Quick Response Freight Manual II

12.0 Case Studies

12.1 Los Angeles Freight Forecasting Model

12.1.1 Purpose and Objective

The Los Angeles freight forecasting model (LAMTA model) was developed as part of the efforts by the Los Angeles County Metropolitan Transportation Authority (Metro) to address the impacts of the growing volume of goods movement in and around the Los Angeles metropolitan area on the overall transportation system in the region. The development of the LAMTA model was prompted by Metro’s decision to upgrade its transportation demand model, with the objective of gaining the ability to assess the potential benefits that could be realized from the implementation of various freight transportation policies and infrastructure projects in the future. Also, it was crucial for the freight model to produce results at the same TAZ system as the passenger model to facilitate an integrated modeling framework for the analysis of goods and passenger movements on the transportation network.

The LAMTA model was developed using a more robust and innovative approach compared to previous modeling undertakings, in order to meet the above technical challenges and model application requirements. Some innovative modeling approaches incorporated in the LAMTA model are outlined below:

• Freight movements are modeled at different levels of detail for long- and short-haul movements. Long-haul freight is derived from commodity flows at a national level with full modal options (truck, rail, and air), and are chained with trips through intermodal terminals. Short-haul freight is derived from socioeconomic data in the region and chained with trips through warehouse and distribution centers.
• Service trucks that do not carry freight are modeled separately and included as part of overall truck movements.
• The model has a separate module for data inputs and modeling of transport logistics nodes/centers (warehouses, distribution centers, and intermodal terminals), which incorporates trip chaining concepts.
• The modeling framework provides forecasts that reflect trends in labor productivity, imports, and exports. Trend forecasts are derived from FAF national data.
• Freight movements coming in through the ports are simulated as special generators based on forecasts from the ports and data collected at the intermodal terminals.

12.1.2 Model Study Area

The study area for the LAMTA model is divided into internal and external zones. Internal zones include detailed representation of the six-county Southern California region, which comprises of Los Angeles, Orange, Imperial, San Bernardino, Ventura, and Riverside counties. The external zones include the rest of California, remainder of the United States (other than California), Canada, and Mexico. The LAMTA model comprises of two levels of zoning systems, namely a fine zoning system and a coarse zoning system. The fine zoning system is based on the LA metro passenger travel demand model zoning system, comprising of 3,000 zones, while the coarse zone system is based on zip code boundaries in the Southern California region. For external zones, both the fine and the coarse zone systems are based on county boundaries for the rest of California region, and state boundaries for the rest of the United States. The entire LAMTA model study area has 3,800 fine zones and 650 coarse zones.

12.1.3 Modeling Framework

The LAMTA model comprises of the following modeling components:

• Trip generation model;
• Trip distribution model;
• Mode split model;
• Transport logistics node (TLN) model; and
• Vehicle model.

The trip generation and distribution models use a combination of the hybrid modeling approach for freight forecasting, while the Transport Logistics Node (TLN) and the Vehicle models incorporate logistics chain/supply chain modeling and tour-based modeling approaches, respectively. The following sections provide a brief description of hybrid, logistics/supply chain, and tour-based models, before proceeding to the specifics of each modeling component of the LAMTA model.

Hybrid freight forecasting models for urban areas are based on a combination of commodity-based freight modeling (which use base year and forecast commodity flow data as inputs, and estimate multimodal freight trips on the transportation network) and three-step truck modeling approaches (which use trip generation, trip distribution, and traffic assignment steps specifically for the trucking mode, to estimate truck trips on the network). Typically, interregional or long-haul freight flows are modeled using commodity-based freight models, while three-step models find applications in modeling local truck trips in an urban area (since most of the intra-urban freight flows are carried by trucks).

Supply chain/logistics chain models simulate logistics choices throughout whole supply chains for specific industries, and those models use that information to model mode choices and the spatial distribution of freight flows through various stages in the supply chain. A typical example of a logistics model is one that combines an economic I-O model, which calculates supply and demand for each economic sector, with a logistics choice model, which assigns goods to logistics families, in order to determine the spatial patterns of supply and demand. A series of logistics models are developed that define the distribution systems that are used by each logistics family and the spatial organization of warehousing and distribution systems for product delivery and supply chain management.

Tour-based models for truck trips derive methods from the relatively new world of activity-based passenger travel demand models. They focus on the tour characteristics of truck trips and are less concerned about what is being carried in the vehicle. These models are particularly applicable for the modeling of local truck trips in urban areas, where significant tour making activity of truck trips is prevalent (for example, truck trips that originate at retail distribution centers, and make multiple stops to deliver goods at many retail outlets in the urban area).

In the LAMTA model, the supply-chain and tour-based modeling concepts are applied selectively in the TLN and vehicle models, respectively, to ensure that these methods apply specifically to those freight or truck movements that will benefit from the additional modeling complexities. Figure 12.1 depicts the steps involved in the freight-forecasting process in the LAMTA model. Each of the modeling components in the LAMTA model are described more comprehensively in the following sections.

Figure 12.1 LAMTA Model Freight Forecasting Process

Source: Los Angeles Cargo Forecast Model Development, Outwater et al., 11th World Conference on Transportation Research, Berkeley, California.

12.1.4 Trip Generation Model

The trip generation model is the first modeling component of the LAMTA model that forecasts total tons of each commodity group produced and consumed in all internal coarse zones. Commodity productions are divided into internal productions, which are destined to internal zones and exports that move to external zones in the model. Similarly, commodity attractions/consumptions are classified as internal attractions that originate from internal zones and imports, which come from external zones. The trip generation component of the LAMTA model has five elements, which are described below.

Trip Generation for Internal Zones

Trip generation for internal zones is estimated by applying linear regression models, which use socioeconomic attributes of zones (population and employment) as inputs to estimate zonal commodity production and attraction tonnages. The parameters and constants of these models (regression coefficients) are estimated using the 2003 commodity flow data from the Caltrans Intermodal Transportation Management System (ITMS) database, and detailed zonal socioeconomic data from the Southern California Association of Governments (SCAG). Different linear regressions are developed for commodity production and attraction models for the 14 commodity categories in the LAMTA model. For production models, regressions are developed separately for each commodity group by modeling outbound commodity tonnages as a linear function of the corresponding industry employment. For attractions, an I-O matrix was used to identify the industries that consume each commodity group for secondary manufacturing and to use the zonal employment for those industries as regression model inputs to estimate the total zonal commodity attraction/consumption tonnages.

Special Generators

Sea and airports are included as special generators in the model, in order to identify and model goods movement flows by each commodity group that travel through these locations. Truck trips related to sea port activity are estimated from the port truck trip tables provided by the Port of Long Beach and Los Angeles model, while the SCAG airport model was used to derive truck trips associated with air cargo. These trip tables were converted into zone system of the LAMTA model and added to the final truck trip table before performing the truck traffic assignment.

Trends in Commodity Production Efficiencies and Other Factors

Trends in labor productivity are estimated by comparing commodity production tonnages with industry employment data at specific points in time. The CFS, conducted in 1997 and 2002, was the main data source used to calculate industrial labor productivity trends, and apply them to the future year LAMTA model.

External Commodity Flows

The external flows (flows between the internal model study area and external regions) are developed from the ITMS data. These are input as user-defined values by commodity class for the base (2003) and future years (2030). These commodity flows represent the amount of production exported to external zones, and the amount of consumption/‌attraction that is imported from external zones. These were compared to truck counts at external stations and updated.

Trends in Import and Export Levels

Trend rates in import and export levels have a significant impact on the growth in commodities and associated truck trips and are represented separately in the model to reflect this. These import and export trend rates, which are only applied to the future year model, are developed from the base and future years of the FAF commodity flow data for California.

The generation model was calibrated by comparing model volumes to total truck counts and making appropriate adjustments to the generation models for productions and attractions to improve this comparison.

12.1.5 Trip Distribution Model

The trip distribution models in the LAMTA freight modeling framework produce trip tables of goods flows by commodity class, which are derived using gravity model techniques. Key assumptions of freight flow distribution in these models are developed using available freight data sources such as the ITMS and FAF.

The trip distribution model assumes percentage splits between short- and long-haul travel by commodity class which are derived from the ITMS data. Since these percentage splits may change over time, based on trends that may be outside the capabilities of the model to predict, these trends are incorporated in the model as external inputs, based on ITMS forecasts. The model assumes all short-haul flows to be carried by the trucking mode, while long-haul movements pass through a mode split modeling component, to determine the mode shares by trucking and rail modes.

A gravity model with a negative exponential deterrence function is used for the distribution of short-haul trips. The function is applied to the square of the distance as shown below:

where,

• c is a commodity group;
• e is the base of the natural logarithm function (approximately 2.718282);
• d is the distance;
• Fc(d) is the deterrence function of distance (impedance) used in the gravity model; and
• p_c is the calibration parameter for commodity c.

Similar to the short-haul trip distribution, a gravity model with a negative exponential deterrence function is applied for the long-haul trip distribution. However, the difference is in the impedance variable due to the multimodal aspect of the trip distribution, which in this case is the generalized cost calculated in the mode choice model. Generalized cost is a linear combination of time, distance, and cost by mode, weighted by the mode choice coefficients.

The mathematical equation for the long-haul trip distribution model is as follows:

where,

• c is a commodity group;
• e is the base of the natural logarithm function (approximately 2.718282);
• G_c is the composite cost derived from the logit model used for modal split;
• F_c(G_c) is the deterrence function of generalized cost used in the gravity model;
• P_c is the calibration parameter for commodity c;
• Г_c is the growth factor for the calibration parameter for commodity c; and
• y is the time, in years, from the base year to the future year for which the model is being run.

Gravity model parameters are calibrated by commodity class for short- and long-haul goods movements using average trip length data from a variety of sources. Trip tables at the coarse-zone level are distributed to the fine-zone system based on an allocation of goods by commodity class using zonal socioeconomic data. The distribution models produce O‑D tables (for short- and long-haul) of annual tons of goods movements by commodity class.

12.1.6 Mode Choice Model

Mode choice models are applicable only to the long-haul goods movements, since all short-haul moves occur by the trucking mode. The mode choice model is a multinomial logit (MNL) model stratified by commodity and distance classes. A generalized cost function is defined for each combination of commodity class and mode. There are three independent variables associated with each mode, which include time, distance, and highway generalized cost. The functions each involve four coefficients, one for each of the independent variables and one for the constant term, as follows:

where,

• c is a commodity group;
• m is the mode;
• d is the distance;
•  is the time;
•  is the highway generalized cost;
• Г_c(d, , ) is the composite cost function for commodity group c and mode m;
• Г_cm(d, , ) is the generalized cost function for commodity group c and mode m; and
• _cm(d, , ): The proportion of tons of commodity group c that will travel by mode m.

Since distance classes were observed to have unique mode choice sensitivities based on calibration data, these models were further segmented based on distance classes. Coefficients for the model were borrowed from the Florida Statewide Freight Model. These were then calibrated from data in the ITMS. The output of the mode choice models include O‑D tables (for long-haul shipments) of annual tons of goods moved by commodity class and mode.

Transport Logistics Node (TLN) Model

An innovative and important component of the LAMTA freight demand forecasting process is the representation and modeling of the long-haul freight logistics system through the Transport Logistics Node (TLN) model. The TLN model is only applied to long-haul freight flows, which are defined as flows between internal zones (within the six-county Southern California region), and external regions (remainder of the United States as well as entry points to/from Mexico and Canada). An example of a typical long-haul shipment treated in the TLN model would be automobiles manufactured in Michigan traveling to southern California. Freight flows that move entirely within the internal study area of the model are not modeled in the TLN model.

TLNs are defined as locations such as major intermodal yards, trucking terminals, transload facilities, and warehouse/distribution centers where trip chaining of long-distance flows occurs. The TLNs considered by the TLN model are only those located within the internal study area, information on which was collected through a shipper survey conducted for 131 locations in Southern California combined with rail operator data obtained at six intermodal yards.

The TLN model is based on two primary elements: the commodity flow matrices from the mode/distribution model and a description of the TLNs. The commodity flow matrices are inputs directly from the mode/distribution model – one table or matrix per combination of major mode of transport and commodity class. The following three parameters are applied where the internal zone has a TLN located within it:

1. Long-distance truck flow splits by direction (inbound and outbound). The following parameters serve as inputs to the model for each commodity class and direction:
   • Amount of goods moved by trucking mode that are shipped in full truckload, partial truckload, and consolidated load (less-than-truckload) shipments; and
   • For each of the above load types, the percentage of shipments that will pass through TLNs.
2. Long-distance rail flow splits by direction:
   • For each commodity class, by direction, the percentage of shipments that will pass through TLNs.
3. Long-distance inland-waterway flow splits by direction:
   • For each commodity class, by direction, the percentage of shipments that will pass through TLNs.

The commodity flows are split into two segments:

1. Long-haul portion of the movement (travels via the input mode: truck, rail, or ship); and
2. Short-haul portion of the movement (travels via truck).

The short-haul portion of the movements is distributed between the TLNs in the internal areas using another set of parameters. These are defined by direction for each commodity class, specifying the percentage of goods that will go to or come from each of the TLNs. At the end of this process, the TLN model produces four matrices per mode per commodity group. These are short-haul direct (do not go via a TLN), long-haul direct (do not go via a TLN), long-haul to or from TLN, and short-haul to or from a TLN. All short-haul flows to or from a TLN are truck only.

Vehicle Model

The vehicle model is used to convert the matrices that contain annual commodity flow tonnages by truck (direct-short-haul flows, short-haul trips to and from TLNs, and long-haul truck flows) into daily vehicle truck matrices. The truck matrices are divided into heavy and light trucks. All long-haul truck flows are assumed to be in heavy trucks.

The main parameters in the vehicle model include the fraction of shipments for each commodity flow category (direct-short-haul flows, short-haul trips to and from TLNs, and long-haul truck flows) that are carried by each truck class, and the payload factors by commodity group and truck classes. The following sections describe the applications of these parameters in the model, in greater detail:

• For the direct short-haul flows, percentage of the goods that move by light trucks by commodity class and average tons loaded in each truck. An additional scaling parameter is used to account for empty short-haul trucks.
• For the short-haul truck flows to and from the TLNs, percentage of the goods that move by light trucks by commodity class and average tons loaded in each truck. An additional scaling parameter is used to account for empty truck short-haul truck flows to and from TLNs.
• For long-haul truck movements, a parameter with the average tons loaded in each truck by commodity class, as well as a parameter to account for empty long-haul trucks.

These parameters are used within two models contained within the vehicle model:

• The standard vehicle model that is used for flows directly from origin to destination and back. This model allows return loads to come from the destination back to the origin and also allows the truck to find return loads within a user specified criteria.
• The touring vehicle model that simulates multi-point pickup and drop-off.
• The standard vehicle model is applied on all origin-destination flows, except those coming to or from TLNs and those specified by the user. Once the models have been run, all matrices for a given mode and commodity group are combined to give a single vehicle matrix, relative to the fine zoning system (i.e., trips to or from a TLN are now assumed to run to or from the fine zone containing the TLN), in vehicles per annum. The matrix is then divided by the value of control data parameter to give units of vehicles per day.

12.1.7 Data Requirements for the LAMTA Model

The following sections describe the model application data (base and future year data inputs) required for inputs to the LAMTA model for the freight forecasting process.

Roadway Network

The source data for the roadway network for the model is the FHWA’s FAF. The roadway network is used to estimate truck travel times and distances, which also consider user assumptions related to average truck pickup and drop-off times, and driver rules related to break and overnight stop times. The roadway network costs are estimated using costs per ton-mile values for each commodity type, in conjunction with roadway distances.

Rail Network

FHWA’s FAF also is the source data for the LAMTA model rail network. The network was used to estimate rail travel times, distances, and costs. User assumptions also are applied to add pickup and drop-off times, transfer times, and average rail speeds. Rail network costs are estimated based on assumptions on costs per ton-mile by commodity type, which are applied to distance. The railway network in the model consists of Class I railroads with other railroad classes retained in order to provide network connectivity.

Socioeconomic Data

Socioeconomic data is a major input for the LAMTA model since it is used to estimate forecast tons produced and consumed in each zone in the model. Socioeconomic data also are used in other parts of the modeling system, including the vehicle and routing model components. As described earlier, the LAMTA model uses a two-tiered zone system, namely the "coarse" and "fine" zone systems. Much of the calculation is done at the coarse zone system since observed matrices of commodity flows (for example, ITMS) are unavailable at a more detailed zonal level. The coarse zone data are then translated to the fine zone level for network assignment. For each zone system, the following socioeconomic data are used in the model:

• Population;
• Households;
• Agriculture, mining, and construction employment;
• Retail employment;
• Government employment;
• Manufacturing employment;
• Transportation employment;
• Wholesale employment; and
• Service employment.

Commodity Flow Matrices

Commodity flow matrices in the model are derived from the Caltrans ITMS database, which has three main regional segregations of commodity flows, including Internal (intercounty flows within California), Inbound (flows from other states in the United States, Canada, and Mexico to California counties), and Outbound (flows from California counties to other states in the United States, Canada, and Mexico). The data from the ITMS were analyzed and aggregated to 16 commodity categories for the LAMTA model, based on the objective of achieving homogenous distance and mode choice characteristics within each category. The commodity classes used in the model are presented below:

• Mining;
• Metal ores and petroleum;
• Raw materials manufacturing;
• Cement and concrete manufacturing;
• Metals manufacturing;
• Processed metals manufacturing;
• Transportation/HH equipment manufacturing;
• Other transportation equipment manufacturing;
• Chemical manufacturing;
• Wood;
• Paper/wood products manufacturing;
• Ranching;
• Farming;
• Grain and specialized;
• Food manufacturing; and
• Other manufacturing.

Transport Logistics Nodes

The TLN model routes a portion of the long-haul commodity flows through transport logistics nodes in order to better model the long-haul freight logistics system, as well as accurately represent and model trip chaining characteristics associated with freight flows through these critical nodes in the freight logistics system. The list of intermodal yards that form a critical nodal component of the freight logistics system in Southern California that are used in the model are presented below:

• Union Pacific Intermodal Container Transfer Facility (UP ICTF);
• Union Pacific East Los Angeles Yard (UP East LA);
• Burlington Northern Santa Fe Hobart Yard (BNSF Hobart);
• Union Pacific Los Angeles Transportation Center (UP LATC);
• Union Pacific City of Industry Yard (UP Industry); and
• Burlington Northern Santa Fe San Bernardino Yard (BNSF SB).

12.2 Portland Metro Truck Model

12.2.1 Introduction

The Portland Metro truck model also is referred to as the Tactical Model System. The Tactical Model System, together with the Strategic Model Database (SMD), form the core elements of the truck freight forecasting model for the Portland metropolitan area. The SMD, which provides commodity flow data inputs to the tactical model, contains aggregate present and future freight flows for different commodity and mode combinations. This database serves as a useful tool providing freight flow inputs required for strategic decision-making concerning the development and operation of Portland’s seagoing and river marine terminal infrastructure, major air, rail, and trucking terminals, as well as its modal transportation networks of freight corridors and access routes, to ensure transportation efficiency, reliability, cost-effectiveness, and economic competitiveness in the region, in the future.

Portland’s SMD, which has been regularly updated, was originally developed by ICF Kaiser (now Kaiser Engineers), and others from a number of data sources, which include the following:

• The Reebie Associates’ (now Global Insight) TRANSEARCH database;
• For air freight, forecasts by commodity and route, based on FAA air freight traffic data and related freight data provided by the Portland airport;
• For seaborne trade, forecasts by commodity and route, based on international trade data showing shipments by customs district;
• Regional macroeconomic forecasts developed by the WEFA Group (now Global Insight);
• PIERS data from the Journal of Commerce showing sea trade movements; and
• Miscellaneous forecasts prepared by the Port of Portland.

The various dimensional characteristics of the SMD are summarized below:

• Year of Data – Every fifth year ranging from 1995 to 2030;
• Commodity Classification – Seventeen commodity groups, including waste by-products and courier services;
• Origin-Destination – Five origin and destination areas defined relative to the Portland region: within the region, northern United States, southern United States, eastern United States, and non-United States (further divided, in some cases, into five major regions; Canada, Asia, Latin America, Europe, and the rest of the world);
• Modes – Eight modes of travel, including private trucking, less than truckload (LTL), truckload, intermodal (truck/rail), rail, barge, sea, and air; and for all international flows by air and sea, the domestic mode also is provided: truck, intermodal, rail, or barge; and
• Volume of Flow – Each cell defined by the full set of dimensions listed above contains four measures of the estimated annual freight flows, which include containerized tons, noncontainerized tons, 40-foot equivalent units (FEU), and total tons.

The Tactical Model takes inputs from the SMD, and other external sources to predict future truck freight flows on the highway network using a set of freight modeling steps involving commodity flow analyses, regional allocations, conversion of commodity flows to truck trips, and the ultimate assignment of truck trips on the highway network. An important objective during development of the Tactical Model was to replicate heavy duty truck trips on the highway network in a way that would be responsive to dynamic changes in the freight market and industry logistics in the region in the future. Such changes typical to the freight transportation supply and demand environment might include, for example, increases or decreases in the volume of goods moving through the Port of Portland facilities, and shifts in market shares of truck and rail. Although the ability to predict these changes are not incorporated within the framework of the Tactical Model System, these changes are reflected in the data that enter the Model at the top level, as external inputs. The Tactical Model also is consistent with Metro’s passenger travel demand modeling system, as it uses the same geographic structure (TAZs), districts, and model study area, and takes into account Metro’s procedures for steps such as time-of-day modeling.

The Tactical Model, as it currently stands, is largely empirical and less behavioral, implying that it has many fixed percentages for data inputs at various stages of the modeling process. However, with better understanding of the regional freight system dynamics and industry shipper behavior, the model is expected to incorporate more behavioral components in various steps of the modeling process. Particularly notable in this regard is the current work being undertaken by Metro to improve the Tactical Model using the data collected from a recently concluded freight data collection project in the Portland metropolitan region, led by Cambridge Systematics, Inc. Apart from these improvement efforts, the Tactical Model has many notable advantages including the following:

• It provides for the sensitivity of heavy truck flows in the region to the level of economic activity and to the shares of this activity by commodity group.
• It provides a direct and consistent linkage to a TRANSEARCH-type commodity flow database that includes not only local flows, but also external domestic and international flows.
• It explicitly deals with reloading and terminal usage for truck trips.
• It retains information on flows by commodity in the sequential modeling steps.
• It provides a general framework within which improved submodels can be added in the future as more knowledge is gained concerning the behavior of commodity producers, carriers, and consumers in the Portland region, as well as those that use the region as a gateway for domestic and international trade.

12.2.2 Model Study Area

The study area of the Tactical Model is comprised of Columbia, Clackamas, part of Marion, Multnomah, Washington, and Yamhill counties in Oregon, and Clark County in Washington. The model comprises of 2,029 internal TAZs and 17 external TAZs.

12.2.3 Modeling Framework

The following steps are involved in the modeling framework of the Tactical Model for the estimation of forecast heavy duty truck trips on the highway network.

• Regional commodity flow inputs by commodity type, market segment and mode;
• Allocation of commodity flows to origins and destinations;
• Linkage of commodity flows to reload facilities or terminals;
• Conversion of commodity flows to vehicle trips;
• Accounting for additional vehicle trip segments;
• Addition of through truck trips;
• Assignment of vehicle trips to the highway network; and
• Each of these modeling steps are discussed in detail in the following sections.

Regional Commodity Flow Inputs by Commodity Type, Market Segment, and Mode

This step involves the preparation of outputs from the SMD which serve as inputs to the Tactical Model system. The outputs from the SMD include a set of summary commodity flow tables at an aggregate geographic level that are used as control totals for the generation of commodity flows between TAZs in the subsequent modeling steps. The primary data inputs from the SMD can be categorized into commodity flows into, out of, and within the Portland region. The SMD also provides through commodity movements in the region, as long as there is a change of mode involved, for example, goods coming into the Port, and moving inland by truck. Since all goods movements that move through the region by the same mode (primarily truck and some rail) are not captured in the SMD, the Tactical Model needs to account for through truck movements using sources other than the SMD. As discussed earlier, the SMD includes 17 commodity classifications, which also are used in the model.

To provide the input data for internal-internal and internal-external freight modeling, the commodity flows from the SMD are categorized into distinct market segments, in order to provide a framework for the allocation and distribution of commodity flows to appropriate zones. The market segmentation approach developed for the optimal translation of data in the SMD to the data input needs of the Tactical Model is presented below:

• Market Segmentation Based on Mode – In this system, the major classification is the mode or a combination of major modes used for the commodity movement into, out of, or through the region. Some examples of these modal classifications include sea and rail, barge and truck, and private truck only shipments.
• Market Segmentation Based on Terminal Facility Usage and Flow Directionality – The classifications included in this scheme include flows from an origin location (within or outside the Portland metropolitan region) to a terminal facility in the region (Inbound), flows from a terminal facility in the region to a destination location (within or outside the Portland metropolitan region) (Outbound), and flows between origins and destinations without the use of a terminal facility (Direct).

Truck flows with specific origin and destination locations are associated with each market segment defined by the two levels of classification defined above. These origin and destination locations are categorized broadly in the SMD as within the Portland region (p), and external to the Portland region (e). To reflect the origin and destination locations of specific market segments in addition to the modes and terminal usage/directionality, the market segments in the SMD are designated with capitalized labels such as STI (sea/truck inbound) along with the location-related information in lower case extensions of the labels, for example, STIpe (sea/truck inbound that is destined to an external region) and STIpp (sea/truck inbound within the Portland region). The MSD also include broad geographic areas for the identification of external regions used by surface modes, which include North (designated by n, and comprising of all external locations reached with I‑5 north, including Canada), South (designated by s, and comprising of all external locations accessible by I‑5 south, including Mexico), and Other (mainly all external locations to the east of Portland).

Within each market segment and origin/destination pair, there are two additional variables required for data inputs to the tactical model that include the commodity category (cc), and the weight unit of the shipment (w). The commodity categories include the 16 commodity classes in the SMD, while the weight units comprise of noncontainerized tonnage (n), and containerized 40-foot equivalent units (f), which, as described earlier, are separately identified in the SMD. Thus, a complete specification of the origin/destination, commodity classification, and weight units in lower case characters that would accompany the upper case notations for the mode and directionality information, will be odccw.

Based on the comprehensive set of market segmentation considerations discussed above, the following sets of market segments form the commodity flow inputs for the Tactical Model:

• Truck Flows – This segment includes all external domestic, intraregional, and Canada and Mexico surface flows, with associated trucking submodes of truckload, LTL, and private trucking. This segment does not designate Inbound, Outbound, and Direct classifications to the trucking shipments since the usage of trucking terminals by trucking shipments are handled separately by the model in a later step. The notations for these truck flow market segments in the SMD include TP (private), TT (truckload), and TL (LTL).
• Rail Flows – This segment includes all external domestic, intraregional, and international rail flows to/from Canada and Mexico, with rail carload or intermodal as the major modes. All these rail flows involve an associated truck move to/from the rail terminal facility in the Portland region, which is indicated by labeling these flows Rail/Truck. However, the trucking submode information for these rail flows is not available from the SMD. The notations for rail flow market segments in the SMD include RSI and RSO, where S denotes the surface trucking mode, and I and O stand for inbound and outbound respectively, with respect to the truck movement to or from the rail terminal facility.
• Sea Flows – This segment includes all international nonair flows, excepting a limited number of surface flows to and from Canada and Mexico. For inputs to the Tactical Model, sea flows are further disaggregated into the following categories, based on the surface mode used to/from the port facility:
  • Sea/Truck – Truck moves associated with oceangoing shipments are an important component of total truck trips in the Portland region. These shipments are designated in the SMD based on the direction and trucking submode into SPI, SPO, STI, STO, SLI, and SLO.
  • Sea/Rail and Sea/Intermodal (SR) – These involve shipments that move by rail carload or intermodal to/from the port facility. Intermodal flows involve containers and trailers on flat cars, while rail flows involve all other rail movements. Some of the SR flows may involve associated truck moves related to drayage activity to/from rail terminals, which are accounted for by applying factors specific to the designated port facility. SR flows also are classified as SRI and SRO, depending on the directionality of truck drayage moves into or out of port facilities.
  • Sea/Barge (SB) – These involve oceangoing shipments that move by barge to or from the port facility. These shipments typically are not expected to generate significant trucking activity in the region. However, for consistency, they are designated by SBI and SBO, to identify directionality.
• Air Flows – This segment involves all air cargo flows in the SMD, both international and domestic that are delivered to or picked up from the Portland International Airport (PDX) by truck. Some portions of the air cargo flows in the SMD have trucking submode information for the surface portion of the flow, which are designated by AP, AT, and AL, to denote associated private, truckload, and LTL truck shipments. As with other market segments, these shipments are disaggregated further based on directionality into Inbound (I) and Outbound (O), depending on the direction of the truck move to or from the air cargo facility. For other portions of the SMD with no information on the trucking submode, the flows are designated by ASI and ASO.
• Barge Flows – This segment includes all intraregional and external domestic flows with barge as the major mode of transport (this does not include flows to and from Canada or Mexico, as no international barge flows to/from these regions are expected to be prevalent or significant). Each of the barge flows is associated with a truck movement to or from the barge terminal facility in the Portland region, which is indicated by denoting each barge flow in the database as Barge/Truck. The SMD does not contain the trucking submode information for barge flows, and consequently, shipments in this market segment are denoted by BSI and BSO.

The commodity flows provided by each of the market segment defined above serve as input to estimate annual flows in the SMD associated with each trucking submode (truckload, LTL, and private trucking). The SMD provides annual commodity flow tonnages while the goal of the Tactical Model is to generate average weekday trucking activity. Consequently, an average conversion factor from annual to average weekday of 1/264 is used for the estimation of weekday truck trips, based on information published in the report Vehicle Volume Distributions by Classification [Hallenbeck et al., Washington State Transportation Center, draft, June 1997].

12.2.4 Allocation of Commodity Flows to Origins and Destinations

The next step in the Tactical Model is the allocation of commodity flows derived from the first step described above to origins and destinations in the Portland metropolitan region. Origin locations in the model to which commodity flows need to be allocated include internal zones (with origins of internal-internal and internal-external flows), highway gateway locations or external zones (with origins of external-internal and through commodity movements), and terminal locations like port facilities, air cargo, or rail terminals, where international or external domestic shipments are offloaded and generate associated truck trip origins). Similarly, destination locations in the model include internal zones (for destinations of external-internal and internal-internal flows), highway gateway or external zones (for destinations of internal-external and through movements), and terminal locations where international exports or external domestic shipments leaving the region are loaded, which generated associated truck trip destinations.

The purpose of this step in the model is to allocate the commodity flows to the origins and destinations described above. At the conclusions of the allocation process, the weekday truck flows defined in Step 1 are converted from a set of variables specific to a few general origins and destination locations to more disaggregate locations comprising zones, highways, and terminal locations. The following sections describe the allocation process for zones, highways, and terminals, respectively:

• Internal Zones – As described earlier, origins or internal-internal and internal-external flows, and destinations of internal-internal and external-internal flows, are allocated to TAZs in the Tactical Model system. This is accomplished by disaggregating flows to zones based on zonal employment shares for specific industry groups associated with each commodity category. For this purpose, the Tactical Model uses base year employment at the two-digit SIC industry level provided by Metro. For future employment forecasts, however, employment data inputs to the model are available only for two employment categories, retail and nonretail. Base year distribution of industry employment across detailed industry groups are applied to the future year total employment by zone to arrive at employment forecasts by zone for detailed industry groups.
• Highway Gateways/External Zones – The Tactical Model allocates commodity flow origins entering the region and destinations leaving the region by truck to highway gateways/external zones. The internal-external and external-internal commodity flows from the SMD by external region (North, South, and Other) are allocated to highway gateways based on a fixed allocation to major roadway facilities as external stations in the Metro travel forecasting network, based on the distribution of current (from observed classification counts) or forecast (based on truck count trends or statewide model results) truck trips on each of the facilities. The same distributions occur for the allocation of all commodity groups, unless there are certain specific restrictions for the use of a gateway by a particular commodity group or if there is specific commodity flow information available at each highway gateway location (for example, from surveys).
• Terminals – Allocation to terminal locations is performed by the Tactical Model for all market segments having their primary mode other than trucking. However, the procedure for the allocation of truck shipments associated with these market segments to specific terminals will depend on the primary mode. For example, all truck shipments associated with the air cargo market segment are allocated to only one terminal location, which is the Portland International Airport (PDX). Where more than one point of entry or exit may exist, the model uses inputs from the Port of Portland or other sources to identify shipment patterns and the use of each terminal location by individual commodity types. This step also allocates drayage truck trips to terminals for sea and rail, and sea and barge market segments, where the associated terminal facilities are at separate locations (thereby leading to a truck drayage move).

12.2.5 Linkage of Commodity Flows to Reload Facilities and Terminals

This step of the Tactical Model links applicable commodity flows to reload and terminal facilities. The model’s current procedure for the linkage of commodity flows to reload/‌terminal facilities is based on an initial assumption on the trucking activity characteristics of each of the trucking submodes – truckload, LTL, and private. The model assumes that commodity flows moving by truckload carriers are not typically associated with a reload or terminal facility (in other words, these flows occur directly from the origin (pickup) to a destination (drop-off) location without having any intermediate reload activity). On the other hand, all LTL shipments are assumed by the model to be associated with reload activity (the reload activity for LTL trips being defined as the activity at LTL terminals, which involves the transfer of cargo between line-haul and pick-and-delivery trucks). For the purpose of determining reload activity associated with private trucks, the model assumes part of the private trucking shipments to behave like truckload shipments, while the rest behave as LTL shipments. Consequently, the fractions behaving like LTL shipments are used in this step to associate the flows to reload/trucking terminals.

The allocation of flows into and out of zones with reload facilities is accomplished by the model by developing a trip-rate factor for reload sites based on employment. The factor is determined using actual counts for some small number of reload sites, and a standard factor of 1.75 trips per employee was used for all other sites. By applying the trip rate to total reload employment in each zone (obtained from the freight facility database), the model estimates the reload truck trips into and out of each zone. Reload flows were allocated to zones in proportion to the amount of reload trips into/out of the zones, calculated as previously described. The model assumes that all flows that use reload facilities or terminals do so only once within the Portland region – that is, there are no reload-to-reload flows. Further, pickup and delivery tours are not represented as tours in the model.

12.2.6 Conversion of Commodity Flows to Vehicle Trips

This step of the Tactical Model converts the commodity flows derived from the previous step to equivalent heavy-duty truck trips. The model defines any truck with three or more axles as a heavy-duty truck (while two-axle six-tire trucks are treated as nonheavy trucks). The conversion factors required to translate commodity flows to equivalent truck trips need to be sensitive to the weight and volume characteristics of the commodity being carried, the type of truck, as well as the need for any specialized transportation equipment. The conversion process from commodity flows to vehicle trips essentially involves two steps, which include the following:

• Heavy-Duty Truck Fractions – This step involves the estimation of the fraction of total commodity flows moving by heavy versus nonheavy trucks. These fractions are derived using vehicle classification counts collected at a number of sites around the Portland region, getting the distribution of heavy versus nonheavy truck trips based on the classification counts, and using these distributions (after converting them to equivalent tonnages) to estimate the tonnage distributions to be allocated to heavy versus nonheavy trucks.
• Flow-to-Truck Factors – This step involves the application of flow-to-truck trip conversion factors by each truck type to estimate heavy duty (and nonheavy duty) truck trips for trip tables. ODOT roadside surveys provide the data for the development of conversion factors for use in this step of the Tactical Model. These factors currently are in the process of being revised using the data collected from the recently conducted roadside intercept surveys in the Portland region as part of the Portland Freight Data Collection project, led by Cambridge Systematics, Inc.

For containerized cargo, the commodity flow data from the SMD includes both TEUs and tonnage by commodity so that an average tonnage per TEU could be estimated for each commodity and multiplied by 2 (2 TEUs per truck) to get an average weight per truck. All containerized cargo are assumed to move on heavy trucks.

12.2.7 Accounting for Additional Vehicle Trip Segments

This step in the model accounts for additional vehicle trips related to empty returns associated with repositioning of tractor-trailers, as well as bob-tail trips associated with tractor-only repositioning. This step is required in the model since the previous modeling steps estimated loaded truck trips based on commodity flows, and did not account for empty truck trips, which are expected to be fairly significant in the region. In the current model, the only adjustments made for empty returns and repositioning are those made for LTL flows through terminal and reload facilities so that the model calculated trips match those from the truck counts around these facilities. In addition, the model also accounts for additional vehicle trips related to imbalanced origin and destination loaded truck trip totals. The predicted loaded flows from the model will be unbalanced in most cases, by commodity, market segment, weight type, as well as trucking submode. This step accounts for the net imbalance in the origins and destinations of these trips, and the additional trip segments associated with this imbalance.

12.2.8 Addition of Through Truck Trips

As discussed earlier, the SMD does not include commodity flows transported entirely by truck that move through the Portland metropolitan region. Originally, during the development of the Tactical Model, it was anticipated that the Statewide Model could be used to estimate through trips in the Portland region. However, in the absence of this, Metro currently accounts for through trips in the model based on a comparison of the assigned trips on the network (excluding through trips) with the truck counts at the external stations. The differences between counts and model assignment volumes are used as targets for an external-external trip table, which is estimated using a function in the truck modeling software. Since these through trip adjustments are made after the conversion of commodity flows to equivalent truck trips, no commodity distinction is available from the model specifically for through truck trips. Also, since the adjustment to account for additional vehicle trips (including empty trips) was done in a previous step, this step inherently also accounts for some empty through trips that might be present in the region although this effect is not expected to be significant.

12.2.9 Assignment of Truck Trips to the Highway Network

This step assigns the truck trip tables derived from previous steps of the modeling process to the highway network, to estimate the average weekday truck trips on each link of the network. Procedures for assigning these truck trip tables are integrated with the Metro passenger trip assignment modeling process. In order to achieve this integration, several issues were first addressed related to consistency with the Metro passenger modeling procedures, development of multi-class or multi-trip table procedures, as well as the development of freight transportation networks. The following trip tables feed into the truck trip assignment process:

• Loaded heavy truck trips developed at the conclusion of the modeling step Conversion of Commodity Flows to Vehicle Trips;
• Additional vehicle trips, obtained at the conclusion of the modeling step Accounting for Additional Vehicle Trip Segments;
• Through trips, developed at the end of the modeling step Addition of Through Truck Trips; and
• All nonheavy truck trips, developed at the conclusion of the modeling step Conversion of Commodity Flows to Vehicle Trips.

Truck trip tables are assigned to the network using a standard multi-class assignment. Truck trip tables are combined into two vehicle classes (heavy and light trucks) and are not assigned by commodity. Trip tables are estimated for average weekday conditions, and no time-of-day or peaking information is provided for the assignment.

12.2.10 Model Calibration and Validation

There were limited data for calibration and validation of the Tactical Model, until the recently concluded Portland Freight Data Collection project, as part of which, vehicle classification counts were specifically collected at 10 primary model screenline locations (screenline locations that account for majority of the truck flows on the regional highway network), to be used for the purpose of model validation. Metro is currently in the process of using the classification count data collected from the study to perform model validation. Owing to the geographic comprehensiveness of the locations included in the vehicle classification count program of the study, Metro also is considering using the data to perform an Origin-Destination Matrix Estimation (ODME) for model calibration. The ODME approach involves the use of the trip tables estimated from the initial model as a seed matrix and to adjust the input trip tables in order to minimize the differences between the model outputs (after assignment) and the vehicle classification counts on the highway network. A sum of square-differences (SSD) minimization approach could be used, or the minimization could be based on a linear-programming approach.

12.3 Florida State Freight Model

12.3.1 Objective and Purpose of the Model

The Florida Intermodal Statewide Highway Freight Model (FISHFM) was designed to support the project-related work of FDOT and Florida’s metropolitan planning organizations. The purpose of the model was to identify deficiencies and needs and to test solutions on major freight corridors throughout the State. These freight corridors suffer from considerable congestion as they pass through metropolitan areas. For example, I‑95 in South Florida is not only a major international freight corridor, it also is the main thoroughfare for local travel in major metropolitan areas, including Miami, Daytona, and Jacksonville. I‑4 in Central Florida is heavily used by both truckers and tourists and is the site of a growing high-technology industry. In addition, the local highway connections between major freight corridors and intermodal terminals – warehouses, seaports, and airports – are often the weakest link in the intermodal highway chain. The truck freight model should be integrated with MPO transportation models to ensure that needs and deficiencies at the local level that impact efficient freight transportation can easily be identified.

Many truck trips in Florida begin or end at intermodal terminals, either as long-distance movements or as short-haul connections between intermodal terminals. Because rail, air, and water serve as important components of the freight system, the model determines how freight traffic is allocated and routed among all freight modes in order to produce truck forecasts. While a primary purpose of the model is to forecast truck volumes on highways, the data and forecasts of other freight modes are important as well.

12.3.2 Model Class

The FISHFM is a four-step commodity forecasting model. Florida has a statewide highway model in which total truck trips are forecasted based on total employment and are assigned together with auto trips. An existing four-step model for passenger auto and total truck traffic provided the state zone structure, highway network, and employment data that served as the structure for developing the commodity model.

12.3.3 Modes

Although the primary purpose of the FISHFM was to analyze freight truck traffic, the model development recognized that over 80 percent of the freight by tonnage serving Florida’s major commercial airports, deepwater ports, and rail container terminals is transported by truck. These intermodal facilities generate significant truck volumes at concentrated locations. The model development further recognized that the rail, water, and air freight systems are important competitors to truck freight. Understanding the demands of other modes was deemed a critical component of the model development.

A primary purpose of FISHFM was to forecast truck volumes on highways. However, the data and forecasts of other freight modes also were determined to be valuable as FDOT prepared to implement the Statewide Intermodal Systems Plan and respond to its Transportation Land Use Study Committee’s recommendation that the Florida Intermodal Highway System (FIHS) be expanded to a Florida Intermodal Transportation System (FITS) covering all modes.

12.3.4 Markets

Trucking in Florida consists of very different markets: long-haul interstate/international, intrastate, private/for-hire, truckload/less-than-truckload, local/metropolitan delivery, and drayage (truck shipment between ports, airports, and rail terminals). These markets have different needs, use different vehicles (combination vehicles versus panel trucks) and are sensitive to different variables. Based on the data available to support the development of the model and the role of MPOs in planning for local/metropolitan delivery, the markets selected for inclusion in FISHFM were interregional freight shipments within Florida, drayage movement to and from intermodal terminals, and interstate freight shipments of all kinds. In order to properly account for the various characteristics influencing the interstate shipment of freight, the model had to cover all of North America, although at a level of zone and network detail that was more geographically aggregated than that for Florida alone, as can be seen in Figure 12.2.

Figure 12.2 Highway Network for Florida Intermodal Statewide Highway Freight Model

12.3.5 Framework

Florida’s Model Task Force decided that the structure of the FISHFM should follow the basic framework of the four-step Florida Statewide Urban Transportation Model Structure (FSUTMS) passenger process. This requires that tons of commodities be generated and distributed and that a mode split component be used to determine which tons are shipped by truck and other modes. Truck trips identified in the mode split process are then assigned to the statewide highway network. All model components operate as part of the FSUTMS software. Following the FSTUMS approach results in a model that is easily understood by users and ensures compatibility with FSUTMS and the statewide passenger model.

12.3.6 Truck Types

The FISHFM focuses primarily on long-distance commodity freight movements. It captures large trucks moving on the FIHS, the shipment of commodities between regions in Florida, and the shipment of freight between Florida and the rest of North America. These truck trips currently represent about 25 percent of the total truck trips in Florida, but 45 percent of the total truck vehicle miles traveled within the State. These freight movements are surveyed as part of Global Insight’s TRANSEARCH database. The FISHFM does not address local delivery or service trucks, which primarily serve regional markets and are best modeled at the regional or urban area level as part of the MPO planning process. As such, FISHFM does not attempt to model the two-axle trucks not commonly used in commodity freight shipments.

12.3.7 Base and Forecast Data

Florida Data – The forecasting data include population and employment that are used as input to the trip generation step of a freight demand estimation model. Base year values for these data are used to calibrate the trip generation (production and attraction) equations. Forecast values for these data are then used in the generation (production and attraction) equations to predict the number of freight trips that will be generated in future years.

Population serves as an input variable in the trip generation (attraction) equations. Population is one of the key variables that determine regionwide consumption of goods originating from other areas of Florida and nationwide. Base year data were collected from the U.S. Census Bureau’s 1998 U.S. Census of population, Florida MPOs, local planning departments, and FSUTMS data (ZDATA1) sets. Future year data were forecast from Florida’s Long-Term Economic Forecast, Florida Population Studies-population projections for Florida counties, MPO forecasts, and FSUTMS data (ZDATA1) forecasts.

Employment by commodity sector serves as an independent variable in trip generation (production and attraction) equations for freight tonnage produced and attracted by commodity group. Employment data by industry code are the principal explanatory variables in the trip generation equations. Base year data were collected from the Regional Economic Information System (employment by standard industrial classification, or SIC), County Business Patterns (SIC employment by county), SIC employees by TAZ, Florida MPOs, local planning departments, FSUTMS data (ZDATA2) sets, and the Florida Department of Labor. Future year data were estimated using the Florida Long-Term Economic Forecast.

Forecast Growth of External Markets – While population and employment were chosen to be the forecasting data for freight shipments to and from Florida TAZs, the data were not available or suitable to forecast freight shipments for the zones located outside Florida. For these zones, freight forecasts were developed by factoring existing flows using the growth rates by industry and state provided by the Bureau of Economic Analysis’s BEA Projections to 2045.

12.3.8 Modal Networks

Freight Modal Networks – Although the FISHFM is a multimodal commodity model, the assignments were only to be made to a highway network. Information from the other modal networks, such as distances, travel times, or costs, were inferred from the highway network. The highway network for Florida was the existing Statewide Model highway network to ensure compatibility with that model. The highway network outside Florida was drawn from the National Highway Planning Network, as shown in Figure 12.2.

Intermodal Terminal Data (seaports, rail yards, airports) – The location of the intermodal terminals (x and y coordinate or ZIP code) and the activity (ton shipments from/to for both base year and forecast year) at the major ports and intermodal terminals by commodity were obtained to locate these facilities in FISHFM as special generators. The locations were obtained from the 1999 National Transportation Atlas Databases for the United States and Florida, the Strategic Investment Plan to Implement the Intermodal Access Needs of Florida’s Seaports (Part II, United States and Florida seaports), FAA Aviation Forecasts for the fiscal years 2000-2011, the North America Airport Traffic Report, the Port Facilities Inventory (United States and Florida water ports), the U.S. Maritime Administration’s Office of Intermodal Development, and published reports from port operators.

12.3.9 Model Development Data

The TRANSEARCH commodity flow database as purchased for Florida was chosen to represent the survey of existing freight flows. The STCC in that database were used to develop commodity groups for the model, the existing mode shares were chosen, flows were treated as revealed-preference surveys, the total tonnage originating in a zone was chosen to be the production of freight, and the total of tonnage destined for a zone was chosen to represent the attraction of freight to that zone. The average trip length between zones was used for the pattern of trip distribution.

12.3.10 Conversion Data

Values per Ton – The TRANSEARCH data used for the model is in the STCC commodity classification code. The dollar value per ton by commodity can be obtained from the CFS records for Florida. However, the 1997 CFS uses the SCTG commodity classification system. To allow the direct use of the value information by STCC commodity, the 1993 CFS, which also used the STCC system, was used to develop values per ton which were adjusted to 1998 dollars using the Consumer Price Index for those years.

Daily Vehicles from Load Weights and Days of Operation – Commodity flow data are given in terms of tons per year. Because transportation planning functions require model output in the form of vehicles (trucks) per day, it is necessary to determine the amount of goods carried in a vehicle and the number of vehicle operation days in a year. Payloads in tons per day were obtained from the U.S. Census Bureau’s VIUS.

12.3.11 Validation Data

Validation data consisted of the truck counts by vehicle class. Classification truck counts on highways are needed to separate truck traffic from passenger car traffic. Truck counts by vehicle class were used for the validation of the model-estimated truck volume. These data are available from the 1999 Annual Average Daily Traffic Report for Florida and Truck Weight Study Data for the United States. These truck counts include all trucks, not just freight trucks. FAF ’s loaded highway network was used to estimate the percentage of freight trucks observed in truck counts.

12.3.12 Software

The Florida Intermodal Statewide Highway Freight Model was designed to run using TRANPLAN software and Florida Standard Urban Transportation Model Structure (FSUTMS) scripts.

Two FORTRAN programs were written specifically to run FISHFM components. The commodity generation program, FGEN, generates production and attraction files representing the number of tons of goods generated in each zone by commodity group. The mode split program, FMODESP, allocates commodities to modes, and converts annual tons of truck commodities to daily truck trips. All other components of the FISHFM run using the TRANPLAN program within the FSUTMS structure.

12.3.13 Model Application

The FISHFM is being considered for use in a variety of applications, including:

• Existing and forecast productions and attractions of annual freight tonnage for each TAZ in Florida for 14 specific commodities;
• The existing and forecast origin-destination table of annual freight tonnage moving between TAZs and the external zones covering North America for 14 specific commodities;
• The existing and forecast table of annual freight tonnage by mode and by commodity derived from the total origin-destination table;
• The existing and forecast table of daily truck trips derived from the origin-destination table of annual tonnage by truck for 14 specific commodities; and
• The existing and forecast daily volumes of trucks moving on the Florida highway system through assignment of the truck table to the highway network.

12.4 Texas State Analysis Model (SAM)

12.4.1 Introduction

The Texas Statewide Analysis Model (SAM) is a traditional four-step model covering passenger and freight flows in Texas. The program is TransCAD-based and mainly utilizes geographic files as its input data. It was developed for TxDOT by Alliance Texas and Wilbur Smith Associates. SAM is a multimodal and intermodal travel demand modeling system with two major components, passenger and freight. The passenger component models highway and rail systems while the freight component models highway, rail, air, and water systems. Passenger and freight models by mode are integrated through common demographic and transportation systems databases within the TransCAD software environment.

The various sources of data provide information for analysis of highway networks, intermodal facilities and traffic, intercity traffic, demographics, and for the purposes of this study, freight. The SAM model is comprised of 4,600 zones within Texas, as well as another 142 external zones, thereby creating a one-county buffer around the State.

TxDOT primarily developed SAM to expand and enhance its travel demand modeling capabilities and process to be state of the practice, to analyze the increase in commercial traffic on Texas highways due to the North American Free Trade Agreement (NAFTA); and to consider passenger and freight modes and quantify the interaction between modes as part of long-distance passenger and freight improvement projects. The model was not designed to replace urban models, but to assist in the development of forecasting activities in nonurban areas. Therefore, the zone structure in urban areas is not as extensive as those of urban models. SAM Traffic Analysis Zones (TAZ) follow census block geography in rural areas, while in urban areas, MPO TAZs are aggregated.

12.4.2 Data

The freight component of the SAM utilizes Global Insight’s TRANSEARCH commodity flow data for the primary commodity movements in the State. Global Insight is a private vendor of commodity flow data. The commodity flow data purchased by TxDOT for the model include origin-destination (O-D) data by two-digit STCC for each county in the State. The commodity flow data were acquired for truck, rail, air, and water. The SAM model supplements the TRANSEARCH data with additional data to cover Mexican freight flows, including data from Wharton Economic Forecasting Associates (WEFA) and the Latin America Trade Transportation Study (LATTS).

TRANSEARCH data are best suited for long-distance commodity movements. It is common for freight models that utilize Global Insight data to supplement the freight flow data with short-distance truck trips based on local socioeconomic data. In the SAM, a second set of commercial vehicle trips are generated as part of the passenger model component utilizing a category called “OTH” with production and attraction rates based on the 1996 Quick Response Freight Manual. The commercial vehicles accounted for in the passenger model include both four-tire commercial vehicles that are purely passenger cars and single-unit trucks with six or more tires. The trucks generated in this component of the model are used to account for the short-distance truck trips not captured in the Global Insight’s data.

While 30 or so commodity types are represented in the base year flow data, the large majority of all tonnage flows is made up of a smaller number (perhaps 10 or so) of commodity types. The commodity types (and tonnages) present within each major flow generation category were reviewed to identify the most significant commodities within each movement group (intrastate, internal-external, and external). The procedure used was to identify the commodities (starting with the largest) making up about 90 percent of the total tonnage in the movement group. Using this definition, 11 commodity groups were established, these are shown in Table 12.1.

Table 12.1 SAM Commodity Groups

1 – Agriculture

2 – Raw Material

3 – Food

4 – Textiles

5 – Wood

6 – Chemicals/Petroleum

7 – Building Materials

8 – Machinery

9 – Miscellaneous Mixed

10 – Secondary

11 – Hazardous

12.4.3 Network

The SAM network covers roadways in Texas only and relies on Global Insight’s data for out-of-state trips. The SAM 1998 roadway network currently contains approximately 54,800 links that have a total length of 87,200 miles. Three hundred sixty-seven links representing future roadway projects were added to the master network for 2025. This adds 730 miles of network and 3,100 lanes miles. In addition, 2,846 links either added lanes or moved up in SAM road class. These links total 4,374 miles in length and equal 9,587 lane miles. The full SAM multimodal network contains approximately 60,500 links.

Texas Network

The biggest piece of the SAM Texas network, state system roadways, was provided by the TxDOT in several ArcView geographic files. These files were converted to TransCAD format and all of the relevant attributes were compiled into one geographic file. The attributes compiled from the TxDOT files include: number of lanes, road name, functional class, and posted speed. The road name and functional class attributes were generally acceptable. However, the number of lanes and speed attributes were inconsistent and required considerable editing before use.

County roads and urban arterials were added to the state system network for connectivity and completeness. County roads (6,604 miles) were added. An additional comprising 1,554 links were used to represent the county roads. Urban arterials (1,555 miles) comprising of 1,330 links also were added.

The 1999 Unified Transportation Program (UTP) and the Metropolitan Transportation Plans (MTP) for each of the 25 MPOs in the State of Texas served as the sources for future roadway projects. The UTP is an annual publication of TxDOT that serves as TxDOT’s 10-year plan for transportation project development. It is the only plan for future roadway improvements that includes the entire State of Texas. The 1999 UTP was chosen over the 2000 or 2001 UTPs so there would be no gap between the SAM 1998 base year and scheduled projects. MTPs are planning documents designed to identify existing and future transportation deficiencies and guide transportation improvements in MPO areas. Typically, MTPs cover a period of 25 years and are updated every 3 to 5 years. Rural projects were identified for a 10-year period form the UTP.

The network contains projects to be built in the next 25 years in urban areas, whereas the rural areas contain projects to be built in the next 10. There is no source for projects past that timeframe in rural areas. However, projects defined by an urban TIP that ended at an MPO’s boundary were extended into the “rural” area to a logical point of termination. These projects were typically extended to the next town or major intersection.

Not all projects in the UTP or MTPs are relevant or represented in the SAM network. Only projects that change the capacity of a roadway are included in the SAM forecast network. A project must add or remove lanes from a roadway already in the SAM network, or add or remove a roadway of statewide significance entirely to be included in the SAM network. Intersection improvements, the addition of turn lanes, and bridge replacements are examples of projects that are not included in the SAM forecast network.

12.4.4 Trip Generation

The freight trip generation models in the SAM are developed at the county level as TRANSEARCH data are organized with origins and destinations by county. As described previously, only primary commodity movements are accounted for in the freight model. The model, therefore, assumes that all of these trucks are combination vehicles.

The model structure used for trip generation was regression equations relating independent variables (employment types and dummy variables representing special freight handling facilities) to the tonnages produced or attracted to individual counties. All trip generation models (and other freight model components) were developed at the county level of geography. Global Insight data flows defined freight origins and destinations as counties. Therefore, no finer level of disaggregation was possible for model development.

Equations were developed for the following freight movement types:

• Internal-Internal Productions;
• Internal-Internal Attractions;
• Internal-External Productions; and
• External-Internal Attractions.

12.4.5 Trip Distribution

Trip distribution is the process of matching trip productions with trip attractions. Trip distribution at the zone level in the SAM is performed by a gravity model that assumes the probability of trips between two locations is inversely related to the trip distance and directly related to the magnitude of activity at the destination. The SAM uses mathematical equations to replicate observed trip length distributions. Adjustment factors also are applied on a district-to-district basis to acceptably reproduce movements between subareas of the State. The SAM uses a standard software plug-in for trip distribution calculations. Friction factors were developed using standard factors extracted from TransCAD, a commonly used transportation software. Additional information on trip distribution can be found in the SAM Theory Report [TxDOT, Statewide Analysis Model Theory Documentation, 1999] and the FHWA primer on trip distribution [Federal Highway Administration, Trip Distribution Modeling, 2002].

12.4.6 Mode Choice

The trip table generated from the TRANSEARCH data in the SAM can be assigned either to road or rail (and in some instances water). A logit model formulation is used to estimate the share of freight that would be assigned for each mode. Coefficients are developed for each commodity group. Those movements that had a rail access distance greater than 25 miles are assigned to the truck mode.

For the truck portion of the SAM model that was developed from TRANSEARCH data, the conversion from tons to number of vehicles was performed using vehicle load factors. These vehicle load factors are adjusted from a single value per commodity group to a set of values related to trip length. Load factors also were increased 15 to 20 percent to provide a better match with traffic count data. Empty truck movements also were reduced to 40 percent of their original values to calibrate the truck volumes with observed truck counts.

12.4.7 Assignment

The truck trips developed by the TRANSEARCH data in the SAM are preloaded as the initial step in traffic assignment. Freight truck assignment volumes are converted to Passenger Car Equivalents (PCE) using a conversion factor of 2.5 minimum travel-time paths are used when assigning truck movements to the road network. An all or nothing procedure is used as there are no other vehicles being assigned at this point. Roadway capacities are then adjusted to account for trucks, and these adjusted values are used to calculate congestion impacts on road speeds and route selection.

The commercial vehicle trips were added to the passenger model. Passenger trips were disaggregated into time periods and assigned to the network using a shortest time algorithm that accounted for potential congestion in the assignment.

Rail Assignment (Passenger and Freight)

Rail trips (passengers and freight tons) produced by the passenger and freight mode choice models are assigned using networks and paths produced specifically for this purpose (they include the appropriate rail and access links of the passenger and freight rail system). Impedance factors are used to increase road travel times so that a rail path is produced (and assigned traffic) if it is a reasonable travel alternative.

Rail passenger trips are assigned at the zone level. Rail freight tonnages are assigned at the county level. Both assignments are made using the all or nothing assignment procedure (no basis exists for defining alternate paths).

Validation

Model validation work consisted of applying the trip generation, distribution, and mode choice models to obtain estimates of freight tonnage and vehicle flows. These were then compared to two sources of actual freight flows; the TRANSEARCH data bases from Global Insight and traffic counts of "heavy" commercial vehicles. As the process proceeded, and comparisons between the estimated and observed data found significant differences, revisions were made to the original forecasting models to improve their performance.

Many of the comparisons made were organized on a geographic basis. The Figure 12.3 shows the areas employed (six internal districts and four external districts). The internal districts were structured to include contiguous areas (so that major origin-destination trip patterns could be examined). The three largest urban areas, Houston, Dallas/Fort Worth, and the San Antonio-Austin Corridor, were defined as separate districts. The eastern, northern (panhandle), and southwestern parts of the State comprised the three remaining internal districts. The four major directions of approach to Texas, east from Louisiana/Arkansas, north from Oklahoma, west from New Mexico, and south from Mexico, were used to define the external districts.

Figure 12.3 Texas SAM Network