4-Step Commodity Model Freight Forecasting

Transcription

1 Freight Forecasting Neda Masoud Kate Hyun Professor Ritchie CEE290A University of California Irvine

2 Generation Distribution Mode Split Assignment Total Production and Consumption Tonnage for each FAZ Productions and Consumptions are distributed between ODs Distributes tonnages of commodities to various modes Assigns trucks or rail cars to the appropriate network to produce vehicle flows 2

3 1. Commodity Classification 2. Zoning Structure 3. Vehicle Classification STCC Code: Standard Transportation Commodity Classification Developed in early 1960s by American Association of Railroads compatible with the SIC (Standard Industrial Classification system) Data Sources: CFS prior to 1997 TRANSEARCH commodity database VIUS (Similar to two digit STCC codes) FAFD (FAF estimates the flows of commodities at the four-digit STCC level for each mode at county level) Reporting system used in the STB s Carload Waybill sample 3

4 1. Commodity Classification 2. Zoning Structure 3. Vehicle Classification STCC digits: First digit: identifies a major Economic Division. Exp. 2: Nondurable manufacturing Second digit: Economic Major Group. Exp. 20: Food Third digit: Industry Group. Exp. 202: Dairy products Fourth digit: Specific Industry. Exp. 2024: Ice Cream and Frozen Desserts. Additional detail provided up to 7 digits. 4

5 1. Commodity Classification 2. Zoning Structure 3. Vehicle Classification 5

6 1. Commodity Classification 2. Zoning Structure 3. Vehicle Classification SCTG code: Standard Classification of Transportation Goods developed jointly by U.S. and Canadian government agencies Based on Harmonized Schedule Classification compatible with the NAICS (North American Industry Classification System) Data Sources Used in 1997 and subsequent CFS 6

7 1. Commodity Classification 2. Zoning Structure 3. Vehicle Classification Compatibility of STCC (and SIC) and SCTG (and NAICS) SCTG code is not directly transferable to STCC Concordance between STCC2 and SCTG2, STCC2 and SCTG4 correspondence between the SIC and NAICS codes at the four-digit SIC and six-digit NAICS level 7

8 1. Commodity Classification 2. Zoning Structure 3. Vehicle Classification Changing from STCC to SCTG code Advantage: Is in accordance with internationally used Harmonized Schedule classification NAICS codes in general cover a larger range of industries, including emerging industries, advanced technology industries, and product and service industries. Disadvantage: Not all data sources have upgraded to SCTG For those who have moved, forming time series would be difficult 8

9 1. Commodity Classification 2. Zoning Structure 3. Vehicle Classification Identifying commodity groups Considering all 43 SCTG commodity groups Costly Increase the dimensions of the problem Not separate explanatory variables recommended criteria Keeping homogeneity of products and related industries in one group Keeping commodities with specific mode of transportation as a separate group Combining commodities with little tonnages Keeping major commodities as a single group State Number of commodity groups Florida 14 Indiana 19 Nebraska 24 Wisconsin 16 production, 16 attraction (Not entirely the same) Pennsylvania 10 Tennessee 10 Texas 11 Michigan 12 9

10 1. Commodity Classification 2. Zoning Structure 3. Vehicle Classification Sate No. of counties No. of zones Inside the state No. of zones outside the state Total zones Note Iowa Indiana nodes or terminals representing adjacent states of Ohio, Illinois, Kentucky and Michigan, and a single zone for other states and the District of Columbia Cross-Cascade Corridor Washington counties within the corridor were rather subdivided into 2 to 4 zones. 1 zone is for Idaho, and 6 external zones 10

11 1. Commodity Classification 2. Zoning Structure 3. Vehicle Classification recommended criteria: land use natural barriers zone size and density number of employment location of connectors to road and rail networks location of major ports and road facilities air-basin TAZ boundaries (exp. Oregon) 11

12 1. Commodity Classification 2. Zoning Structure 3. Vehicle Classification Classification In terms of size, weight, and class Based on the number of axles, axle spacing and vehicle weight Caltrans :15 classes by axle including motorcycles and passenger vehicles FHWA : 13 classes scheme VIUS : 72 classes associated with truck and axle type on specific truck models FHWA Vehicle Classification Description Category Description 1 Motorcycles 2 Passenger cars, Light vans, Light pick-ups 3 2 Axle 4 Tire, Full size Pick-ups, Full size Vans, Limos 4 Buses 5 2 Axle, 6 Tire single-unit trucks 6 3 Axle single-unit truck 7 4 or More Axle single-unit trucks 8 3 or 4 Axle single-trailer trucks 9 5 Axle single-trailer trucks 10 6 or more Axle single-trailer trucks 11 5 or less Axle multi-trailer trucks 12 6 Axle multi-trailer trucks 13 7 or more Axle multi-trailer trucks (Source : Federal Highway Administration, vehicle classification monitoring,) 12

13 Data Source CFS (Commodity Flow Survey) FAF (Freight Analysis Framework) TRANSEARCH Description A national survey, conducted every 5 years as part of the economic census. Provides flow data by commodity, origin, destination, and mode. The reported flows are in two forms: state to state, and between 86 Metropolitan areas. Combination of modes in CFS is defined as a separate model. Advantages: Is the only survey shippers are obligated to reply to and therefore is less likely to be biased Disadvantages: Is not consistent from year to year (methodology, sample size, etc.) making it hard to build time series Is a sparse matrix (lots of zeros) that should be filled out Does not include some commodities A national database, providing freight flow by origin, destination, commodity group, and mode of transport. FAF3 provides aggregated data for 123 zones inside the united states. Advantages: A more complete version of CFS where the data that is missing from CFS is obtained from other data sources A proprietary national database of freight flows that uses mode-specific data sources to provide flows by origin, destination, commodity, and mode at the county aggregation level. TRANSEARCH is an unlinked trip table that reports trip proportions by each mode. Advantages: Accepted and used by FHWA and many states and MPOs. Disadvantages: Is a proprietary database and price depends on the number of records purchased The complexity of the sources and the proprietary nature of the database prohibit its owners to reveal all the data sources used to build this database 13

14 Data Source Census population CBP (County Business Pattern) VIUS (Vehicle Inventory and Use Survey) WIM PMA Description Obtained every 5 year as part of the US census. The population estimates are provided at the state, county, city and town level. CBP annual reports show the economic activity by industry in county level. Employment by industry used in development of the generation models in the first step of the 4-step commodity model is obtained from CBP. Disadvantages: Does not include all employment data A survey that is conducted every 5 years, as a part of the US Economic Census. This survey contains a 2000 sample of truck data per state, including physical and operational characteristics. VIUS is usually used to convert tons to truck trailer for each commodity class in the network assignment step of the 4-step commodity model. Data provides 24hour traffic information including the speed, weight, and count on the highway at the state level, and the traffic count can be used for validating the freight forecast model. Disadvantages: Does not provide the arterial or local traffic condition and has erroneous records due to the WIM stations performance. Obtained by the Pacific Maritime Association every year. This database provides waterborne cargo loaded and discharged in California, Washington and Oregon ports. Advantages: The data includes international flows and domestic trades assessed by the tonnages and values Disadvantages: The data are obtained only from the membership carriers of PMA and does not include liquid or LPG carrier cargo. Cargo tonnage can be over-estimated by reporting all loaded and discharged cargo, regardless of actual transport. 14

15 Data Source Waterborne Trans-border STB Carload Rail Waybill sample surveys Interviews Description A national data source that reports point of loading and unloading for each commodity. The reports are usually made to the Army Corps of Engineers on the basis of vessel movements. Disadvantages: Military cargo that are moved in department of defense vessels are reported Is obtained from documents collected by the U.S. Customs and Border Protection, and shows flows by commodity and mode between united states and its NAFTA partners. Advantages: The data is obtained from a complete enumeration of documents obtained from US customs and border protection and therefore is free of sampling errors. Disadvantages: Other sources of error, including underestimation of low-valued transactions, undocumented shipments, and reporting errors are present Is a national stratified sample of Carload Waybill, containing origin, destination, commodity, tons, etc. from railroad terminating more than 4500 cars per year. Disadvantages: Is not available for public use. However, there is an aggregated publically available version Local surveys where necessary. Interviews where necessary. 15

16 4. Network Assignment Forecasts the productions and attractions of freight movements in a zone Input: exogenously supplied zonal employment or economic activity Methodology: Regression models Output: annual or daily trip generation equations by commodity Explanatory variables: Production: employment by the industry that produces that commodity Attraction: two main markets for each commodity at a given zone: Direct consumer : population Industrial consumer : employment 16

17 4. Network Assignment Intermodal Facilities An assumption of trip generation equations: freight shipment to and from a zone is a function of industrial activity of that zone. But: it is possible to have freight activity in a zone when there is little or no activity in related industries Sign: outlier zones in regression models Solution: Special generators Whether special generators should be considered or not depends on the commodity flow database used Unlinked commodity trips: TRANSEARCH special generators Linked commodity trips: CFS special generators only for international gateways 17

18 4. Network Assignment Example : Florida Airport Sea ports Warehouses (source : Oak Ridge National Laboratory,2007) 18

19 Case study 4. Network Assignment Indiana Commodity group production model attraction model Pro Adj RR 2 Att.-Adj RR 2 Farm Products Prod01 = Agser Cash Attr01 =.819 Prod Coal Prod11 = 7.6 Coal Attr11 = 3.1 Coal Min Nonmetallic Minerals Prod14 =.078 Man Attr14 =.997 Prod Food and Kindred Products Prod20 =.282 Food Attr20 =.832 Pop Food Basic Textiles Prod22 =.016 Tex Attr22 =.003 App All Apparel Prod23 =.004 App Attr23 =.002 App Pop Lumber and Wood Products Prod24 =.668 Lum Attr24 =.728 Prod Furniture and Fixtures Prod25 =.017 Furn Attr25 =.033 Pop Furn Pulp and Paper Products Prod26 =.103 Pulp Lum Attr26 =.085 Pulp Furn Chemicals and Allied Products Prod28 =.150 Chem Pet Attr28 =.077 Chem Pet Pop Petroleum and Coal Products Prod 29 = Pet Attr29 = Pet Pop Stone, Clay and Glass Products Prod32 = Pop Attr32 = Pop Primary Metal Products Prod33 =.085 Met Attr33 =.093 Met Fab Fabricated Metal Products Prod34 =.013 Met Fab Attr34 =.035 Fab Machinery (except Electrical) Prod35 =.013 Mac Attr35 =.010 Mac Electrical Machinery Prod36 =.004 Met Fab Elec Attr36 =.005 Fab Pop Transportation Equipment Prod37 =.040 Tran Attr37 =.027 Tran Waste and Scrap Material Prod40 = Pop Attr40 =.0067 Man Other Manufactured Products Prod50 = Attr50 Attr50 =.245 Pop Source: NCHRP report 19

20 Case study 4. Network Assignment Florida-production No. Commodity Groups Coefficient Variable (Employment) 1 Agricultural SIC07 2 Nonmetallic 6, SUM(SIC10-14) 3 Coal No Production Employment 4 Food SIC20 5 Nondurable Manufacturing SUM(SIC21,22,23,25,27) 6 Lumber SIC24 7 Chemicals SIC28 8 Paper SIC26 9 Petroleum Products SIC29 10 Other Durable Manufacturing SUM(SIC30,31,33-39) 11 Clay, Concrete, Glass SIC32 12 Waste 0.5 TOTEMP 13 Miscellaneous Freight TOTEMP 14 Warehousing SIC50 + SIC51 Source: NCHRP report 20

21 Case study 4. Network Assignment Florida- attraction No. Commodity Groups Coefficient Variable Coefficient Variable 1 Agricultural SIC20 2 Nonmetallic Minerals SIC28 3 Coal SIC49 4 Food SIC51 5 Nondurable Manufacturing SIC51 6 Lumber SIC Pop 7 Chemicals SIC51 8 Paper SIC51 9 Petroleum Products Pop 10 Other Durable Manufacturing SIC Clay, Concrete, Glass Pop 12 Waste SIC33 13 Miscellaneous Freight SUM (SIC42,44,45) 14 Warehousing Pop Source: NCHRP report 21

22 4. Network Assignment State Method Independent variables Data source Florida Regression, growth rate Employment by the SIC Population TRANSEARCH freight database Intermodal ports or terminals CFS Indiana Regression Employment by the NAICS Population (to represent the consumer market) 1977 CFS 1977 census 1977 county business pattern 1993 CFS 1993 Census 1992 Woods & Poole (population and employment forecast-future year) Nebraska Regression Employment by the SIC Population 1993 CFS (production) IMPLAN (attraction coefficients) Vermont NA NA TRANSEARCH Road-side surveys Motor carrier surveys Interviews with shippers Employment by the SIC Iowa Fratar TRANSEARCH Population Wisconsin Regression Employment by SIC 2001 TRANSEARCH 22

23 4. Network Assignment Input: productions and attractions by trip generation, relative impedances between zones Output: a trip table for each commodity 23

24 4. Network Assignment Gravity Models: Number of trips: TT iiii = PP iiaa jj FF iiii AA jj jj FF iiii Friction factor: FF iiii = ee cc iiii kk Friction factor could be in terms of travel time, cost or distance average distance for travel (k) is commodity specific State Method impedance Data source Florida Gravity model Distance and time (time for short distance trips) TRANSEARCH (average trip length) Indiana Gravity model Straight line distance 1993 CFS 24

25 4. Network Assignment Fractional split models Have shown to give better results compared to gravity models estimate the fraction of commodity at the destination zone from all the origin zones. 25

26 Data Conversion 4. Trip Assignment Overview of Mode choice Estimate freight tonnage flows by commodity group between the FAZ shipped by all freight modes; highway, rail, water, and air mode Input data : tonnage of commodity group Output of the mode choice step = An input to the assignment Intermediate step : Data conversion flow units in tons Tons -> vehicle 26

27 Data Conversion 4. Trip Assignment Characteristics of modes Goods characteristics and selecting modes (Source : Quick Response Freight Manual II, FHWA, 2007) Rail, air, and water Lack of data (need more effort on gathering the data) Take up considerable part in transporting freight, serve important role in estimating truck trips Future competitors of the truck freight 27

28 Data Conversion 4. Trip Assignment Mode shares Georgia Freight Model TRANSEARCH Tonnage Mode Split Mode split by weight (Cambridge Systematics 2002) Water 31% Air <1% Rail 4% Truck 65% Commodity tonnage by mode and by commodity group (FISHFM) Commodity Group Truck % Carload % Intermodal Rail % Air % Water % Total Agricultural , ,964 Non-metallic Minerals , ,817 Coal , ,270 Food 22, , , ,474 Non-durable manufacturing 4, ,134 Lumber 12, , ,420 Chemiclas 27, , , ,495 Paper 11, , , , ,894 Petroleum Products 7, , ,958 Other Durable Mfg 16, , ,195 Clay/Concrete/Glass 26, , , ,566 Waste 4, ,141 Miscellaneous Freight , ,399 Warehousing 83, ,020 TOTAL 216, , , , , ,747 (Source : Quick Response Freight Manual II, FHWA, 2007) 28

29 Data Conversion 4. Trip Assignment Discrete Logit Choice Model Most commonly used Based on random utility maximization theory Utility = f(generalized cost) Generalized cost = distance, time, cost and reliability for mode Fundamental assumption : each shipping unit should serve as a decision maker considering utility of a mode In freight, decisions are involved with a few of individuals even though freight movement provides the millions of tons of commodities Florida model 29

30 Data Conversion 4. Trip Assignment Simple Approach Existing mode splits remain in future Estimate future forecast : apply the current mode share by commodity Assumption : freight movement may not be sensitive to travel time, costs and reliability Indiana, Wisconsin 30

31 Data Conversion 4. Trip Assignment CARGO model in CUBE : Logit model z ( dtx,, ) e = e G ( dtx,, ) cm G ( dtx,, ) z cm m cm cm = proportion of commodity group c by mode m = Generalized cost for commodity group c by mode m G (,, ) cm dtx Generalized cost G( dtx,, ) = k + k * d+ k * t+ k * x m ocm 1cm 2cm 3cm d = distance, t = time, x = cost 31

32 Data Conversion 4. Trip Assignment Case Study 1. Indiana Assumption : the mode choices of the base year remain the same Single Modes Parcel/Courier U.S. Portal Service Private Truck For-Hire Truck Air Rail Inland Water Great Lakes Deep Sea Water Multiple Modes Private Truck and For-Hire Truck and Air Truck and Rail Truck and Water Truck and Pipeline Rail and Water Inland Water and Great Lakes Inland Water and Deep Sea Source: Indiana commodity transport model, William R. Black, Indiana University) 17 mode categories, 9 distance-based sub-categories less than 50 miles, 50 to 99 miles, 100 to 249 miles, 250 to 499 miles, 500 to 749 miles, 750 to 999 miles, 1,000 to 1,499 miles, 1,500 to 1,999 miles, and 2,000 miles or more 32

33 Data Conversion 4. Trip Assignment Case Study 2. Florida Incremental logit model for truck and rail (TRANSEARCH data) S S ' i S i U i v b = Si Ui ' i J S j = 1 j *exp( ) *exp( U ) j = new share of mode i = original share of mode i = utility of mode i in the choice set J (j=1,2,3 J) v = Modal constant + b * Explanatory variable = coefficient for explanatory variable Explanatory variables (= f(distance)) : travel time, commodity value per ton, and cost Coefficient for the model : from the survey in New York and calibrated with the Florida database Simple Approach for waterborne and air modes 33

34 Data Conversion 4. Trip Assignment Case Study 3. Wisconsin Assumption : The existing mode share is identical to the future Database : TRANSEARCH database of the tonnage matrices for each mode Trendlining freight forecast method in Wisconsin Year Origin Destination SIC Tons 1992 ABC DEF Rail Share Truck Share Rail Tons Truck Tons XX % 60% YY % 40% TOTAL % 47% ABC DEF XX % 60% YY % 40% TOTAL % 46% (Source : Multimodal Freight Forecasts for Wisconsin, Wilbur Smith Associates in association with Reebie Associates, 1996) 34

35 Data Conversion 4. Trip Assignment Freight Rail : America on the move 35

36 Data Conversion 4. Trip Assignment Data Conversion 1. Conversion from tonnage to daily freight truck trips Included in mode choice step to assign the commodity flow on network The consistency of the unit between freight flow and the passenger flow Payload factor Average tonnage of freight carried By commodity groups for long-haul trips Data Sources : VIUS from U.S. Census Bureau or carrier survey Not applicable in waterborne and air freight 2. Annual to daily conversion Conversion factor : working days that trucks are expected to move 36

37 Data Conversion 4. Trip Assignment Holguin-Veras & Gopal Patil Issues regarding empty trucks Empty truck percentage of Florida (source : technical report for FISHFM, model specification(2002), Cambridge Systematics, Inc.)

38 W o l, W e l Introduction Data Conversion 4. Trip Assignment Holguin-Veras & Gopal Patil A model combined a commodity-based model with empty trips Model 1 : Loaded + empty truck trips Model 2 : Loaded and empty trips, respectively Objective function V W U o l o l o l e, Vl, Wl e, U l e Model 1 Model 2 o e min( β, p) F = ( V V ) v i i l 2 min( β, p) F = ( V V ) = ( W W ) + ( U U ) o e 2 o e 2 o e 2 v i i i i i i l l l l l l l e e e e e e V = z p = ( x + y ) p = x p + y p i ij ij ij ij ij ij ij ij ij i j i j i j i j = W + U Observed and estimated total truck traffic on link i Observed and estimated loaded link volumes Observed and estimated empty link volumes e l e l x y z ij ij ij Loaded trips of commodity k from i to j Empty trips from i to j Total number of trips from i to j 38

39 Data Conversion 4. Trip Assignment Holguin-Veras & Gopal Patil (Cont) Results Model 2 outperformed to reduce errors associated with estimated traffic and observed counts Explain the pattern of empty trips of urban area that comprises 30% to 40% of total truck trips Can be applied to statewide model to improve replicating the travel pattern 39

40 Data Conversion 4. Trip Assignment Case Study 1. Indiana Rail and truck flows Data Source : Carload waybill sample Rail density factor Truck assumed to carry 40% of the rail-load based on the different capacity in rail and truck Annual to daily conversion factor : Highway Capacity Manual Special Report working days weekend truck traffic of multiplying 0.4 by weekday truck traffic 40

41 Data Conversion 4. Trip Assignment Case Study 1. Indiana Commodity STCC Import rail traffic Export rail traffic Weighted rail Weighted truck density (tons) density (tons) *10.00 * ,27,30,31 38 and * Weighted average of import and export rail density factor = weighted rail density(tons) * Estimated values (Source: Indiana commodity transport model, William R. Black, Indiana University) 41

42 Data Conversion 4. Trip Assignment Case Study 2. Florida Truck trips with empty trips Data Source : Vehicle Inventory and Usage Survey (VIUS) data by commodity groups and distance classes - The distances are categorized to five class; less than 50 miles, miles, miles, miles, and over 500miles trip Payload factor by the commodity and the distance class = average pound miles /average miles Annual to daily conversion factor days 42

43 Data Conversion 4. Trip Assignment Case Study 2. Florida Tonnage to truck conversion factors Annual tons to annual trucks conversion factors Average payload in pounds Commodity group < 50 miles miles miles miles 500+ miles Commodity Group On road average < 50 miles miles miles miles 500+ miles Agricultural Agricultural Minerals/Coal 32,725 18,408 36, , , , Minerals/Coal 41,637 41,237 35,000 42,138-46,000 Food Products Food Products 36,456 17,283 37,194 44,574 42,209 42,465 Non-durable manufacturing Non-durable manufacturing 17, , , , , ,579 Lumber Lumber 28, , , , , ,314 Paper Paper 30, , , , , ,960 Chemicals Chemicals 33, , , , , ,329 Petroleum Petroleum Products Products 42, , , , , ,653 Other Durable Mfg 22,761 10,237 13,944 37,440 38,416 34,464 Other Durable Mfg Clay/Concrete/Glass 36,931 31,647 40,617 39,934 45,413 44,802 Waste Clay/Concrete/Glass 25, , , , , ,066 Miscellaneous Freight Waste 24, , , , , ,853 Warehousing Miscellaneous Freight 18, , , , , ,125 Average Warehousing 28, , , , , ,602 (source : technical report for FISHFM, model specification(2002), Cambridge Systematics, Inc.) 43

44 Data Conversion 4. Trip Assignment Description Input data : Freight truck trips by mode between FAZ 1. Rules-based ( fixed path ) assignment 2. Dynamic path assignment Freight truck only assignment : an optimum path and estimates the flow on the highway considering only in trucks Multiclass assignment : Considering the congestion effect all-or-nothing, equilibrium, capacity-restrained equilibrium, and stochastic multi-path capacity restraint techniques 44

45 Data Conversion 4. Trip Assignment Description 1. Fixed path assignment rail, water, and air assignment : the fixed route that is not flexible in the new facility or changes in performance long-haul traffic : as those traffic patterns are not easily affected by changes in network or performance and may remain fixed over long period of time Data source : TRANSEARCH, MapQuest, etc 45

46 Data Conversion 4. Trip Assignment Description 2. Dynamic path assignment Consider the congestion and may alter routes responding to the travel time Freight Truck only Multiclass network analysis All or nothing technique Assumptions : dominant long-haul trips of large trucks, no significant effects in route choice in longhaul trip from the congestion Limitation : load too many vehicles and produces high travel time To respond the congestion by adding an additional step of adjusting speed Equilibrium technique Assumptions : familiar with the alternatives and respond to congestion Conversion into PCE (passenger car equivalents) : performance factor such as acceleration, deceleration, and braking times of each mode which can significantly affect the travel time on the network 46

47 Data Conversion 4. Trip Assignment Case Study 1. Indiana Step 1: All or nothing technique Step 2: adjustment procedure is added to calculate the new speed For truck: New speed = Old speed + (2* (65 Old speed )) For rail: Index of spatial separation for minimizing the distance and maximizing the usage of rail-line I 1 = L( ) D + 1 Where, I = the index of spatial separation; L = the length of the line segment of the network; D = the traffic density of the line in millions of gross ton-miles-per year 47

48 Data Conversion 4. Trip Assignment Case Study 2. Florida All-or-nothing technique using the free flow path Trucks are assigned prior to loading the passenger vehicles Airway Intermodal Highway Railway Multimodal Waterway (source : Oak Ridge National Laboratory,2007) 48

49 Description Examine the reliability of future forecasts Consider the accuracy of the forecasts while maintaining the balance between the accuracy and economic feasibility 1. Calibration : parameter adjustment parameters estimated in regression on trip generation step impedance factors in distribution model parameters estimated in logit model for mode choice - Data source : socioeconomic data and commodity flow census 2. Validation : estimated and observed values are matched Vehicle classification counts and the estimated truck traffic volumes on highway are compared in assignment validation step - Data source : WIM data, Truck Count data 49

50 Case Study 1. Indiana Data source : 1993 CFS data and InDOT truck count data Rail flows were validated by visual examination due to the limitation on data availability 2. Florida Truck count data from 1999 AADT Report for Florida and Truck Weight study data FAF s loaded highway network is used for estimating the freight truck trips including empty trips 50

51 Case study in Florida Comparison of total observed and assigned VMTs (2242links) 1. Missing intra-county flow 2. AoN assignment without considering the congestion effect Percentage difference between assigned and observed VMT by function class (source : Oak Ridge National Laboratory,2007) 51

52 1. Data limitations 2. Modeling limitations 3. Other issues Data availability at the FAZ level Exp. Truck payload factors from VIUS at the state level Solution: adjustments based on the locally available data, such as WIM Exp. SCAG: cordon surveys Exp. Seattle truck model: land use-based trip rate method to obtain commodity flows at the FAZ level validation of the available data vehicle counts or cordon surveys Forecast year data 52

53 1. Data limitations 2. Modeling limitations 3. Other issues Trip distribution: Neglecting the relationship between suppliers and manufacturers Handling empty trips Probabilistic models (Hautzinger) Model validation Intra-zonal trips, service sector trips, and household moving vans are not generated by models 53

54 1. Data Limitations 2. Modeling Limitations 3. Other Issues Other Issues Seasonality Certain commodities have seasonal trend in transporting or producing Applying a fixed factor might overestimate (or underestimate) the actual flow Daily tons of all commodities are converted based on the working days per year which is 312 days per year (6 working days per week) or 250 days per year (5 working days per week) 54

55 Vehicle-based models: Disadvantages: fail to recognize that freight travel is related to commodity movement, not truck movement are not policy sensitive are not able to reflect changes in growth rates by commodity class. Advantages: useful in estimating short distance service vehicle trips (package delivery) Economic activity models use an economic/land use model to forecast employment or economic activity prior to the trip generation step 55

56 Main References Beagan, D., Fischer, M., Kuppam, A. Quick response freight manual II FHWA-HOP , Federal Highway Administration, Black, W. R, Indiana commodity transport model, Transportation Research Center, Indiana University, Cambridge Systematics, Inc., 2002, Technical memorandum; model specification for FISHFM (Freight Intermodal State wide Highway Freight). Cambridge Systematics, Inc., Forecasting statewide freight toolkit, National Cooperative Highway Research Program (NCHRP) Report 606.Transportation Research Board, Washington, DC. Wilbur Smith Associates in association, Multimodal freight forecasts for Wisconsin 56

57 Thanks for your attention!