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.
118 This chapter summarizes the current literature on data collection techniques and data sources. The review focuses on two main topics: â¢ Data collection techniques. This section reviews surveys and other data collection techniques that are relevant to the study of FG/FTG. It includes the various survey tech- niques and newly emergent technologies that can be used for FTG data collection, such as Global Positioning Sys- tems (GPS). â¢ Data Sources. This section reviews data sources that are currently available for FG/FTG analysis and modeling. It describes primary data sources that are directly related to freight activities, and secondary data sources, such as GPS data and other Intelligent Transportation Systems (ITS) related data. Data Collection Techniques Freight data collection techniques are intended to address multiple needs. Data of various kinds must collected, including: (1) types and amount of commodities shipped; (2) modes of conveyance utilized; (3) origins and destinations; (4) shipment and terminal travel times; (5) loading and berthing require- ments; (6) daily and hourly variations in shipments; and (7) frequency of shipments. The number, type, and weight of commodities carried in relation to the nature and magnitude of the activities served are important in developing FG/FTG models, and in accessing the need for roadway and access improvements. Data needs and collection is probably âthe biggest single issueâ challenging the development of freight modeling (Kuzmyak 2008, 38), and has been the focus of sev- eral NCHRP studies. For example, the data needs, sources, and collection techniques for FTG are the main focus of NCHRP Synthesis 298: Truck Trip Generation Data (Fischer and Han 2001). NCHRP Synthesis 384: Forecasting Metro- politan Commercial and Freight Travel also discusses the data issues of freight modeling, and identifies cost and confiden- tiality as the two most challenging issues. Under the freight demand modeling framework that includes FTG as a sub- component, Holguin-Veras et al. (2001) and HolguÃn-Veras et al. (2010) investigated the freight data issues in a broader sense. This section summarizes previous efforts, and current data collection practice. Data Needs for FTG Similar to other modeling frameworks in transportation, the accuracy, resolution, and coverage of relevant data elements are important for FG analysis and modeling. Tables 1 and 2 obtained from HolguÃn-Veras et al. (2010) provide a frame- work for FTG data collection, analysis and modeling. The type, location, and intensity of various activities are a basic input to the data category shown in Table 81. A description of the role and importance for each of the various data classes is provided in HolguÃn-Veras et al. (2010). Table 82 presents a summary of data needs pertinent to FG and FTG, as adopted from HolguÃn- Veras et al. (2010). In the table, âCâ stands for models for cal- ibration and âFâ stands for models for forecasting purposes. Review of Data Collection Procedures Data collection is necessary in order to supplement the data sources currently available. This section describes the key findings from a comprehensive review of freight data collec- tion approaches. As previously discussed, there are key issues involved in freight transportation that affect the efforts of conducting freight transportation surveys, and the different means of collecting data. These key issues are: (1) multiplicity of metrics to define/measure freight; (2) multiplicity of factors to determine freight/freight trip generation, distribution and the other factors that determine demand; (3) multiplicity of economic agents involved; and (4) agents that only have a par- tial view of the freight system. All of these aspects complicate A p p e n d i x F Review of the Literature on Data and Surveys
119 Data class Items Freight generation data (amount of commodities, vehicle trips, deliveries) Production Consumption Delivery tours Sequence of stops Location of deliveries Commodity, vehicle-trip OD flows Empty trips Economic characteristics of participating agents Shippers, warehouses, forwarders Carriers Receivers Spatial distribution / Location of participating agents Shippers, warehouses, forwarders Carriers Receivers Network characteristics Travel times, costs Use restrictions Capacity Traffic volumes Other economic data Production functions Demand functions Input-Output technical coefficients Source: HolguÃn-Veras et al. (2010) Table 81. Data categories for FTG modeling. ytido m mo C noitareneg sledo m tuptu O-tupnI sledo m pirt ytp m E sledo m pir T noitareneg sledo m Information/insight into logistical pattern of flows C* Production C C, F** C Consumption C C, F C Sequence Location OD flows C, F C, F Empty flows C Shippers C, F C, F Carriers C, F C, F Receivers C, F C, F Shippers C, F Carriers C, F Receivers C, F Travel times and costs C, F C, F Use restrictions C, F C, F Capacity C, F C, F Traffic volumes Production functions Demand functions IO tech. coeffs. C, F *: "C" stands for data for Calibration purpose **: "F" stands for data for Forecasting purpose Freight generation data Aspect: Other economic data Delivery tours Economic characteristics of participating agents Spatial distribution / Location of participating agents Network characteristics Source: HolguÃn-Veras et al. (2010) Table 82. Data needs for alternative FTG modeling approaches.
tremendously the data collection process. Consequently, it seems clear that a comprehensive approach to freight data collection is best, and to fully describe what happens in the system, a combination of methods may be required. In general, the different data collection techniques or sur- veys could be grouped depending on how the sampling frame is defined (i.e., on the basis of the establishments at the origin or the destination of the shipment, the truck traffic, cargo tour). This translates into collection procedures that focus on the origin or destination of the cargo; en-route, as in a truck intercept survey; or along the supply chain the ship- ment follows. Table 83 presents a summary of the different data collection methodologies depending on their sampling frame. For each frame, the table discusses its application, and the type and collection method generally used, together with the strengths and limitations of each. A clear representation of the level of detail of the data provided by each unit or sam- pling frame is shown in Table 84. As said before, no single sampling frame can provide a good representation of all the data categories required for freight demand modeling. As shown in Table 84, these sampling frames do not provide information on freight traffic volumes, which are also needed, for instance, to assess the impact of freight volume on traffic congestion. Collection of freight volumes is mainly performed via Automatic Vehicle Classifier (AVC), or manual counting. Manual counting involves a trained observer collecting vehicle classification counts at a location based on direct observation of vehicles. Alternatively, this can be done using videography, which involves collecting vehicle classification counts using video tape recorders and tallying them manually by observing vehicles on the video with the ability to stop time and review data, if necessary (Beagan et al. 2007). On the other hand, AVC is usually based on techniques such as Weight-In-Motion (WIM), consisting of loop detectors, video cameras, or other types of detectors to automatically classify vehicles and collect freight volume (Sharma et al. 1998). The full installation of 120 Description Application Type/Collection Methods Strengths Weaknesses reppihS Provides measures of total sales, market share, materials quantity/cost, modes, produc- tion hours, and location data. Examples include: Commo- dity Flow Survey, and Annual Survey of Manufacturers. Self-administered or staff- assisted surveys to agents that ship out the cargoes. Ability to capture data about the characteristics of the cargo. May be complemented with shipment tracking. Questionable validity about routes, intermediaries, pro- cessing /transfer points, etc. data. -iece R rev Targets the receivers of the shipments. Freight/Freight Trip demand generation models. Self-administered or staff- assisted surveys. Can provide excellent data about the goods received. Receivers are unaware of the cargo transportation aspects. reirra C Most widely used approach to collect freight data. Examples include: Freight Movement Survey and the Highway Carrier Attitude Survey (17) Based on Vehicle Registra- tion samples. Provide vehicle detailed travel information (trip diary forms). Mail-out or CATI surveys (18) . Target population relatively easily defined. Collects good travel patterns data. Vehicle list obtained from Depart- ment of Motor Vehicles. Questionable quality of car- go related data. Mismatch bet- ween vehicle registration lists and commercial vehicle popu- lation in urban areas (18). Travel diaries. Collect travel diaries for a period of time from a sample of trucks operating in the region. Useful for understanding internal-internal truck trips in an urban area. Difficult sampling process. Using vehicle registration samples may produce biased results. Low response rates. Assisted by GPS to track the routing patterns inside the study area. Spatial /temporal movement data could be collected; real time data. GPS cannot provide the data collected by traditional surveys. Focus on data collection along the supply chain. Individual shipments are tracked long a supply chain. Use shipments as transpor- tation measurement units. Capture economic relations vital for transport policy. Longitudinal surveys: individual shipments are tracked along a supply chain. Provides a comprehensive description of supply chains. Tracks each shipment from shipper to final receiver. Expensive, budget may condition their success or failure. Requires a very specific survey design (20). Focus on truck /vehicle trips. Collect: routing patterns, OD locations, commodity/ truck type, weights, shipper/recei- ver/carrier data (16) Freight modeling and planning applications. Roadside Interviews. Low costs (16) . High response rates. Best statistical control and reliability. Capture trucks entering/ exiting, and passing through the study area. Limited locations may lead to sampling bias. Potential traffic disruption. Cannot collect tours data. Not effective for internal-internal truck traffic data (16). Roadside postcard survey distribution to be mailed back. Less likely to disrupt traffic than roadside interviews; re- quires fewer field personnel. Response rate is usually lower, which could result in significant nonresponse bias. License plate recording /matching with a survey mailed out to be returned (19). Does not disrupt traffic. Lag between observation and survey reception may lead to low response rates (and bias) and high recollection errors. desab tne mhsilbats E Individual vehicles as the sam- pling unit. Collect: origin, des- tination, trip mileage, travel time, routing, purpose, time of day, commodity, shipment size, truck type, land use, activity at trip end (16) Trip chaining, trip genera- tion, and trip routing (16) Collect travel pattern data; origins / destinations at the perimeter of a region; routing patterns; truck/ commodity type; vehicle/ cargo weight; and, shipper / carrier / receiver information. Freight modeling/planning applications such as: the development of OD freight flow matrices, commodity tonnage distribution to truck classes, empty and through truck factors (16) desab elcihe V desab ruo T rc ep t etni pir T desab nodro C Sources: (Beagan et al., 2007), (Jessup et al., 2004), (Cambridge Systematics, 1996), (Miller et al., 1993), (Rizet et al., 2003) Table 83. Summary of data collection methodologies.
121 WIM, however, may be expensive, and is only deployed at lim- ited locations. Other AVC methods include: pneumatic tubes, loop detectors (or other types of magnetic detectors), and video cameras. Pneumatic tubes are easily portable, and need only to be placed across travel lanes to automatically record vehi- cles. However, the classification accuracy degrades where there is simultaneous crossing of multiple vehicles, such as on high- volume, high-occupancy road segments. Loop detectors involve embedding one or more loops of wire in the pavement, which are very useful under all weather conditions, and used mainly as permanent recorders at locations where counts are required for a longer time duration (Beagan et al. 2007). Role of Global Positioning Systems (GPS) on Data Collection In recent times, there has been a great deal of interest in the use of GPS for freight demand modeling. Among other benefits, these data are: very accurate; increasingly common as the number of companies using GPS devices multiplies; and free, as they are the byproduct of vehicle tracking and navigation systems. However, a fundamental limitation that has not been overcome is that GPS cannot collect the key data that traditional surveys provide (e.g., commodity type, ship- ment size, trip purpose). This presents a number of issues. First, there is no guarantee that the data are representative of the region, as in most cases, the data are biased toward medium and large firms. As a result, the data lack observa- tions for the small companies that transport the bulk of the freight in urban areas. Second, although delivery tours can be estimated from GPS data, shipment sizes and the purpose of the stop are unknown. These are important implications that severely hamper the use of GPS for freight demand modeling. As a result, the maximum utility of GPS is realized when it is combined with other data collection methods. For example, origin, destination and routing information received from GPS receivers can be used to validate and improve the information provided by truck drivers in manually completed travel diaries. Also, combining GPS truck trip information with Geographic Information System land use data can yield useful information on truck activity characteristics at trip ends (Beagan et al. 2007). Data Sources The development of FG/FTG models requires information on freight movements, and the characteristics of the activities that are served. This section discusses the principal freight data sources. Overview of Data Sources Several TRB synthesis reports, the Bureau of Transporta- tion Statistics, and a variety of research studies provide useful information for freight data sources: â¢ NCHRP Synthesis 298 (2001) summarizes and discusses main data sources for FTG. Data categories include: Compendia of Trip Generation Data, Engineering noitcudorP noitp musno C ecneuqeS noitaco L s wolf D O s wolf ytp m E sreppihS sreirra C sreviece R sreppihS sreirra C sreviece R stsoc ,se mit levar T snoitcirtser es U yticapa C se mulov ciffar T eciohc edo M e mit yrevile D setubirtta edo M snoitcnuf noitcudorP snoitcnuf dna me D .sffeoc .hcet OI Shipper Carrier Receiver Unit/ Sampling Frame Trip intercepts Vehicle Tour Excellent level of detail Good level of detail Some level of detail Low level of detail Only general information Establishment atad ci monoce reht O noitareneg thgier F atad sruot yrevile D ci monoc E fo scitsiretcarahc stnega gnitapicitrap / noitubirtsid laitapS fo noitaco L stnega gnitapicitrap kro wte N eciohc laicepS sessecorp Table 84. Sampling frame of different data collection procedures.
122 Primary Data Sources There are a large number of primary data sources on freight activity, though the coverage they provide is still lacking. For a comprehensive review of data sources, the reader is referred to HolguÃn-Veras et al. (2010). Among them, the CFS data and the ZIP code Business Pattern (ZCBP) data are of great importance to FG/FTG modeling, although they are not currently used for FG/FTG. The team is developing a proposal to the United States Census Bureau to obtain access to these two types of data. CFS and ZCBP data are briefly reviewed in this section. Commodity Flow Survey (CFS) Data Among all data sources, the CFS data collected by the Cen- sus Bureau and the Bureau of Transportation Statistics every 5 years since 1993 is of particular importance to FG/FTG model- ing. The CFS is a collaborative effort among the Census Bureau, U.S. Department of Commerce, the Bureau of Transportation Statistics, and U.S. Department of Transportation. The CFS collects data regarding cargos originating from selected types of businesses located in the 50 states and the District of Colum- bia (U.S. Census Bureau, 2010a). Table 85 summarizes the CFS data elements that were collected for the 2007 CFS survey. The table shows both the general information that is collected by CFS, and additional information that is required if the ship- ment is an export or hazardous material. For shipments that include more than one commodity, respondents are instructed to report the commodity that makes up the greatest percentage of the shipmentâs weight (U.S. Census Bureau 2010a). The CFS has collected a massive amount of data. For instance, the CFS 2007 collected data from 100,000 establishments that were required to provide information about all shipments for four different periods of 1 week of the year. The CFS data are used to produce the standard tabulations of commodity flows Studies, Special Generator Studies, Port and Intermodal Terminal Data Resources, Vehicle-Based Travel Demand Models, Commodity-Based Travel Demand Models, and Other Critical Data Resources. It also summarizes the data sources presented in each category. â¢ NCHRP Synthesis 384 (2008) identifies commodity flow related data resources (e.g., Freight Analysis Framework, Commodity Flow Survey, TRANSEARCH database, Vehi- cle Inventory and Use Survey, Vehicle Travel Information System, Carload Waybill Sample and WaterBorne Com- merce Statistics Database). For urban freight data, NCHRP Synthesis 384 indicates that vehicle classification counts and OD Survey data may be used. â¢ The Appendix II in a report for NYMTC by Holguin-Veras et al. (2001) summarizes the main freight data sources (up to year 2001), using a standard format that has been adopted by the Directory of Transportation Data Sources (Bureau of Transportation Statistics 1996). The summary provides the geographical coverage for the data sources, defines the cor- responding transportation modes, and identifies the list of attributes in each dataset. The data sources are also classified according to their potential use with information on col- lecting agency, contact person, means of contact, and other useful information about content, methodology, significant features and/or limitations, distribution media, availability, price and web site. A usefulness-ranking is included, and provided for each data source including the categories: very useful, useful, marginal and specialized. â¢ A report by HolguÃn-Veras et al. (2010) updates the 2001 NYMTC report. The update was presented as notes to the original data sources to avoid confusion, since some sources may have been subject to name changes or were discontin- ued. In addition to updates for the original sources, HolguÃn- Veras et al. (2010) documents a description of new data sources or reference material and tools. Shipment Types Data collected Domestic destination or port of exit Commodity Value Weight Mode(s) of transportation Date of shipment Indication of whether the shipment was an export or hazardous material Mode of export Foreign destination city Country Hazardous materials UN/NA code * *: UN: United Nation number; NA: North America number General Exports Table 85. CFS data elements.
123 â. . . until these technologies are in wider use, their applica- tion to truck trip generation data will be limited.â The freight industry has been a pioneer in using GPS for dynamic manage- ment of large fleets. As a result, over the years, a large amount of GPS data has been accumulated. However, the use of these data for FTG modeling has been relatively sparse. Similar issues apply to ITS data. There are a number of underlying reasons: (1) the proprietary and commercially sensitive nature of the data, which prevents the sharing of the data among different interest groups (e.g., data owners, decision-makers, practitio- ners, and researchers); (2) although highly accurate, the data do not contain much information of the kind transportation modelers need to build meaningful models (e.g., trip purposes, company characteristics); and (3) the level of penetration is still relatively small toward large and sophisticated companies. However, it is important to recognize that data from sec- ondary sources could be useful. GPS provides accurate infor- mation about a delivery tour (e.g., number of stops, dwell times) and travel times of individual vehicles that, as noted by NCHRP Report 298, are important, and usually hard to obtain via other data collection means. GPS data can also provide estimation of the temporal distribution of freight generated at an origin. For instance, GPS logs can provide an exact count of FTG produced by the distribution center if all vehicles are equipped with GPS. This may significantly reduce data col- lection efforts as the only thing needed is data describing the company characteristics. On the other hand, since GPS data are usually samples, they only provide a partial picture of the freight traffic. As small carriers are less likely to equip their fleet with GPS monitoring systems, GPS data can be biased since they tend to reflect the delivery patterns of large companies. Table 86 summarizes the use of ITS data and GPS traces in FTG modeling, as well as their advantages and disadvantages. In general, ITS and GPS data can be used to calibrate various FTG models such as estimation of model parameters. The GPS traces can also be used to estimate the tours of delivery vehicles, such as the stops along the tour, dwell times, and paths of a specific tour, which are useful for developing FTG models that integrate both commodity-based and trip-based formulations. at the current stage. The data from the CFS are currently also used by public policy analysts, and by transportation planners and decision-makers to assess the demand for transportation facilities and services, to assess energy use, safety risks and envi- ronmental concerns (U.S. Census Bureau 2010a). The CFS data however have not been widely used for freight modeling. Expanding the use of the CFS micro-data to estimate parame- ters of freight demand models would, at the same time, enhance the usefulness of the CFS, making it easier for practitioners to estimate freight demand models, and providing a significant boost to freight transportation modeling research by making available high quality data. ZIP Code Business Patterns (ZCBP) ZIP code Business Patterns (ZCBP) is one of the three pro- grams developed by the Census Bureau to cover most of the countryâs economic activities. (The other two are County Busi- ness Patterns and Metro Business Patterns.) As described by U.S. Census Bureau (2010b): âCounty Business Patterns is an annual series that provides sub-national economic data by industry. The series is useful for studying the economic activity of small areas; analyzing economic changes over time; and as a benchmark for statistical series, surveys, and databases between economic censuses.â The ZCBP data is crucial since âBusi- nesses use the data for analyzing market potential, measuring the effectiveness of sales and advertising programs, setting sales quotas, and developing budgets. Government agencies use the data for administration and planning.â ZCBP data include the number of establishments, number of employees, and payroll data by NAICS industry. ZCBP data were first available in 1994, and the most recent release for this dataset was 2007. Secondary Data Sources Secondary data sources include data related to ITS and GPS, which are of increasing importance to FTG and other types of freight-related modeling. There are, however, some obvious challenges, as stated in the NCHRP Synthesis 298: Data class Use in FTG modeling Advantages Disadvantages Accurate Aggregated information Can be automatically collected Hard to infer individual behavior Model calibration Accurate Privacy/proprietary Estimation of delivery tours Can be automatically collected Not representative (sparse samples or biased) Computationally expensive (large amount of data) GPS traces ITS â vehicle classification data (i.e., vehicle mixes), truck counts, traffic travel times, network travel time or delay Model calibration Table 86. ITS and GPS data, and their use in FTG modeling.
124 When using ITS or GPS data, one should recognize that the data do not have direct connections to characteristics of freight trips such as shipment purposes, and shipment type or size. Therefore, they should be used with care; particularly one should provide appropriate freight contents to ITS and GPS data, and make sure these data can be âexplainedâ from freight activity perspectives before being used in FTG modeling. Ideally additional data that are directly related to freight trips should be collected, for use with the ITS and GPS data, so that the connec- tions between ITS/GPS data and freight trips can be established. One example is to collect GPS traces in combination with travel diaries. In this way, not only stops and dwell times of a tour are collected, but also the purpose and delivery amounts at each stop along the tour. This will provide a more complete picture of a delivery tour, one which could be valuable in developing more advanced/accurate freight FTG models. Summary There is lack of primary FG/FTG data. This lack of pri- mary data is a major issue because of the need to effectively incorporate freight transportation into the planning process. The fact that many of the sources of FTG models are now dated, and that one of the most important primary freight data sources, the CFS, has not been widely used for freight modeling only exacerbates this problem.