Freight Demand Modeling and Data Improvement (2013)

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.

22 • Lower population growth in the Northeast and Rust Belt states, with higher growth rates occurring in Southern and Western states. • A gradual shift in location advantage for traditional indus- tries. The need for access to inland waterways (including the Great Lakes system) and freight rail connections to domestic sources of raw materials and subassembly loca- tions has diminished in importance, while deep-water ports and Class 1 rail connections to ports and North American Free Trade Agreement trading partners have played a bigger role in freight transportation. • Minimal capacity additions to the highway system. Since the establishment of the U.S. Interstate Highway System in the 1950s and its development and expansion in the fol- lowing decades, there has been little expansion in the high- way system in the past 15 to 20 years. • The growing use of the shipping container as a standard means of moving many forms of freight has made freight transportation across multiple transport modes increas- ingly modularized and improved transportation efficien- cies over longer distances. • The deregulation affecting the railroad, trucking, and air- line industries. • Consolidation in all sectors of the freight transportation industry (particularly rail) has resulted in longer supply chains and driven the development of massive economies of scale that have reduced transportation costs on a unit (ton-mile) basis. • The application of advanced information and communi- cations technology in many areas of manufacturing and freight transportation has enabled shippers and carriers to increase efficiencies at multiple steps in their supply chains. • The increasing dominance of the service sector emerging in the mature economy of the United States, along with online retailing and the resultant implications for over- night shipping, small-parcel deliveries, and terminal-based truck activity within smaller geographic regions. Passenger models have advanced throughout the past five decades, but freight modeling and data are still in their forma- tive stages. However, freight is critical to the national, regional, and local competitiveness of the United States. Accommodat- ing, or at least considering, freight within everyday planning practices at the federal, state, regional, and local levels is neces- sary to benefit the public as consumers, employees, and busi- ness owners. The findings within this chapter reflect extensive research and outreach to freight stakeholders. The following sections lay out the current state of freight demand modeling and data, as well as potential innovations to advance the state of the practice. Freight Industry Trends Overview The past several decades have been marked by major changes in domestic and global freight transportation. These changes have been driven by population growth and other demo- graphic changes, adjustments in consumer behavior, dynamic market and economic forces, changing business practices, and advances in transportation and information technology. Public and private decision makers responsible for under- standing the implications of these trends for the transporta- tion infrastructure planning process must contend with the influence of increasingly complex supply chains and logistics processes. Several key issues related to the changes and increasing com- plexity in freight transportation in the United States include the following: • The decline of manufacturing occurring in Northeastern and industrial Midwestern states, accompanied by the growth of industry and manufacturing in the Sun Belt, and a shift in manufacturing activity for many products from the United States to other locations around the world. C h a p T e r 3 Findings and Applications

23 The various factors described here, coupled with the increased complexity of supply chains and logistics processes, have resulted in an environment in which incorporating freight movement considerations in the transportation plan- ning process has become increasingly difficult at the very time that these considerations are more critical to the ability to forecast long-term transportation trends and plan for future needs. This situation is exacerbated by the inherent political dilemma faced by decision makers involved in almost any freight-related aspect of the transportation sys- tem: the widespread disconnect between users of the system (shippers and carriers) and those who benefit most from the system (the general public). Taxpayers can be more attuned to the real and perceived negative impacts of freight move- ment than to the broad benefits they receive from freight movement. The significant changes that the freight industry has under- gone over the past two decades and the key trends influenc- ing freight transportation fall into four general categories (Kuzmyak 2008): 1. Globalization of trade—Freight movements range in geo- graphic scale and scope, and the supply chains that span the entire globe can often have very localized impacts. 2. The economy—The cyclical nature of economic trends results in changes in freight transportation characteristics over time. 3. Private sector inventory practices—Many of today’s national and international businesses rely on manufacture- to-order and just-in-time inventories to meet customer demands, which makes reliability, speed, and flexibility crucial to both maximize efficiencies and maintain indus- try profit margins. 4. Warehousing—Freight transportation processes have become cost-efficient across different transport modes over time, and the close physical proximity of supply points (e.g., raw material sources, production facilities) to consumers no longer offers the advantages it once did. The traditional warehouse, which was primarily used for storage of raw materials and finished products, has been replaced by a dis- tribution center whose primary functions include efficient consolidation and distribution activity aimed at reducing shelf time for materials and enhancing the efficiency of the overall logistics process. Current practices As a result of these major changes in freight transportation, the planning process for freight-related capacity needs has become increasingly complex. However, public sector trans- portation decision making remains relatively uninformed with respect to freight transportation due to the limits of the current models. These models are unable to accurately repli- cate current conditions or forecast the impacts of freight on future transportation systems, thus limiting the possibilities for policies and improvements to solve expected problems and address future capacity needs. The practice of freight demand forecasting has received greater attention with the growing recognition that efficient freight and commercial truck travel is essential to national, state, and local transportation infrastructure planning and the economic well-being of the nation as a whole, as well as the prosperity of individual states and regions. Incorporating freight movement considerations in the transportation plan- ning process is difficult, but these considerations are increas- ingly critical to the ability to forecast long-term transportation trends and plan for future needs. Current practices in freight modeling and data develop- ment used by various organizations and planners in the United States are documented in the following sections. This review of the practice addresses stakeholder and user needs along with the strengths, weaknesses, opportunities, and threats regarding critical issues and the knowledge, data, and tools that should be addressed through innovation and fur- ther research. This review includes descriptions of various types of freight demand models, the methods of freight demand forecasting currently in use, and promising methods of freight forecasting emerging from research. This review also summarizes the variety of public and private data sets commonly used by model developers and practitioners to estimate, validate, and apply state-of-the-practice forecasting methods for freight movement. A basic understanding of the current state of the practice provides an important founda- tion for future improvement. Models Freight planning practitioners use different models and anal- ysis tools depending on the purpose of the analysis and data availability. These tools assess a range of measures, from com- modity flows to economic impacts. It is important to under- stand existing tools in order to determine whether using them as building blocks, components, or discrete tools is fea- sible and advantageous to advancing innovations in the state of the practice. Each model is described below, along with its strengths and weaknesses. Economic Flow Models Economic flow models estimate the flow of goods and ser- vices between households and firms, balanced by the flow of

24 payments made in exchange for them. Economic flow models are built on four economic activities: 1. Production—The use of economic resources in the cre- ation of goods and services; 2. Consumption—Consumer or raw materials purchasing; 3. Employment—The use of economic resources for labor in production or economic activity; and 4. Income generation—Maximum amount an individual can spend during a period without being worse off (Valdehueza 2008). Economic flow models are used in freight modeling to esti- mate the flow of goods based on these economic activities and then applied to estimate modal flows through a network. This estimation can be completed for national, regional, and local geographies, depending on data availability. StrengthS • Estimates goods movement from the origins of freight activity; • Estimates the volume of physical goods while considering other economic indicators and influences that may affect current and future movements; and • Could be used to better estimate local freight touring trips, as well as regional truck flows, depending on the level of detail in the supporting data. WeakneSSeS • Requires significant data collection for disaggregated inputs; and • Requires an understanding of markets for both goods and services, as well as monetary policies. Land Use and Economic Input–Output Models Input–output analyses are used to understand an economy by describing flows to and from industries and institutions. These models are best suited to predict changes in overall economic activity as a result of changes in underlying economic forces. They can be applied to freight trip generation models by using input–output data (with employment and population data) to estimate the zonal level of commodity production and attrac- tion. Input–output models have three basic components: 1. Transactions—The monetary flows of goods and services in a local economy for a given time period, including goods and services purchased and used in the production process, purchases for consumption, and payments for factors or inputs outside intermediate production processes. 2. Direct requirements—The purchases of resources (inputs) by a sector from all sectors to produce one dollar of output based on a multiplier effect. This measures the total change throughout the economy (output, employment, and income) from one unit change for a given sector. 3. Total requirements—The relationships between the dif- ferent input and output requirements, recognizing that if output for final demand increases, not only must pur- chases of indirect inputs increase, but firms supplying those direct inputs must increase their purchase of inputs. This analysis is done through a relational table of all industries being examined. Input–output models (depending on the software pack- age) have comprehensive and detailed data coverage of the entire United States by county and the ability to incorporate user-supplied data at each stage of the model-building pro- cess. These options provide a high degree of flexibility for both geographic coverage and model formulation. StrengthS • Estimates changes to various industries as a result of con- sumer spending, raw materials consumption, and other economic indicators or scenarios; and • Offers flexibility to change various parameters for scenario- based analyses. WeakneSSeS • Requires significant effort to translate results into goods movements and commodity flows and modal movements; and • Requires significant effort to generate accurate multipliers in order to yield accurate truck volumes. Commodity-Based Models Commodity-based models estimate the amount of freight moved by weight. This method simulates the economic basis for freight movements, focusing on commodity attributes (e.g., shape, unit weight), and includes the following steps: • Generation—An estimate of total tons produced and attracted by zone; • Distribution—An estimate of goods exchanged between origin–destination (O-D) pairs; • Mode split—An estimate of the weight moved by the vari- ous modes; and • Assignment—Loaded, partial, and empty trips applied to origin–destination matrices by mode and assigned to a network (Jack Faucett Associates 1999). This approach yields a region-to-region commodity ton- nage table based on economic forecasts and historic trade pat- terns. Flows are then disaggregated to zones based on historic

25 and forecasted activity levels of production and consumption within each zone for each commodity. The disaggregated flows are converted into trucks and assigned to a network (Holguín-Veras et al. 2001). StrengthS • Provides sound estimations of national, statewide, and regional movements; and • Provides a more robust method for estimating truck trips than a vehicle-based model. WeakneSSeS • Provides low-quality local commodity flows as a result of lack of data; and • Uses a similar method to the four-step passenger modeling process, which does not consider the entire logistics chain in its estimation. Trip-Based Models Trip-based models focus on modeling vehicle trips, which implies that mode selection and the vehicle selections have been completed using other methods. One advantage of trip- based models is that traffic data are readily available; for example, ITS applications are able to provide data on vehicle movements on highway networks. Trip-based models also con- sider empty vehicle trips (Holguín-Veras and Thorson 2000). The trip-based model generates truck trips as a function of different land uses and trip data from trip logs or shipper surveys. The generated trips are distributed using spatial interaction models (such as a gravity model), which are cali- brated using trip lengths obtained from trip logs (Jack Fau- cett Associates 1999). StrengthS • Uses readily available data; and • Easily calibrates truck movements to current volumes. Is easily incorporated into existing statewide and MPO mod- eling processes. WeakneSSeS • Does not consider the entire logistics chain in its estima- tion because of its simplistic design; and • Does not provide an ideal method for distributing truck trips across a network because of limitations in the gravity model method. Estimation Routines Estimation routines apply localized, regional parameters to localized, regional zonal and network data to produce truck size, trips, and vehicle miles traveled (VMT) estimates for each vehicle category, and in some cases, goods movement by commodity and modes (Cambridge Systematics 2007). These techniques include • Three-step model—Various techniques that estimate com- mercial vehicle trips from intercept surveys or trip-based, regional commercial vehicle surveys. These models, which typically include a trip generation, distribution, and assignment mode-chain structure, estimate commercial vehicle zonal trips, categorized by type, and link-based net- work volumes. • Three-step plus port model—Various three-step model techniques that include a separate trip generation and dis- tribution routine for a marine port. • Tour-based microsimulation—Various techniques that estimate commercial vehicle tours from tour-based, regional, commercial vehicle surveys, considering trans- shipment and distribution center movements from the moment the vehicle leaves until it returns. These models provide estimates by type of establishment (e.g., manufac- turing, construction) of the number of light, medium, and heavy commercial vehicles; the purpose of each trip on the tour; and each stop location. • Sample enumeration—A technique that repeatedly sam- ples regional large-scale survey data to develop multiple- class truck trip matrices as input to a multistep freight model (Donnelly et al. 2008). StrengthS • The three-step model is computationally easy and consis- tent with the common practice in most four-step models. It can be based on local surveys, as well as national data sources (such as the Quick Response Freight Manual [QRFM]), and is easy to understand. • The three-step plus port model has all of the strengths of the three-step model and includes unique characteristics of marine ports. • Tour-based microsimulation provides more detailed information on truck distribution patterns than the three- step methods, is more behaviorally based than other methods, and can be integrated with economic input– output models. • The sample enumeration technique uses locally collected survey data as the basis of truck movement patterns and is more behaviorally based than the three-step models. WeakneSSeS • The three-step model has little behavioral relationship to the actual decision-making process in freight movement, fails to consider the unique characteristics of some genera- tors, lacks multimodal goods movement characteristics, and introduces some aggregation error.

26 • The three-step plus port model has little behavioral rela- tionship to the actual decision-making process in freight movement, lacks multimodal goods movement character- istics, and is more dependent on certain local survey data than the three-step model. • Tour-based microsimulation is computationally complex, is more dependent on some local survey data than the simple three-step model, and lacks multimodal goods movement characteristics. • The sample enumeration technique is computationally complex, requires local survey data, lacks multimodal goods movement considerations, and does not address true logistical considerations. Aggregate Measures Aggregate measures apply national default growth factors or parameters to localized, regional data to produce fleet size, trip, and VMT estimates for each vehicle category (Cam- bridge Systematics et al. 2004). These techniques include • Factored trip matrix—A technique that applies national growth factors to an existing, often dated, localized, regional truck trips matrix; • Simple matrix estimation—Various techniques that apply a single seed, or best guess, truck trips matrix, calibrated using the most up-to-date truck counts, to develop a likely truck trip matrix, which is then factored using national growth trends; • Elegant matrix estimation—Various techniques that apply multiple seed (weighted by quality and level of confidence) data, calibrated using the most up-to-date truck counts, to develop multiple-class truck trip matrices, which are then factored using national growth trends; • Polenske–Roberts (PR) variant—A technique developed in the 1970s that uses basic input–output models to allo- cate Commodity Flow Survey (CFS) or Transearch® data into freight zonal-level trips; and • PR variant plus matrix estimation—A technique that uses basic input–output models to allocate CFS or Transearch data into freight zonal-level trips and is calibrated using either simple or elegant matrix estimation (Donnelly et al. 2008). StrengthS • The factored trip matrix relies on readily available data sources and avoids some of the aggregation errors in the three-step methods; • The simple matrix estimation and elegant matrix estima- tion techniques rely on readily available data sources, avoid some of the aggregation errors in the three-step process, and incorporate local truck traffic patterns; • The PR variant relies on readily available data, avoids the aggregation errors in the three-step process, and adds an economic component to the analyses; and • The PR variant plus matrix estimation relies on readily avail- able data, does not require the three-step methods (and thus avoids their aggregation errors), adds a national economic component to the analyses, and uses local survey data. WeakneSSeS • The factored trip matrix has no behavioral basis, no multi- modal components, and is based on national factors that may not consider local development nuances; • The simple matrix estimation has no behavioral basis, no multimodal components, is based on national factors, and cannot be readily used for forecasting in areas with rapid development; • The elegant matrix estimation has no behavioral basis and cannot be readily used for forecasting in areas with rapid development; • The PR variant is not based on locally collected data and is not readily forecasted; and • The PR variant plus matrix estimation cannot be readily used for forecasting in areas with rapid development. Quick Response Procedures Quick response procedures typically apply national default parameters to localized, regional zonal and network data to produce truck fleet size, trip, and VMT estimates for each vehicle category (Cambridge Systematics et al. 1996). These techniques include • QRFM model—QRFM does not recommend or supply a particular modeling technique; however, it provides a wealth of generalized urban freight patterns, compiled from several sources, to build a multistep freight model; and • QRFM plus matrix estimation—A technique applying either single or multiple seed (weighted by quality and level of confidence) data, calibrated using the most up-to- date truck counts, to develop multiple-class truck trip matrices, which are then factored using national growth trends, and included in a multistep freight model devel- oped using the generalized urban freight patterns outlined in QRFM (Donnelly et al. 2008). StrengthS • The base QRFM method relies on readily available national data sources, is easy to implement, and requires no local data collection. • The QRFM plus matrix estimation technique relies on readily available national data, is easy to implement, and uses some locally collected data.

27 WeakneSSeS • The QRFM method has little behavioral basis and does not incorporate locally observed characteristics. • The QRFM plus matrix estimation method has little behavioral basis and cannot be readily used for forecasting in areas with rapid development. Summary and Implications • Public and private decision makers involved in freight trans- portation must deal with an increasingly complex landscape involving rapid changes, both domestically and globally, related to population growth, economic forces, and techno- logical advances. • Significant changes in freight transportation documented in key research papers include the globalization of trade, underlying economic forces, private sector inventory and logistics practices, and centralized warehousing. • Basic types of techniques used in freight planning, forecast- ing, and modeling include economic flow models, land use and economic input–output models, commodity-based models, trip-based models, estimation routines, aggregate measures, and quick response procedures. • The strengths and weaknesses of these various techniques relate to the ease of understanding the method in question, the ability to use readily available data, the complexity of the modeling process, the relationship of the freight movement measurements to economic influences and land use, the flexibility to use different parameters for scenario-based testing, the accuracy of the modeling technique based on local data, the model’s ability to incorporate behavioral considerations, and the consideration given to complex logistics processes in freight movement (e.g., multimodal transportation, local touring and delivery). • Data typically used in freight planning and forecasting include local data sources (e.g., truck counts, land use data), the National Transportation Atlas Database (NTAD), CFS, Freight Analysis Framework (FAF), Transearch data, private sector data sets, and federal resources such as U.S. Census data, the Surface Transportation Board’s Carload Waybill Sample, or the U.S. Army Corps of Engineers Waterborne Commerce Statistics Database. • The strengths and weaknesses of these data involve their availability and frequency of updates, the cost of collecting data to fill gaps, their accuracy and suitability for planning and modeling on different geographic levels, potential errors in aggregating or disaggregating data for appropri- ate geographic scales, and the ability to establish relation- ships of attribute data with model networks. Moreover, it must be established that any of this information provides value in informing the transportation decision-making process. • The traditional four-step technique for modeling freight, which is based on the basic modeling method for passen- ger travel, has served the industry for some time but has significant shortcomings related to freight modeling: 44 The use of multiple freight transport modes is a standard business practice, but it is not captured by the models; 44 The approach does not accommodate the varying needs related to data and forecasting tools for different geo- graphic scales; 44 It does not reflect variations among transportation pat- terns for different types of commodities; 44 There is difficulty in obtaining certain proprietary data from private sources, and a lack of private data stan- dardization among the various sources; 44 Peaking characteristics for freight activity differ substan- tially from passenger travel activity, which serves as the foundation for most travel demand forecasting tools; 44 These models do not include international trends and economic considerations, yet these trends heavily influ- ence freight transportation activity in the United States; 44 The four-step technique does not capture the influence of time sensitivity in the mode choice process for vari- ous types of cargoes; 44 It is difficult to quantify local deliveries (touring) in metropolitan areas; and 44 The four-step technique falls short in identifying and quantifying the complex relationships between land uses and freight generation and attraction. Data Understanding how freight moves into, out of, and through a modeled area (i.e., nation, state, region, corridor) is an impor- tant first step to forecasting and planning for the movement of both goods and people. Providing easy-to-comprehend infor- mation on current and future freight movements helps inform decision makers about freight volumes and trends in relation to system capacity and impacts. Yet there is no single definitive source from which model developers forecast freight patterns to paint the picture of freight impacts on the transportation system. Forecasting and understanding the movement of goods within the United States requires assembling informa- tion from a variety of sources. Several national, state, regional, and local data sets provide information on goods movement at any geographic level, as well as information on how that area fits into the larger local, state, national, or global perspective. Generally, depending on the size of the planning area, freight models developed and maintained by public agencies use the following data sources: • Local data sources; • NTAD;

28 • CFS; • FAF; • Transearch data; • Other federal resources; and • Private sector data sets. Local Data Sources Local plans and studies provide information about truck traf- fic counts and forecasts of truck and passenger car travel. These sources are often used by freight modelers to fill in data gaps and identify hot spots where truck traffic causes or is entangled in traffic breakdowns. Similarly, local land use plans frequently identify the location of current or future freight development (i.e., industrial sites, freight transfer cen- ters), which is important to developers. Efficient access to and from these high-freight-traffic areas can be a major contribu- tor to future economic development. Local population and employment data provide a basis for performing simple regression analyses and likely growth sce- narios for regional goods movement. With an understanding of the current production and consumption of goods per capita and goods per job, baseline forecasts of future goods production and consumption can be developed. Changes in productivity rates are often examined, particularly on the local production forecast elements, which could affect longer- term trends. Local sources of data related to freight movement are ori- ented primarily toward the trucking industry, as trucks oper- ating on public roadways are a key consideration in most local decisions involving traffic operations and infrastructure investment. Local data commonly used in the planning and forecasting process include vehicle classification counts, which provide reliable information about vehicle size and operating characteristics, but no insight into trip origins and destinations or commodities carried. For more detailed data needs, classification data that may be readily available from state or local DOTs are often supplemented by a more detailed data collection program to obtain the needed O-D and com- modity data. Depending on the data needs and the geo- graphic scale of the study in question, these data collection efforts could include roadside intercept surveys and surveys of local business establishments involved in the transporta- tion of freight (shippers and carriers). StrengthS • Local data are sometimes readily available through exist- ing resources (local planning departments, state or local DOTs); and • Local data tend to be very accurate at smaller geographic scales and are ideally suited for freight planning at this level. WeakneSSeS • Availability of data for a specific area or project can be uncertain; and • Local data do not adequately reflect broad economic influ- ences in local freight activity. National Transportation Atlas Database NTAD is a set of nationwide geographic databases of trans- portation facilities, transportation networks, and associated infrastructure. These data sets include spatial information for transportation modal networks and intermodal terminals, as well as the related attribute information for these features. Metadata documentation, as prescribed by the Federal Geo- graphic Data Committee, is also provided for each database. The data support research, analysis, and decision making across various modes (highway, rail, and air). This database is most useful at the national level, but it has major applications at regional, state, and local scales (Research and Innovative Technology Administration 2012). NTAD also includes information on the following related transportation infrastructure: • Automated traffic counter locations; • Highway Performance Monitoring System data; • Highway and rail at-grade crossings; • Intermodal terminal locations; • National Bridge Inventory; • Ports; • WIM station locations; • FAF; and • Hazardous materials routes. StrengthS • Is best suited for larger geographic scales; and • Includes detailed descriptive data from a variety of differ- ent sources in a single database. WeakneSSeS • Is primarily descriptive in nature and not directly usable for network-based analyses; • Establishing relationships of attribute data with model net- works can be cumbersome; and • Does not include any commodity flow information. Commodity Flow Survey CFS is a primary source of national and state-level data on domestic freight shipments by U.S. establishments in mining, manufacturing, wholesale, auxiliaries, and selected retail industries. It is used in the development of the FAF and Tran- search databases. Data are provided on the types, origins and

29 destinations, values, weights, modes of transport, distance shipped, and ton-miles of commodities shipped. CFS is a shipper-based survey that is conducted every 5 years as part of the Economic Census. It provides a modal picture of national (highway, rail, air, and pipeline) freight flows and is a publicly available source of commodity flow data. CFS was conducted in 1993, 1997, 2002, and in 2007. The final version of the 2007 CFS was released in December 2009 (Research and Innovative Technology Administration 2009). CFS does not include the following information: • Forestry, fishing, utilities, construction, transportation, and most retail and services industries; • Farms and government-owned entities (except government- owned liquor stores); or • Foreign-based businesses shipping goods to the United States (domestic portions of imported shipments are cap- tured once at a U.S.-based establishment) (Research and Innovative Technology Administration 2009). Most data are available, expanded, and summarized at the national, state, or county level as long as the data are not con- fidential. Unlike the FAF databases, CFS reports flow using the North American Industrial Classification System. This method of reporting is helpful as it adds a dimension to the under- standing of freight flows in those states that report employ- ment using that classification scheme. As with the FAF data, CFS information is reported on a regional level: metropolitan statistical areas (MSAs), combined statistical areas (CSAs), and states or balances of states outside MSAs and CSAs. StrengthS • Comprehensive data include origins–destinations, value, tonnage, and transport modes; • North American Industrial Classification System–based commodity flows allow for correlation with industry- based employment data; and • Is best suited for larger geographic scales, but can be dis- aggregated for smaller regions. WeakneSSeS • Does not include all business sectors, and must be supple- mented for international freight, which is particularly impor- tant given the impact of imported goods on the transportation network; and • Is not well suited for local freight analyses; disaggregation process for subregional level can be cumbersome. Freight Analysis Framework FAF is a product of the FHWA Office of Freight Management and Operations. According to FHWA, “FAF is based primarily on data collected every five years as part of the Economic Census. Recognizing that goods movement shifts signifi- cantly during the years between each Economic Census, the FHWA produces a provisional estimate of goods movement by origin, destination, and mode for the most recent calendar year. These provisional data sets are extracted and processed from yearly, quarterly, and monthly publicly-available publi- cations for the current year or past years and are less complete and detailed than data used for the base estimate” (South- worth et al. 2010). FAF integrates data from a variety of sources to estimate commodity flows (using Standard Clas- sification of Transported Goods codes and categories) and related freight transportation activity among states, regions, and major international gateways. The FAF commodity O-D database estimates tonnage and value of goods shipped by type of commodity and mode of transportation (highway, rail, air, water, and pipeline) among and within 123 areas (MSAs, CSAs, and states or balances of states outside MSAs and CSAs), as well as to and from seven international trading regions throughout the 123 areas plus 17 additional international gateways. The 2007 estimate is based primarily on CFS and other components of the Eco- nomic Census. Forecasts are included for 2010 to 2040 in 5-year increments. StrengthS • Data has similar characteristics of CFS data, with added value of future forecasts; and • Data translated from geographic basis to transportation network includes National Highway System and National Network roadways, along with limited coverage of inter- modal connectors. WeakneSSeS • Is not well suited for local planning efforts; • Commodity flows are based on O-D pairs only; • Does not consider full supply chain activity; and • Forecasting methodology is not clear. Transearch Data Transearch, a proprietary data set developed and owned by IHS Global Insight, describes goods movement, usually at a coarse level of geography, for various modes, commodities, and industries. Transearch is an annual, nationwide database of freight traffic flows between U.S. county or zip code mar- kets, with an overlay of flow across infrastructure. The data- base draws from a variety of data sources covering commodity volume and modal flow, including a long-term, proprietary motor carrier traffic sample; proprietary railroad data; and numerous commercial and federal government surveys, samples, and census data. To compose the database, these

30 multiple and diverse information sources are placed in a single, consistent format (IHS Global Insight 2010). Most of the Transearch national data database is at the county level, except for major metropolitan areas, which are available in a zip code format. The database includes all domestic shipments and international traffic moved on U.S. infrastructure for rail, inland water, and air (for Canada only). Shipments by truck are captured for all U.S. domestic traffic of manufactured goods, and for inland international traffic including nonmanufactured goods, such as agricul- tural products, coal, ores, and nonmetallic minerals. Inter- modal truck drayage is included for international marine, domestic air, and all railroad trailer-on-flatcar or container- on-flatcar moves. Drayage for inland waterways, pipelines, international air, and rail carload transfers is not included. Examples of other excluded domestic truck traffic are • Nonmanufactured goods (e.g., from logging activities, waste); • Small-package and mail shipments moved exclusively by truck; • Military and other government trucks; and • Household goods and local service trucks (e.g., utilities, repair) (IHS Global Insight 2010). Transearch provides a variety of summary levels of data for the two-digit commodity code. This is the commonly accepted degree of resolution to understand the modal choice of certain commodities while acknowledging there may be some sup- pressed proprietary data. Should more detail be required for certain types of analyses, Transearch includes up to six-digit Standard Transportation Commodity Codes for certain commodities. StrengthS • Is a more refined geographic scale than CFS and FAF data, and thus better suited for freight planning on a more local- ized level. WeakneSSeS • As with CFS and FAF, Transearch does not consider full supply chain activity; • Delivery and touring trips and drayage activity are not cov- ered; and • Forecasting methodology is not clear. Other Federal Resources Various federal agencies have compiled databases of informa- tion related to freight activity and vehicles over the years. Population and employment information from the U.S. Department of Commerce and the U.S. Department of Labor are basic sources of information used in various freight plan- ning processes. Data from the nation’s truck weigh stations can be obtained from FHWA’s Vehicle Travel Information System. The Vehicle Inventory and Use Survey, which was compiled by the U.S. Census Bureau as part of its Economic Census every 5 years, included characteristics of the nation’s commercial vehicle fleet (e.g., vehicle size and type, average daily miles traveled, commodities carried). The latest data available through the Vehicle Inventory and Use Survey are from 2002, and the U.S. Census Bureau lists this as a discontinued data source. This type of information is useful for validating models and esti- mating VMT over large geographic areas, but it lacks any of the O-D data that are provided by other data sources dis- cussed here. The primary federal source of rail data is the Carload Waybill Sample, which is compiled by the Surface Transpor- tation Board. This database is a sample of rail waybill data provided by rail carriers, with detailed information about the shipper and receiver, O-D points, and other information about these loads. The U.S. Army Corps of Engineers Waterborne Commerce Statistics Database is the most notable source of data for mar- itime freight data compiled by the federal government. Sum- mary reports on these data, which are based on U.S. Census Bureau trade data and vessel data from U.S. Customs and Border Protection, are published annually. StrengthS • These data sources usually provide very detailed informa- tion by mode or geographic area, or both; • Accessing this public data is free or inexpensive; and • Depending on the data source, updates may be frequent. WeakneSSeS • These data sources tend to be mode specific, and do not consider full logistics chain activity; • Federal budget constraints may result in a cessation of the data-gathering and reporting process; and • These sources generally do not include commodity flow, routing, or intermodal transfer information. Private Sector Data Sets Numerous private shippers’ data sets may be used to analyze goods movement. Data are maintained within each enter- prise, and data sets are small. Because the data are proprietary, private companies are generally perceived to be reticent to share data openly and publicly in common databases. Public sector agencies succumb quickly to this blanket perception that the private data will not be shared. Data sets proliferate among producers and receivers along complex supply chains

31 in a plethora of industries, from the transaction level to the container level. Each enterprise stores its own data for use in internal applications. Subsets of these data are shared for spe- cific purposes between trading partners and shippers to pro- vide visibility in monitoring the goods and when intervening to resolve disruptions in the supply chain. The overall pur- pose is to meet delivery expectations while optimizing overall logistics and distribution costs. Supply chain data can be shared among firms, including shipping companies, because domestic and international reporting formats have been standardized and refined over the past 30 years. These widely used formats are based on traditional value-added network transmissions via electronic data interchange standards defined by the American National Standards Institute (ANSI), ANSI X12, and the United Nations Directories for Electronic Data Interchange for Administra- tion, Commerce and Transport. Many firms use other media, such as the Internet, using extensible markup language for transaction data that are based on the same data definitions as electronic data interchange. Typical data formats in wide use are • Purchase order—Information for goods; • Bill of lading—Detailed shipment bill; • Advance ship notice—Prior notification of shipment details and contents; and • Shipment status—Current status in terms of dates, times, locations, and routes. Private carriers use either internally developed proprietary models or models embedded in many top-tier software pack- ages for analysis. StrengthS • These private or combined private and public data are often more detailed than the data behind other resources discussed in this report (e.g., CFS, FAF); • Private data may offer more visibility to a full logistics process. WeakneSSeS • Data sources are often industry specific, which may not translate well to planning efforts across multiple modes and industries; • The cost of obtaining data from proprietary sources can be high; • A high level of cooperation with private interests across regions or multiple modes is necessary; and • Private data are usually gathered and stored in a variety of different formats; data processing and analyses can be cumbersome when data require aggregation among differ- ent sources. Summary and Implications Forecasting and understanding the movement of goods, regardless of geographic scope, requires assembling informa- tion from a variety of data sources, all of which are incomplete or contain inaccuracies. However, despite the current data deficiencies, several state-of-the-practice modeling methods and techniques have been developed and successfully applied within a variety of planning processes. As a result • Data issues related to the analysis of freight movements are now being discussed among the freight planning commu- nity. This much-needed dialogue can spur improvement. • Progress is being made through the development of new data sources. • Good national data exist; however, there are substantial data gaps for supporting regional and local analyses. Many agencies are now turning to developing better data sets for local movements and delivery tours. • Freight forecasters are hindered by data deficiencies, and thus have an insufficient understanding of complex supply chains to successfully develop forecasting models that address the information needs of elected officials, transportation officials, and the public regarding the impact of goods movements. • Dialogue and partnership between the public and private sectors regarding freight capacity are limited. • Public sector transportation decision making is relatively uninformed with respect to freight transportation, though several MPOs throughout the United States have engaged the private sector through freight working groups. Uninformed decision making is due to the limits of the state-of-the- practice data, which make it difficult to accurately forecast the impacts of freight on future transportation systems and also limit the potential policies and improvements that might solve expected problems; • Existing data resources are best suited to large geographic scales and do not translate well to local planning efforts. • Current planning tools and data do not accurately reflect the nature of supply chains and increasingly complex logis- tics practices in freight-dependent industries. • Documenting the various factors that influence freight transportation needs is challenging because establishing links between disparate data resources (e.g., land use, demo- graphics, employment by industry) and the freight activity that relates to these measures (e.g., truck counts, vessel activ- ity, rail activity) is extremely difficult. • Transportation forecasting and modeling practices tend to focus on average trip generation rates, but freight activity is heterogeneous and does not lend itself to average rates of production and consumption. • The growing role of third-party transportation providers makes freight less visible, which makes it more difficult to

32 document pricing and cost variables for various legs of multimodal freight transportation processes. • There are very few freight modeling and data university research centers, freight planning consultants, and freight data providers, which limits both the development and use of tools and data and the incentive to innovate. Best practices Of the current models and data being used, only a few stand out based on their technical merit. Some methods are effi- cient as a result of minimal data needs and their ready avail- ability, but their outputs and analytic capabilities are not necessarily robust. These methods therefore have limited use for in-depth analysis of freight. The best practices in models and data that currently most fully address analysis needs are presented in this section. The common underlying objective of model and data use is to analyze and document baseline conditions related to freight movement and estimate future activity based on met- rics involving economic activity, demographic changes, employment by economic sector, supply and demand of raw materials and finished products by consumers and industries, commodity flows, and other factors. Different tools and data are used by practitioners for different geographic scales, depending on the issues and scale of needs. This section lays out those identified practices that most accurately address these metrics and offer potential innovations for future prac- tices. The identified best practices • Provide a baseline assessment of models and data; • Find innovative approaches to better understand goods movement in a variety of contexts for a variety of users; • Provide a springboard for future data and model devel- opment. The underlying methodology for most tools used in freight planning and forecasting includes using resources to 1. Document existing demographic and employment condi- tions and characteristics of freight transportation (includ- ing tonnage, geographic origins and destinations, and mode of transport); and 2. Estimate future measures of freight transportation for these same parameters (tonnage, origins, destinations, modes of transport) based on changes in population and employment, productivity improvements by industry, and other economic drivers. Depending on the geographic scale, the ultimate objective of freight planning and forecasting is to forecast freight activity and its effects on local or regional conditions related to eco- nomic activity, traffic congestion, air quality, and other impacts. In addition to FAF commodity forecasts and other national freight forecasts, which have been shown to drastically under- or overestimate freight demand (Hancock 2008), researchers and practitioners have developed freight forecasting methods that use freight demand factors in various ways (Bhat et al. 2005; Sivakumar and Bhat 2002). Each method is best suited for describing different aspects of freight demand. Model developers must select appropriate methods based on how they define freight demand, their data sources, their assump- tions regarding the factors affecting freight demand, and their modeling focus. Defining Freight Demand and Factors The best current practices characterize freight demand by several dimensions, including volume, geographic scale, time period, source, transportation mode, and commodity. • Volume—The amount of freight demand being moved, typically described in terms of tons, ton-miles, or value. • Geographic scale—The spatial extent of the origins and des- tinations of freight being moved, which can be framed within a local, regional, state, national, or international market con- text. Time period—The temporal dimension of freight demand, which can constitute seasonal, annual, or short-, medium-, or long-term time frames. • Source—The basis of freight demand estimates, either as a specific area estimate (e.g., coal tonnage produced at a spe- cific mine or the volumes moved through a specific port, rail intermodal terminal, airport, or border point of entry) or as an O-D flow. Both specific area estimates and O-D flows are commonly found in regional freight plans or corridor stud- ies. These estimates describe the movement of freight within a specific area and between two specific locations; • Transportation mode—The method of transport being used. • Commodity—The freight (or goods) being shipped. The U.S. DOT’s Bureau of Transportation Statistics describes freight demand based on the following dimensions: multiple quantities (value, tons, and ton-miles), a national spatial scale, annual time periods, a general area source, and across all trans- portation modes and commodities. Freight demand is intrinsically interrelated with regional, national, and international economic and demographic char- acteristics; operational factors and logistics; infrastructure; public policy and regulations; technology; and environmen- tal factors. Changes in factors within these categories can not only cause changes in other factors, but also affect the quanti- ties and method of transport of freight demand (Cambridge Systematics 1997). Among these categories, the infrastruc- ture, public policy, and environmental factors have an indi-

33 rect impact on freight demand; in contrast, the economic, demographic, and operational factors more directly affect freight demand. Theoretically, researchers should compre- hensively consider all of these factors in freight demand models. But quantitative measures for factors are not always available, and this missing information must be accounted for by making assumptions or by narrowing the modeling focus (Eatough et al. 1998). Researchers must take these limi- tations into account when selecting a freight demand model. Depending on the type of research or planning being com- pleted, the factors have varying levels of sensitivity. For exam- ple, a study to determine the need for an intermodal facility is more sensitive to data related to mode than an analysis of warehousing and distribution facilities. The sensitivity of these factors to data has been taken into account within these best practices, which makes them more robust than methods that use existing data and perform analyses based on the limi- tations of those data. Data Best Practices The data used in the best freight planning and forecasting processes are predominantly drawn from public resources. Although national data sets are generally the most complete and accessible, they lack the detail required for local, regional, or specific freight analysis. Local data sources provide a more comprehensive scale for these analyses, but some of the data require expensive, ongoing updates. Although these sources are the best in terms of current general practices, a critical challenge in the development of freight models remains insufficient and inferior-quality data. The principal data for predicting freight transportation demand are the commodity flows by truck, rail, and water, and through selected border ports of entry and marine ports available from FHWA in the 2007 CFS (Research and Innova- tive Technology Administration 2009). In addition, FHWA has recently released FAF Version 3, an improved version of FAF that estimates commodity flows (tonnage and value) within, to, and from states and select regions by mode based on 2007 data, as well as freight movements among major metropolitan areas, states, regions, and international gate- ways (Southworth et al. 2010). Based on new estimation methods developed for this version, the forecasts developed using older versions will be updated in the near future. The most commonly used database for statewide analysis of freight movements is the commercial Transearch database developed by IHS Global Insight. Transearch estimates freight flows (i.e., commodity tonnage) by truck (i.e., for-hire truck- load, for-hire less than truckload, and private truck), rail car- load, rail–truck intermodal, water, and air at the county, business economic area, and state or provincial level (Prozzi et al. 2006; Bhat et al. 2005; Cambridge Systematics 2007). The Transearch database is a proprietary source of detailed freight data available for purchase that includes assumptions (undisclosed) to estimate and forecast movements (Prozzi et al. 2006). Some research relies on smaller freight data sets compiled by facility operators and owners, data collected by public and pri- vate entities, and data collected as part of a customized survey. Sources for these data sets range from the Waterborne Com- merce and Vessel Statistics database, to the U.S. Census Bureau’s County Business Patterns and Economic Census databases, to mail-out–mail-back surveys of freight shippers. Unfortunately, many of the data sources and databases available for statewide or MPO-level freight planning have considerable limitations as they focus on certain modes or commodities and are available at different geographic levels. Consequently, combining or integrating the data sources into a comprehensive, coherent, and consistent database is a challenging task. The Ontario Ministry of Transportation has conducted roadside vehicle surveys every 5 years since 1978 to develop truck travel and commodity flow information on intercity movements throughout the province. The Ministry is able to track Ontario-based trucks throughout North America as part of this program. Modeling Best Practices The development of models is generally constrained by the data available to populate them. If a specific model is required for an analysis, the pertinent data must either be available or collected. The models identified as best practices range from complex to simplistic and have been used successfully for their given purpose. There is no one tool that is ideally suited for every application, and the benefits and limitations for each have been identified. Trend and Time Series Analyses Trend analysis and time series analysis methods forecast freight demand through longitudinal extrapolation of histori- cal trends. Depending on the data available, this category of freight demand modeling can consider varying levels of com- plexity and aggregation. The simplest trend analysis model involves the computation of a growth factor that represents the annual compound growth rate of freight shipments, which is computed from historical aggregate freight data and applied to project future freight shipments. In order to account for temporal variations and temporal interdependencies, trend analysis is often implemented using more advanced statistical time series analysis techniques, including smoothing, autocorrelation, autoregressive mov- ing average models, and the use of neural networks. The first technique involves smoothing out various short-term or

34 random fluctuations in demand by determining patterns in the data and extrapolating into the future. These tech- niques remove random fluctuations through the use of parameters that dictate the extent to which more recent observations are weighted in isolating the trend (Cambridge Systematics 1997). Autocorrelation predicts future demand through time series regression models with temporal correlation across error terms. The correlation in these models attempts to account for the fact that freight demand at a specific time is dependent on previous time periods and needs to be treated accordingly. Autoregressive integrated moving average models are sophisticated time series modeling approaches that build forecasts from more inclusive and simultaneous analysis of complex past patterns in the time series than is achievable using simple smoothing models or models of autocorrela- tion. Autoregressive integrated moving average models pro- cess all of the information available in a time series data set with very limited information required from the researcher (Cambridge Systematics 1997). Neural network methods represent the most advanced time series models for predicting freight demand. These models assume a probabilistic progression of demand through time (e.g., freight demand in 1990 is dependent on freight demand in 1980, which is further dependent on freight demand in 1970, and so forth) and related factors ( Dougherty 1995). Although these procedures can provide short- or long- term forecasts based on projection of the smoothed underly- ing patterns in the data, they are most appropriate for short-range forecasting. Trend and time series analysis is sim- ple to use, not data intensive, and builds on historic trends to predict the future. These methods can also support modeling for freight shipments by mode, commodity, O-D pair, origin, destination, or a combination of these parameters. Due to their simple nature, however, these methods have a number of limitations. First, freight projections become less accurate when researchers use data covering shorter periods of time. Second, they assume that “the underlying economic conditions on which the forecast is based remain the same throughout the duration of the time series data and continue forward through the forecast” (Cambridge Systematics 2008). Most importantly, such models often do not explicitly incor- porate explanatory factors that affect freight transportation demand, such as changes in market factors, freight logistics, pricing, or policies. BenefitS • Is simple to use; • Requires historical information but is not data intensive; and • Supports mode, commodity, and O-D analyses. LimitationS • Accuracy is suspect if based on short-term historical data; • Assumes past trends are indicative of future activity; and • Lacks the dynamics of explanatory factors that affect freight demand. Elasticity Methods Elasticity methods are specifically used to estimate how freight commodities are split among transportation modes. These models assume that mode-choice decisions are based on the total logistics costs (TLC) associated with using vari- ous modes or modal combinations that are practical for a set of freight shipments. TLC includes the actual transport costs (or carrier charges) and other logistics costs (e.g., inventory costs, stock-out costs) incurred. The models assume that increases in TLC result in the diversion of some freight traffic to competing modes (Cambridge Systematics 1997). Researchers use price elasticity, defined as shippers’ sensi- tivity to TLC associated with a mode, to study how changes in TLC affect the quantity of freight demand shipped by each mode. Elasticity is calculated in two ways: change in demand for a mode with respect to its own price, known as a direct elasticity; and change in demand for a certain mode with respect to a change in price of a competing mode, referred to as a cross elasticity (Wilson 1980; Miklius et al. 1976). Either way, researchers can calculate point elasticity, arc elasticity, or shrinkage factors from field observations on price and quan- tity before and after a price change or from knowledge of the functional relationship between quantity and price. Because this method separates freight demand by mode, elasticity measures must be used in conjunction with other models of total freight demand. Modal diversion may be esti- mated using disaggregate data for a sample, or by using more aggregate data when the total volume of movements is sum- marized by key variables (such as commodity). The diversion estimates can then be derived from estimated changes in TLC, or when other logistics costs are unaffected by cost changes (Cambridge Systematics 1997). Elasticity measures derived from recent data sources can be useful, particularly for sketch planning applications. Elastic- ity may be computed using observed data directly (often leading to aggregate elasticity) or by estimating models on data (leading to aggregate or disaggregate elasticity). In either case, elasticity can be used to determine changes in freight traffic by mode, commodity, and corridor in response to a change in one explanatory factor. Elasticity can be short-run or long-run in nature, depending on the time period over which changes in demand are observed. Differences between short- and long-run elasticity can be substantial, as consider- able adjustments in behavior can be made in a long-term time frame.

35 Still, elasticity models face a number of challenges. Plan- ners must be careful when dealing with results from elasticity studies in the 1970s, before deregulation, which may not be appropriate in today’s context of a deregulated and highly competitive business environment. Collecting data to update elasticity studies to current practices can be equally challeng- ing, as many researchers find it difficult to collect proprietary cost information due to alliances and contracts. This problem is further complicated by the fact that many elasticity studies do not explicitly incorporate intermodal combinations. Per- haps the most serious limitation is the inability to simultane- ously account for multiple factors in predicting changes in freight traffic. As a result, it can be challenging to apply elas- ticity methods for comprehensive freight transportation planning (Hancock 2008). BenefitS • Observed or estimated data may be used; • Can be applied for short- or long-term analyses; and • Can be useful for sketch planning applications. LimitationS • Not applicable for using data prior to deregulation in the 1970s; • Difficult to collect data for inputs; and • Difficult to correlate multiple factors that affect demand. Logistic Network Models Logistic network models forecast how freight demand is divided between modes (or carriers) and travel corridors between a specific origin and destination. Like elasticity meth- ods, logistic network models can be used in conjunction with other models of total freight demand. These models are recog- nized because they consider the freight transportation system as a whole, defined by interactions among producers, con- sumers, shippers, carriers, and the government. In particular, logistic network models “assign commodity flows to a mode (or combination of modes) and specific route within a net- work that minimizes total transport costs, taking into account the location of activities within the network” ( Cambridge Systematics et al. 2008). Depending on the factors in which they are most inter- ested, researchers have two main options for modeling logis- tic networks. The first approach, known as freight network equilibrium modeling, focuses on shipper–carrier interac- tions (Tavasszy 2008). In these models, the generation of trips from each region is assumed to be known; shipper transpor- tation needs are determined and are then routed so that the carrier’s costs are minimized (Friesz et al. 1983). The second approach, known as spatial price equilibrium, focuses on producer, consumer, and shipper interactions. These models estimate trip generation by including com- modity supply and demand functions. Transportation costs are fixed values or functions of the flows on the network. Pro- ducer and consumer behaviors are incorporated through a supply and demand function for each zone. The shippers are assumed to behave according to the following two equilib- rium principles: (a) If there is a flow of commodity i from region A to region B, then the price of commodity i in A plus the transporta- tion costs from A to B will equal the price of the commod- ity in B; (b) If the price of commodity i in A plus the transportation costs from A to B is greater than the price of commodity i in B, then there will be no flow from A to B (Friesz et al. 1983). Neither approach is considered technically superior, as both face limitations due to their underlying assumptions (Cambridge Systematics et al. 2008). For example, the freight network equilibrium models’ treatment of shipper and car- rier decision-making processes presupposes that carriers will provide commodity routings that give levels of service per- fectly consistent with those levels of service perceived and anticipated by shippers. This is possible only if the shippers have perfect foresight, which is difficult due to unpredictable congestion and independent carrier routing decisions. In contrast, carriers must know shipper demands before route establishment. Therefore, shipper and carrier decisions must be modeled simultaneously (Friesz et al. 1983). Logistic network models have been shown to be best suited to larger geographies, such as intercity freight flows. They are more complex to implement than other modeling methods and often have more intensive data requirements. The implementa- tion of network models of logistics should be used as a viable long-term strategy for statewide freight demand forecasting. BenefitS • Is good for larger geographic areas; and • Considers the defined interactions among producers, consumers, shippers, carriers, and the government. LimitationS • Requires an assumption of shared perfect knowledge of freight movements between the shipper and carrier; and • Is complex to implement. Aggregate Demand Models Aggregate demand models estimate freight traffic using aggregate data that include limited information on the multi- tude of factors affecting freight transportation demand. They attempt to model the aggregate volume of commodity flow rather than the number of individual trips. These methods

36 support modeling for freight shipments by mode, commod- ity, O-D pair, origin, destination, or a combination of these parameters. The simplest aggregate demand models use a total flow approach, which uses regression-based statistical methods to calculate an overall aggregate measure of freight travel demand in an economy. The main factor considered in this model is the predicted output of economy (commonly pre- pared in conjunction with time series or cross-sectional data) (Bayliss 1988). Total flow measures of demand are typically measured in tons or ton-miles for a specific mode over a given period of time. Another approach to aggregate freight demand models is to consider relative flows, attempting to determine the pro- portion of total traffic carried by each discrete mode (Bayliss 1988). An advantage of this type of model is that, in some contexts, it may be more appropriate to use a single equation that estimates a single aspect of freight traffic demand. This method uses regression techniques to model the relative flow of one mode when compared against another. The aggregate demand model has several limitations, pri- marily due to its highly aggregate nature. Total flow approaches are much more satisfactory than the fully aggregated model when applied in a disaggregate industry or commodity con- text by mode, because the analyst would deal with a possibly more homogeneous data set. However, no attempt is really made to construct a demand model. It is also noteworthy that national output figures are usually on an industry basis, but ton or ton-mile figures are usually on a commodity basis. Rec- onciling these two data sources often creates problems. Still, aggregate models are extremely useful for freight travel demand modeling. First, aggregate data are commonly available for national as well as local scales. Second, the model can easily be applied by commodity, thus providing an esti- mate of freight demand by commodity or industry classifica- tion. It also incorporates relative modal attributes (time and cost) in determining freight traffic demand. This joint demand model is appealing for statewide freight traffic demand modeling. Most aggregate joint demand models contain two separate sets of variables with interaction effects embodied in the coefficients rather than explicitly specified in the model. As a result, aggregate freight demand models can be applied in most planning scenarios (Hancock 2008). BenefitS • Availability of required data; and • Easily applied to commodity, time, and costs to determine freight demand. LimitationS • Does not consider route choice; and • Does not estimate overall freight demand. Disaggregate Demand Models Disaggregate demand models take the methods of the aggre- gate models one step further, which offers several theoretical and empirical advantages. Specifically, these models attempt to estimate the number of individual trips on modes and links of the freight transportation network. Unlike aggregate models, they can distinguish freight demand across different routes and trips. In addition, disaggregate demand models are more accurate at identifying freight shipments by mode, commodity, O-D pair, origin, and destination. Researchers have a variety of disaggregate models to choose from that parallel the four-step urban transportation modeling process (Cambridge Systematics et al. 2008). The market survey approach involves the administration of detailed market surveys to shippers. Shippers are asked to rank various factors with respect to their importance in the modal decision-making process. These factors include such items as certainty of delivery time, charge, speed, safety, regu- larity, service to customer, packing requirements, length of haul, location of firm, method of payment, and intermodal capability. In addition, shippers may be asked to rate different modes on ordinal ranking scales with respect to these factors. The survey results are used to construct a modal preference matrix to indicate the mode chosen for a shipment of certain characteristics. This matrix is then used to determine freight shipments by mode for various O-D pairs. As this approach does not involve the use of a model per se, it is not considered useful for freight transportation planning efforts. However, the information from such surveys may be useful for con- structing disaggregate demand models. Alternatively, the behavioral mode split model predicts freight demand by focusing on the mode choice decisions made by the manager of the receiving or shipping firm. The advantage of this approach is that choice is observed at the most disaggregate level possible, namely, with respect to indi- vidual shipments dispatched by individual firms. In contrast to the market survey approach, these models are estimated using revealed choices without depending on the shipper explaining how he or she chooses a mode. Behavioral mode split models are based on the assumption that the shipper is concerned with maximizing utility (i.e., satisfaction) with respect to the various explanatory variables that affect the mode choice decision-making process. These decisions incor- porate mode characteristics, consignment characteristics, firm characteristics, and shipper characteristics. The empirical model used to estimate the demand for freight transportation within this framework is known as a random expected utility model. Because the framework assumes that the random com- ponents of the total utility function of the alternative modes are independently and identically distributed with a Gumbel distribution, the behavioral choice model takes the form of

37 the well-known multinomial logit model. Due to the shifting nature of freight logistics, this model is best suited for analyz- ing small windows of time. Additional research is needed to develop improved ways of correlating multiple decisions and predicting freight demands over longer time periods. A third inventory-based approach attempts to integrate the mode choice and production decisions made by a shipper. Variables related to production, such as shipment size, mode choice, and frequency of shipments, are treated as internal decisions. The rationale of the inventory approach is that freight in transit can be considered to be, in effect, an inven- tory of goods on wheels, similar to goods in process in the factory. The model predicts the expected total annual variable cost of hauling the commodity (Winston 1983). The basic difficulty with implementing inventory theoretic models is the acquisition of data. Several approximations have to be made in order to estimate the TLC and its role in modal choice behavior. These approximations often lead to the inventory theoretic model being very similar to the behavioral mode choice model. However, there is merit to simultaneously modeling the choice of mode, shipment size, and shipment frequency. By far the most advanced disaggregate models of freight demand are agent-based microsimulations. These micro- simulations track individual vehicles and commodities over an entire network for a given period of time. Each individual agent is assigned a set of decisions and behavior using carrier and shipper characteristics, network design, and other factors as exogenous variables. The microsimulation allows inter actions between agents and adjusts carrier and shipper behavior accordingly. Agent-based microsimulations reflect the actual process with which carriers and shippers contend (Tavasszy 2008). Researchers are also able to incorporate a variety of factors into the models to evaluate how changes in supply or operations will affect freight movements (Jinhua et al. 2003). In the end, simulations provide a comprehensive summary of how freight commodities are distributed and where vehicles are routed (Cambridge Systematics 2008). Although this method is the most accurate means researchers have for fore- casting freight demand, it also requires the most in-depth data collection, financial investment, and technical expertise. Nevertheless, many areas, such as Calgary, Alberta, have suc- cessfully implemented tour-based microsimulation of freight demand (Stefan et al. 2005). Disaggregate freight demand models have several advan- tages over aggregate freight demand models. Disaggregate models rely on microeconomic theories and richer empirical specifications that attempt to reflect real decision making. By incorporating actual modal attributes for freight movements and actual characteristics of commodities, disaggregate mod- els allow for a better understanding of intermodal competi- tion (Winston 1983). Current applications of disaggregate freight models are detailed in the Quick Response Freight Manual (Cambridge Systematics et al. 1996). Unfortunately, disaggregate freight demand models continue to be used sparingly for freight transportation applications (de Jong et al. 2004) due to the expense and challenges associated with collecting comprehensive survey data from shippers and car- riers (Cambridge Systematics 2008). BenefitS • Estimates freight demand by mode over a network; and • Is more accurate than aggregate demand models. LimitationS • Cost-intensive; and • Difficult to collect the necessary data inputs. Input–Output Models Input–output models are the simplest and, consequently, least descriptive methods for forecasting freight demand. They are used primarily in sketch planning applications, regional planning studies at an aggregate level, and when data are extremely scarce. Input–output analysis involves using economic input and output indicators to determine the levels of economic activity that may drive freight transportation demand. Inputs (e.g., capital, labor, land) are entered into an input–output analysis matrix to determine the various economic outputs. These may include the quantity of goods and services produced by type, geographic location, and temporal frame; the demand for goods and services by type, geographic location, and tem- poral frame; and other such measures of economic output. The outputs are converted into estimates of freight transpor- tation demand that would satisfy the demand for goods and services. In today’s context, when data are generally available (at least at an aggregate level), these methods are not used very often for comprehensive statewide or local freight transpor- tation modeling and planning. Instead, input–output models are used when data are scarce and time is very short. For com- prehensive statewide and metropolitan freight transportation planning, the modeling methods discussed earlier would be more appropriate as they can quantitatively estimate freight transportation demand as a function of various explanatory factors. However, such modeling approaches may benefit from peer interaction and qualitative reviews by different agents involved in freight transportation. In such a situation, input–output methods may be used to complement quanti- tative modeling approaches. BenefitS • Simple and quick to implement.

38 LimitationS • Is not comprehensive. Summary and Implications • Freight demand is characterized by a variety of factors, including quantity, geographic scale, time period, source, transportation mode, and commodity; • Freight demand models are emerging as tools to inform transportation policies; however, insufficient and inferior- quality data remain a critical challenge in the development of these tools; • Freight demand is intrinsically interrelated with regional, national, and international economic and demographic characteristics, operational factors and logistics, infrastruc- ture, public policy and regulations, technology; and envi- ronmental factors, all of which have varying data sets that are incomplete or contain inaccuracies, or both; • Freight demand model developers use a variety of methods to account for this missing information, such as making assumptions or narrowing the modeling focus when select- ing a freight model; and • Current best practices in freight demand model development include trend analysis and time series analysis methods, elas- ticity methods, logistic network models, aggregate and dis- aggregate demand models, and input–output models. Forecasting and understanding the movement of goods, regardless of geographic scope, requires assembling informa- tion from a variety of data sources, all of which are incom- plete or contain inaccuracies, or both. Despite the current data deficiencies, several best practice modeling methods and techniques have been developed and successfully applied within a variety of planning processes. Nevertheless, the lack of useful freight forecasting data has several serious implications: • Freight forecasters are hindered by data deficiencies and thus cannot completely analyze complex freight supply chains. This limits the development of forecasting models that answer the questions asked by today’s elected officials, trans- portation professionals, and public regarding the impact of goods movement. • There is limited dialogue and partnerships between the public and private sectors about freight capacity due to the lack of common understanding of the conditions and range of solutions. • Because state-of-the-practice models are limited and can- not accurately forecast the impacts of freight on future transportation systems—and the potential policies and improvements that might solve expected problems— decision makers are not adequately informed. The body of recent and ongoing research in freight data and modeling tools is extensive, including research into data development, modeling methods, and freight operations. Important practices that relate to the desired innovative ele- ments of the Strategic Plan include • Freight modeling that reflects the transportation system, land use, and economic factors—Examples of such freight modeling include the Oregon Statewide Integrated Model (SWIM2) and a series of transportation–economic models that culminated in the development of the MOBILEC model for Flanders in Belgium. Papers on both of these models were presented at the 2010 Innovations in Freight Demand Modeling and Data Symposium conducted for this SHRP 2 C20 research effort. • Modeling that reflects logistics patterns—Excluding logis- tics is one of the shortcomings of otherwise advanced fore- casting techniques. The combined PINGO and logistics models developed in Norway represents an innovative attempt to combine the transportation and economic ele- ments of traditional freight modeling with a logistics-based module that reflects real-world decision making in freight transport. These models were also presented at the 2010 Innovations in Freight Demand Modeling and Data Sym- posium. Beyond this important development, there is an enormous body of research involving private sector trans- portation practices that can support future research in this area. This research often gets little exposure in traditional public venues because it is primarily oriented toward sup- porting enhancements in private sector logistics practices, is often specific to certain industries, and is rarely used to support and inform current public sector freight forecast- ing techniques. The examples are too numerous to list, but several have been referenced in this document for illustra- tive purposes (Zsidisin et al. 2007; Cruijssen et al. 2007; Cooper et al. 1997; Bolumole 2001; Lieb and Bentz 2005; Belman and White 2005; Mello et al. 2008; Wiegmans 2010). • Integration of local touring and trip chaining—This is an important element of local freight transportation, com- prised primarily of local truck distribution and deliveries. Local touring activity has been documented as a key research need in NCFRP Report 8: Freight-Demand Modeling to Sup- port Public-Sector Decision Making (Cambridge Systematics and GeoStats 2010). This type of local truck activity is not captured in national data sets and is not modeled accurately in regional freight models. Research conducted at the Uni- versity of Illinois at Chicago (Ruan et al. 2010) is particularly innovative in that it incorporates various commodity types and various combinations of direct and peddling touring with single-base and multiple-base delivery systems. Addi- tional research at the State University of New York at Buffalo and Rensselaer Polytechnic Institute developed an entropy

39 maximization technique (Wang and Holguín-Veras 2010) to address a key limitation of traditional four-step modeling as it pertains to local deliveries. Another example of an inno- vative current practice related to local touring and chaining is the truck element of the model developed for the City of Calgary in Alberta, Canada (Stefan et al. 2005). This tool is incorporated in the Calgary EMME/2 transportation demand forecasting model, making Calgary the first major city to incorporate a tour-based microsimulation element in a regionwide transportation model. Decision-Making Needs and Gaps To establish a strategic direction for innovative freight research, an extensive review and outreach process was undertaken. The focus of the outreach was state DOTs, MPOs, county and municipal planners, toll road authorities, and port infrastruc- ture owners and operators. The purpose of this effort was to identify the data and tool needs of various decision makers in the public and private sectors and to lay the foundation for a programmatic approach to meeting these needs. Needs com- mon to both the public and private sectors were of particular interest to researchers. The outreach elements of this effort included • State DOT workshops in Washington and Ohio; • A regional freight stakeholders workshop for the Northeast held in Newark, New Jersey; • Stakeholder engagement at various conferences, including the American Planning Association’s National Planning Conference, the Innovations in Travel Modeling Confer- ence, the TRB Toward Better Freight Transportation Data Conference, and the meeting of the TRB Visualiza- tion in Transportation Committee at the TRB Semi- Annual Meeting; • A special stakeholders workshop in Washington, D.C., which included representatives of public agencies, consul- tants, and transportation industry representatives to vali- date previous outreach results; and • The 2010 Innovations in Freight Demand Modeling and Data Symposium in Washington, D.C., which was con- ducted as part of this research effort. Decision-Making Needs Although the research and outreach efforts identified a vari- ety of different needs for the wide array of participants, there were common threads and recurring themes. Freight forecasting and analysis should be enhanced through a recognized and valid inventory of standardized data sources with common definitions. One of the common items of discussion among stakeholders is the need for stan- dardization of data sources across different geographic levels and transport modes. There is little consistency among data sources for truck, rail, marine, and air transport, and they are not all ideally suited for comparable geographic scales. This makes multimodal freight planning extremely difficult. A number of stakeholders expressed great interest in devel- oping a statistical sampling of truck shipment data, similar to the Carload Waybill Sample for railroads. This would enable planners to get a microscopic view of trucking activity that would be comparable to the level of detail available for the railroad industry. Not surprisingly, a range of standardized analytic tools and applications is needed to address diverse decision-making needs. There is a generally recognized need for some stan- dardization of planning tools and methods for different geo- graphic scales, including large regions, states, metropolitan areas, and corridors. Behavior-based facets of freight decision making must be incorporated into modeling, or at least better understood as important context. One of the major deficiencies in current freight planning practice is that the tools and data are based on the movement of freight as measured in unit loads (i.e., trucks, railcar loads, tonnage) transported between origin and destination points. Freight planning and forecasting must undergo a dramatic transformation to include provi- sions for all of the complex factors that are involved in deci- sion making by freight shippers and carriers. This relates to a general need to expand the knowledge base of public sector planners and decision makers to include a more thorough understanding of private sector decision-making processes. Better information is needed to understand the nature, volume, and trends of intermodal transfers. This item relates to the need for developing real-world logistics-based plan- ning tools. One particular element of freight planning that is not always covered in current tools and data sources, but is of great interest to some decision makers, is the movement and repositioning of empty trucks, vessels, and rail equipment. Industry-level freight data are needed at the subregional level, and there is also a need to better understand local deliv- eries in urban areas. The current practice in freight planning is best suited for large geographic scales that do not translate well to local planning efforts. In addition, even the best tools and data do not accurately model the local touring aspect of freight deliveries. Freight models should incorporate local land use policies and controls to increase the accuracy of freight forecasting at the local level. Since freight transportation is a derived eco- nomic activity that is ultimately driven by consumption and production at a local level, local land use decisions have an enormous impact on freight transportation demand. The current planning tools based on population and industry

40 employment trends should be enhanced by incorporating a wide variety of land uses, especially those that are major gen- erators of freight traffic (e.g., manufacturing, warehousing, retail sales, transportation terminals). There is a need to better understand the correlation between freight activity and various economic influences such as fuel price, currency valuation, and macroeconomic trends. One of the major challenges facing many public agencies is their inability to accurately predict important changes in freight transportation activity that result from external influences and underlying economic forces. In addition, the influence of passenger traffic on shipper and carrier decisions related to routing, mode choice, time-of-day freight shipments, and other freight activity in a region needs to be understood more clearly. Conversely, decisions in industries involved in freight transportation (e.g., manufacturing, trucking, warehousing), such as site selection, production schedules, and mode choice, produce demographic and economic impacts that need to be quantified. Enhanced tools are needed to help review and evaluate freight forecasts. Any evaluation process has to account for a myriad of factors that drive freight demand. One of the inherent weaknesses of freight modeling today is that the field is so new that long-term planning horizons have not yet been reached for any of the models developed in the last 10 to 15 years. An overarching need for the freight planning practice is to develop and cultivate a process to routinely generate new data sources and problem-solving methods. This challenge points to an underlying need for innovation in freight planning and modeling and a recognition that major advances in the state of the practice are likely to be tied to the industry’s ability to harness creativity and technological advances. Attention should be given to using ITS resources and related technologies, such as GPS and IntelliDrive, to gener- ate data to support freight planning and modeling. This item was usually discussed in the context of the planning commu- nity’s understanding of the need to promote advances in tech- nology that have become commonplace in other industries. The need for enhanced visualization tools for public outreach related to the freight planning process was also mentioned frequently. There is a need to develop a full multimodal, network- based freight demand model that incorporates all modes of transport (vehicle, railcar, vessel) to a similar level of detail for various geographic scales. To be truly effective, this ambi- tious effort would have to address some of the other needs identified in this section: namely, the need to more fully understand the underlying economic drivers in freight trans- portation and the need to incorporate real-world supply chain and logistics practices in the planning process. The development of such a model is the ultimate goal of the freight planning and modeling community. Of great interest are benefit–cost analysis tools that go beyond traditional financial measures by including other direct and indirect benefits and costs (public and private). Tools would include metrics to assess environmental and eco- nomic development policy initiatives on a comparable basis with standard financial measures. More effective methodologies are also needed to apply freight forecasts to funding and finance analyses, such as revenue pro- jections. These types of tools are of great interest to toll road authorities and owners and operators of freight infrastructure such as port terminals, whose future needs and financial stabil- ity are tied to the ability of the owners and operators to develop accurate forecasts of demand by mode and commodity. Highway authorities have a strong interest in the develop- ment of tools that would let them use freight forecasts to sup- port their infrastructure design processes. This need relates particularly to the relationship between truck volumes and weights and highway infrastructure (e.g., bridge and pave- ment design). Agencies with oversight responsibilities for inland waterway systems that serve as important freight links have similar needs. Stakeholders consistently emphasized the importance of a concentrated effort to develop the requisite knowledge and skills to support freight analysis. The factors that drive freight transportation demand are complex and require an under- standing of a wide range of topics, such as economics, political science, demographics, transportation planning, engineering, finance, information technology, and organizational skills. This need for knowledge and skills also relates to the need to understand the goals and objectives of shippers and carriers in the private sector and planners in the public sector. Bridging the gaps between the needs of the public and private sectors would help facilitate more effective planning and forecasting. Decision-Making Gaps Table 3.1 shows the decision-making needs, the gaps between the needs and the current modeling and data practices, and the data and modeling requirements to meet those needs. Articulating the capabilities of the current state-of-the-art models and data sets and comparing them with the needs of decision makers sets the stage for identifying the modeling and data needs to fill the gaps. These needs are the foundation for the actions incorporated in the Strategic Plan. research program The SHRP 2 C20 Freight Demand Modeling and Data Improve- ment Strategic Plan advances a broad new direction for improv- ing freight planning, promoting continuous innovation for breakthrough solutions to freight analytic and data needs, and fostering a collaborative approach for private, public, and aca- demic stakeholders.

41 Table 3.1. Freight Decision-Making Needs and Gaps Decision-Making Needs Gaps Between Needs and Current Practices Data or Modeling Requirements to Close Gaps Standardized data sources with common definitions • Various data sources collected through different programs result in extensive inconsistencies. • Homogeneous data for ease of incorporation into freight models and for consistency of freight models in different regions. • Reduction in data manipulation to improve accuracy. Statistical sampling of truck shipments • Detailed knowledge of truck movements in local areas. • Understanding of current truck activity by different industry segments (long-haul, local, drayage). • An ongoing standard data-collection program to gather local truck movements. • Compilation of truck data to a level comparable to rail industry data (i.e., Carload Waybill Sample). Standardized analytic tools and applications • Wide range of various tools that require unique data sets. • Consistency in modeling approaches and data needs for similar geographic scales. Inclusion of behavior-based elements into freight models • Current practices use truck movements and com- modity flows, but should be based on the behav- iors, economic principles, and business practices that dictate the movement of freight. • Current modeling tools do not accurately reflect real-world supply chains and logistics practices. • Determination of the influencing behavioral factors that affect freight movement and ongoing data collection to inform models. • Behavior-based freight modeling tools to take advantage of newly collected data sets for various geographic analyses. • Incorporation of intermodal transfers, consolida- tion and distribution practices, and other shipper and carrier practices in modeling tools. Data development to under- stand the nature, volume, and trends of intermodal transfers • Public sector access to intermodal transfer data of containers, bulk material, and roll-on–roll-off cargo is lacking for most transfer facilities other than those of large ports and rail yards. • Data sets developed through collaboration with the private sector to inform the planning practice knowledge base and models on intermodal transfers. • Protocols to collect data on a regular basis. Industry-level freight data development at a sub- regional level and within urban areas • Freight data are generally not industry-specific, which translates into forecasts that are not sensitive to the unique industry trends that are critical to regions that rely heavily on specific industries. • Industry-level forecasts that are sensitive to the unique factors of different industries. • Tools and data at a disaggregated level (local) that can be aggregated for larger geographic analyses. • Tools and models to take advantage of the new data sets. Incorporation of local land use policies and controls for better local forecasting accuracy • Current freight data and models lack local detail related to the generation of freight activity, which hampers local efforts to effectively plan for the last mile. • Enhanced understanding of land use decisions and their implications on freight activity. • Resources for local organizations to incorporate land use considerations into freight planning data and models. Development of a correlation between freight activity and various economic influences and macroeconomic trends • Freight models are typically based on population-, employment-, and industry-level productivity forecasts, with no consideration for the impacts of other economic factors. • Enhanced models that incorporate a wide array of economic factors in forecasting freight demand. Better accuracy of freight forecasts • Freight models rarely (if ever) are reviewed to see how accurately they are forecasting, calling into question their reliability and validity. • A systematic approach for freight model and data owners to review and evaluate forecasts (every 3 to 5 years) and adjust models and data methods accordingly. Development of a process to routinely generate new data sources and problem- solving methods • The improvement of freight planning nationally depends on continuing innovation and steady progress in the development of models, analytic tools, and knowledge acquisition. • A value-adding and sustainable process to generate new and innovative ideas. • Acknowledgment of failed practices that can contribute to the knowledge base of practitioners. Use of ITS resources to gener- ate data for freight modeling • Technologies that can be used to collect freight data have not been used to their potential. • Data can provide a wealth of information related to current conditions and diversions as a result of traffic incidents. • An understanding of the information needed by the modeling community and the standard to which it can be used. • An accessible data bank for freight modeling developed with the cooperation of GPS device providers, ITS infrastructure owners, and other data providers. (continued on next page)

42 Table 3.1. Freight Decision-Making Needs and Gaps Decision-Making Needs Gaps Between Needs and Current Practices Data or Modeling Requirements to Close Gaps Development of a universal multimodal, network-based model for various geo- graphic scales • The fragmentation of modeling techniques and data means that practitioners typically must develop or improvise data and models for their own applications. • Agencies with fewer resources are not able to adequately analyze freight movements. • Some freight transport modes are analyzed more than others because they have more data avail- able for analysis. • An open-source data bank and universal freight modeling tool is the ultimate goal. • A level playing field among different modes of freight transportation in terms of quantity and accuracy of data and complexity of modeling tools. Development of benefit–cost analysis tools that go beyond traditional financial measures • Analysis of the benefits of project-based scenar- ios lacks the precision required for those deci- sions, including direct and indirect impacts, costs, and benefits. • Tools that incorporate a comprehensive analysis of the factors associated with infrastructure development, expansion, and enhancement spe- cifically related to freight. Development of funding assessments resulting from freight forecasts • Transportation funding scenarios and what-if analyses are limited in their ability to forecast rev- enues associated with freight movement. • Estimated costs and potential funding sources that can be justified based on credible freight forecasts. Creation of tools to support the infrastructure design process • Infrastructure design, unless specific to freight, rarely focuses efforts on how best to accommo- date freight movements. • Incorporation of freight forecasts into infrastruc- ture design related to vehicle size and weight and future freight activity (i.e., tonnage) by mode. Development of knowledge and skills among the freight planning community as a foundation for improved analysis • The freight planning community is relatively small and knowledge transfer is challenging. • Talented innovators who can lead new approaches to freight transportation planning are pursuing careers in other industries. • A comprehensive knowledge base for planning professionals that includes the wide range of sub- ject areas related to freight transportation. • Greater recognition or formal standing of freight planning as a profession with an associated body of knowledge. (continued) This research program is built on a foundation of seven strategic objectives that have been identified as the basis for future innovation in freight travel demand forecasting and data. The desired direction for enhanced freight planning, forecasting, and data analysis expressed by the many stake- holders who participated in this project are reflected in these objectives, which are aimed at stimulating innovation through the avenues laid out in the accompanying strategic plan. The seven strategic objectives are 1. Improve and expand the knowledge base for planners and decision makers. 2. Develop and refine forecasting and modeling practices that accurately reflect supply chain management. 3. Develop and refine forecasting and modeling practices based on sound economic and demographic principles. 4. Develop standard freight data (e.g., CFS, FAF, and possible future variations of these tools) to smaller geographic scales. 5. Establish methods for maximizing the beneficial use of new freight analytic tools by state DOTs and MPOs in their planning and programming activities. 6. Improve the availability and visibility of data among agen- cies and between the public and private sectors. 7. Develop new and enhanced visualization tools and tech- niques for freight planning and forecasting. Building on the foundation of the seven strategic objectives listed above, the SHRP 2 C20 research effort culminated in the development of 13 research areas, described in this report as sample research initiatives. Collectively, these sample research initiatives constitute a programmatic approach for systemati- cally improving freight modeling and data availability and fore- casting tools. Each of these initiatives is tied to one or more of the seven strategic objectives, with the ultimate goal of promot- ing and cultivating innovation through Strategic Objectives 2 and 3, supported by the innovations in data development in Strategic Objective 4 and visualization in Strategic Objective 7. Each of the 13 research initiatives also relates to one or more of the three main research dimensions identified at the 2010 In - novations in Freight Demand Modeling and Data Symposium: • Knowledge relates to a general understanding of freight transportation issues and the extensive array of elements involved in planning and forecasting freight demand; • Models are the tools used to carry out planning and fore- casting activities at various geographic levels; and • Data are the underlying information resources for model- ing and planning efforts, and often represent an important limitation of modeling. The ultimate long-term goal for the research documented is to build on Strategic Objectives 2 and 3 to promote the devel-

43 opment of a full network-based freight forecasting model that incorporates all modes of freight transport and accurately reflects the various factors related to the supply of freight infra- structure and services (Strategic Objective 2) and the underly- ing demand for these services (Strategic Objective 3). This model will effect a dramatic change in current freight planning and forecasting. It is a highly ambitious endeavor because of the complexity of freight transportation and the numerous ele- ments that are necessary to achieve this long-term goal. The other five strategic objectives are tied to this goal through the development of the applicable knowledge base needed to further it (Strategic Objective 1), the development and dissemination of data necessary to support it (Strategic Objectives 4 and 6), and the development of enhanced meth- ods for disseminating information from these analytic tools for public stakeholders (Strategic Objective 5) and decision makers (Strategic Objective 7). Although development of a full multimodal network-based freight forecasting model is the ultimate long-term goal, it is important to note that freight transportation has not tradition- ally lent itself to innovative planning and forecasting methods. This is because freight transport has historically been a relatively uncomplicated, low-tech process. In addition, past experience in freight transportation does not necessarily correlate well with future freight activity due to short-term changes in the forces of supply and demand. As a result, developing accurate forecasts for freight transportation will require a radical paradigm shift in the way the practice is currently conducted. These research initiatives are based on the SHRP 2 C20 research conducted for this effort, but they should be viewed in their proper context as steps in support of the seven strategic objectives. The specific research initiatives are initial recom- mendations for potential research to help move this process forward. These recommendations will likely change as a result of funding availability, industry needs, and developments that spring from some of the other elements of the Strategic Plan, such as the Global Freight Research Consortium (GFRC); future data and modeling symposia recommended in this study; other data and modeling innovations featured in TRB conferences; and NCHRP and NCFRP research. Identification of Freight Modeling and Data Innovations Innovations in freight modeling and data have primarily been borrowed from technologies developed for other purposes and then applied to facilitate a particular process related to freight movement. For example, GPS, after being developed by the Department of Defense and made available to the general public, provided drivers with point-to-point route navigation tools. Only recently have captured truck route–related data been used to provide valuable information to planners to better understand truck movements. Another example involves the data collected by weigh-in-motion (WIM) technology, used for the enforcement of safety and trucking regulations, and the subsequent use of that data to track truck movements along major highways throughout the United States. For the purposes of this research, innovations in the freight modeling and data community are defined as significant (or potentially significant) movements toward the betterment of freight models, tools, data, or knowledge in freight planning practices. Innovations were identified through the review of past practices, research reports, studies, and entries for the symposium held as part of this research. Innovations are ongoing and dynamic. Many innovations in the freight industry are not published because their appli- cations for freight movements have not yet been realized— similar to the way GPS and WIM data took some time to be applied to freight modeling and data collection practices for use in freight planning. Other innovations that seem promis- ing at first may actually be lacking in detail or applicability. The innovations identified as part of the research effort are by no means exhaustive. Ongoing and future initiatives may hold great promise for freight planners. The Strategic Plan is the starting point for setting the course to foster new innova- tions, allowing the industry to pursue promising areas of research and develop methods for improved decision making by learning from innovations that are not quite ready for planning applications or are found to be subpar for such applications. Early identification of potential initiatives and allowing for setbacks assists in the growth of the practice and should not be discounted outright. Assessment of Current Freight Transportation Technologies and Innovative Programs Several promising technologies in use and under development may affect freight forecasting tools. This section includes high- level assessments of these technologies and programs, includ- ing how each technology is used, by whom, the opportunity costs associated with not fully integrating these technologies into the modeling process, and what prevents developers from using them. To successfully implement these techniques, sev- eral challenges must first be overcome. As seen in Table 3.2, these institutional and technical issues limit how technologies are used in, and developed for, freight modeling. Table 3.2 acknowledges the current challenges in freight model development, but as Table 3.3 shows, it is equally clear that many potential opportunities exist to improve how the process of freight planning can be addressed to meet both public and private sector needs. Technological advancements offer some of the greatest opportunities to improve the reliability and accuracy of goods

44 movement planning. Some technologies are familiar to almost anyone with a modern cell phone, and others are emergent from the freight industry itself. Taken together, existing and emergent technologies offer possibilities to improve processes and close knowledge gaps. Current Freight Technologies Global positioning systems (GPS) and intelligent transporta- tion systems (ITS) can collect freight data via transponders that trace individual trip activities, but they do not collect key trip characteristics, such as commodities hauled, shipment size, or trip end activity. However, these technologies serve as a basis of how such data collection might be accomplished when combined with other information systems. gLoBaL PoSitioning SyStemS GPS technology is based on a geosynchronous global satellite system that provides location and time information anywhere Table 3.2. Institutional and Technical Issues Affecting Freight Modeling Advancements Institutional Issues Lack of a freight analysis national vision Because state-run freight models are affected by out-of-state policies and activities, it is important to estab- lish a national vision for interstate freight analysis for efficiency and improvement purposes. National guide- lines regarding model structure, data requirements and collection, and calibration and validation would help MPOs, regardless of their size. A national freight analysis system would identify states that are primarily through states for freight movement and fully use their freight movement. Insufficient data and data collection Data limitations are the primary obstacle to developing and upgrading freight demand models. Inability to analyze market- driven changes Freight demand models do not always include changes in market demands or economic trends. For example, the Texas DOT analyzes mode shifts from a policy perspective, but does not analyze market-driven mode shifts, such as trackage rights. Inability to analyze policy- driven changes Decision makers do not always have quick access to required information (e.g., fuel costs, climate change– related information, energy usage) when considering policies. Quick response analytic tools are important for policy-driven changes, but they are not always available. Lack of multimodal modeling Multimodal flows and the interactions between truck, rail, water, and air modes are important for efficient modeling, but they are not always available. Technical Issues Lack of data Detailed models require detailed data, which are not always available. Truck distribution errors Truck route forecasting is limited because it incorrectly distributes trucks across all roadways evenly and thus does not accurately account for common trucking routes. Limited modeling scopes Models only provide a small portion of a larger transportation picture. Because freight demand models have not been fully integrated with economic models, they do not sufficiently relate freight transportation improvements to economic developments. Inconsistencies in modeling No clear standards exist regarding data input and validation, methodologies, model validation, calibration, or updating and recalibration. This lack of standard procedures contributes to the black box effect of freight demand modeling. Missing relevant information Certain relevant information, such as time-of-day and seasonal demand, are not incorporated into freight demand models. Lack of multimodal diversion estimates Many freight models are limited to truck movements, even though other freight modes are sometimes used. Table 3.3. Opportunities for Model Improvement Behavioral Better consideration of private sec- tor decision-making criteria. Multimodal network Public sector applications of private sector network models. Network design Improved applications of private sec- tor terminal and facility location models for public sector purposes. Supply chain and logistics Methods of forecasting cargo chain- ing may be of particular interest to private sector and economic agen- cies while freight vehicle chaining may be of more interest to public transportation agencies. Improved routing and scheduling A consideration of internal freight trips, possibly related to supply chain topics. Hybrid commodity models or applicationsa Improved real-time freight data. a Beagan 2009.

45 on Earth. Location estimates are achieved through triangula- tion of a time differential as measured by the time it takes for a signal to leave a satellite and reach a ground-based receiver and the time stamp indicating when the signal was sent. By measuring the difference in time stamps between the various received satellite signals, GPS receivers can estimate location, trajectory, and speed. A minimum of four (or more) line-of- sight satellite signals are necessary to estimate location, but the more satellite signals received, the more accurate the estimate of location and speed will be. Although the GPS satellite sys- tem only provides one-way information to receivers, onboard GPS devices can translate time and location to information that can be transmitted to ground-based data management sys- tems. This technique allows highly accurate real-time location data to be transmitted and collected by shipping companies or even a GPS-enabled cellular phone. The data transmitted or collected by ground-based GPS devices have entered both the public and private sector in day-to-day operations. Because privacy concerns related to GPS-based data have become less pronounced over time, GPS combined with other technologies may offer a rich source of information to support future data and model development efforts. • Safefreight Technology uses SecurityGuard™ GPS tracking devices that can either be mounted to a vehicle or embed- ded in cargo (Safefreight Technology 2010). The system uses real-time tracking to manage fleet systems. UPS uses a GPS tracking system to track the progress of shipments with a user-friendly web application that plots the custom- er’s shipment progress (Belt 2008). • Integrating GPS and freight technology has resulted in greater efficiency, as well as increased customer services and satisfaction. This technology could be developed to track more cargo options, which could result in better carrier per- formance, which in turn would decrease lost revenue and make the shipment process more efficient and smoother. inteLLigent tranSPortation SyStemS ITS is a name given to “the application of advanced informa- tion, electronic, communications, and other technologies” to maximize the efficiency of the transportation system and address surface transportation needs (Donnell et al. 1998). ITS technologies include those as basic as signal preemption and as complicated as dynamic weather and integrated de icing systems. Some of the ITS technologies below offer particular promise for the advancement of freight movement data col- lection or model development efforts in the coming years. CommerCiaL VehiCLe oPerationS A commercial vehicle operation is an automated preclearance system that checks a commercial vehicle’s weight, safety, and credential status and monitors individual commercial vehicle activities. This system would typically be used by the manag- ers of a trucking company. Vehicle information is transmitted in real-time to a centralized computer system under the con- trol of a team of dispatchers and stored as a concise electronic record called a snapshot (Cutchin 2005). Information is con- veyed through a satellite navigation system, a small computer, and a digital radio in every truck. This system allows for the central office to know where the trucks are at all times. Addi- tional functions include commercial vehicle clearance (on-board electronic tag identification that provides vital credentials, such as vehicle weight, safety status, and cargo); automated roadside safety inspection (automatically gener- ated vehicle credentials that are stored in the vehicle’s elec- tronic tags and can be read electronically); on-board safety monitoring (detects vehicle problems such as load imbalance, shifts in the load, load temperature changes, an open door, low tire pressure); hazardous materials incident response (timely electronic cargo information in the event of an acci- dent); automated administrative processing (data processing such as tax information and automated fuel reporting); and basic commercial fleet management (vehicle, driver, carrier, and cargo information). Individual loads are tracked using a bar-coded container system, and pallets to track loads com- bined into a larger container. If a truck is lost or delayed, the system can divert the truck to a more efficient route. Auto- matic vehicle identification readers are installed at exit gates; these gate systems detect the vehicle’s electronic tag while the software references the container and related information and posts it to a secure web-based database. FedEx, which operates one of the best proprietary systems, tracks its shipments through the method described above and achieves better than 99.999% on-time delivery. Load-tracking systems use queues, linear programming, and minimum spanning tree logic to predict and improve arrival times. Efforts to implement these new technologies may fail because they may not be used properly or not to their full potential. Benefits of using this technology include • Safety enhancements, such as reduced congestion at weigh stations (reduced accident risk) and freeing law enforcement to concentrate their efforts on high-risk and uninspected carriers and operators; • Simplicity, such as automated screening (improved enforce- ment efficiency), automated administrative filing, and quicker information gathering; and • Savings, through reduced costly paperwork, eliminating unnecessary weight and safety inspections, eliminating man- ual filing, more efficient government license processing and revenue and tax collection, and less cost to implement the commercial vehicle operation system than to maintain and construct new weigh stations (Pratyush 2003).

46 automated Container tranSPort Automated container transport (ACT) refers to an automated system of controlled vehicles on dedicated lanes between ports and terminals. The framework of this system would be a (most likely underground) network of conveyer belts that would transport goods. Similar to how underground net- works transport sewage, gas, oil, and water, ACT would carry goods in an underground freight network. This system would be an expansion of the already-existent road, rail, air, and water modes of transport. A simpler ACT-like system was used in Paris and Berlin until the end of the twentieth cen- tury, when it was damaged by floods. Benefits of the ACT system include a reduction of labor and operating costs (Decker 2008). The European Combined Terminal in Rotterdam, Nether- lands, was the first fully automated container terminal in the world. All containers in the terminal are handled by auto- mated stacking crane and automated guided vehicle systems. These automated vehicles can carry 20-, 40-, 45-, and 50-foot containers that weigh up to 40 tons. The system uses free- ranging-on-grid (FROG) technology to navigate the terminal (Spasovic 2004). radio frequenCy identifiCation Radio frequency identification (RFID) uses a passive, active, or semipassive transponder device with memory for data storage, a small battery, and an antenna for receiving and transmitting a radio signal used to establish real-time loca- tion. RFIDs are used in supply chain management and manu- facturing, as well as in processing on-time deliveries. RFID systems consist of interrogators (which read information) and tags (which are the readable labels). These tags can be applied to or incorporated into a product, animal, or person for the purpose of identification and tracking. RFID technol- ogy can be used for toll collection, machine-readable travel documents, and airport baggage tracking logistics, as well as to track retail goods, people, and animals. For example, microSD cards are being used by DeviceFidel- ity in Dallas, Texas, to store bank account information and to electronically pay mobile phone bills (D’Hont 2004). A per- sonal identification number is used to secure sensitive infor- mation stored on the RFID system (Swedburg 2009). The Housing and Development Board of Singapore implemented a parking ticket system using RFID that replaced the paper season parking ticket (Tay 2007). The first E-Passport was used in Malaysia in 1998 to log information about a citizen’s travel history. Many toll roads, both nationally and inter- nationally, use RFID technology to improve efficiency. An E-ZPass system—which is currently being used in Massachu- setts, Delaware, New Hampshire (Reino 2010); Maryland (Maryland Transportation Authority 2008); New Jersey (Tri-State Transportation Campaign 2002); and a few other states—uses RFID to create an easier alternative to the typi- cal, manual toll road framework. In South Korea, T-money cards can be used to pay for public transit and, in some stores, used as cash. In Hong Kong, mass transit is paid for almost exclusively through the use of RFID-enabled Octopus Cards (Mas and Rotman 2008). RFID is also used to track animals on large ranches and in rough terrain and to identify crucial animal information. If a packing plant condemns a carcass for safety purposes, RFID information can identify the animal’s herd of origin. RFID technology was used after the outbreak of mad cow disease. Potential problems and concerns about RFID include data flooding (which requires a means of filtering raw data), the need for global standardization, security concerns (because of world-readable private, corporate, and military informa- tion), exploitation, passport hacking (the encryption on U.K. chips was broken in under 48 hours), shielding (to block unwanted reading of data), and hardware susceptibility (vibration and high temperatures may loosen an RFID con- nection). Privacy concerns are a variable because of the per- ceived threat of RFID implantation in humans and potential government control (BBC News 2004). Because of these con- cerns, an anti-RFID campaign was launched in Germany (Hansen and Meissner 2007). maChine ViSion Machine vision refers to automated systems of digitization, manipulation, and analysis of images used for traffic monitor- ing, navigation, and transport safety, as well as for detecting lane markings, vehicles, pedestrians, road signs, traffic condi- tions, traffic incidents, and driver drowsiness. These systems could potentially offer a source of generalized freight data. Emergent Technologies and Processes Perhaps one of the most exciting areas of freight data and methods development will come from emerging technologies that at present have limited real-world application but are suf- ficiently mature to merit testing and evaluation. These tech- nologies, built to improve safety or delivery efficiency, can provide new sources of data to improve the goods movement planning processes used by both public and private sectors. automated guided VehiCLeS An automated guided vehicle uses an on-board automated system of vehicle cruise control, lane departure warnings, collision avoidance, and obstacles detection. A FROG vehicle uses sensors to navigate and orient itself on a grid (or map), such as posted calibration points along its route. An operator uses a call button, similar to an elevator call button, to activate the FROG. When the vehicle arrives, the operator enters the vehicle and selects a destination via an onboard

47 touch-screen computer, and the FROG vehicle automatically drives toward the destination, leaving the operator free to perform other tasks. Statistics (pickups, deliveries, move- ments, events) are all recorded over a wireless network and stored on a management system. Automated guided vehicle systems typically have only one or two vehicles, only a few pickup and drop-off locations, and simple road systems (John 2009). When operating, an automated guided vehicle uses rear laser bumpers, side safety bumpers, and optical sen- sors to detect obstacles; that is, the safety bumpers are active in all directions of vehicle travel (Deaton 2008). PiCkuP CenterS Pickup centers are convenient local collection and distribu- tion depots, or boxes, where consumers can pick up goods they have ordered. Scheduling, online booking, detailed ship- ping manifests and reports, shipment tracking, and confir- mation and delivery notices are all provided at the pickup center. Freight agents locate qualified carriers with proper insurance and licensing, negotiate preferred rates, and man- age all the processes from initial pickup through final delivery to ensure a timely and cost-effective shipment. Platinum Worldwide Logistics (a division of Pak Mail Centers of Amer- ica, Inc.) is an example of a pickup center that provides inte- grated, global freight forwarding solutions for businesses and consumers via ground, air, and water, as well as shipping logistics (Pak Mail WorldWide 2010). freight forWarderS, BrokerS, and third-Party LogiStiCS ProViderS These relatively new companies use new communications technologies to directly link worldwide business-to-business and business-to-consumer transactions through an out- sourced one-stop shop. Third-party logistics providers typi- cally specialize in integrated operation, warehousing, and transportation services that can be customized based on the needs of the customer and the demands of the market ( Murray 2012). Internet access to management information is becoming more popular; it provides intranet and extranet access, administration, support, and management. Third- party logistics providers also provide legal assistance, such as liability concern analysis and contract and risk assessment (Marsh 2007). Innovative Freight Programs Several innovative approaches to programs and studies have been initiated around the world to harvest some of the fruits of new technologies. U.S. DOT has several initiatives that have introduced freight data and freight management and goods movement optimization to several urban areas around the United States. These programs have evolved from a variety of needs ranging from safety, U.S. Customs and North American Free Trade Agreement requirements, pavement design, and web-based coordination. These initiatives are described below. Smart roadSide initiatiVe The Office of Freight Management and Operations partnered with the Federal Motor Carrier Safety Administration (FMCSA) to introduce advanced roadside technologies to conduct inspections and measurements that traditionally have been delivered at weigh station sites. The Smart Road- side Initiative is a system deployed at strategic points along commercial vehicle routes to improve the safety, mobility, and efficiency of truck movement and operations on the roadway (IntelliDrive Program 2010). Smart Roadside Initia- tive technology includes wireless roadside inspections, elec- tronic truck size and weight enforcement, electronic driver credentialing, customs and borders preclearance, and advanced traveler information systems to establish communication pro- tocols (Trucking Industry Mobility & Technology Coalition 2008). Although a standardized system would be used, the initiative would not be exclusively used by a government. Four applications would be deployed: (1) E-screening a vehicle while it is in motion to detect safety issues; (2) improved truck size and weight enforcement; (3) direct information transmitted from the vehicle to the roadside carrier system to a government system; and (4) commercial vehicle parking information, including advanced route planning decisions that would survey hour-of-service constraints, location and supply of parking, travel conditions, and loading and unload- ing. The Smart Roadside Initiative, as a concept, would create efficient data handling between private and public sector motor carrier systems while maintaining current operational systems (IntelliDrive Program 2010). CroSS-toWn imProVement ProjeCt The Cross-Town Improvement Project is a high-level concept program that incorporates an intermodal move database for coordinating crosstown traffic to reduce empty moves between terminals. The program also tracks intermodal assets and dis- tributes information to carriers wirelessly. The intermodal move exchange facilitates the exchange of load data and avail- ability of information between railroads, terminal operators, and trucking companies. Chassis utilization tracking pro- vides a means for chassis owners and users to accurately account for asset use. Real-time traffic monitoring provides a means to obtain the up-to-the minute information about roadway conditions, travel speeds, and predicted travel times that is captured by traditional roadway sensors and traffic probes. Dynamic route guidance uses input from real-time traffic monitoring and a geographic information system source along with simulation tools to provide real-time visual

48 routing around congested areas. Wireless drayage updating provides a means to wirelessly and inexpensively exchange information with drivers regarding trip assignments, traffic congestion information, trip status, and location information through a truck-mounted driver interface device (Cross- Town Improvement Project 2010). e-Permitting and VirtuaL Weigh StationS The Office of Freight Management and Operations recently completed two program initiatives that addressed electronic permitting and virtual weigh stations (VWS). VWS can improve the operational efficiency and effectiveness of states’ roadside enforcement programs by targeting commercial carriers that have a history of poor safety performance or commercial vehicles that are known to be overweight. The program would provide additional information on tracking, weather, and traffic conditions for system managers. Cur- rently, however, VWS technology can only be applied with the current state of technology, which is primarily license plate information, vehicle identification numbers, and U.S. DOT numbers. Research is creating a more efficient and more reli- able system of identification through camera recognition; however, some states are concerned that such optically based VWS identification will be unable to achieve a perfect identi- fication system with current technologies. Because human interaction is still required to screen, enforce, and issue cita- tions for compliance issues, VWS technology could be slowed by manual operations, thereby reducing its efficiency. These potential problems may be remedied by the development of architecture for e-permitting and virtual weight, determining which vehicle identification technology is best suited to iden- tify all commercial vehicles, conclusively documenting the benefits of VWS, and investigating the deployment of direct enforcement concepts in the United States (Federal Highway Administration 2009a). uniVerSaL truCk identifier ProjeCt The Office of Freight Management and Operations initiated a program to identify the advanced technologies capable of uniquely identifying commercial vehicles subject to U.S. Code Titles 23 and 49 inspections and measurements. WIM devices are designed to capture and record truck axle weights and gross vehicle weights as they drive over a sensor. Unlike older static weigh stations, the vehicle can be in motion while information is being gathered, which makes WIM devices more efficient (Federal Highway Administration 2009b). Benefits of WIM include quicker processing rates, safety improvements through decreased vehicle accumulation, con- tinuous data processing, increased coverage with lower costs, minimized scale avoidance (i.e., monitoring trucks without alerting truck drivers, thereby providing more reliable data), and dynamic loading data (Norikane 2008). WIM short- comings include a reduction in accuracy, reduced informa- tion (information that is usually collected at static weight stations, such as fuel type, state of registry, year model, loaded or unloaded status, origin, and destination cannot be col- lected), and susceptibility to damage from electromagnetic transients (Washington State DOT 2010b). eLeCtroniC freight management Electronic freight maintenance (EFM) is a U.S. DOT– sponsored program that applies web technologies to improve data and message transmissions between supply chain part- ners, enabling process coordination and information sharing for supply chain freight partners through public–private col- laboration. A common electronic freight framework would improve the efficiency and productivity of the transportation system. The initiative would first test the concept at the truck–air freight interface, and then move on to other modal interfaces, such as truck–truck, truck–rail, rail–sea, and truck–sea. EFM provides shippers (the supply chain owners) with visibility to meet very tight performance standards and improve operational efficiencies by offering uniform access to existing customized MySQL database formats, computing platform independence, and adaptable services. The value and operational efficiencies grow as more supply chain part- ners link into EFM. Because international trade accounts for a quarter of the U.S. gross domestic product, the trend toward globalized trade places new burdens on those organizations involved in freight movement. EFM provides improved data quality, administrative cost reduction, more efficient opera- tions, better supply chain agility, extended supply chain visibility, improved supply chain security and resiliency, and the ability to coordinate business processes across organiza- tions (Intelligent Transportation Systems Joint Program Office 2009). CoLumBuS efm ProjeCt The Columbus EFM project was a successful 2007 deployment test that implemented web services and other components to support an existing international import truck–air–truck sup- ply chain. The Columbus EFM project encompassed a broad, worldwide air cargo supply chain and was successfully deployed from overseas suppliers in China to distribution centers in Columbus, Ohio. The deployment test focused on the pilot test of a portion of a single supply chain. The evalu- ation of the deployment is especially important because it tests potential government impacts and wider industry impacts while quantifying the benefits. The goal of the EFM program was to provide a platform to exchange information among trading partners on a many-to-many basis over the web (Intelligent Transportation Systems 2008).

49 Innovative Freight Studies Several ongoing freight studies have explored how data from modern technologies can be integrated to advance freight planning. These efforts bridge various technologies to solve specific problems, identify trends, and gauge the willingness of private shippers to embrace technology in their day-to-day business. euroPean truCk teChnoLogy SCan imPLementation ProjeCt Ten transportation professionals participated in a scan of six European nations (Slovenia, Switzerland, Germany, the Netherlands, Belgium, and France) sponsored by the Office of Freight Management and Operations, the American Associa- tion of State Highway and Transportation Officials (AASHTO), and NCHRP. The scan focused on the use of advanced tech- nologies employed to support truck size and weight enforce- ment. The technologies observed included “a high speed WIM system, video/photograph capture, [and] handheld/portable equipment” for real-time selection of noncompliant vehicles (Federal Highway Administration 2009c). Unique enforcement technologies such as heavy goods vehicle control sites in Switzerland (including a three- dimensional vehicle profile scanner, a full gross vehicle weight static scale system, an automated citation issuance system, and full safety inspection facilities) and bridge weigh-in- motion systems in Slovenia and France were tested (Federal Highway Administration 2009c). WaShington State freight PerformanCe and moBiLity imProVement Study The Washington State Transportation Research Center tested commercial vehicle information systems and networks (CVISN) electronic truck transponders using software to link the transponder reads from sites anywhere in the state to col- lect specific truck movement data in order to benchmark when and where the monitored trucks experienced conges- tion. The goal of the CVISN program is to improve safety and security, simplify operations, improve efficiency and freight mobility, and move toward nationwide deployment. To achieve these goals, the U.S. DOT is targeting high-risk oper- ators, integrating various systems and improving the creden- tialing and screening process (Federal Motor Carrier Safety Administration 2010). CVISN uses WIM scales and tran- sponder readers to electronically screen trucks from about a half-mile away as they approach a weigh station. The weigh stations use carrier and vehicle snapshots (electronic records) to support screening decisions. These snapshots contain information such as the carrier’s current safety rating based on the state’s computer files, a historic review of the carrier’s safety records, an overview of the last safety inspection for the vehicle being screened, and the expiration date of any Com- mercial Vehicle Safety Alliance decal on the vehicle being screened. In 2009, Washington State DOT reported that “transponder- equipped trucks were pre-cleared and received more than 1,048,000 green lights at Washington weigh stations” with an average weigh station stop of 5 minutes. The DOT estimates that this yielded industry savings of approximately 87,000 hours of travel time and $6.5 million dollars (Washington State DOT 2010a). Oregon uses a similar Green Light weigh station preclearance program, which is arguably the nation’s best screening program. The program has precleared truckers 10.5 million times since January 1999, which has resulted in an estimated $9.8 million in savings per million preclearances (Oregon DOT 2008). nationaL roadSide teChnoLogy SurVey FMCSA conducted a perception survey that documented nationwide motor carriers’ input regarding the use of road- side technologies to electronically identify commercial vehi- cles and the sharing of commercial driver data. The project included perspectives on GPS and transponder benefits, cost implications, and information sharing. The study focused on what motor carriers would like to see in the future. FMCSA states the Roadside Technology Corridor’s goal as having “a series of specially-equipped testing facilities at weigh stations to demonstrate, test, evaluate, and showcase innovative safety technologies under real-world conditions in order to improve commercial truck and bus safety. Data gathered from experiments and field tests along the corridor will be used to support FMCSA enforcement and compli- ance programs, state safety programs, policy research, and future rulemaking activities” (Federal Motor Carrier Safety Administration 2011). Summary and Implications Model developers forecast the impacts of freight patterns to support the development and implementation of policies and infrastructure improvements to enhance safety, effi- ciency, and the overall effectiveness of goods movement on national, state, regional, and local transportation systems. Forecasting and understanding the movement of goods, regardless of geographic scope, requires assembling informa- tion from a variety of data sources, all of which are either incomplete or contain inaccuracies. Recent studies have shown that technology is playing an increasing role in data collection, policy development, and private shipper response to market stimuli. Current and emergent technologies will make the ground more fertile for such endeavors and, with the right framework, will provide opportunities for data

50Table 3.4. Sample Research Initiatives Sample Research Initiativesa Research Dimensions Strategic Objectives Knowledge Models Data 1. Improve and expand knowledge base. 2. Develop modeling methods to reflect actual supply chain management practices. 3. Develop modeling methods based on sound economic and demographic principles. 4. Develop standard freight data to smaller geographic scales. 5. Maximize use of freight tools by public sector for planning and programming. 6. Improve availability and visibility of data between public and private sectors. 7. Develop new and enhanced visualization tools and techniques. A: Determine the freight and logistics knowledge and skill requirements for transportation decision makers and pro- fessional and technical personnel. Develop the associated learning systems to address knowledge and skill deficits. l n B: Establish techniques and standard practices to review and evaluate freight forecasts. l n M M C: Establish modeling approaches for behavior-based freight movement. l l n D: Develop methods that predict mode shift and highway capacity implications of various what-if scenarios. l l n n E: Develop a range of freight forecasting methods and tools that address decision-making needs and that can be applied at all levels (national, regional, state, metropolitan planning organization, municipal). l l n n M F: Develop robust tools for freight cost–benefit analysis that go beyond financial considerations to the full range of ben- efits, costs, and externalities. l l M n G: Establish analytic approaches that describe how elements of the freight transportation system operate and perform and how they affect the larger overall transportation system. l l n M H: Determine how economic, demographic, and other factors and conditions drive freight patterns and characteristics. Document economic and demographic changes related to freight choices. l n I: Develop freight data resources for application at subregional levels. l M M n J: Establish, pool, and standardize a portfolio of core freight data sources and data sets that supports planning, programming, and project prioritization. l n n M K: Develop procedures for applying freight forecasting to the design of transportation infrastructure, particularly pavement and bridges. l n L: Advance research to effectively integrate logistics practices (private sector) with transportation policy, planning, and programming (public sector). l M n n M: Develop visualization tools for freight planning and model- ing through a two-pronged approach of discovery and addressing known decision-making needs. l l l n Note: Directly Addresses Objective n; Indirectly Addresses Objective M. a The sample research initiatives outlined as part of the SHRP 2 C20 research project demonstrate how the strategic objectives could be advanced. Each initiative also applies to one or more of the three research dimensions (indicated by l).

51 mining and evaluations that can effectively serve the needs of both public and private decision making. Some of the major findings from this section of the research follow: • The majority of freight technologies and innovative pro- grams use GPS or ITS technologies, or both; • A multitude of current freight technologies, programs, and innovative studies are not fully used by freight model developers due to institutional challenges that limit the use of the data; • Institutional challenges that might explain why these tech- nologies are underused include a lack of national direction and modeling standards, the proprietary nature of the data sets, and the cost and time required to collect (or purchase) the data and develop the models. In addition, the economic downturn has reduced the urgency (and resources) with which elected officials and transportation planners had pre- viously supported the development of accurate, complex freight models; and • The need for effective and efficient transportation systems to move freight will most likely become an issue once the economy rebounds. Shrp 2 C20 Sample research Initiatives Thirteen research areas were identified for future pursuit that flow from the strategic objectives described above. The pri- mary goal of the proposed research initiatives is to promote, cultivate, and support innovative research related to the modeling of freight activity based on real-world supply chain and logistics practices and tied to an enhanced understanding of the demographic and economic influences of freight activ- ity. These efforts would be supported by new data development methods for small geographic scales, as well as improved visualization techniques for freight planning. This research is aimed at advancing an industrywide vision for developing a comprehensive network-based freight forecasting model that incorporates all freight transportation modes and can be applied at different geographic scales. An overview of the sample research initiatives and their relationships to the strategic objectives is shown in Table 3.4. Detailed information for each of the sample research initia- tives, along with potential research projects and implementa- tion time lines, is provided in Chapter 4.