Office of Operations Freight Management and Operations

Quick Response Freight Manual II

6.0 Hybrid Approaches

6.1 Introduction

State-of-the-practice metropolitan truck models are hybrids that blend commodity flow modeling techniques with freight truck modeling techniques. Commodity flow databases tend to be relatively accurate for inter-county flows, but undercount intra-county flows because commodity flow databases rely partly on economic input-output data that ultimately are based on financial transactions between producers and consumers of goods. However, in an urban area, many truck moves are not easily traced to such transactions. Moves from warehouses and distribution centers, repositioning of fleets, drayage moves, parcel delivery, and the like are generally short-distance trips in which there may not be an economic exchange of the goods from one party to another. To compensate for the lack of inclusion of the shorter distance trips in commodity flow data, and to account for types of trucks that do not carry freight, local truck trips are generated based on local employment and economic factors using trip generation rates. These trips are usually generated at the zone level and trip distribution uses methods such as gravity models. The trip rates are calibrated so that the truck traffic volumes that are generated from the combined commodity flow and locally generated truck trips match those from available truck counts. Several terms are used to refer to these two trip types, including commodity-flow trips versus locally generated trips, external versus internal truck trips and long-haul versus local truck trips. Taking advantage of the relative strength of the commodity long-haul approach and the truck short-haul approach within the same model has been called a “hybrid approach.” The two modeling frameworks – freight truck models and commodity-flow models – are described briefly in the following sections. These two models form the basis for the freight/truck hybrid forecasting procedures.

6.2 Three-Step Freight Truck Models

Freight truck models develop highway freight truck flows by assigning an O‑D table of freight truck flows to a highway network. The O‑D truck table is produced by applying truck trip generation and distribution steps to existing and forecast employment and/or other variables of economic activity for analysis zones. This method involves estimating the O‑D table directly using trip generation rates/equations and trip distribution models at the TAZ level. This is similar to the four-step passenger models. The mode choice step is eliminated since truck trips are estimated directly without consideration of other possible modes for moving freight. The components required for this modeling technique include existing and forecast zonal employment data, methods to generate zonal freight productions and attractions by using freight truck trip generation rates, methods to generate truck O‑D flows by applying trip distribution procedures to truck productions and attractions, and methods to assign the O‑D freight truck flows to a highway network.

Freight truck models usually attempt to account for shipment of goods, including local delivery. Because these models are focused exclusively on the truck mode, they cannot analyze shifts between modes. Truck models are usually part of a comprehensive model that forecasts both passenger and freight movement and, consequently will often use a simultaneous assignment of truck trips with automobile trips.

As noted above, freight truck models follow a three-step process of trip generation, trip distribution, and traffic assignment. Trip generation estimates the number of trips either produced in each zone or attracted to each zone and is usually a function of socioeconomic characteristics of the zone (employment by industry, population, or number of households). Trip generation is accomplished using truck production and attraction equations whose coefficients are estimated based on local surveys or by using parameters borrowed from other sources such as the Quick Response Freight Manual (QRFM). Trip distribution determines the connection between trip origins and trip destinations. Trip distribution is generally accomplished using a gravity model similar to that used in a passenger model. In the gravity model, the number of trips that travel between one zone and another is a function of the number of trip attractions in the destination zone and is inversely proportional to a factor measuring the impedance between the two zones. The gravity model is usually related to the travel time between two zones, i.e., the longer it takes to get from one zone to another, the less attractive trips to that destination zone become. Parameters in the gravity model can be developed from local surveys or borrowed from other sources such as the QRFM. The route that trucks use to get from origin to destination is a function of network characteristics, taking into account traffic conditions on each route. Network assignment of the truck trips is usually based on a multiclass equilibrium highway assignment that includes passenger cars; in other words, the model looks for the shortest time path for all trips simultaneously. Freight truck models can take into account the size of trucks and their impact on congestion compared to automobiles (large trucks cause more congestion because they occupy more space than automobiles). In addition, the networks can be coded so that any prohibited routes are not available for truck trips.

6.3 Four-Step Commodity Flow Models

The four-step commodity flow model is similar in structure to the four-step passenger model. Both the four-step commodity flow models and the four-step passenger models require the development of a network and zone structure. Since a larger percentage of freight trips in an urban area are long haul than is the percentage of passenger trips that are long haul, a skeletal highway network external to the region is usually appended to a local passenger network to allow for assignment of these long-haul freight trips. Commodity models can analyze the impact of changes in employment, trip patterns, and network infrastructure.

The commodity-based “trip” generation model actually estimates the tonnage flows between origins and destinations. These flows are converted to vehicle trips after the mode choice step in the process. The trip generation models include a set of annual or daily commodity tonnage generation rates or equations by commodity group that estimate annual or daily flows as functions of TAZ or county population and disaggregated employment data. Base year commodity flow data at the zonal level are used to estimate the trip rates or trip generation equations. The O‑D tables for these flows are typically estimated using gravity models similar to the trip distribution step in four-step passenger models. Trip distribution models are estimated separately for each different commodity group. The unit of flow in the O‑D table is typically tons shipped. The distribution of freight is to a national system of zones, recognizing the large average trip lengths in this class of models. Mode split is a necessary component because O‑D patterns are developed for particular commodities rather than for trucks. Quite often the mode-split step simply assumes that the base year mode share of each commodity flow stays the same in the future. The conversion of commodity truck tonnage to daily freight truck trips uses the application of payload factors (average weight of cargo carried per vehicle load). Payload factors can be estimated on a commodity-by-commodity basis using locally collected survey data (e.g., roadside intercept surveys) or national surveys (e.g., the U.S. Census Bureau VIUS). The assignment of truck freight will typically use either a freight truck only or multiclass assignment model.

6.4 Case Studies

The Southern California Association of Governments (SCAG) in Los Angeles, the San Joaquin Valley in Central California, and the Puget Sound Regional Council in Seattle employ the hybrid method for their truck forecasting models. These three models are discussed in the ensuing sections.

6.4.1  SCAG HDT Model – Los Angeles

The SCAG heavy-duty truck (HDT) model is in the process of being updated based on new truck travel surveys and commodity flow data. In the “current” SCAG truck model, the external trip model is based on a commodity flow database and forecast developed by DRI/McGraw Hill and Reebie Associates (now Global Insight). The external model estimates truck trips for which at least one trip end (either origin, destination, or both) occurs outside of the region.

The “new” updated truck model uses the commodity flow data that were originally compiled for the California Intermodal Transportation Management System (ITMS) developed by the California Department of Transportation (Caltrans). This commodity flow data have an original base year of 1995 and these were based on 1993 county-level commodity flow data developed for the SCAG Interregional Goods Movement Study and the original Caltrans ITMS. Caltrans has since updated the ITMS commodity flow data to a 1996 base year with forecasts to 2006 and 2016 based on the FHWA FAF and Caltrans employment forecasts by industry at the county level. Cambridge Systematics has recently estimated 2003 base year commodity flow data, forecasted from the 1996 ITMS database. In addition, the national CFS for 2002 and the data available for the Southern California metropolitan area was used to conduct a limited validation of the 2003 base year estimate developed from ITMS. This provided an important update to a key data input to the external model of the “new” SCAG truck model.

The framework of the “new” external HDT modeling methodology is determined by the direction of flows (inbound/outbound), commodity type (agricultural, manufacturing, mining, etc.), and shipment type (TL/Private Carrier or LTL), since these factors affect the input parameters and the procedure for commodity flow disaggregation from county-level flows in the ITMS database to the SCAG TAZ level. County-level commodity flows in the SCAG truck model were disaggregated to TAZs using zone-level employment data. For outbound truck moves, commodity flows were allocated to TAZs in the origin county based on the employment share of the producing industry in each TAZ. For inbound flows of manufactured goods and farm products by truckload and private truck modes, economic input/output models were used to determine the portion of each commodity that moves to a manufacturing facility and the portion that moves directly to a warehouse for eventual distribution to a retail facility. For commodities carried by less-than-truckload carriers, these flows were disaggregated from county to TAZ level based on the exact locations of LTL facilities in the SCAG region using a list of LTL terminals.

The SCAG truck model converted commodity flows into truck trips using data from a combination of O‑D surveys (2002 SCAG Truck Count Study) and data from the Census Bureau’s 2002 VIUS. [Note that this data collection effort has been expanded to include all types of vehicles, and the name of the survey has changed to the Vehicle Inventory and Use Survey (VIUS).] First, the tons were allocated to the three truck classes in the model (light-heavy duty trucks, medium-heavy duty trucks, and heavy-heavy duty trucks) using the data from VIUS. Next, the tons in each of the truck classes were converted to truck trips using the payload data from the intercept surveys and VIUS. Weigh-in-motion data were used to convert annual truck trips to daily truck trips. This disaggregation process converted the annual truck tons in the commodity flow database into a daily zone-level truck trip table for the SCAG region.

The internal component of the SCAG truck model is being updated in 2007 based on new truck travel surveys. This component will estimate truck travel for trips where both the origin and the destination are within one of the six SCAG counties. The “new” internal model will be a three-step freight truck model just like the “current” model.

In the “current” model, the trip rates for internal truck trips were estimated using data on daily truck activity collected from a shipper-receiver survey and zone-level employment data. The land use/employment categories were agriculture/mining/construction, transportation/‌communication, wholesale trade, retail trade, financial/insurance/real estate/services, government, and households. Samples for the shipper-receiver survey were drawn by industry group from the American Business Directories’ Southern California Business Directory, a listing of 725,000 businesses, their addresses, telephone numbers, seven-digit Standard Industrial Classification (SIC) Code, and sales and employment figures. The sampling frame did not include households or government facilities.

The survey of shippers and receivers divided trips into two major categories: trips that delivered something to a facility (including services) and trips that removed something from a facility. Essentially, the survey distinguished between pickups and deliveries. Respondents estimated the number of truck trips per day made to their facility and noted whether shipments were truckload or partial truckload deliveries. For several of the categories, insufficient survey data were available to estimate trip rates, so rates were borrowed from other metropolitan area models (Phoenix and San Francisco). Special generator models were used to add truck trips to the table from the major sea ports, intermodal transfer facilities, and airports. The final truck trip table was the sum of the external truck trip table, the internal truck trip table and the truck trip table developed from the special generators.

Trip distribution for the “current” internal trips was accomplished through a gravity model based on a limited number of truck trip diaries. The traffic assignment was done by first allocating the truck trips to the four time periods in the SCAG passenger model using truck count data collected by weigh-in-motion equipment at California’s weigh stations. A multiclass assignment was then performed using both the passenger car and truck trip tables. The model was calibrated and validated using 11 screenlines in the region.

While state of the art for its time, the SCAG model suffers from four weaknesses:

  1. The data used to develop the trip generation and trip distribution elements of the “current” internal model are extremely limited. The “new” model will use the ongoing new truck travel survey data and will try to overcome some of its limitations.
  2. The behavioral basis of the “current” internal model is crude and based on a considerable simplification of different types of truck operations. The “new” model is based on stratifying trucks into trip purposes or sectors, and the surveys are being collected by targeting different economic sectors.
  3. There is no direct linkage between the external commodity flow model and the internal trips generated in the “current” model, and this will continue to be a problem in the “new” model as well.
  4. It is not multimodal.

6.4.2  FASTruck Model – Seattle

The freight action strategy truck (FAST) forecasting model was developed to provide an analytical basis for evaluating the benefits of transportation investments that impact the movement of goods throughout the Puget Sound region. The truck model defines a truck based on relative weight classes and separates light, medium, and heavy trucks for analysis purposes. Medium and heavy trucks are defined to match the definitions used for collecting truck counts by the Washington State Department of Transportation (WSDOT).

The development of the truck model was based on using different forecasting methods for internal and external truck trips because the factors that influence these truck trips are very different. In the case of the external trips, defined as those truck trips that begin and end outside the region, truck trips are affected by economic factors beyond the region borders. In the case of the internal trips, defined as those truck trips that begin and end within the region, truck trips are affected by economic factors within the region borders. Truck trips that have either an origin or destination outside the region and an origin or destination inside the region are affected by both external and internal factors. These three types of truck trips are, therefore, estimated separately using unique methods for each type.

The socioeconomic data used in the FASTrucks Forecasting Model are consistent with those data used in the passenger model, except that the employment data are stratified into more employment categories. This process provides more accuracy for truck travel and allows for a direct relationship between the commodities being estimated in the external trip model and the allocation of these commodities to TAZs within the region.

The trip generation rates for the internal truck model were developed from two primary sources of existing truck models: the QRFM [U.S. Department of Transportation, Quick Response Freight Manual, developed by Cambridge Systematics, Inc., with Comsys Corporation and the University of Wisconsin for the Travel Model Improvement Program, September 1996] and the Vancouver BC truck model [Jack Faucett & Associates, Draft Report for the Lower Mainland Freight Study, for the Greater Vancouver Regional District, May 2000]. The QRFM was selected because it provided trip rates based on national averages. The Vancouver trip rates were selected to provide stratifications of trip rates for more employment categories. The QRFM was used to derive trip rates for light trucks, while both the aforementioned sources provided trip rates for medium and heavy trucks, although the QRFM defines these categories as six or more tire trucks and combination trucks, respectively. These trip rates were originally developed using the two primary sources of data, but were adjusted during model calibration.

One additional source of data that was available to use in adjusting the internal model heavy truck trip rates for manufacturing and wholesale sectors was the TRANSEARCH commodity flow dataset. These data were processed to identify internal, county-to-county commodity flow and converted to average daily truck flows for comparison with other trip rates. The TRANSEARCH commodity flow dataset did not contain any commodities for internal truck trips other than manufacturing and wholesale trade, so these were the only sectors that were adjusted based on these data.

For the external FASTruck model, three primary types of external trips were represented: 1) trips that begin in Puget Sound region and leave the region; 2) trips that begin outside the region and are destined to someplace within Puget Sound region; and 3) trips traveling through the region. The two sources of data for these trips are the TRANSEARCH commodity flow data, which was converted to truck trips, and the traffic counts at external stations. Both of these sources provided some, but not all, of the data needed to develop comprehensive truck trip tables so some adjustments were made to these sources to fill in the gaps in these data sources.

WSDOT purchased TRANSEARCH data from Reebie Associates (now Global Insight) for commodity flows that traveled into, out of or through the Puget Sound region. The commodity flow data provided tons of goods moved by commodity and truck type (private carrier, less than truckload, and truckload). These data were converted to truck flows by applying payload factors (average tons per truck by commodity category) that were derived from the 2002 VIUS. VIUS is a national database of trucks that was used to derive payloads for all trucks registered in Washington State.

The truck trip tables developed from the TRANSEARCH data were further processed to evaluate the origin and destination of the commodities with respect to the Puget Sound region. These tables were compared to total volumes of truck trips at the external stations and to total internal volumes from the trip generation model. The truck trips for external trips (both internal-external and through trips) compared favorably to the total truck volumes at external stations for heavy trucks. The internal truck trips represent 32 percent of the total internal heavy truck trips estimated in the trip generation model, so these were used to estimate trip rates for manufacturing and wholesale trade, as mentioned in the previous sections.

The TRANSEARCH data identifies the origin and destination of commodity flows for 30 geographic markets. These regions were associated with appropriate external stations and internal Puget Sound counties to disaggregate these data into traffic analysis zones. Modifications to the original dataset were made to eliminate those commodities that would not likely travel through Puget Sound. The TRANSEARCH data provided a direct calculation of external (through) trips. These through trips were subtracted from the total heavy truck counts to provide an estimate of internal-external and external-internal trucks at each station. It was assumed that all TRANSEARCH commodities were moving on heavy trucks. The internal-external and external-internal trucks were distributed to internal zones using the same allocation by industry as the internal truck trips.

Some of the critical issues in the FASTruck model are:

  • The internal truck model is entirely based on borrowed trip rates and not on any local survey data.
  • The internal model is based only on GVW ratings and not trip purposes or sectors.
  • The external trips are derived from TRANSEARCH data that were purchased for years 2000 and 2020. When the model was updated to year 2005, these external data were interpolated using the two years data.
  • The external commodity flow data were available only for manufacturing and wholesale inside the four-county Puget Sound region, which enabled cross-checking the internal model heavy truck trips associated with these two categories only.
  • It is not multimodal.

6.4.3  San Joaquin Valley Truck Model – Central California

The approximate bounds of the San Joaquin Valley region are the Sacramento metropolitan area to the north, the San Francisco Bay Area and California coast to the west, the Sierra Nevada Forest to the east, and the Los Angeles metropolitan area to the south. The purpose of developing a truck model for the region was to provide an analytical framework for evaluating how changes in the transportation system of the Valley would impact goods movement.

The San Joaquin Valley truck model was developed using the Caltrans road network. The truck model reported truck volumes in two truck classes: medium heavy-duty trucks and heavy heavy-duty trucks. These truck classes are defined based on gross vehicle weight rating and are consistent with the California Air Resources Board truck definitions. The model utilizes a truck trip table that was generated from two separate truck trip tables. The first of the truck trip tables was developed using the Caltrans Intermodal Transportation Management System (ITMS) commodity flow data. These truck trip tables were developed entirely from commodity flow data. The second truck trip table was developed from local socioeconomic data.

An automated procedure was developed to calculate the number of truck trips associated with the ITMS commodity flow data and to assign these truck trips to TAZs. The ITMS database includes O‑D detail for freight flows for each of the major modes and each of the major commodities at the county level. The first step towards creating the truck trip table was to convert the truck tons into truck trips. This was done by developing a ton per truck ratio, referred to as the average payload, for the ITMS truck tonnage data. Average payloads were calculated for each commodity using the 1997 VIUS data. The commodity classification used for the payload matrix is the Standard Transportation Commodity Code (STCC) system. Application of the payload matrix to the ITMS data created a county-level truck trip table for the State of California from the truck tonnage data.

The truck trip table generated from the ITMS data was then disaggregated geographically to create relevant regions for the truck model. Internal regions were based on the eight counties that constitute the San Joaquin Valley study area. Regions external to the Valley were developed to correspond to each of the external cordons that can be used for trucks exiting the study area.

Next, the county-level ITMS commodity flow truck trip data were allocated to zip codes. This allocation was performed using Dun & Bradstreet employment data from 2000. These data include the number of employees by zip code for each of the eight counties in the San Joaquin Valley for thousands of different employment categories based on the SIC system at a four-digit level. The truck trips were allocated to zip codes based on matching the STCC codes in the truck trip table with the employment categories in the Dun & Bradstreet database for each STCC and each zip code. For outbound flows, one-to-one correspondences were made between commodity codes in the two databases. For inbound flows, tons were allocated based on employment in the consuming industries for each commodity.

The zip code-level trips were then allocated to the TAZs in the truck model. This allocation was done based on a combination of employment data from the statewide model and the areas of geographic overlap between the zip code and zone boundaries. This process developed the final zone-level truck tonnage table for the 1996 ITMS data. This truck trip table was then projected to the year 2000 based on the freight tonnage growth derived from the FHWA FAF data for the State of California.

The second truck trip tables or the non-ITMS truck trip table was used to supplement the truck trip table developed from ITMS data. It is typical for truck trip tables developed from commodity flow data to underestimate total truck activity because of an underestimation of local truck trips. Therefore, secondary truck trip tables are generated to improve the match between truck volumes generated by truck models and truck count data. These secondary truck trip tables are typically generated from socioeconomic data.

The trip production rates for the secondary truck trip tables were developed primarily from the QRFM [U.S. Department of Transportation (DOT), Quick Response Freight Manual, developed by Cambridge Systematics, Inc., with Comsys Corporation and the University of Wisconsin for the Travel Model Improvement Program, September 1996]. The QRFM provides trip rates based on national averages for medium and heavy trucks. These rates were scaled back during model calibration. Truck trip consumption rates were developed to estimate the relative number of trucks that are attracted to each zone in the Valley. These consumption rates were developed by evaluating the industries that are present in the Valley (based on employment data) and estimating the inputs required for these industries based on input-output data. The input-output data were available at the national level and scaled to represent the input-output characteristics of the State of California. The tables for the State of California were then disaggregated to represent truck trip rates for medium and heavy truck trips.

For this model, the socioeconomic data available are stratified into the following 10 industry groups: 1) agriculture/farm/fishing, 2) mining, 3) construction, 4) manufacturing – products, 5) manufacturing – equipment, 6) transportation, 7) wholesale, 8) retail, 9) finance, and 10) education/government. The availability and use of multiple industry groups increases the accuracy for truck travel generation because each industry group can be assigned different truck trip generation rates.

Trip distribution was performed using a standard gravity model. Model calibration was performed using a reasonableness check of the average truck trip lengths estimated by the model.

The truck model is designed to generate truck volumes based on average daily traffic. The truck model output reports truck volumes based on truck classes that the CARB defines as medium-heavy duty and heavy-heavy duty for regulatory purposes (more than 14,000 pounds gross vehicle weight rating (GVWR)). Medium-heavy duty trucks (MHDT) have a GVWR between 14,001 and 33,000 pounds. Heavy-heavy duty trucks (HHDT) have a GVWR of 33,001 pounds or more. A multiclass equilibrium assignment was performed and validated by comparing model truck volume outputs to observed truck counts collected by Caltrans.

Some of the issues in the San Joaquin Valley truck model that are being addressed in the ongoing model update include:

  • There were no calibration procedures adopted to validate the ITMS commodity flows to observed truck counts.
  • Flows of nonmanufactured commodities (especially farm and mining products), flows between major city pairs (e.g., flows between the urbanized portions of Southern California and the San Francisco Bay Area), and flows disaggregated to the zip code level need more careful scrutiny and adjustment using a variety of other sources.
  • The secondary truck trip tables were developed using QRFM trip rates that were found to be too high and needed to be scaled back during calibration. The new model update will derive trip rates from the National Cooperative Highway Research Program (NCHRP) Synthesis Report 298 [Cambridge Systematics, Inc., Truck Trip Generation Data, National Cooperative Highway Research Program Synthesis 298, Transportation Research Board, 2002] on truck trip generation data.
  • It is not multimodal.

6.5 Issues with Hybrid Approaches

6.5.1  Conversion of Commodity Flows in Tonnage to Truck Trips

After the commodity flows have been distributed or allocated to various TAZs based on socioeconomic data, they need to be converted to truck trips before any assignments can be done. There are a few ways to do the conversion, but using nationally available databases is the most popular, easiest, and cheapest method. These databases include the Truck Inventory Use Survey (TIUS), which is now called the VIUS. The other methods of conversion include conducting external cordon surveys that provide information about truck payloads by commodity type. From all these methods, the information necessary derived to convert flows to trucks are commodity type, number of axles, and weight of trucks. These data are then used to compute average tons per truck by commodity category also known as payload factors.

The major drawback of using national databases, such as VIUS, is that it provides data by state and not by any specific region. So the payload factors are an average of all trucks across the state. Usually adjustments are made to these based on locally available data either from weigh-in-motion (WIM) data or intercept-based cordon surveys.

The truck models in Seattle and San Joaquin Valley used the 1997 VIUS data to estimate the payload matrices. The external truck trips in the Seattle model were recently updated using the observed data on certain key external stations.

In a recent study in Los Angeles, the payload matrices in the new SCAG external HDT model were updated using the 2002 SCAG Goods Movement Truck Count Study. This study was conducted at external cordons that provided new information about truck payloads by commodity. The data from these external surveys suggested that the payload factors in the old model that were derived from the 1992 TIUS data were too high for heavy-heavy trucks. In addition, the data showed that the allocation of tonnage carried by weight class that was used in the model did not allocate sufficient amounts of tonnage to heavy-heavy trucks. This led to an underestimation by the model of the number of heavy-heavy trucks at the external cordons.

6.5.2  Intra-County Flows Are Underrepresented

The commodity flows are usually estimated and available at the county level and are the strongest for county-to-county freight movements. However, the flows within a county are underrepresented in a commodity flow database and it precludes the ability to disaggregate these flows to TAZs that have both the origin and destination within a county. This is, however, not an issue if the intra-county truck movements are captured using travel survey-based trip rates.

The truck model for Seattle did not use the TRANSEARCH-based commodity flow data for flows within the four-county region but instead used a land use-based trip rate method to generate truck trips internal to the region. The commodity flow data were used only for flows that traveled into, out of, or through the Puget Sound region. A similar approach also was used in the San Joaquin Valley truck model and the SCAG HDT model where the Caltrans ITMS data was used for the external freight movements.

Another drawback of intra-county commodity flows is that it does not include trucks that do not carry freight such as trucks related to the service industry. A significant portion of the truck movements within a county are attributed to this sector that encompasses safety, utility, public service, and business and personal service vehicles.

6.5.3  Overlap of Commodity- and Truck-Based Estimates of Truck Trips

In a hybrid model, both the commodity- and truck-based models predict truck trips in a certain region but it is very difficult to separate the two estimates from each other. That is, commodity-based estimates already might be picking up the trucks in the region that the truck-based estimates include, and vice versa. This overlap is crucial and needs to be dealt with in those models that do not define the study area by the county boundaries. Usually, the commodity-based truck trips are used for those trips with at least one external trip end that is outside the study area. Since the commodity-based truck trips are county-to-county, and if the study area includes partial counties, then the overlap of truck trips from the two estimates can result in overestimation of truck trips. This is, however, not an issue if the study area is defined by its counties’ boundaries.

6.5.4  Lack of Correlation of Truck Trip Purposes or Sectors between Commodity- and Trip Rate-Based Models

In a hybrid model, after the commodity- and truck-based estimates are developed, they are all added together irrespective of what commodity type or sector they belong to. This happens after the trip distribution stage. The only stratification that is carried through the assignment process is the truck class which is either in GVW ratings or number of axles such as the FHWA classes. However, this becomes an issue if the external truck trips from the commodity-based estimates need to be included in the trip distribution stage where distributing truck trips by economic sector is a necessity. The reason for this is the poor correlation between the commodity type carried by trucks from the commodity-based approach and the economic sector of truck trips derived from the truck-based model.

6.5.5  Hybrid Approaches Are Not Multimodal

The commodity-based approaches estimate flows by different modes of travel such as surface, rail, air, and water, whereas the truck-based approaches estimate truck trips only. Therefore, the hybrid approaches are appropriate only for trucks, and as a result, planning and policy analyses needed for multimodal studies are not possible. The freight flows carried by nonsurface modes (rail, air, water) need to be modeled and forecasted through other modeling tools.

6.5.6  Limitations in Validating Multimodal Commodity Flow Models

The commodity flow models are usually developed based on commercially and nationally available databases such as the TRANSEARCH and the CFS. Once the trip tables are developed from these databases, they are added into the model either during or after trip distribution. These trip tables are assumed to be accurate and normally validation of these trips are not done. Moreover, there is very limited data collected to validate these trip tables and their trip distribution patterns. Even if observed data need to be gathered, it has to be through external cordon surveys to get O‑D information on truck flows coming into, going out of, and passing through the region. Vehicle classification counts at certain key locations or corridors also can be used to validate the entire truck flows passing through those locations but there is no way to separate the trucks that are external to the region from the internal truck trips.

6.5.7  Commodity Flow Databases Are Expensive

The commercially available commodity flow databases such as TRANSEARCH and Claritas are very expensive. They could cost up to $50,000 for one year of commodity flows in a particular state or region. The nationally available databases such as CFS data and FAF data are produced by the U.S. Bureau of Census and are free of cost. However, these have many drawbacks and are not very comprehensive. A series of checks and adjustments need to be made to these data before they can be used and applied to a region. The ITMS data that was used for the SCAG external HDT model development, was thoroughly reviewed and preprocessed before actually using it to develop external trip tables. The preprocessing involved calibrating and validating the flows at certain key external stations which had vehicle classifications counts from the Caltrans Traffic Count book.

6.5.8  Mode Choice Models Are Required to Separate out Truck Flows from the Rest (Air, Water, Rail)

A mode choice analysis needs to be done in order to separate out truck flows from other modes of travel as the hybrid models predict the truck flows or trips in a region. These mode choice models can be done in a couple of ways and are usually data intensive. The market segmentation-based mode choice model is simple and inexpensive, but it does require detailed commodity flow and length of haul information. However, this approach does not consider modal characteristics and, hence, is not policy-sensitive.

An alternate and more robust method that is behaviorally sensitive is the logit choice method which is the most comprehensive. These models examine the characteristics of each individual shipment and the available modes. However, a number of data items need to be gathered to develop these models such as the travel-time data by mode, price of shipment through different modes, schedules and routings, and reliability data of various modes. Surveys can be done to gather these data but they are expensive and time-consuming.

6.5.9  Commodity Flow Forecasts Are Required/Purchased

The hybrid model that uses commodity flows to represent the external truck movements also needs future year flows to forecast future year tuck volumes. These forecasts are usually purchased for a certain year in the future and growth factors are developed based on the base and future year flows to develop trip tables for any interim years. The forecasts for the SCAG external HDT model were derived from growth factors that were developed using different years of ITMS flow data. These data are available for every 10 years from 1996 to 2026. In the case of the hybrid truck model in Seattle, base and future year TRANSEARCH databases were purchased.

If the future year forecasts are not available or purchased, then relationships among external truck flows and socioeconomic data need to be developed using the base year trip tables. The future year socioeconomic data can then serve as the input to the forecasting model and external truck trips can be forecasted.

6.5.10 Special Generators (Ports/Airports) Not Well Represented in Commodity Flow Models

The commodity flows are developed based on the economic activity, consumption rates, and the input-output characteristics in a region. However, these flows do not adequately capture the freight/truck flows related to certain special generators such as airports and seaports. The reasons for this are the nonlinear relationships or a lack of relationship or hard to establish relationships at these special facilities among the freight/truck flows and the corresponding economic activity.

In the SCAG HDT model, two separate models were developed, one for the air cargo shipments and the other for the port truck flows. Separate surveys were conducted at these two facilities and they have different inputs and networks to capture the truck flows coming into and going out of these generators. In the LAMTA Cube Cargo model, surveys were conducted at various intermodal terminals to capture the trip chaining of truck trips.

6.5.11 Issues with Logistic Nodes

Logistic nodes are used in supply chain/logistic chain models that use economic input-output characteristics to calculate supply and demand for each economic sector with an assignment of goods to logistics families to determine the spatial patterns of supply and demand. The logistic nodes are used as means to distribute or disseminate the external movements to internal zones. These nodes are places such as major goods yards, multimodal terminals, railway stations, and distribution centers where trip chaining of long-distance flows occurs.

The LAMTA freight forecasting modeling process involves the representation and modeling of the long-distance logistics system in the Transport Logistics Node model (TLN). The TLN model is only applied on the long-distance flows. These are defined as flows from the internal area (for example, in the Los Angeles study, this was defined as the greater southern California area) to the external area (in the Los Angeles study, this was defined as the remainder of the United States as well as entry points to/from Mexico and Canada) and flows from the external area to the internal area. Data on TLNs was collected through a shipper survey conducted for 131 locations in Southern California combined with rail operator data obtained at six intermodal terminals.

The following are some of the critical issues that need to be addressed before using such an approach for modeling external freight/truck flows:

  • The commodity flows that move wholly within the internal area are not modeled or captured using the logistic nodes approach, unless they are flows that move from one node to another. These are referred to as short-haul movements.
  • Although the concept of using logistic nodes is well established in industrial engineering processes, it has not been applied until recently to the truck flows in a travel demand model.
  • The long-haul commodity flows are split at the logistic nodes by mode, commodity type, and direction. So a lot depends on the placement of these nodes in the internal parts of the region and the right logistic nodes need to be picked to ensure precise distribution of flows among modes and zones.
  • Shipper/receiver surveys need to be conducted in as many logistic nodes as possible to ensure proper representation of the distribution points in the region. This can lead to the whole process being very expensive.

Office of Operations