Guidebook for Developing Subnational Commodity Flow Data (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.

59 Collecting Subnational Commodity Flow Data Using Roadside Truck Intercept Surveys 3.1 Introduction This section provides a comprehensive examination of methods for developing subnational commodity flow data using roadside truck intercept surveys. Roadside truck intercept surveys are conducted by interviewing truck drivers while they are in the process of operating their vehicle. Data is gathered from each truck driver regarding recent stops, future stops, goods being carried, and other specialized information related to the trip. These surveys typically occur at the side of a roadway, but also can occur at the entrance or exit of a major freight facility such as a port terminal gate or an intermodal rail yard. Unlike establishment surveys, which were described in Chapter 2.0, roadside intercept surveys cannot be used to develop comprehensive commodity flow surveys or studies because road- side surveys focus exclusively on truck movements and miss shipments by any other mode. If the shipments that are intercepted are actually part of an intermodal shipment (for example, intercept surveys conducted at the gates of an intermodal terminal) they will not capture the true origin and destination of the shipment. Because the respondents to the survey are carriers instead of shippers or receivers of goods, the truck drivers may have only limited information about the characteristics of the shipment. This is especially true if the carriers are carrying mixed shipments (as would be the case with a less-than-truckload carrier). Even with these disadvantages, there are some advantages to conducting roadside surveys. The biggest advantage is that these surveys provide more accurate data about corridor-level flows. As noted in the introduction to the Guidebook, there are a number of applications that benefit from this corridor-level detail. Roadside survey data also can be used in combination with other sources of commodity flow data to validate the results of efforts to model the routing of commodity flows. The Guidebook identifies the following nine general steps involved in administering a roadside truck survey data collection program: • Step 1—Site Selection • Step 2—Questionnaire Design • Step 3—Selecting Survey Dates and Times • Step 4—Sampling Issues • Step 5—Interviewer Training • Step 6—Site Preparation (including traffic control) • Step 7—Utilizing Uniformed Officers and Vehicles • Step 8—Data Quality Control Procedures • Step 9—Preparing Data for Usage in Commodity Flow Development. C H A P T E R 3 . 0

60 Guidebook for Developing Subnational Commodity Flow Data Many of these steps are interrelated, but the discussion of each step is ordered as shown in the above bulleted list. The description of each step is structured to focus on the following four key elements described in the Playbook (Chapter 6.0): 1. Key Considerations—A brief description of the main issues encountered and tradeoffs that will need to be made for the step. 2. Implementation Process—A detailed description of how to implement the step. 3. Example—An example of how this step has been implemented in other studies. Many of the examples in this chapter are taken from two roadside intercept surveys conducted in Washington, the Eastern Washington Intermodal Transportation Study (EWITS) and the Strategic Freight Transportation Analysis (SFTA) study. 4. User’s Guide Worksheet Punch List—Simple bulleted instructions that Guidebook users can check off to ensure that they have implemented each of the major steps involved in developing an establishment survey. Each of these four elements is designed to focus on different aspects of conducting a roadside truck intercept survey and to reflect the types of activities that might be undertaken by a state or local transportation agency. For transportation agencies that are considering hiring a contractor to conduct a roadside survey, reading the “Key Considerations” section of each step will likely provide enough information for the generation of a request for proposals (RFP) on the topic. Transportation agencies that want to understand the details of how to conduct a roadside survey should begin by focusing on the “Implementation Process” sections, and then move on to the “Example” section. The “Example” section will describe previous specific efforts that have been undertaken in other regions. These examples can provide important lessons to improve existing efforts in an area or help to evaluate contractor responses to RFPs. After transportation agencies have a sufficient background in all of the aspects related to developing a roadside truck intercept survey, the “User’s Guide Worksheet Punch List” sections can be used to walk the agency through all of the specific steps that need to be completed to implement the survey. This section also can be used to help evaluate RFP responses. 3.2 Step-by-Step Process for Conducting Roadside Truck Intercept Surveys Step 1—Site Selection Key Considerations The choice of locations for conducting roadside surveys is primarily driven by the practical constraints of feasible data collection locations. The ideal loca- tions for roadside truck surveys are truck weight and inspection stations, because they have the space to accommodate parking for large trucks and they have equipment that can safely intercept trucks from the traffic stream. However, the locations of these facilities in most states is such that they tend to over-represent intercity trucks when compared to the universe of commod- ity flow patterns that might be encountered in urban areas. Additionally, using these facilities also requires cooperation from the facility operators, which can be challenging to secure if there is any sense that the surveys could interfere

Collecting Subnational Commodity Flow Data Using Roadside Truck Intercept Surveys 61 Implementation Process Because the major constraint in identifying roadside surveys is finding feasible locations, the first step in this process is to identify all of the feasible locations in the transportation agency’s area of concern. This identification process will likely include truck weigh stations and truck inspection stations, but it also can include truck pull-out areas, public rest areas, and large parking lots located immediately adjacent to the corridor of interest. Potential sites will need to be screened to ensure that there is sufficient physical space to conduct the surveys. This screening can often be done using a satellite mapping service, but in some cases site visits may be needed to ensure that there are sufficient truck and automobile parking spaces, roadways with sufficient turning radii and acceleration/deceleration distances to get trucks safely in and out of the survey site. Another requirement of the site is that it provides surveyors with the capability to communi- cate with truck drivers while they are on the road and request that they participate in the survey. This can be done through variable message signs, static arrow signs, and, in some low volume/ speed cases, a flagman or flagwoman directing trucks to the survey location. Most weigh stations have this technology built in as they often need to visually inspect vehicles. Similarly, gate sur- veys have natural stopping points for trucks that enable surveys to be conducted easily. At other types of locations, such as rest areas and truck pull-out locations, a variable message sign that can be quickly changed at the discretion of the survey site manager will be needed to communicate with trucks regarding their participation in the survey. After potential site locations have been screened, they should be mapped and compared to the origin-destination patterns for the types of trucks for which information is desired. This will help ensure coverage of key corridors at critical points and ensure that key origin-destination pairs are captured. One potential method of identifying key truck corridors is to locate the top truck count locations in the region and ensure that there are survey locations that capture trucks at or near these locations. An origin/destination check involves identifying already known, concentrated generators of freight in the region and determining whether the roadside survey locations will capture trucks moving between these locations. This determination tends to be more of a judgment call that does not have specific quantitative rules to follow. Truck count data can also be helpful for site selection. Ideally, this count is collected with suf- ficient detail to indicate time of day and day of week variability. Having this information can ensure that the highest volume locations are incorporated into the study. Additionally, it ensures that the specific data collection time periods match with the most important days and times for truck activity on the corridor. The next step is to contact the facility operator to obtain approval to conduct roadside truck surveys. For weigh stations and rest areas, the facility operator is often a department within the state DOT, but in some cases, state safety departments or highway patrol agencies operate these facilities. International land border crossings are operated by U.S. Customs and Border Protection with the inspection and enforcement activities that are the primary function of these facilities. It also is worth noting that the operating hours of weigh stations may be limited, and this can make these locations less than ideal for conducting robust roadside surveys. Finally, surveys at weigh stations need to include trucks that are cleared to bypass the weigh stations in advance using a system such as NCPass.

62 Guidebook for Developing Subnational Commodity Flow Data and often require security issues to be addressed prior to approval. Marine ports can be oper- ated by state or local agencies or authorities, but many terminals are privately operated, with the public entity acting as the “landlord.” In these cases, it is important to obtain approval from both the terminal owner and operator. Intermodal rail terminals are privately operated by the railroads. The approval process in some instances may simply involve permission from the day-to-day operator at the facility. In other instances, approval may be required from higher- ranking officials at an organizations’ headquarters. Example In the early 1990s, the state of Washington commissioned the Eastern Washington Inter- modal Transportation Study (EWITS) to better understand freight flows in the eastern por- tion of the state. This was followed by the Strategic Freight Transportation Analysis (SFTA) project with the goal of better targeting investments in freight infrastructure. To support both of these efforts, a series of roadside truck origin-destination surveys were conducted throughout the state. To identify potential survey site locations for the EWITS and SFTA, preliminary data analysis was conducted using the state’s traffic count and vehicle classification data in order to iden- tify the primary freight concentrations using the state highways. These data were collected at weigh-in-motion stations and permanent traffic-recorder stations across the state. These data indicated that the heaviest truck volumes were on the long-haul Interstate corridors, in particular, I-5 and I-90. Secondarily, the major state highways had large volumes, particularly highways connected to the state’s agricultural regions. This included U.S. 395, U.S. 97, and U.S. 12. The state also was interested in understanding the travel patterns of trucks traveling between Washington and Canada. Therefore, survey locations at or near to border crossings were included as well. Figure 3.1 shows the freight economic corridor map for the Washington DOT (WSDOT). The color of each highway segment is based on tonnage estimates: the orange-colored highways carry 10 million tons or more per year and the green-colored highways carry between 4 and 10 million tons per year. Alternatively, truck count data can be used to identify high truck volume loca- tions along the corridor. For the EWITS and SFTA surveys, the feasible roadside truck survey locations were found to match up well with the high truck volume locations that were indicated by truck count data. Note that for cases where potential survey sites are not located near count sites, supplemental classification count data may need to be collected to expand the raw roadside truck origin-destination survey data. Another check on the data was to determine whether the survey locations captured all of the key origin/destination patterns. Washington’s largest metropolitan region, Seattle, also is the state’s largest freight generator. Intercity truck trips from Seattle were well covered via the stations at I-5 in Everett, I-5 at Kelso South and I-90 in Cle Elum. Spokane is the second largest metropolitan region in Washington, and it likely also generates a relatively large frac- tion of freight. Spokane was well covered by the survey locations on I-90 in Tokio and at the border with Idaho. It also was covered to the north by a survey location at Deer Park on U.S. 395. However, there were no surveys to the south of the Spokane region, so truck trips between Spokane and points dead south of Spokane were not covered. This would include truck trips from the Spokane region to Boise, Idaho, and Salt Lake City, Utah. It is important to note these deficiencies at the point of site selection, so that a decision can be made regarding whether enough of the key origin-destination pairs are being captured to justify implementa- tion of the full survey effort. For the roadside surveys in Washington, it was determined that the lack of coverage of southward, Spokane-generated truck trips was not problematic enough to discontinue the study.

Collecting Subnational Commodity Flow Data Using Roadside Truck Intercept Surveys 63 As noted earlier, it is oftentimes possible to conduct roadside truck surveys at rest areas, truck fuel stops, and designated truck parking areas near metropolitan areas. One of the disadvantages of using truck stops and truck parking areas is that trucks are not required by law to enter these areas, whereas they are required to enter weigh stations. For the Washington roadside truck surveys, it was determined that truck stops and rest areas would not be used due to the addi- tional work that would be needed to develop representative samples at those locations and the availability of more suitable locations. The survey locations were restricted to weigh stations and ports of entry. The final locations are shown in Figure 3.2. To gain permission to use weigh stations, ports of entry, and border crossings to conduct roadside surveys, it was first necessary to make an official request and obtain approval from the respective agencies overseeing each site. In Washington, this began with scheduling a meeting with senior Washington state DOT officials, Washington State Patrol officials and U.S. and Canadian customs officials to explain the purpose of the survey, describe the benefit of data obtained from the study, and provide assurance that the survey would not interfere with the typical inspection activities of the state patrol and customs officials. It is not a given that all state organizations would approve of similar surveys. Coordinating information was then sent to all regional offices explaining the survey process and asking staff members to comply. In addition, information on operating hours for each location was obtained so that surveys could be scheduled on days when weigh stations already would be staffed. Figure 3.1. Washington freight economic corridor map.

64 Guidebook for Developing Subnational Commodity Flow Data Figure 3.2. Roadside survey sites for the SFTA origin-destination roadside survey in Washington. User’s Guide Worksheet Punch List Conduct the following activities for the area of interest: • Locate all truck weigh stations, rest areas, truck pull-out areas, and points of entry through contact with operating agencies or Internet research. • Locate major port terminals and rail intermodal yards. • Confirm the feasibility of all of these locations for conducting roadside surveys using a satellite mapping service, site visits, and/or contacting facility operator. • Determine truck corridors of interest using truck volume information and other relevant information specific to the region. • Identify corridors (or segments along corridors) of interest that are not well covered by the existing locations. • Identify concentrations of freight generation that represent truck trip origin/ destination pairs in the region. Determine which of these origin/destination pairs are not well covered by the potential locations listed in the bullets above.

Collecting Subnational Commodity Flow Data Using Roadside Truck Intercept Surveys 65 Step 2—Questionnaire Design Key Considerations Designing roadside truck survey questionnaires requires a clear definition of the purpose of the data collection effort and also involves anticipating truck driver response to questions and a knowledge of the surveyors’ familiarity with truck- ing operations. Additionally, there are tradeoffs that involve the number of questions asked relative to the response rate and the number of questions asked relative to the number of surveys that will be conducted over a fixed period of time. To the extent that a smooth flow of questions can be preserved, it is best to ask the most important questions first, followed by questions of less importance. It is ideal that someone with intimate knowledge of the trucking industry be involved in the survey design process, including potentially having the question- naire reviewed by the state motor trucking association. Implementation Process The general structure of a roadside truck survey questionnaire is to start with questions that the surveyor can answer by visually observing a truck as it approaches the survey location. These questions are followed by requests for information that must be collected from the driver about the current trip, and are followed by requests for general information about the trucking company or broad questions about services provided. The survey usually concludes with open-ended ques- tions of interest to the transportation agency. Items that can be noted as the truck is approaching include the following: • Time of day • Vehicle and trailer configuration • Number of axles • Hazardous material placard (if any) • Name of trucking company • Identifying information on trailer (e.g., DOT registration number) This information is typically easy to identify. There are a few cases where some or all of the hazardous material placard information may not be legible from a distance. Information that can be collected from the driver about the current trip includes the following: • Location of last stop and next stop of the truck • Location of last pickup/delivery and next pickup/delivery • Ultimate origin and destination of cargo being transported (can be read from a manifest or bill of lading) • Locate major truck stops and large parking lots nearby locations along corri- dors of interest and in between major freight generation locations that could potentially be used to fill in gaps for roadside surveys where other types of locations are not readily available. • Contact facility operators to confirm their willingness to participate.

66 Guidebook for Developing Subnational Commodity Flow Data • Facility type (home, intermodal facility, warehouse, port, etc.) at last stop, next stop, last pickup/delivery, next pickup/delivery, ultimate origin and destination • Commodity being carried • Weight of cargo • Empty, partially loaded, or fully loaded • Identification of routes used or to be used on this trip An important distinction is between a location where a truck stops, a location where a truck picks up or delivers goods, and the locations where the cargo begins or terminates the shipment. A location where a truck stops can include diesel filling station, hotel, home base, or restau- rant. The location of last pickup or delivery is the last point at which the cargo inside the truck changed and may have occurred a few days prior to the survey or can occur a few days following the survey. The ultimate origin and destination of the goods may be different than the last pickup or delivery, particularly for less-than-truckload trucks or intermodal goods. It is important to note that the driver may or may not know the ultimate origin and destination of the cargo, because these moves may be outside of the responsibility of the truck driver. Each of these pieces of information is important for different reasons. Typically, information on last stop and next stop is the most important for truck modeling. Commodity flow mod- els tend to benefit most from information on ultimate cargo origin and destination, because that provides useful information on producers and consumers of goods, which is the basis of commodity flow modeling. However, it is common for truck drivers to not know whether the ultimate cargo origin and destination is different from the last pickup/delivery and next pickup/ delivery. Therefore, the pickup/delivery information tends to be the best information available from roadside truck surveys regarding the movement of the cargo. Another key element of truck roadside surveys is the determination about the geographic specificity requested for origin and destination information. The most commonly geographic information requested through these surveys is city and state. However, there are several circum- stances that require more refined trip information at one or both of the truck trip ends being surveyed. In these circumstances, street address information can be requested. However, truck drivers are sometimes not willing to provide such specific information about their customers. Another option is to ask for zip code information at trip ends. Truck drivers may or may not have this information readily available, and they may not want to spend the time to look up the information on the bill of lading or other documents related to their shipments. Other options include getting information on freeway entrance or exit points, which can serve as a proxy for final origins or destinations. Alternatively, the survey instrument can include a map with pre- specified zones that are at the subcity level, and the truck driver can pick from one of the zones on the map. In cases where multiple stop information is needed, the survey questionnaire can be modified to reflect the need to collect this information. For these surveys, it is recommended that an electronic data collection device be utilized, because hard copy surveys would need to be very lengthy to allow for collecting information on several stops. It also is important to determine the amount of commodity detail that is needed from the roadside truck origin/destination survey. Truck drivers tend to know in general the goods that are being carried, but, on occasion, additional probing is needed to specify this to the level that is needed for the survey. Additionally, there may be instances in which commodities are predefined into categories for the truck driver to choose from. Just as truck drivers may not know the specifics of the goods being carried, truck drivers may not have information regarding cargo weight or whether or not the truck is partially or fully loaded. Inquiries regarding routes selected on the trip are typically best conducted by preselecting no more than five routes of interest and having the truck driver identify the ones that will be used. When detailed informa- tion is needed, it is typically best to provide the truck driver with a map and let the driver sketch the route directly onto the map.

Collecting Subnational Commodity Flow Data Using Roadside Truck Intercept Surveys 67 General information about the trucking company or broad questions about the trucking services offered include the following: • Location of truck’s home base • Type of trucking firm (truckload, less-than-truckload, private) • Number of vehicles in fleet • Number of times that the truck driver passes the survey location (or alternative location) in a day, week, or year • Number of years of experience of the truck driver Open-ended questions can include questions on infrastructure, operations, and/or on policy that span a wide range of topics. Survey design decisions about these kinds of questions require balancing the information needs by the transportation agency with the knowledge of the truck driver and the time needed to conduct the survey. Typically, these questions should be carefully phrased so that they do not imply that any immediate action is going to occur. Examples of these open-ended questions are the following: • What improvements should be made to this corridor and where should they be made? • Are there any safety issues with this corridor in regard to truck operations? • Would you be willing to pay a toll to use this corridor if additional capacity were added? • In your opinion, in the last 5 years has congestion on this corridor gotten worse, better, or stayed roughly the same? There are several interesting pieces of information regarding the cargo that the truck driver is not likely to have or would be unwilling to answer, such as the value of the cargo, the date that the product was produced, or the amount paid by the customer to move the goods. After deciding which information is sought through the surveys, the next step is the physical construction of the questionnaire. In recent years, it has been common for many of these surveys to be conducted electronically using handheld devices. The advantage of an electronic method is that it can automatically capture information in a database format that allows for easier pro- cessing at a later time. The advantage of the traditional paper survey format is that it is easier to review the previous questions answered by the truck driver to ensure consistency throughout the survey. The paper survey format also makes it easier for the survey manager to monitor the responsiveness of truck drivers and make field adjustments, if necessary, to improve the survey process. If a traditional paper survey is conducted, then the paper survey should be no more than two pages long to allow for an easy data collection process. Printed questionnaire cards have been used in several roadside surveys as a method to slightly increase response rates. The cards can be filled out by drivers at their convenience and then be mailed back to the survey team. Additionally, the mechanics of the field survey need to be considered, especially if the surveyor will be referencing a map, trying to hold paper and a clipboard, and writing or typing questions all at the same time. Upon completion of the survey questionnaire, a practice survey should be conducted in the office with one surveyor pretending to be a truck driver. This will assist in fine-tuning the survey. It also will provide an estimate of the length of time it takes to complete the questionnaire, which assists in the overall survey development process. Example The SFTA roadside truck surveys asked questions in the following categories: • Time of day • Vehicle and trailer configuration (e.g. dry van, flatbed, bulk, tanker) • Number of axles • Hazardous material placard (code) • Name of trucking company and home base

68 Guidebook for Developing Subnational Commodity Flow Data • Empty weight of the vehicle/trailer • Loaded weight (current payload weight) • Commodity on-board • Cargo origin (city, state) • Facility type for cargo origin • Cargo destination (city, state) • Facility type for cargo destination • Identification of highways used on this route (two maps, state level, and urban map for Puget Sound) • Number of times in past 7 days this route has been traveled. Data on the first five items listed above (time of day, vehicle configuration, number of axles, hazardous material placard, and trucking company home base) were observed visually by the surveyor while the truck was approaching the survey location. The first question addressed to the truck driver was whether they were loaded or empty, and the survey continued through the questionnaire from there (see Figure 3.3). For vehicles that were making many stops/deliveries, the SFTA survey questionnaire was modified slightly to include recording each stop along the route at the city and state level within Washington or province level for Canada. This modification was found to allow for questions that were easy for truck drivers to answer, but the responses were likely to lack adequate detail for evaluating freight activity within large cities such as Seattle. User’s Guide Worksheet Punch List Conduct the following activities for the roadside truck survey: • Identify the most important pieces of information to be obtained from the survey. • Specify the survey questions that will solicit this information. • Identify additional questions that are likely to provide information that will assist in freight planning efforts over the next 10 years. These questions should be limited to those that truck drivers are likely to answer in a time-efficient manner, thereby focusing the resources of the survey effort on the most impor- tant questions to be answered. • Develop a survey questionnaire based on the responses to the above bullets. • Based on the length of the survey questions and available technology for the survey, determine whether it will be conducted using electronic or paper survey instruments. • Conduct a practice run of the questionnaire to estimate the length of time it takes to complete the survey questionnaire as this has implications for future steps in the roadside truck survey process. Step 3—Selecting Interview Dates and Times Key Considerations The primary considerations for selecting interview dates and times are any restrictions inherent in the survey locations, likely sampling issues, the resources available to conduct the survey effort, and/or the need to collect data during different periods within typical truck cyclical behavior (e.g., time of day, day of week, and season in year).

Collecting Subnational Commodity Flow Data Using Roadside Truck Intercept Surveys 69 Figure 3.3. Roadside survey questionnaire used in Washington state (EWITS). (continued on next page)

70 Guidebook for Developing Subnational Commodity Flow Data Figure 3.3. (Continued).

Collecting Subnational Commodity Flow Data Using Roadside Truck Intercept Surveys 71 Implementation Process In practice, the selection of interview dates and times is typically constrained by the resources available for the survey effort, restrictions inherent at survey locations, and the number of loca- tions that will be surveyed. Therefore, it is important to first understand whether there are limitations at any of the survey locations of interest. At weigh stations, dates and times can be limited by the operating hours of enforcement activities. Often times, weigh stations are only open during daytime hours, and they may be closed on weekends as well. If evening and night- time surveys or weekend surveys are desired, that often requires restricting surveys to the few dates when the weigh stations are operating during these hours or requesting special permission from the weigh station operator. Rest areas and truck pull-out locations may become infeasible during nighttime hours as reduced visibility creates safety concerns for truck drivers, surveyors, or other users of the rest areas. Similarly, truck stops also may be limited in terms of their hours of operation. Addition- ally, truck stops have safety and security considerations that must be addressed when conducting nighttime surveys. Many roadside truck survey efforts also are constrained by the resources available. If a road- side survey is funded to the level of $500,000 for field data collection, and there are 40 locations that are to be included in the survey, that leaves $12,500 to collect data at each location. If each location is staffed with three surveyors at a cost of $40 per hour per surveyor, then that allows for a little over 100 hours of survey activity at each location. The selection of appropriate survey dates and times is then made using these 100 hours per location as a general guidepost. Once the number of hours available at each location has been estimated, it is important to con- sider the cyclical nature of the trucks that are to be surveyed at different locations in order to deter- mine at which points of the cycle the survey is desired. For instance, volumes of some agricultural commodities peak in late summer, but this can be highly variable depending on the crop. Interna- tional intermodal container volumes tend to peak in the early fall in advance of the Christmas shop- ping season. Other types of goods have other types of volume peaks throughout the year. To guard against the potential of seasonal bias in the collected data, it is ideal to conduct one survey in each of the four seasons. Based on resource availability, this may be done only at locations where seasonal variability is considered to be most the important, such as agricultural locations. There also is the potential for variation by day of week and time of day in collected survey data. Typically, nighttime and weekend truck movements include a higher proportion of long-haul trucks relative to daytime, weekday truck movements. Therefore, a fully representative sample of data would include data collected during day and night and during the week and on weekends. However, this may again be restricted by resource availability. Additionally, if the purpose of the data collection effort is primarily to support the development of a travel demand model, then the most important data to collect may be data on typical weekday activities. Another alterna- tive is to collect survey data during the nighttime and weekends at select locations and then use the off-peak data at these select locations to factor the data at locations where more limited data were collected. Similarly, if focus is on trucks most likely to be impacted by congested traffic conditions, then it may be important to collect data during the daytime. Example The dates and times for data collection in Washington were selected to ensure the collection of information on the seasonality of freight shipments and to ensure that fluctuations in the large and diverse agricultural economy across the state could be tracked in the collected data. To accomplish these aims, it was decided to survey each site during each of the four seasons. Each survey was conducted on a Wednesday, in order to capture midweek flows that were expected to be “normal” freight flows and avoid unusual traffic peaks or valleys at the beginning or end of the week.

72 Guidebook for Developing Subnational Commodity Flow Data The survey schedule also was designed to minimize double counting of trucks on a single day. Sites surveyed on a given day were selected to be somewhat mutually exclusive. The hours of operation of each weigh station also influenced when data could be collected. In Washington, most weigh stations on heavy freight corridors (I-5 and I-90) are open 24 hours during days of operation. Other stations were only open for short portions of the day. At these locations, data collected from nearby vehicle classification counters were used to estimate truck activity for the full 24-hour period. During the implementation of the survey, there were multiple unfore- seeable events that resulted in rescheduling of data collection at a particular site. The need to reschedule could be caused by highway accidents, last-minute state patrol officer reallocation, and issues with assembling the survey team. User’s Guide Worksheet Punch List • Estimate the approximate budget available per survey by dividing the antici- pated resources for data collection by the number of locations as identified in Step 1. Note that the resources available for data collection will be only part of the total resources for the survey because some resources must cover the costs associated with survey setup, travel costs, data analysis, and report writing. • For each survey location, determine whether the effort will be to have the survey capture seasonal variation. • For each survey location, determine whether the effort will be to have the survey capture time of day or day of week variation. • Estimate the number of surveys that can be conducted at each location by dividing the total number of hours per location by the estimated number of surveyors at the location and the estimated hourly cost per surveyor. • Adjust the available budget at each location to optimize the data collection at each location relative to available survey hours while maintaining a constant budget for total data collection activities. • Optional Task: Redo the above four bullets assuming that all weigh stations in the sampling plan are open only during the weekday between the hours of 8:00 a.m. and 6:00 p.m. Note that this step in the worksheet will need to be repeated following completion of Step 4 on sampling issues. Step 4—Sampling Issues Key Considerations Sampling is perhaps the most complex quantitative issue related to setting up a roadside truck survey program. The main issue with sampling is determining the necessary number of surveys and the necessary timing of those surveys that will ensure statistical confidence in the survey results. This determination depends on several factors, including the key survey variables of concern, presurvey assumptions about variability in truck behavior, and feasible hours of surveying at each location. In practice, the overall sample size tends to be determined by resource availability rather than statistical analysis. However, the statistical analysis can still be used to assist in determining which survey locations should feature additional samples and whether additional surveys are needed for particular times of the day, days of the week, or seasons in the year.

Collecting Subnational Commodity Flow Data Using Roadside Truck Intercept Surveys 73 Implementation Process Steps 1 and 3 have covered identifying a set of potential survey sites along with the dates and times of potential surveys at each site. This step addresses how many trucks to survey at each site, given the fiscal constraints of the roadside intercept survey process, and the types of freight questions that are usually answered through truck roadside intercept surveys. From a statistical perspective, to determine the appropriate number of samples it is impor- tant to isolate a few key variables collected from the survey questionnaire that are deemed the most critical. One example of a variable that is often considered important for origin- destination surveys is commodity. To determine the sample size needed to estimate for a par- ticular commodity, the first step is to generate an estimate of the commodity percentage and variability through the location. This estimate can be developed using nearby or state-level commodity distribution percentages such as those available in the FAF database. If resources allow, pilot surveys can be conducted to develop estimates of the mean and variation for key variables. Alternatively, these estimates can be confirmed for accuracy at the conclusion of the roadside survey effort. The next step is to develop a confidence level for the estimate. The confidence level typically used is 95 percent. Then, the approach used in Step 4 of Chapter 2.0 can be applied to this scenario. As stated in Step 4 of Chapter 2.0, there are three key equations used to estimate sample size: • Equation 1 is the standard normal expression for sample mean stating that with 95 percent confidence, the mean will lie within two standard deviations. • Equation 2 is derived from the first equation. • Equation 3 is achieved by solving for the sample size, n. W is the width in units of the confidence interval. So the wider the confidence interval, the lower the sample size necessary in order to maintain the 95 percent confidence level. The mean is represented by x and the standard deviation is s. 2 , 2 (Eq. 1)x n x n )( − σ + σ 4 (Eq. 2) n W σ = 16 (Eq. 3)2 2n W= σ As in Step 4 of Chapter 2.0, an alternative approach would be to develop the sample size based upon some acceptable threshold of error, in which case, Equations 4 though 6 would be used reflecting the probability that pˆ lies within two standard deviations of the mean. Again, W refers to the width in units of the confidence interval and B refers to the allowable error. Thus, for a 10 percent error (90 percent confidence) in a normal population, a sample of at least 100 would be needed. Likewise, in order to achieve 99 percent confidence (allowing only 1 percent error), a sample of 10,000 would be needed. ˆ 2 0.25 , ˆ 2 0.25 (Eq. 4)p n p n )( − + 4 0.25 (Eq. 5) n W= 4 1 (Eq. 6) 2 2 n W B = =

74 Guidebook for Developing Subnational Commodity Flow Data With the sample size determined for key variables, the next step is to determine the survey periods to consider as part of the analysis. Survey periods are generally determined based on observed fluctuations of truck count data at a particular location, but could vary based on a number of different factors that are known at a location. For each period, the truck activity is assumed to be different. Therefore, it is important to collect the correct number of samples from each time period. Dividing the time period by the time it takes to conduct the survey, then mul- tiplying it by the number of surveyors at each location will enable a determination of whether the sample size can be collected over the course of a day or whether a multiday period will be needed. Ideally, the number of surveyors can be altered during the design phase to ensure that the optimal number of surveyors is present at each location to meet the sample size requirements at each location. However, the number of surveyors at a location may be limited by the physical constraints of the survey site and labor availability of the survey team. It is very likely that origin-destination patterns and potentially commodity distributions change over the course of the day. Nighttime truck operations tend to include a higher propor- tion of long-haul truck trips. The difference between daytime and nighttime operations may be more pronounced at very urban locations, where daytime trucks are often used for the purposes of urban distribution and service facilities that are open during the daytime hours. Roadside intercept surveys are often used to estimate through truck trips relative to internal truck trips at a particular point. To most accurately answer this question, collecting nighttime surveys is preferable. For this reason, it is ideal to collect 24-hour data, at least for a handful of locations, to enable an estimation of the variability between daytime and nighttime operations. This vari- ability can be established based purely on time of day or also can be developed separately for single commodities or commodity groups. If it is believed that there is a significant seasonal variability in origin-destination patterns or commodity distribution in the state, then the survey process should be repeated during each of the four seasons at a minimum. One indication of the potential of seasonal variability at a location is the occurrence of wide fluctuations in truck volumes during different seasons. The potential for seasonal variability also can be obtained by talking to experts in the industries that produce the commodities that are identified in initial intercept surveys. An alternative to conducting full survey efforts during each season to capture seasonal variability is to conduct seasonal surveys at a handful of locations. From the data collected at these locations, seasonal vari- ability factors can be developed for each commodity or for commodity groups. This is similar to the approach for capturing seasonal variability using a limited set of survey locations. As mentioned previously, in practice most roadside intercept locations are constrained by budgets. Therefore, the vast majority of roadside intercept surveys are conducted such that each location is surveyed over the course of a single day. Sometimes the survey period is roughly during the midday peak hours of the truck operations. Surveys with larger budgets can capture trucks over a 24-hour period at each location. The goal of most surveys is to have some data available in as many locations as possible given the limits of the budget. Example One of the specific purposes of the roadside intercept surveys conducted in the EWITS and SFTA studies was to create a truck origin-destination matrix for the state of Washington. To achieve this, survey team staffing was one of the critical factors examined in the survey design process. Generally, survey team staffing throughout the day was determined based on time-of-day freight traffic distributions from WSDOT traffic count data, knowledge concerning each survey site, and physical parking limitations, which limit the number of trucks that can be interviewed at any one time. Increasing the number of days that the survey is conducted doesn’t impact the sample size because the population of traffic likewise increases for the additional days surveyed.

Collecting Subnational Commodity Flow Data Using Roadside Truck Intercept Surveys 75 If the maximum number of questionnaires collected at a given site is 20 percent of the total freight traffic, it is most likely to be the same on the additional days that the survey is conducted (assuming the traffic conditions are similar). Sampling on additional days certainly enriches the data and provides increased confidence in the proportion of freight traffic that is obtainable. Given the dynamic and variable nature of freight traffic conditions, the proportion of freight traffic that may be sampled varies through- out the day and across different data collection sites. Realizing these constraints, the stated goal of the Washington roadside surveys was to obtain as many completed survey questionnaires as physically and safely possible during the dedicated data collection periods. Once the samples were collected, a weighting scheme would be developed to expand the sample averages so as to represent characteristics of the total traffic population. In almost all cases, available parking at each weigh station limited the number of vehicles that could be surveyed at a given time, so regardless of whether survey staff were available, only vehicles that arrived when parking was available were surveyed. At the lower volume survey sites, available parking was less of a problem, and between 60 and 80 percent of freight truck traffic was interviewed (at a few very-low-volume sites along the Canadian border, almost 100 percent of trucks were interviewed). At the very-high-volume survey sites (I-5 and I-90), a range of 5 to 20 percent of the freight truck traffic was captured in the survey sample. The variation in sample size is practically unavoidable given the fluctuations in traffic conditions and available parking areas, but represents the best compromise given the prevailing surveying conditions and the goal to obtain as much information as possible. As was previously mentioned, the state of Washington, similar to many states, has a pro- gram where freight shippers and truckers may register their vehicle with the state electronic bypass program, Commercial Vehicle Information Systems and Networks (CVISN), and install vehicle transponders that allow shippers to register trip and shipment details online and bypass weigh stations that have CVISN capability. Currently, there are 12 of these weigh stations in Washington, but during the EWITS origin-destination survey, this technology was not available, and during the SFTA origin-destination survey, the technology had only just become available and only two weigh stations possessed CVISN capability. However, the data obtained from the CVISN program is available (via request to the CVISN data office), and this information for each survey site can be used in determining sample characteristics and the population of trucks that used the program during the survey time period. In addition, many of the questionnaire attributes are available from the CVISN data and can be incorporated into the data validation/ verification process at relevant sites. User’s Guide Worksheet Punch List • Identify a key variable collected through the roadside survey effort (e.g., pay- load, origin-destination pairs, commodity). • Estimate the mean and variance of the key variable for truck trips through a survey location using FAF data or a pilot survey. • Determine the desired confidence level for the sample size. Ninety-five percent can be used as a default value. • Estimate the sample size needed to estimate the variable using Equations 1 through 3 provided in the implementation section of this step. • Repeat this process for as many variables and locations that are of interest to the survey effort.

76 Guidebook for Developing Subnational Commodity Flow Data Step 5—Interview Team Recruiting and Training Key Considerations The structure of the interview team will impact both the quality of the data and the cost of collecting the data. Using an interview team with significant experience in conducting roadside truck origin/destination surveys tends to provide the best data quality. Additionally, the time and resources needed to train experienced data collectors should be minimal. However, the hourly rate for experienced data collectors is likely to be higher than the hourly rate for other potential data collectors. In contrast, using general temporary labor will decrease the hourly cost, but will increase the need for upfront training and potentially result in somewhat lower data quality. Regardless of the experience level of the team involved, training will be needed to ensure safety during data collection operations and effectiveness in asking truck drivers questions. Implementation Process The data collection field team can include transportation professionals, data collection spe- cialists, or temporary employees. Typically, the most cost-effective strategy is to use temporary employees to collect the bulk of the survey data and have a transportation professional manage the operation at each survey location to provide quality control of the data collected. In addition to being cost-effective, using temporary employees to collect data also has the advantage of providing the most bandwidth of all of the options. With a small number of transportation professionals, it is possible to conduct dozens of surveys simultaneously or within very short periods of time. Site managers also can collect data, but may have their collection interrupted by other activities. In the case of conducting multiple surveys simultaneously, the site manager may be responsible for several locations and may need to travel between locations to monitor operations. Temporary employees will need to be tested for the minimum levels of knowledge required to conduct a survey, including geography, basic math, reading, and writing. This testing can be provided by the temp agency from which employees are recruited. Using temporary employees will require the most upfront training of staff. It also requires a higher level of monitoring of the data collection process. Temporary employees will need instruction on the purpose of the survey effort, a detailed description of each question that is included in the questionnaire, and detailed safety training for each type of survey location. An option related to using temporary employees would be the use of volunteer data collectors. These can potentially come from local charitable organizations. An alternative structure would be to staff the team entirely with transportation professionals. This option would incur a higher cost relative to using temporary employees, but it will save time. It also has the potential to streamline the entire process, because team members who col- lect the data in the field also can do data entry in the office. Data collection specialists are firms that specialize in collecting data across a number of differ- ent topics such as market research and polling. Using one of these firms tends to cost less than using transportation professionals, but cost more than using temporary employees. While the employees of firms that specialize in data collection are trained in the data collection aspect of conducting a survey, they will still need detailed training for each question in the questionnaire and safety training on the sites. These employees will typically require much less monitoring than temporary employees.

Collecting Subnational Commodity Flow Data Using Roadside Truck Intercept Surveys 77 Regardless of the structure of the team, each data collection location should have a site super- visor that manages the coordination needed with facility operators, monitors the quality of the collected data, ensures that safety guidance is being adhered to at each location, and collects com- pleted survey forms or data from electronic devices. This site supervisor should be an experienced transportation professional that has participated in previous roadside truck survey efforts. For surveys that are done at a handful of locations, it is typically most effective to use transpor- tation professionals to save time and money on training and ensure data quality over the survey effort. Transportation professionals or data collection specialists also are recommended if the survey is asking complex questions of truck drivers or questions that will require a certain amount of probing or pivoting from one answer to the following question. This would include asking for detailed origin/destination data such as assigning each trip end to a prescribed set of zones. For surveys that occur at several dozen locations, temporary employees are likely to be the most effective. These employees can be trained once and then be used repeatedly over a short period of time. If temporary employees are used, then the questionnaire should be made relatively straight- forward with a finite set of responses that are expected to be captured through the survey effort. If the transportation agency elects to hire contractors to perform the data collection services, then proposals can be accepted from firms offering alternate team structures, and the agency can select the team that provides the best overall services in terms of cost and data quality. Staff training should include a description of the purpose of the survey, including what is driving the need to conduct surveys and what the data will ultimately be used for. Then, each question in the survey should be reviewed with the field staff. For temporary staff, this will need to be a detailed question-by-question review of the survey questionnaire. For each question, staff should be trained in how to phrase the question, how to record the answer, and what types of responses are likely. For staff not familiar with freight terminology, basic descriptions of key items (such as rail intermodal yard) will be needed. For electronic surveys, staff will also need to be trained in using the actual devices. As with paper surveys, the actual survey instrument should be available and reviewed by the staff being trained. There also will need to be special alerts for responses to questions that will require follow-up or additional probing. This is the case most commonly with responses to origin and destination questions where either not enough specificity is given or landmarks rather than addresses or city/state combinations are provided. If special maps will be used in the survey, then these maps should be developed and presented during the training session. If there are predefined categories for any of the questions, then these need to be reviewed at the training session as well. As part of the training session, a dry run of a survey should be conducted so that all surveyors can get a sense of timing and protocol and have a chance to raise any other logistical questions. The training session also should include training on site safety. This aspect of the training will include a typical site layout for each of the different types of survey facilities and will include a dia- gram showing how the site will operate, including where trucks are diverted from the traffic stream, how they approach the survey location, where the surveyors will be stationed, how trucks will pull out from the survey location, and the approximate time interval between surveys. The training will need to cover how surveyors should approach a vehicle and where they should position themselves as a truck approaches and pulls away. Surveyors should also be trained in how to handle truck drivers that decline to participate in the survey or decline to answer specific questions. The final portion of training will address how to store and transfer the data for future analysis. This training will be needed whether the data are collected using paper and pen or electronic devices. Example The SFTA roadside surveys were conducted using state service clubs (Lions Clubs and Kiwanis Clubs). Recruitment of these teams began by contacting club presidents near the selected data collection sites so that survey teams would not have long travel times to and from survey sites.

78 Guidebook for Developing Subnational Commodity Flow Data Using service club members was ideal in several respects. First, they proved to be a relatively large and geographically diverse short-term labor force, which matched the needs of the survey. Use of these service club members allowed for 15 to 18 surveyors at each data collection site for the 24 hours of data collection, equating to 5 to 6 surveyors per 8-hour time period. Because the service club members lived near the survey locations, they also had extensive local knowledge of the roads and cities being accessed by the truck population. In lieu of individual salaries, contribu- tions were made to the service clubs. Once interview teams were identified, each team received classroom training, typically lasting about 2 to 3 hours, which included the following agenda (see Figure 3.4): • Explanation of the overall project goals of the study • Description of each question on the questionnaire, including typical and atypical answers to each • Safety requirements • Site setup/operation • Items to bring to the survey location such as comfortable clothes/shoes, hats, sunscreen, rain gear, etc. • A question/answer session. Each survey member was provided with a hard copy of a safety manual that outlined key safety considerations such as the need to always wear safety/reflective vest. Other safety instructions provided in the safety manual were never approach a moving truck, wait for it to stop instead; do not enter or climb onto the truck; never allow traffic congestion at survey sites; and always be mindful of traffic flow and other moving vehicles. Training of each team continued during the actual implementation of roadside surveys, as each survey site had a site manager that oversaw all operations and monitored completed surveys in order to quickly identify problems in conducting the surveys. As teams became more experi- enced, the role of the site manager became less important, except for survey members who may not have participated on earlier roadside surveys. User’s Guide Worksheet Punch List • Determine the ideal mix of transportation professionals, data collection specialists, and temporary staff desired for each survey location. • Develop a training session agenda for field data collectors. • Develop all needed training materials for the training session. This should include a description of how to phrase each question, the types of responses that are likely to be heard, how to record the response, and typical responses that require follow-up by the surveyor. • 9:00 a.m. – 9:15 a.m. Introductions • 9:15 a.m. – 9:30 a.m. Overview of project and survey process • 9:30 a.m. – 10:00 a.m. General site setup and safety requirements • 10:00 a.m. – 10:15 a.m. Items to bring to survey locations • 10:15 a.m. – 11:00 a.m. Question-by-question instructions including typical responses • 11:00 a.m.– 11:30 a.m. Mock surveys • 11:30 a.m. – 11:45 a.m. Wrap-up and Q&A Figure 3.4. Hypothetical agenda for roadside truck survey training session.

Collecting Subnational Commodity Flow Data Using Roadside Truck Intercept Surveys 79 Step 6—Site Preparation and Traffic Control Key Considerations Site preparation and traffic control needs vary for each survey site. Coordination with the facility operator will be needed to ensure adherence to site-specific safety guidelines. Additionally, a transportation professional should determine whether there is a need for a traffic control plan on site. Additional preparations will be needed if off-site equipment such as variable message signs is used. Implementation Process Weigh stations are considered to be the ideal roadside truck survey location because they are designed specifically for managing truck operations. This includes the ability to intercept any truck from the traffic stream and inspect it for long periods of time if needed. Weigh stations require the least amount of site preparation and in some instances will not require a formal traffic control plan. The survey manager should meet with the on-site operator of the weigh station to determine what the specific survey operations will be. This includes the process for ensuring that the trucks selected for the survey are random, deciding who will select the trucks, and settling on how communication will occur between the survey site manager and the weigh station operator. From this meeting, the survey site manager should, at a minimum, develop a sketch drawing detailing the location of the survey, the number of truck parking spaces available for the survey, and the path of the trucks as they enter and leave the survey location. Locations that are not weigh stations will require the development of a traffic control plan. This plan should be developed with oversight of the facility operator and take into account the sample size requirements that were identified in Step 4. The plan should also be reviewed with the transportation agency that is responsible for the roadway where trucks are being intercepted to ensure that the plan does not create truck queues or speed deceleration events within the mainline traffic area. If the site is not capable of handling the number of trucks expected, then Step 4 may need to be revisited to ensure that sufficient data can be collected and to determine whether different survey dates and times are required. At locations that are not weigh stations, it is likely that variable message signs will be needed to intercept trucks. The survey site manager will spend much of their time at these sites manag- ing the selection of trucks from the traffic stream, including changing the information on the variable message sign as needed. These locations also will require standard signs to direct trucks to the survey location and assist them in navigating away from the survey site. The survey site location also should be isolated from vehicular activity not associated with the survey, so there is no comingling of traffic at the site. It is likely that a series of traffic cones, signs, tape, and other markings will be needed to ensure a safe data collection site. The site preparation and traffic control plans may need to be altered during non-daylight hours depending on the availability and quality of lighting and any different traffic patterns that occur during this time period. Example The site-specific preparation for conducting roadside surveys at weigh stations and ports of entry generally followed physical layouts for weigh stations of two sizes (see Figure 3.5). The smaller sites on lower volume highways often only allowed enough space to park two to four vehicles, as is depicted in Figure 3.5. A large reflective sign indicating that there would

80 Guidebook for Developing Subnational Commodity Flow Data be a survey team ahead was first placed at the entry of each weigh station to alert approaching drivers of the survey team’s presence. Then, depending upon the individual officer(s) operat- ing the weigh station and how busy they happened to be with their normal enforcement and inspection activities, one of two approaches was used to request that drivers participate in the survey. One approach was to place a survey team member on the weigh scales or just beyond the scales; once the officer indicated the truck was “OK” to continue, the team member would approach the vehicle and ask if the driver would be interested in completing a short survey. If the driver agreed, the survey team member would then ask the driver to pull ahead and park, and one of the other survey team members would immediately conduct the survey. Once the survey was complete, the team member would thank the driver and the driver would continue on. If the driver indicated no interest in participating in the survey, then the team member would ask the next truck driver. Figure 3.5. Typical traffic control plan for Washington roadside truck surveys.

Collecting Subnational Commodity Flow Data Using Roadside Truck Intercept Surveys 81 The person who is first interacting with the driver has the added responsibility of maintain- ing traffic flow, asking drivers if there is an available place to park, and finding an available staff person to conduct the survey. It is important not to interfere with the normal operations of state patrol personnel. The larger sites generally operated in the same manner, but could often have enough space to park 6 to 10 vehicles or more. Another approach for indicating which vehicles to survey would be for the state patrol officer and one of the survey team members to work together inside the weigh station. The survey team member inside the weigh station would monitor the activities of the other team members as well as parking availability to determine when there was availability for another truck to be surveyed. When parking and a survey team member were available, the team member inside the weigh sta- tion requested that the state patrol officer stop a particular truck for a survey. The officer would turn the switch of the weigh station light to indicate that the designated truck should pull into the survey site location. In some cases, the vehicle that was being asked to park was also being stopped for inspection or weight violations, in which case the officer would tell the team member inside the weigh sta- tion to communicate to the survey team member conducting the survey that once it was com- pleted, the driver should bring their log book inside the station. Generally, the second approach worked much more efficiently and safely. There were periods during survey sampling when the weigh station would temporarily shut down or stop accepting trucks for inspection. This would occur when there was only one state patrol officer at a specific site and it was necessary to pursue a truck driver who had avoided the scales. There also were periods during which the highway was closed due to accidents nearby or issues such as brush fires. In each case, the survey team would note the exact time of the clo- sure, and loop detector data was used to determine the number of freight vehicles that passed during this time. User’s Guide Worksheet Punch List • Contact facility operators at each location to determine whether there is a need for a traffic control plan. • Develop a traffic control plan for one weigh station and one non-weigh station location, if applicable. • Compare traffic control plans with needs identified in the sample size step (Step 4) and the survey dates and times step (Step 3). • Submit these plans for review by facility operators and relevant transportation agencies. Step 7—Using Uniformed Law Enforcement Officers and Their Vehicles Key Considerations Trucks traveling on the Interstate system are accustomed to stopping at enforce- ment locations. The use of enforcement officers at data collection sites is therefore useful from both a response rate perspective and a site safety perspective.

82 Guidebook for Developing Subnational Commodity Flow Data Implementation Process It is recommended that uniformed law enforcement officers and/or their vehicles are used at every roadside survey site location. This will happen automatically if weigh stations are used as survey site locations because uniformed law enforcement officers are always stationed at these facilities. Truck drivers tend to feel more comfortable participating in surveys if they are aware that law enforcement is involved. Additionally, law enforcement officers are trained to develop and administer safe vehicle control plans for all types of sites. At non-weigh-station locations, visible uniformed law enforcement officers are even more useful than they are at weigh stations. Most truck drivers are aware that mobile inspection stations may divert them out of the traffic stream on virtually any Interstate or state highway. This familiarity will extend to visible uniformed officers at non-weigh-station locations for the purpose of administering roadside truck surveys. The uniformed officers also should review the traffic control plans at all sites to confirm their agreement and where they will be stationed throughout the survey effort. There is typically an additional cost to using these officers that will need to be incorporated into final budgets. Another option is to have an unmanned patrol car stationed adjacent to the survey site to indicate that the survey has the approval of local law enforcement, even if an officer is not available to monitor the site dur- ing the survey effort. Example For roadside truck intercept surveys conducted at weigh stations in California, several unique site locations were used, including truck stops, rest areas, truck pull-out locations, and large park- ing lots located adjacent to Interstates and state highways. At each of these locations, the survey team found it extremely helpful to locate law enforcement vehicles just upstream of where truck drivers were requested to pull off of the road. Truck drivers are trained to anticipate truck inspec- tions happening not just at weigh stations, but at other locations as well. The presence of the law enforcement vehicle put truck drivers on notice that they might be requested to exit the traffic stream. Using this method generated a high survey response rate for trucks at nonstandard loca- tions throughout the state and enabled the identification of sites based on data needs rather than on weigh station locations. User’s Guide Worksheet Punch List • For all survey locations that are not weigh stations, determine whether or not use of uniformed law enforcement officers will occur. • Contact the relevant law enforcement agency and record the protocol in terms of the timing, cost, and role of uniformed enforcement officers. Step 8—Quality Control Procedures Key Considerations Quality control procedures need to be considered as early as the questionnaire design phase and be carried through the analysis of data collected in the field. There are key intervention points when collected data should be reviewed to ensure that data quality is preserved throughout the process.

Collecting Subnational Commodity Flow Data Using Roadside Truck Intercept Surveys 83 Implementation Process Quality control procedures need to be interspersed throughout the survey effort. One of the initial steps in quality control is for the survey site manager to review the survey process for each surveyor while they are in the field. This will ensure that questions are being asked in a similar format for each truck driver and that there are no issues with the survey questionnaire. At the end of the first day, the survey site manager should meet with other senior staff on the survey team and review the mechanics of the survey process. Issues that tend to arise at this stage could include the following: • Identification of questions that are complex to ask or answer in the questionnaire • Adjustments in how one or more surveyors ask questions during the survey process • Improvements in the ease of use of the data collection device, including paper surveys, electronic data collection software, or supporting materials such as maps • Confirmation of the time needed to complete the survey • Clarification of confusing questions that frequently are omitted or answered incorrectly by truck drivers When the field data collection is complete, the raw data should be analyzed to identify miss- ing data items and incorrect data items. Missing data items are typically the result of truck drivers electing not to answer a question, truck drivers not knowing an answer to a question, or a field data entry error. Due to the data quality check that should have happened early in the survey process, missing data items should be minimal at the end of the survey. Data items with unexpectedly high missing elements (perhaps more than 20 percent) indicate that there are ques- tions that should be reconsidered for future roadside truck survey efforts. In some cases, missing survey items can be filled in following the survey. For example, if the city was recorded, but state information was not recorded, then it can usually be inferred and corrected in the database. Incorrect data items can sometimes be identified through responses that are inconsistent with other known information about the survey. For example, a truck driver may state an origin/ destination pair that is inconsistent with the driver’s direction of travel. Alternatively, the truck driver may state a payload amount that is inconsistent with legal limits or inconsistent with its noted truck configuration or practical loading considerations. In other instances, incorrect data items can be the result of data entry errors such as recording the number of axles as being different than what is implied in the truck configuration. Some incorrect data items can be easily corrected such as backwards origin/destination pairs. Other incorrect data items should be completely deleted such as a payload weight that is not feasible. It is possible to develop programs that check the data automatically for accuracy. It is rare for an entire truck record to require deletion, but there may be multiple missing items for a single truck record. The raw data collected through the survey process should be preserved for future reference. This includes the paper copy of survey responses and the entry of the raw data from paper into spreadsheet form. For electronic surveys, the raw electronic data should be preserved for future reference as well. Data edits should be done in a separate file. Example Efforts to increase data quality and accuracy began with survey team training and continued through the survey itself and on to the inputting of the data into computer databases. Having a site manager at each survey site significantly reduced the likelihood that someone conducting the surveys was doing so incorrectly or completing the survey form incorrectly or illegibly. Not every survey questionnaire was double checked by the site manager, but the site manager did double check many questionnaires and was especially diligent when new team members began or when there was a shift change. Once problems were addressed and corrected, oversight became less rigorous for each survey shift and team.

84 Guidebook for Developing Subnational Commodity Flow Data The completed questionnaires were then thoroughly reviewed prior to being input into a relational database. During this process, much of the missing or incorrect data was corrected. In most cases, miscommunication between the driver and the survey team member caused incor- rect information to be recorded on the questionnaire. But by thoroughly reviewing each ques- tionnaire, the research team could evaluate the data to ensure that the information was logically and rationally consistent and correct any problems. Some of the common errors on completed questionnaires included the following: • Incorrect weight (payload and empty weight considerably above legal limit) • Incorrect state or province associated with shipment origin/destination • Missing commodity or some other missing attribute • Number of axles didn’t agree with truck configuration specified • Hazardous material code incorrect (didn’t match real code) In almost all cases, corrections to the data were developed from reviewing other informa- tion on the survey and using deductive, logical reasoning or by reviewing other survey ques- tionnaires that came from the same carrier, same origin-destination, or some other common attribute. Most of the cargo weight errors involved the driver providing the gross vehicle weight or maximum legal weight of the vehicle. It was fairly easy to deduce the cargo (payload) weight once the research team had the vehicle weight and the gross weight. If the vehicle weight was not known, but the exact truck combination was known, then the research team could verify the empty vehicle weight by using known empty weights (provided by state patrol officers) of each configuration. Most of the incorrect or missing information from the origin state could be ascertained from the highlighted route the drivers provided, and the specific type of hazardous material being shipped could often be obtained by the name of the shipper or the business. User’s Guide Worksheet Punch List • Document the protocol for site survey managers to check field data collection for each survey location. • Make sure to include data checking and editing in the survey schedule and budget. Step 9—Preparing Data for Usage in Commodity Flow Database Development Key Considerations The key element in the data preparation process is data expansion. There are several data expansion methods that can be used, but hourly truck count data are typically the best alternative for roadside truck intercept surveys. Implementation Process This step takes the data sample collected through the survey effort and expands it to estimate the full set of truck activity at each location. Truck count data are used to expand the sample, because the truck count data can be used to determine the percentage of trucks that are captured at the location throughout the day.

Collecting Subnational Commodity Flow Data Using Roadside Truck Intercept Surveys 85 The first step is to divide the day into time periods that are believed to be consistent with fluctuations in truck activity. An hourly truck count curve at a location is a good indicator of these fluctuations. Figure 3.6 shows an example of hourly truck counts near the Port of Savan- nah. Based on this curve, there appear to be four time periods of activity for trucks—a morning buildup in truck activity, a midday peak, an evening drop off in truck activity, and a nighttime low in truck activity. These fluctuations may vary at each site, but for simplicity it is reasonable to use the same time periods at each location. For each time period, the sampling factors are developed by calculating the percentage of vehicles that were captured by the survey, then multiplying that by each of the variables that are being expanded. As a simple theoretical example, consider a scenario with two time periods, daytime and nighttime. Fifty percent of the survey data are collected in each time period, while 80 percent of the truck traffic occurs in the daytime and 20 percent of the truck traffic occurs at night. The variable to be expanded is average truck trip length through the survey location. In the daytime, the survey data indicate an average truck trip length of 100 miles, and at nighttime the survey data indicate an average truck trip length of 200 miles. To calculate the average truck trip length at the location during a typical 24-hour period, the calculation would be the following: Average truck trip length 80% 100 20% 200 120 miles) )( (= ∗ + ∗ = Note that the average truck trip estimate does not depend on the number of samples collected during each period. Rather, it is dependent on the expansion factors that are developed using the periodic truck count data. The next section provides three detailed examples of how to expand surveyed data. Examples of Data Expansion There are three examples of data expansion that are presented in this section. First, is the expansion of collected data to a 24-hour time period. This example is presented using hypotheti- cal truck count data and origin-destination data. The second example is how to expand data 12:00 A.M. 1:00 A.M. 2:00 A.M. 3:00 A.M. 4:00 A.M. 5:00 A.M. 6:00 A.M. 7:00 A.M. 8:00 A.M. 9:00 A.M. 10:00 A.M. 11:00 A.M. 12:00 P.M. 1:00 P.M. 2:00 P.M. 3:00 P.M. 4:00 P.M. 5:00 P.M. 6:00 P.M. 7:00 P.M. 8:00 P.M. 9:00 P.M. 10:00 P.M. 11:00 P.M. Hour of Day SR21, Savannah NB SR21, Savannah SB SR25, Savannah NB SR25, Savannah SB 0 20 40 60 80 100 120 140 Truck Volume Figure 3.6. Hourly truck count data near the Port of Savannah.

86 Guidebook for Developing Subnational Commodity Flow Data collected in different seasons to develop annual averages. This example is provided using an example from the EWITS survey using payload data. This example also shows how to develop systemwide averages of data collected at numerous sites. The third example describes uses of commodity flow data collected using a roadside intercept survey. Hourly truck count data are typically used to expand data collected at a single site. For this hypothetical example, it is assumed that the hourly data are as shown in Table 3.1. Examination of the hourly count data indicate that four time periods of analysis make sense: 1. Morning period—6 a.m. to 9 a.m. 2. Midday peak—9 a.m. to 5 p.m. 3. Evening period—5 p.m. to 9 p.m. 4. Late night period—9 p.m. to 6 a.m. If hourly truck count patterns are drastically different across locations, then it may be neces- sary to develop different truck count time periods for different survey locations. Based on the time periods defined above, the origin-destination patterns collected in the survey can be aggregated into time periods for the data expansion analysis. An aggregation of these data is shown in the column labeled “Truck Count” in Table 3.2. The expansion factors shown in the third column of Table 3.2 represent the percentage of total daily truck trips that are in each time period. The next step is to calculate the percentage of through truck trips during each time period. This is shown in the fourth column of Table 3.2. In this example, through truck trips represent truck trips that travel through the state of Washington without conducting pickup- or delivery-type activities. The percentage of truck trips can be multiplied by the expansion factor and summed to arrive at the expanded estimate for the percentage of through truck trips at the location. The percent- Hour Beginning Truck Count 12:00 a.m. 250 1:00 a.m. 200 2:00 a.m. 200 3:00 a.m. 250 4:00 a.m. 250 5:00 a.m. 300 6:00 a.m. 400 7:00 a.m. 500 8:00 a.m. 600 9:00 a.m. 600 10:00 a.m. 700 11:00 a.m. 800 12:00 p.m. 850 1:00 p.m. 800 2:00 p.m. 850 3:00 p.m. 800 4:00 p.m. 850 5:00 p.m. 650 6:00 p.m. 550 7:00 p.m. 450 8:00 p.m. 400 9:00 p.m. 250 10:00 p.m. 250 11:00 p.m. 250 Total 12,000 Table 3.1. Hypothetical truck count data.

Collecting Subnational Commodity Flow Data Using Roadside Truck Intercept Surveys 87 age of through truck trips for each time period is shown in the fifth column of Table 3.2. The total expanded estimate of through truck trips is 40 percent. Any variable can be substituted for through truck trips and a similar data expansion process can be conducted. Due to the vast number of variables included in origin-destination surveys, this operation is typically carried out in a spreadsheet format. For surveys that are conducted across multiple days, there is also the need to develop estimates of truck characteristics and activity across each of the time periods. In the EWITS survey, each location was surveyed four times—once for each season. Given that the sample was collected for each season, an origin-destination matrix for each season could be developed or one aggregated matrix could be developed in which each season’s sample was aggregated into a total using a weighted average of each season’s sample. The choice of matrix type largely depends upon the kinds of questions to be addressed and whether seasonal separation is necessary. Table 3.3 shows the calculation of average payload for the EWITS survey across different seasons and locations. A similar method could be used to develop an origin-destination matrix by expanding the data across each of the potential origin-destination pairs in the study area (including external regions). Once the data from the Washington surveys was compiled into a relational database, it was used in a variety of applications. One use was to develop a matrix of shipments (or tonnage of freight) between all origin and destination pairs in the database. Table 3.3 illustrates how this can be achieved with roadside data. The volume of tons shipped between any pair is calculated by taking all observations across all data collection sites with that origin and destination, applying the site-specific expansion factor that accounts for the proportion of freight traffic captured at each respective site, and multiply- ing by the average payload weight for that site to arrive at the average daily total across all sites. This type of matrix may be developed for each season or averaged (seasonal weighted average) across all seasons, depending on how the matrix will be used. This process is shown in Table 3.4. In addition, for each origin-destination cell, it is possible to identify specifically which com- modities comprised those shipments. In most cases, at the commodity-level detail there will not be very many observations per commodity type for each origin-destination pair. Ideally, the goal should be to have at least 30 observations per origin-destination cell in order to overcome small sample size issues, but often this is not possible because of limitations on the proportion of traffic that may be sampled. But even when the variance may be wide for cells where the freight shipments are based on two or three observations, it is informative to know that some shipments were captured with some information regarding commodity. In addition to developing a freight origin-destination matrix of shipments, one may wish to identify the relationship in terms of freight traffic between any one point on the transportation network and all other areas, similar to stream flow analysis (up and down stream flows). Given Time Period Truck Count Expansion Factor Percent of Through Truck Trips Survey During Each Period “Expanded” Percent of Through Truck Trips Morning Period 1,500 13% 30% 4% Midday Peak 6,250 52% 25% 13% Evening Period 2,050 17% 45% 8% Late Night Period 2,200 18% 80% 15% Totals 12,000 100% 40% Table 3.2. Data expansion calculation to estimate through truck percentage.

88 Guidebook for Developing Subnational Commodity Flow Data that each route in the database has been geocoded to represent all highway arcs comprising each observation, it becomes easy to provide this powerful analytical capability that develops a con- nection between all data attributes collected in the roadside questionnaire and the state geogra- phy or highway network. As an example, Figure 3.7 depicts the entire database of information (all routes) collected from the SFTA roadside survey in 2003/2004, across all sites. This graphic, when compared with the actual data from permanent traffic recorders, mirrors actual freight traffic intensity, indicating that in aggregate across the entire region the statewide coverage was adequate. In Figure 3.8, all observations that are common to Weigh-In-Motion site P21 (selected at random) may be depicted across the entire network, illustrating the geographical reach (in this case both origins and destinations) and intensity of flows that pass through this one particular Survey Site Season Expansion Weight Relevant Observations Avg Payload (tons) Total Expanded Number of Trucks Average Daily Total Payload Table 3.3. EWITS seasonal expansion of sample data.

Collecting Subnational Commodity Flow Data Using Roadside Truck Intercept Surveys 89 Origins / Destinations Aberdeen, WA Addy, WA Adna, WA Alderdale, WA Algona, WA Amanda Park, WA Anacortes, WA . . . . Aberdeen, WA Addy, WA Adna, WA Alderdale, WA Algona, WA Amanda Park, WA Anacortes, WA . . . . Abbotsford, BC Abilene, TX Acton, Ontario . . . . Internal External Internal Origins / Destinations Survey Site Expansion Weight Observations Avg Payload (tons) Average Daily Total Table 3.4. Matrix of freight tons between origins and destinations. Figure 3.7. Intensity of freight routes collected from statewide origin-destination survey.

90 Guidebook for Developing Subnational Commodity Flow Data area on the network. This same type of analysis also may be performed for any attribute, includ- ing (but certainly not limited to) the following: • All freight flows common to a specific origin or destination • All freight flows of a given commodity or value • All freight flows of a specific vehicle type or hazardous material type • All freight flows by a specific time of day These analyses could even become more advanced geographical analyses that include com- binations of many attributes (depict all statewide flows of a specific commodity type, using a straight truck configuration, which passed through King County). In addition, this type of analy- sis could be used to identify/estimate the economic linkages between cities or counties or regions. The possibility of different combinations is limitless, but further illustrates the value of route data that is perhaps best captured in roadside surveys. Figure 3.8. Relationship between freight traffic at one location on the network and all other areas. User’s Guide Worksheet Punch List • Identify the most important variables to expand in the survey. • Obtain hourly truck count data for the region, ideally as close to potential sur- vey locations as possible. At a minimum, the hourly truck count data should be located on one of the same roadways as the roadside truck intercept survey.

Collecting Subnational Commodity Flow Data Using Roadside Truck Intercept Surveys 91 3.3 Next Steps This chapter provides a detailed description of several elements involved in developing a road- side truck intercept survey. Nine basic steps have been identified that can be followed to fully implement a survey. Each step has been divided into the following four elements: 1. Key Considerations. Refer to this section to identify key concepts to keep in mind when initially considering conducting a roadside truck survey. 2. Implementation Process. Refer to this section to get information on the step-by-step process for conducting each step. 3. Example. Refer to this section to see how other agencies have conducted this step in separate roadside truck intercept survey programs. 4. User’s Guide Worksheet Punch List. Refer to this section when you are ready to begin to work through several of the specific items that are needed to develop a roadside truck inter- cept survey. Refer the Playbook section (Chapter 6.0) to identify the next portion of the Guidebook that will be most relevant to where your transportation agency is in the data collection process. • Develop expansion factors at each survey location based on the observed fluc- tuation in the hourly truck count data. • Expand data to estimate truck activity for a 24-hour period at each survey location. • Analyze and map the distribution and origin-destination pairs for key truck trip variables at individual locations and for the roadside survey program as a whole.