Below is the uncorrected machine-read text of this chapter, intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text of each book. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.
91 This chapter summarizes the current literature on FG and FTG modeling. It provides a comprehensive review of the state- of-the-art research and practice in FTG, with critical examina- tion of the technical merits, advantages, and disadvantages of different FTG methods and models. Freight Trip Generation (FTG) Modeling This section reviews the state-of-the-art research and prac- tice in FTG analysis and modeling. It includes both domestic and international FTG modeling efforts. The FTG models are reviewed in terms of factors that serve to classify a specific FTG model. The various factors that should be considered in developing and analyzing freight modeling techniques are given in Table 66. They include: dependent and independent variables, levels of aggregation and geography, estimation techniques, and model structure. Table 67 provides a sum- mary of various FTG models that are reviewed in this docu- ment. To facilitate interpretation and analysis, Table 68 shows the breakdown of the features of the models by level of geo- graphy. As shown, the majority of the models are for states (35%), and metropolitan areas (39%). Corridor and facility specific applications represent the remaining 26%. A p p e n d i x d Review of the Literature on Freight Trip Generation Modeling Vehicle-trip Cargo weight Value Employment Building area (square footage) Land Use Other economic data (sales, establishments, industry segments, type of business) Vehicle type Commodity type Other Disaggregate Aggregate Metropolitan Statewide National Corridor Special Facility (e.g. Ports) Trip - Growth rates Ordinary Least Squares (regression) Spatial regression Multiple Classification Analysis Time series Input-Output Neural Networks Others Linear Non-Linear Dependent variable Independent variable Level of aggregation Level of geography Estimation technique Model structure Table 66. Review factors.
pi r t - e l c i h e V t h g i e w o g r a C e u l a V t n e m y o l p m E e g a t o o f e r a u q S e s U d n a L a t a d p i r T a t a d c i m o n o c E e p y t e l c i h e V s w o l f y t i d o m m o C r e h t O e t a g e r g g a s i D e t a g e r g g A n a t i l o p o r t e M l a n o i g e R l a n o i t a N r o d i r r o C ) s t r o P . e . i ( . l i c a F l a i c e p S s e t a r h t w o r G - p i r T ) n o i s s e r g e r ( s e r a u q S t s a e L y r a n i d r O n o i s s e r g e r l a i t a p S s i s y l a n A . s s a l C . t l u M s e i r e s e m i t d n a d n e r T t u p t u O t u p n I s k r o w t e N l a r u e N s e t a r h t w o r G s r e h t O r a e n i L r a e n i L - n o N NCHRP Synthesis 298 Lower Mainland Truck Freight Study x x x x x x NCHRP Synthesis 298 Denver Regional Council of Governments (DRCOG) x x x x NCHRP Synthesis 298 Development of a Statewide Truck Trip Forecasting Model Based on Commodity Flows and Input-Output Coefficients x x x x x NCHRP Synthesis 298 Assessment of Market Demand for Cross-Harbor Rail Freight Service in the New York Metropolitan Region x x x NCHRP Synthesis 298 Transport Flows in the State of Indiana: Commodity Database Development and Traffic Assignment: Phase 2 x x x x x x NCHRP Synthesis 298 Multimodal Freight Forecasts for Wisconsin x x x x x x x NCHRP Synthesis 298 Analysis of Freight Movements in the Puget Sound Region x x x x x x NCHRP Synthesis 298 Portland Commodity Flow Tactical Model System: Functional Specifications x x x x x NCHRP Synthesis 298 The Second Generation Michigan Statewide Truck Travel Demand Forecasting Model x x x x x x x x NCHRP Synthesis 358 Virginia Freight Component x x x x NCHRP Synthesis 358 Wisconsin Freight Component x x x x x x NCHRP Synthesis 384 Atlanta (ARC) Commercial Vehicle and Truck Models x x x x x x x NCHRP Synthesis 384 Baltimore Metropolitan Council (BMC) Models x x x x x x NCHRP Synthesis 384 Southeast Michiganâs regional freight system x x x x x x x NCHRP Synthesis 384 Los Angeles regional Cube Cargo Model x x x x x x x NCHRP Synthesis 384 New York Best Practice Model x x x x NCHRP Synthesis 384 Delaware Valley Regional Planning Commission Model x x x x x x NCHRP Synthesis 384 Maricopa Association of Governments MAG x x x x x x x Dependent variable Independent variable Level of aggre- gation Level of geography Estimation technique Model structure Table 67. FTG models classified by review factors.
pi r t - e l c i h e V t h g i e w o g r a C e u l a V t n e m y o l p m E ( S q . f t . ) a e r a . d l i u B e s U d n a L a t a d c i m o n o c e r e h t O e p y t e l c i h e V e p y t y t i d o m m o C s r e h t O e t a g e r g g a s i D e t a g e r g g A n a t i l o p o r t e M S t a t e w i d e l a n o i t a N r o d i r r o C ) s t r o P . e . i ( . l i c a F l a i c e p S s e t a R p i r T s e r a u q S t s a e L y r a n i d r O ) n o i s s e r g e r ( n o i s s e r g e R l a i t a p S s i s y l a n A . s s a l C . t l u M s e i r e S e m i T t u p t u O t u p n I s k r o w t e N l a r u e N s e t a R h t w o r G s r e h t O r a e n i L r a e n i L - n o N Jack Faucett Associates (1999) Research and Development of Destination, Mode, and Routing Choice Models for Freight x x x x x Cambridge Systematics (1996) Quick Response Freight Manual (QRPM) x x x x Marker and Goulias (1998) Truck Traffic Prediction Using the Quick Response Freight Model Under Different Degrees of Geographic Resolution: A GIS Application in Pennsylvania x x x x x Garrido (2000) Spatial interaction between trucks flows through the Mexico-Texas border x x x FHWA (1999) Guidebook on Statewide Travel Forecasting x x x x Bastida and HolguÃn- Veras (2009) Freight generation models: comparative analysis of regression models and multiple classification analysis x x x x x x x Brogan (1980) Improving Truck Trip-Generation Techniques through Trip-End Stratification x x Middleton et al. (1986) Trip Generation for Special-Use Truck Traffic x x x Tadi and Balbach (1994) Truck Trip Generation Characteristics of Nonresidential Land Uses x x x Wegmann et al. (1995) Characteristics of Urban Freight System x x x Guha and Walton (1993) Intermodal Container Ports: Application of Automatic Vehicle Classification System for Collecting Trip Generation Data x x x HolguÃn-Veras et al. (2002) Truck-trip generation at container terminals x x Al-Deek et al. (2000) Truck Trip Generation Models for Seaports with Container-Trailer Operations x x x x Bartlett and Newton (1982) Goods vehicle trip generation and attraction by industrial and commeical premises x x x x x x x Dependent variable Independent variable Level of aggre- gation Level of geography Estimation technique Model structure (continued on next page)
pi r t - e l c i h e V t h g i e w o g r a C e u l a V t n e m y o l p m E ) . t f . q S ( a e r a . d l i u B e s U d n a L a t a d c i m o n o c e r e h t O e p y t e l c i h e V s e p y t y t i d o m m o C s r e h t O e t a g e r g g a s i D e t a g e r g g A n a t i l o p o r t e M e d i w e t a t S l a n o i t a N r o d i r r o C ) s t r o P . e . i ( . l i c a F l a i c e p S s e t a r p i r T s e r a u q S t s a e L y r a n i d r O ) n o i s s e r g e r ( n o i s s e r g e R l a i t a p S s i s y l a n A . s s a l C . t l u M s e i r e S e m i T t u p t u O t u p n I s k r o w t e N l a r u e N s e t a R h t w o r G s r e h t O r a e n i L r a e n i L - n o N Al-Deek (2001) Comparison Between Neural Networks and Multiple Regression Approaches for Developing Freight Planning Models with Specific Applications to Seaports x x x x x Kawamura et al. (2005) Business and Site specific trip generation methodology for truck trips x x x x Novak et al. (2008) Nationwide Freight Generation Models: A Spatial Regression Approach x x x x Maruyara and Harata (2005) Incorporating Trip-Chaining Behavior into Network Equilibrium Analysis Waliszewski et al. (2004) Comparison of Commodity Flow Forecasting Techniques in Montana x x Sorratini and Smith (2000) Development of a Statewide Truck Trip Forecasting Model Based on Commodity Flows and Input-Output Coefficients x x x Boyce (2002) x x x x x Giuliano et al. (2007) Estimating freight flows for metropolitan area highway networks using secondary data sources x x x x x Fisher and Han (2001) External Urban Truck Trips Based on Commodity Flows. x x x x Al-Battaineh and Kaysi (2005) Commodity-based truck origin-destination matrix estimation using input-output data and genetic algorithms x x x x x x x x Sorratini (2000) Estimation Statewide Truck Trips Using Commodity Flows and Input-Output Coefficients x x x x x Hewings et al. (2002) Combined model of interregional commodity flows on a transportation network x x x x x Zhao and Kockelman, 2004 The Random-utility-based Multiregional Input- Output Model: Solution Existence and Uniqueness x x x x Ham et al. (2005) Implementation and Estimation of a Combined Model of Interregional, Multimodal Commodity Shipments and Transportation Network Flows x x x x Dependent variable Independent variable Level of aggre- gation Level of geography Estimation technique Model structure Table 67. (Continued).
pi r t - e l c i h e V t h g i e w o g r a C e u l a V t n e m y o l p m E ) . t f . q S ( a e r a . d l i u B e s U d n a L a t a d c i m o n o c e r e h t O e p y t e l c i h e V s e p y t y t i d o m m o C s r e h t O e t a g e r g g a s i D e t a g e r g g A n a t i l o p o r t e M e d i w e t a t S l a n o i t a N r o d i r r o C ) s t r o P . e . i ( . l i c a F l a i c e p S s e t a r p i r T s e r a u q S t s a e L y r a n i d r O ) n o i s s e r g e r ( n o i s s e r g e R l a i t a p S s i s y l a n A . s s a l C . t l u M s e i r e s e m i T t u p t u O t u p n I s k r o w t e N l a r u e N s e t a r h t w o r G s r e h t O r a e n i L r a e n i L - n o N Iding et al (2002) Freight trip generation by firms x x x x x x x x x x Taniguchi and Thompson (2002) Modeling city logistics Patier and Routhier, 2008 How to improve the capture of urban goods movement data Russo and Comi (2002) Urban Freight Movement: a quantity attraction model x x x x x x x x x x Routhier et al (2002) Mesurer lâimpact du transport de marchandises en ville: le modèle de simulation FRETURB v1. x Wagner (2010) Regional traffic impacts of logistics-related land use x x x x x x x x DeVries and Dermisi (2008) Regional Warehouse trip production analysis: Chicago Metro Analysis x x x x x x x x Orsini et al. (2009) Logistics Facilities impacts on the territory, Ratio for French warehouses x x x x x x x x Bröcker (1998) Regional/national SCGE model x x x x x x x Tavasszy et al (1998) Multistep freight model x x x x x x x x x x x x x x Oosterhaven et al. (2001) Regional/national SCGE model x x x x x x x x WSP (2005) Multistep freight model x x x x x x x x x x x x x x Ivanova et al. (2002) Regional/national SCGE model x x x x x x x x Gentile and Vigo (2009) Movement generation and trip distribution for freight demand modeling applied to city logistics x x x x x x x x Swahn (2001) Multistep freight model x x x x x x x x x x x x x Al-Deek (2001) Regression Analysis Model for the port of x x x x x x Al-Deek et al. (2002) BPN Model to Port of Everglades x x x x x x Dependent variable Independent variable Level of aggre- gation Level of geography Estimation technique Model structure (continued on next page)
pi r t - e l c i h e V t h g i e w o g r a C e u l a V t n e m y o l p m E ) . t f . q S ( a e r a . d l i u B e s U d n a L a t a d c i m o n o c e r e h t O e p y t e l c i h e V s e p y t y t i d o m m o C s r e h t O e t a g e r g g a s i D e t a g e r g g A n a t i l o p o r t e M e d i w e t a t S l a n o i t a N r o d i r r o C e . P o r t s . i ( . l i c a F l a i c e p S s e t a R p i r T s e r a u q S t s a e L y r a n i d r O ) n o i s s e r g e r ( n o i s s e r g e r l a i t a p S s i s y l a n A . s s a l C . t l u M s e i r e S e m i T t u p t u O t u p n I s k r o w t e N l a r u e N s e t a R h t w o r G s r e h t O r a e n i L r a e n i L - n o N NCHRP Syn. 606 Heavy Truck Freight Model for Florida Ports x x x x x x x NCHRP Syn.606 Cross-Cascades Corridor Model x x x x x x NCHRP Syn.606 Minnesota Truck Highway 10 Truck Trip Forecasting Model x x x x x x NCHRP Syn.298 New York Metropolitan Region Model (Freight tunnel) x x x x x NCHRP Syn.384 Atlanta Commercial Vehicle and Truck Models x x x x x x x NCHRP Syn.384 Baltimore Metropolitan Council Models x x x x x x x NCHRP Syn.298 Bangor Area Model x x x x x x x NCHRP Syn.298 Chicago Commercial Vehicle Travel Demand Model x x x x x x NCHRP Syn.384 Delaware Valley Regional Planning Commission Model x x x x x NCHRP Syn.298 Denver Regional Model x x NCHRP Syn.298 Greater BuffaloâNiagara Regional Model x x x x x x x x NCHRP Syn.384 Los Angeles Regional Cube Cargo Model (2004) x x x x x x x NCHRP Syn.384 Maricopa Association of Governments Model x x x x x x x x NCHRP Syn.384 New York Best Practice Model x x x x NCHRP Syn.298 Portland Commodity Flow Tactical Model System x x x NCHRP Syn.298 Puget Sound Region Freight Model x x x x x NCHRP Syn.298 Skagit Countywide Model x x x NCHRP Syn.606 Southern California Association of Governments Heavy Truck Model x x x x x x x x x x NCHRP Syn.298 Vancouver Truck Freight Model x x x x x x NCHRP Syn.298 San Francisco Bay Area Freight Model x x x x x x NCHRP Syn.384 Southeast Michigan' s Regional Freight System x x x x x x x x x NCHRP Syn.298 Connecticut Model x x NCHRP Syn.606 Florida Intermodal Statewide Highway Freight Model x x x x x x x NCHRP Syn.298 Florida Model (2001) x x x x x x NCHRP Syn.606 Indiana Commodity Transport Model (1993) x x x x x x x Dependent variable Independent variable Level of aggre- gation Level of geography Estimation technique Model structure Table 67. (Continued).
pi r t - e l c i h e V t h g i e w o g r a C e u l a V t n e m y o l p m E ) . t f . q S ( a e r a . d l i u B e s U d n a L r e c o n o m i c d a t a e h t O e p y t e l c i h e V s e p y t y t i d o m m o C s r e h t O e t a g e r g g a s i D e t a g e r g g A n a t i l o p o r t e M e d i w e t a t S l a n o i t a N r o d i r r o C S p e c i a l F a c i l . ( i . e . , P o r t s ) s e t a R p i r T t s a e L y r a n i d r O n ) o i s s e r g e r ( s e r a u q S n o i s s e r g e R l a i t a p S . A n a l y s i s s s a l C . t l u M s e i r e S e m i T t u p t u O t u p n I s k r o w t e N l a r u e N s e t a R h t w o r G s r e h t O r a e n i L r a e n i L - n o N NCHRP Syn.298 Indiana Commodity Transport Model (1997) x x x x x x NCHRP Syn.298 Kansas Model x x NCHRP Syn.298 Kentucky Model x x x NCHRP Syn.298 Massachusetts Statewide Model x x x NCHRP Syn.298 Michigan Statewide Truck Travel Demand Forecasting Model x x x x x NCHRP Syn.298 Multimodal Freight Forecasts for Wisconsin (1996) x x x x NCHRP Syn.606 New Jersey Statewide Model Truck Trip Table x x x x x x x x NCHRP Syn.298 New South Wales Australia Model x x x NCHRP Syn.298 North Carolina Model x x x NCHRP Syn.298 Ohio Model x x x x NCHRP Syn.298 Oregon Model x x x NCHRP Syn.358 Virginia Freight Model x x x x x x x NCHRP Syn.358 Wisconsin Statewide Freight Component (2006) x x x x x x x NCHRP Syn.298 Wisconsin Statewide Truck Trip Forecasting Model (2000) x x x Dependent variable Independent variable Level of aggre- gation Level of geography Estimation technique Model structure
98 pi rt el ci he V y ti do m mo C wo lf C & V no it an ib mo C e ul a V tn e my ol p m E . qS ( ae ra . dli u B ).t f es U dn a L r eh t O at ad c i mo no ce ep yt el ci he V y ti do m mo C ep yt sr eh t O et ag er gg as i D et ag er gg A se ta R pi r T t sa e L yr an id r O s er au qS )n oi ss er ge r( l ait ap S no is se rg e R . ss al C . tl u M si sy la n A se ir eS e mi T tu pt u O t up nI l ar ue N sk ro wt e N se ta R ht wo r G sr eh t O ra en i L ra en il -n o N F & C 11 10 1 3 1 1 1 4 10 1 1 3 1 1 3 3 4 M 27 19 7 1 17 4 6 11 1 3 3 12 16 4 7 4 5 4 15 S 24 4 19 1 11 1 6 2 5 4 9 7 1 11 2 4 1 N 7 1 3 6 3 5 6 2 3 7 1 6 3 6 6 Total 69 34 27 5 6 34 6 12 23 3 9 12 26 33 5 17 2 4 1 23 3 3 12 22 11 n oi ta ge rg g A le ve L Modeling Technique er ut cu rt S Note: F & C: Facility and Corridor; M: metropolitan; S: Statewide; N: National y hp ar go e G se sa c fo r eb mu N Dependent Variables Independent Variables Table 68. Model characteristics by level of geography. In terms of the dependent variable, 47% use vehicle trips; 38% use commodity tonnage; and 15% use a combination of vehicle trips (usually for internal-internal trips); and commodity tonnage (for the rest of the flows). About 38% of the models are aggregated, 48% are disaggregated, and others (14%) cannot be derived from the review. The inde- pendent variables used include: employment by industry sector (49%); building area (9%); commodity type (13%); land use (2%); and other variables (27%). As for model- ing techniques: 25% use least square; 10% use trip rates; 6% use multi-classification analysis; and 33% use IO analysis. These three modeling methods constitute the majority of FTG models used in practice (or about 74%). In addition, from the model information that is known, most of the models are linear (22 out of 33), while a small fraction are nonlinear. Review of FG/FTG Models This section reviews some key publications. The models are grouped into vehicle-trip-based models and commodity- based models. The section is followed by descriptions of the advantages and disadvantages of the different modeling techniques. Trip-Based Models The FHWA Guidebook on State Travel Forecasting (Fed- eral Highway Agency 1999) uses land use and trip data from travel diaries and shipper behavior to estimate truck trips; these are then distributed using a form of gravity models that are calibrated with trip length frequency distributions obtained from trip diaries. Another trip-based model, the Quick Response Freight Manual (Cambridge Systematics Inc. 1996), calculates the number of commercial vehicle trips at the zonal level, commercial vehicle volumes at external sta- tions, and commercial trips between zones, by applying trip generation rates using economic activity data for the traffic analysis zone. After the trips have been estimated, the model uses mode shares for each trip and then loads the O-D matrix to the network. The estimated vehicle miles traveled (VMT) were compared with control VMT for calibration. This model was implemented in a truck flow survey study that investi- gated the effects on traffic assignment when using different degrees of geographic resolution (Marker and Goulis 1998). The study showed that applying a very aggregated model (e.g., the one suggested by the Quick Response Freight Manual) to a study area using extremely disaggregated Travel Analysis Zones (TAZs) results in no noticeable loss in model accuracy. The ITE Trip Generation Manual, 8th Edition (Institute of Transportation Engineers 2008) contains a comprehen- sive compilation of estimated FTG rates for a broad range of land use types. Although the focus of the ITE Manual is on all vehicle types, some of the results can be applied directly to FTG, e.g., truck terminals. The ITE Trip Generation Hand- book, 2nd Edition, provides guidelines on how these rates (for all vehicle types) may be used for a given trip genera- tion study. Appendix A of the ITE Handbook provides some information on truck trip generation. The Appendix also provides a number of cautionary notes to keep in mind when conducting FTG studies. The most noticeable ones are related to the need to: use consistent definitions of trucks and truck trips; consider the age of the existing FTG data; avoid land use classes that are too broadly defined; and think carefully about the selection of independent variables. Iding et al. (2002) estimated linear regression models of truck trip generation at industrial sites. The sample included 1,529 firms within the Netherlands with more than 5 employ- ees. Parameters (slope and intercept) were obtained for two
99 to auto-regressive moving average models (Garrido 2000). Sorratini (2000) estimated truck flows for the state of Wis- consin, using data from the 1993 CFS and IO coefficients. The authors derived production and attraction rates in tons for heavy truck mode for 28 economic sectors; the annual tons for the county level were converted to daily truck trips using aver- age tons-per-vehicle load factors. The trips were then assigned to the network and the results were compared to real counts. It was found that the production and attraction values were underestimated since not all truck trips were included. Bartlett and Newton (1982) studied FTG using regression models based on four independent variables: total employment, site area, gross floor area, and non-office employment. The firms were grouped based on FG intensity. It was found that the model results matched very well with actual vehicle-trip counts. It was also found that haulage firms, fuel distributors, waste disposal firms, and ready-mix concrete/bulk distribution firms were the most intensive generators, while manufacturers and printers were the least intensive. Freight generation intensity, however, varies significantly within the same industry sector. Commodity-Based Models Waliszewski et al. (2004) estimated zonal commodity generation using commodity-type specific growth rates and assumed that land use characteristics do not change over time. Novak et al. (2008) estimated FG at the national level, using commodity flow data using spatial regression meth- odologies. These models explicitly consider the spatial auto- correlation among variables based on spatial proximity. The authors found that the spatial autocorrelation violates the independence assumption usually imposed by ordinary linear regression models but contains valuable information that can improve model fit. The authors concluded that at the national level, spatial regression is the preferred specification for FG. Input-output (IO) models are generally implemented for large-scale systems at the regional, national or international level, since they require a great amount of data on regional economic activity and interregional flows. They have been used to estimate commodity-based generation models (Sor- ratini and Smith 2000). These models are basically macro- economic models that start from IO tables, which describe, in monetary units, what each sector of the economy delivers to the other sectors. Boyce (Boyce 2002) formulated and ana- lyzed a model of interregional commodity flows, incorporating regional IO relationships and the corresponding transpor- tation network flows. Using a local-area IO model combined with import-export commodity flow data from secondary sources, Giuliano et al. (2007) estimated intra-metropolitan freight flows on a highway network. Al-Battaineh and Kaysi (2005) used IO data with employment and population infor- mation to estimate commodity production and attraction different classification types (18 sectors and 5 types of heavy industry site) and two independent variables (area and employment). The results indicate that which independent variable is better depends on the industry sector and on the direction of freight (in- or outbound). The logistics and transport services sector was found to have the highest aver- age level of outbound trips produced. Other vehicle trip models estimate FTG rates for produc- tions and attractions using cross classification [Bastida and HolguÃn-Veras (2009) compared the use of cross classifica- tion and OLS for FTG modeling]. The authors estimated disaggregated freight trip delivery rates taking into consider- ation company attributes. Using cross classification analysis, the authors identified the groups of company attributes that best explain FTG. When using linear regression models, the authors identified that commodity type, industry segment, and employment are strong predictors for FG. Brogan (1980) analyzed different stratification strategies for improving trip- end generation models, identifying land use as the most effec- tive stratification scheme for improving model significance. Middleton et al. (1986) analyzed trip generation character- istics for special land use truck traffic in Texas; their study included an assessment of each special land use class in terms of FTG. Data collected included trip generation rates, trip length and vehicle type. Tadi and Balbach (1994) estimated trip generation rates based on vehicle type stratification for nonresidential land uses in Fontana, California; traffic counts were used on their estimations. Kawamura et al. (2005) took into consideration the supply chain decisions made by indi- vidual businesses in the estimation of FTG. Among other findings, the authors concluded that store floor space and the number of employees are poor indicators of truck trips at retail stores. At the city level, different freight models were developed in Europe that include some form of trip generation modeling [see Taniguchi and Thompson (2002) and Patier and Routhier (2008) for overviews]. Models are generally linear and based on zonal aggregates or survey data. Examples are Russo and Comi (2002) for Italy, and Routhier et al. (2002) for France. FTG models of various kinds have been developed for spe- cial facilities such as ports (Guha and Walton 1993; Wegmann et al. 1995; HolguÃn-Veras et al. 2002). Al-Deek et al. (2000) and Al-Deek (2001) used regression analysis and neural net- works respectively to develop trip generation models. Wagner (2010) carried out an analysis of trip generation around the port of Hamburg, Germany. Regional warehouse trip production rates were published in DeVries and Dermisi (2008) for the Chicago metro area, and in Orsini et al. (2009) for France. Other methodologies that have been implemented for production and attraction include: time series models, IO, and related models. Time series data have been used to develop models that range from growth factor models
100 based travel demand models were presented. The report also reviews numerous projects related to FTG, especially on the topics of truck trip generation data needs and survey methods. It lists three major methods to estimate truck trip generation data: estimation of simple rates, linear regression models, and commodity flow models. In Chapter Three of the report, data sources that were used to estimate truck trip generation in practice are compiled. This report also sum- marizes seven most commonly used approaches to collecting data for truck trip generation, including trip diaries; classi- fication counts; published commodity flow data; collected commodity flow data; shipper/carrier/special generator sur- veys; intercept surveys; and published rates. NCHRP Synthesis 384: Forecasting Metropolitan Commercial and Freight Travel The report reviews methods of freight and commercial vehicle forecasting in practice, together with promising meth- ods emerging from ongoing research. The primary focus of the report is on metropolitan-level forecasting, although some consideration is also given to statewide freight forecast- ing models. The report reviews application of the four-step model process to freight demand modeling, including the process of FTG. Major sources of planning information for freight and commercial vehicle forecasting are presented in this report. Besides the four-step model process, the report also summarizes seven emerging methods in freight demand models: time series modeling of freight traffic growth; behav- iorally focused demand models; commodity-based forecasts, including interregional I-O models; methods that forecast flows over multimodal networks; micro-simulation and ABS techniques; and models that incorporate supply chain/ logistics chain considerations. The report also lists several methods acquiring FTG results, including developing truck trip generation rates, borrowing trip rates from one or more other regions, introducing special generators, and using external stations. On urban freight data collection, the report presents two major methods: vehicle classification counts; and origin-destination surveys, which include roadside inter- cept surveys, mail and telephone surveys, establishment sur- veys, and carrier surveys. NCHRP Synthesis 358: Statewide Travel Forecasting Models The report examines statewide travel forecasting mod- els, including passenger vehicles and freight components. It reviews the types and purposes of models being used. Data requirements, survey methods, funding, and staff resources are also reviewed to investigate the limitations and benefits of the models. In the survey of statewide freight forecast- at the zonal level. Interregional commodity flows have been formulated and analyzed incorporating regional IO relation- ships and the corresponding transportation network flows (Hewings et al. 2002). Zhao and Kockelman (2004) not only estimate freight traffic generation and attraction, but also the flows between regions and the mode share using inter- regional versions of IO models. In Europe, commodity-based freight models are opera- tional in several countries. Most of these start from IO tables or Make/Use tables (IO tables that include an additional segmentation of type of goods). Value-to-weight ratios and regional employment statistics are used to convert macro level IO data to regional commodity flow data. Trip generation rates are, to an increasing degree, becoming endogenous variables in these commodity-based freight models, either implemented as variants of the Lowry type land useâtransportation inter- action (LUTI) models, or as advanced Spatial Computable General Equilibrium (SCGE) models. Several countries in Europe (and the United States and Canada) have transferred their freight models from IO-based, fixed coefficient models to flexible coefficient models, either in the shape of LUTI or of a full SCGE model. For a broad inventory of international experiences in integrative commodity-based trip generation and distribution modeling, the reader is referred to Tavasszy et al. (1998). Examples include the Dutch freight models SMILE and RAEM (Tavasszy et al. 1998; Oosterhaven et al. 2001); the UK Freight model (WSP Policy & Research 2005), the Swedish freight model SAMGODS (Swahn 2001), the Norwegian model PINGO (Ivanova et al. 2002) and the Ger- man SCGE model, which later became known as CGEurope (Bröcker 1998). Review of TRB Synthesis Reports FTG has been a focus of several NCHRP studies includ- ing NCHRP Synthesis 298, NCHRP Synthesis 384, NCHRP Synthesis 358, NCHRP Synthesis 606, NCHRP Synthesis 410, and NCHRP Report 404. Brief summaries of these studies are provided in this section. NCHRP Synthesis 298: Truck Trip Generation Data The synthesis report mainly identifies available data sources and data collection techniques, and assesses the current state- of-the-practice in truck trip generation. The report discusses key considerations in the development of truck trip genera- tion data needs, which include uses of truck trip generation data, trip purposes, estimation techniques, and data col- lection. Two types of trip generating models are discussed: vehicle-based models and commodity-based models. Twelve vehicle-based travel demand models and 14 commodity-
101 Analysis Framework 2 (FAF2), TranSearch Commodity Flows Database, freight databases from local and regional studies, socio-economic data from regional studies, and other data sources. It also conducted interviews with potential data users and providers. NCHRP Synthesis 410: Freight Transportation Surveys This review examines the sample size, data accuracy, data comprehensiveness, and survey objectives for freight trans- portation. It also includes a discussion of the feasibility and benefits of linking survey data with data from roadway and sensors. NCHRP Report 404: Innovative Practices for Multimodal Transportation Planning for Freight and Passengers This report reviews innovative agency practices and meth- ods in multimodal planning. The report monitors the per- formances of public involvement planning effects on rural areas of the studied cases. It also mentions fiscal constraints in planning and programming. Comparison of FTG Methods and Models Table 69 presents a summary of the advantages and dis- advantages of various methods and models used to estimate freight transport production and attraction. The table com- bines the review results in Jong et al. (2004) and those in Bastida and Holguin-Veras (2009). Summary This section summarizes the findings of conducting the literature review regarding FG and FTG models. The Bulk of the Studies Have Focused on FTG, Not FG. As illustrated in the literature review, the bulk of the models are based on vehicle trips, though a handful of studies con- sider FG in the context of IO models. This stands in contrast with European practices that emphasize commodity-based approaches that incorporate FG modeling as an endoge- nously determined variable. It is Not Yet Clear Which Modeling Techniques Are the Best. Although extensive research has been conducted in the last several decades on developing FG/FTG, there is no study to compare specifically the performances of these techniques; there is no consensus yet regarding which models ing practice, the report defines two fundamentally different styles of freight forecasting: Direct forecast of vehicle flows without reference to commodities; and forecasting of com- modities, then using the commodity flow forecast to esti- mate vehicle flows. The report includes five case studies: two are on freight components, including the Virginia freight component and the Wisconsin freight component; two are on passenger components; and one is a combined passen- ger and freight component. The report concludes that most statewide models are similar in structure to four-step urban transportation planning models, and that there exists no well-accepted definition of best practice in statewide mod- els. The report points out several distinct trends in recent statewide model development, such as the emergence of commodity-based models, and more effective use of GIS to manage data, among others. NCHRP Synthesis 606: Forecasting Statewide Freight Toolkit This report presents an analytical framework for forecasting freight movements at the statewide level to develop forecast- ing models. The framework includes a tool kit of data collec- tion techniques, analytical procedures, and computer models. It includes management approaches, decision-making pro- cedures, and performance evaluation methods, which help improve statewide transportation under the increment of freight demands. The report also summarizes several classes of data sources, including model development (local and national surveys, compilations); flow conversion (tons to vehicles and tons to value); network data (modal network and intermodal terminals); forecasting data (population and employment); validation data; and classification schemes (commodity classification and industry classification). Mean- while it presents five forecasting models and performance measures. Ten case studies of statewide freight modeling projects are reviewed, including FTG models, and model application and validation. NCHRP Project 08-36/Task 79, âScoping Study for a Freight Data Exchange Networkâ This report investigates the feasibility of building a freight data exchange network to provide access to higher quality freight data. It considers a centralized data repository from which data providers and users can access freight datasets, metadata, or reports of data quality. In this network, data pro- viders can upload data while end-users can download them in the form of summary tables, reports, and customized tabu- lar data. The report describes various types of freight-related datasets, and suggests potential ways to utilize them, includ- ing the CFS, Rail Waybill Data, foreign trade data, Freight
102 There Are No Consistent Definitions of Trucks, Truck Trips, and Land Use Classes. The lack of consistent industry terms is a point made by the ITE Trip Generation Handbook, and other publications. The inconsistent definitions of these important variables contributes to shaky results regarding which factors are the most important in explaining FG/FTG, and which modeling techniques are most effective. There is thus a need to standardize these definitions so that more con- sistent FG/FTG modeling approaches could be developed. can produce the most accurate results. This is reflected by the fact that different agencies are applying a variety of dif- ferent freight (trip) generation models (see Appendix E) due to the lack of a commonly agreed upon âbest prac- ticeâ model. However, based on previous research experi- ences, the research team does believe that certain modeling techniques, such as disaggregated models and regression analysis, have advantages that stand out among all modeling techniques. Type of model Advantages Disadvantages Time series Require multiple data points over time for the same facility. Limited data requirements for independent variables. Little insight into causality and, limited possibility to study policy effects Simple to calculate Unable to connect the effect of business size on FTG which may lead to significant errors Limited data requirements (zonal data) Little insight into causality and, limited scope for policy effects Linked to the economy Need input-output table, preferably multi- regional Can give land use interactions Need to identify import and export trade flows Policy effects could be considered if coefficients are elastic Restrictive assumptions if fixed coefficients Need conversion from values to tonnes Ordinary Least Squares (regression) Able to identify not so obvious relations pertaining to demand generation; can be used not only to forecast future demand, but also to establish the dynamics between variables Violations of the Ordinary Least Square (OLS) assumptions could lead to inaccurate parameters; especially using aggregated data Spatial regression Improve model fit; eliminate problems associated with the spatial autocorrelation Choice of a spatial model depends on actual data and it is hard to pre-determine which structure is more appropriate Multiple Classification Analysis Can overcome the disadvantages associated with cross classification analyses May overestimate the future number of trips if the number of observations by category is not exactly the same Neural networks Can produce accurate results; do not need to preselect independent variables; the learning capability of the model can discover more complex and suitable interactions among the independent variables. Need a sizeable database to develop and calibrate the model Input-output Trip rates Table 69. Advantages and disadvantages of FG/FTG models.
