Notes: Presented by Anne Goodchild, University of Washington. This model examines carriers and characterizes statistically significant and predictable transportation patterns at the regional and statewide levels. The data were applied to existing statewide transportation models and the goal was to improve the state of the practice, not to try to capture all the complexities of the world. The model represented carriers and recognizes that carriers have different behaviors based on their category. Data collection included short survey phone interviews and received 522 responses from Washington and Oregon. Responses to one question on whether the respondents owned or operated facilities were used to classify carriers into node carriers or link carriers. The model identifies significant transportation characteristics and determines clusters of behavior. Differences related to time-of-day operations, length of trips, and scheduling were modeled. Current models are trip-based, and the percentage of truck trips is based on spatial and industrial classifications, while this model modifies assumptions based on link and node classifications. As a result, time-of-day distribution, frequency, and intermediate facilities serve as the fifth step to the model. Abstract Transportation decisions in the private sector are made within a supply chain framework. Decisions made in the public sector, however, are based on models with limited supply chain considerations (Tatineni and Demetsky 2005). If policies are to effectively manage congestion, they must be informed by models that reflect at least those most fundamental of industry’s supply-chain practices. Carriers are expected to respond differently to policy changes depending on variables such as type of cargo, trip distance, and fleet size. The investigation of this hypothesis requires data on carrier demographics and transportation characteristics. In this paper, through an evaluation of survey data, the statistical validity of classifying carriers based on their role within the supply chain is demonstrated. Carriers associated with firms that own or operate a supply-chain node (raw production facility, manufacturing facility, storage center, distribution center, and/or retail facility) are termed supply-chain node carriers; those carriers that are associated with firms that do not own or operate a supply-chain node are termed supply-chain link carriers. Node carriers can be broken down further, depending on which nodes they are associated with. Transportation characteristics such as shipment size, pickup/delivery time of day, delivery time windows, and locations served vary, depending on the type of carrier. Existing freight models are typically limited to characteristics such as vehicle classification or a small set of commodity codes. Commodity flow data used for these 70

classifications, however, lack the important operational detail that is necessary to understand the implications of policy changes. Modeling the effects of disruptions will require additional knowledge of logistical practices (Goodchild et al. 2013). Without this data, models used to evaluate policies cannot capture the varying uses of the network. This research is intended to help address this gap by developing a categorization of motor carriers based on their supply- chain and transportation characteristics. The goal is not to capture all of the complexity of supply-chain logistics but to identify discriminating categories from a transportation perspective. Literature Review of Modeling Frameworks Modeling freight proves to be more complex than modeling passenger vehicles, due to the number of stakeholders influencing decisions and the physical limitations on trucks. For instance, truck trips are influenced not only by the road network and time of day but also by the physical qualities of the commodities being carried (e.g., perishable, hazardous, or bulk), the value of the shipments, and the availability of intermodal facilities (de Jong et al. 2004). As a result, many modeling techniques are being explored to capture the complexity of freight flows. In an attempt to address the limitations of both trip-based and commodity-based modeling approaches, Holguin-Veras suggested modeling empty truck trips on the basis of commodity (2000). In 2005, Cheng suggested using logistics-based modeling for agriculture, petroleum, forestry, and mining industries (i.e., bulk goods) while using tour-based modeling for textiles, apparel, electronics, furniture, and services (i.e., packaged goods). Logistics-based modeling focuses on how shipments move from the producer to the consumer, while tour-based modeling focuses on forming a single tour from a series of legs. Tatineni and Demetsky (2005) stress the integration of supply chain considerations into truck models. They surveyed manufacturing companies and split freight traffic into business-to- business links and business-to-customer links of the supply chain. Business-to-business links are less time sensitive and have larger shipments than business-to-customer links. They found that road pricing needs to incorporate commodity because different industries travel at different times and are dependent on customer service requirements and production schedules. While these studies have addressed components of supply chain decision making, these have been hampered by limited data and have often only considered modifications to existing freight modeling frameworks (de Jong et al. 2004). More detailed data are required to build an understanding of transportation characteristics. This paper contributes to closing this gap. Survey The survey designed for this research asks general business demographic questions and freight- related questions aimed at capturing how the respondent moves freight. The survey ends with questions concerning number of vehicles, travel locations, travel distances, delivery/pickup types, vehicle types, time windows, travel times, delivery/pickup locations, facility locations, facility size, and company revenue. 71

The principal question asked if the respondent owns or operates a facility that produces raw materials, a facility that manufactures goods, a storage center, a distribution center, and/or a retail store. Respondents who responded “no” to all the facility types were classified as link carriers. Those who responded “yes” to at least one of the facility types were classified as node carriers. Node carriers were then further classified to create four subcategories: raw and/or manufacturing, distribution and/or storage centers, retail facilities, and multi-node. This paper addresses the hypothesis that link and node carriers demonstrate substantially different transportation characteristics. Results This section describes the transportation differences between link and node carriers. Conclusive differences are drawn among factors that are statistically significant as determined by a Welsh two-sample t-test for continuous data or the Fisher comparison of proportions test for categorical data. The transportation characteristics investigated include delivery/pickup type, frequency, location, style, time of day, and time windows. • Both less than truckload (LTL) and full truckload deliveries/pickups had a significant difference between all subgroups except when comparing retail and multi-node carriers. • Link carriers and node carriers had significantly different results for the range of delivery frequencies (multiple times a day, daily, weekly, monthly, and less than monthly) to a single facility. • Raw/manufacturing had a much lower percentage of activity in urban areas. Retail carriers had much lower percentages of activity in suburban and rural areas. The majority of all deliveries/pickups were to businesses, regardless of carrier type. • Except for link carriers, manufacturing facilities were the most visited and intermodal facilities the least. Link carriers most often visited distribution centers. • On average, 75% of pickups/deliveries by link carriers are scheduled and 15% are made on a first-come-first-served basis. In comparison, node carriers average 55% scheduled and 37% first-come-first-served. The p-values for these results indicate that both delivery styles differentiate link and node carriers. Within the node carrier breakdown, retail and multi-node are the two extremes, with retail carriers having 58% first-come-first-served and 23% scheduled deliveries and multi-node carriers having 28% first-come-first-served and 68% scheduled deliveries. • Between link and node carriers, all times of day (morning, daytime, evening, and overnight) were significantly different, except the daytime period. Within the node carrier breakdown, all comparisons yielded significant results. The morning and daytime periods were the most active periods for all carrier types. Link carriers had the highest morning and lowest overnight activity. 72

• All time windows (less than 30 minutes, one to two hours, half day, and all day) were significantly different between link and node carriers. Although the percentages of carriers that use other time windows differ greatly, they are not found to be statistically significant. The significance of the differences between subgroups of node carriers varied. The majority of carriers use time windows of less than 30 minutes or one to two hours. Retail carriers had the highest percentage of one-to-two-hour time windows. Conclusions This paper presents travel behavior trends in the classification of link carriers and node carriers as well as the further breakdown of node carriers into raw/manufacturing, storage/distribution center, retail, and multi-node carriers. Through the analysis, it is evident that there are travel behavior differences among these classifications and that the classifications provide a useful framework for truck modeling. The research team proposes a modeling methodology that classifies carriers by their place along the supply chain. Carriers would be classified based on the types of facilities they own or operate within a supply chain. Using different modeling parameters according to the results presented here will provide a freight model that connects the industries’ supply-chain practices with government policy plans. References Bureau of Transportation Statistics, U.S. Department of Transportation. 2012 Commodity Flow Survey. http://www.rita.dot.gov/bts/sites/rita.dot.gov.bts/files/publications/ commodity_flow_survey/index.html Cheng, L. Building a Multimodal Comprehensive Truck/Freight Modeling for Los Angeles Metropolitan Area. CitiLabs, 2005. de Jong, G., H. F. Gunn, and W. Walker. National and International Freight Transport Models: An Overview and Ideas for Further Development. Transport Reviews, Vol. 24, Issue 1, 2004, pp. 103–124. Federal Highway Administration. U.S. Department of Transportation. Freight Analysis Framework. http://www.ops.fhwa.dot.gov/freight/freight_analysis/faf/ Garrido, R., and A. Regan. Modeling Freight Demand and Shipper Behavior: State of the Art, Future Directions. Institute of Transportation Studies, University of California, Irvine, February 2002. Goodchild, A., A. Gagliano, and M. Rowell. Characterizing Oregon’s Supply Chain: Final Report, Oregon Department of Transportation SPR 739, March 2013. 73

Holguin-Veras, J. A Framework for an Integrative Freight Marketing Simulation. Proc., 3rd Annual Intelligent Transportation Systems Conference ITSC-2000, IEEE, Dearborn, Mich., 2000, pp. 476–481. Tatineni, V., and M. Demetsky. Supply Chain Models for Freight Transportation Planning. Center for Transportation Studies at the University of Virginia. Research Report No. UVACTS- 14-0-85, August 2005. 74