California Statewide Freight Forecasting Model (CSFFM) September 2012 Progress Report Stephen G. Ritchie Professor of Civil Engineering and Director, Institute of Transportation Studies University of California, Irvine Presented at Freight Project Working Group Meeting September 24, 2012 1
Outline Introduce the team Overview California Statewide Freight Forecasting Model (CSFFM) Proposed FAZs and TLNs Primary modules Summary of Progress Overall Progress Task Schedule Completed Tasks Progress Update on Q1 for FY-13 Key Decisions for the CSFFM Modules Upcoming Tasks Upcoming Tasks for 2012-13 Q2 Peer Advisory Committee Meeting 2
Project Team from ITS-Irvine Principal Investigator: Professor Stephen Ritchie University of California, Irvine Professor Michael McNally, University of California, Irvine Assistant Professor Joseph Chow Ryerson University, Toronto (formerly ITS-Irvine Postdoc) Andre Tok, Ph.D. Postdoctoral Scholar, University of California, Irvine Soyoung Iris You, Ph.D. Assistant Project Scientist, University of California, Irvine Jaeyoung Jung, Ph.D. Assistant Project Scientist, University of California, Irvine Graduate Students: Miyuan Zhao Fatemeh Ranaiefar Pedro Camargo Daniel Rodriguez-Roman Kyungsoo Jeong Paulos Lakew Neda Masoud 3
OVERVEIW 4
California Statewide Freight Forecasting Model Commodity-based model Forecasts the long haul flow of commodities and commercial vehicles on freight infrastructure as a function of socioeconomic conditions and infrastructure parameters Primary Modules Mode Truck Class Commodity Module : Predicts commodity flows (Mode Split Module : Predicts commodity flows by linked mode) Transshipment Module : Assigns commodities to specific moderoute through Transshipment nodes Truck Rail Water Air (+Truck-Air) Multiple Modes Caltrans Class : FHWA Class 5 to 13 CARB Class* : EMFAC 2011 Network Module : Assigns truck and rail modes to routes Truck Rail * Because of limited data, some truck classes might not be addressed (i.e. Medium-Heavy Duty Diesel CA International Registration Plan Truck by Gross Vehicle Weight Rate (GVWR), Medium-Heavy Duty Diesel instate Truck by GVWR, Medium-Heavy Duty Out-of-state Truck by GVWR, etc..) 5
Continued.. Policy-sensitive model within California (base year 2010) Initially defined freight policy scenarios for the CSFFM: Categories Future Year Scenarios Network Investment Scenarios Pricing Policy Scenarios Future Trends Scenarios (Domestic) Future Trends Scenarios (International) Target year: 2020 Target year: 2040 Scenarios Highway system modifications Railroad system modifications Transshipment facilities projects Fuel price, carbon pricing, tax Road pricing Change in land use or supply chain Changes in domestic industrial, agricultural or energy sectors Increase/Decrease in domestic manufacturing Change in commodity import/export Increase in domestic manufacturing, reduction of imports Increase/decrease in trade with CAN/MEX Widen of Panama Canal 6
Proposed FAZs and TLNs FAZ (222) Zones in CA: 96 Other States (FAF regions) : 118 International : 8 Santa Barbara Kern San Bernardino TLNs (48) Airport: 13 Seaport: 11 Rail terminal: 24 LA Orange Gateways (?) Airport: 13 Seaport: 11 Land port:? Total OD Nodes (270+?) FAZ + TLNs + Gateways Figure. A part of CSFFM network : Los Angeles, Ventura, and San Bernardino counties 7
CSFFM Primary Modules Socioeconomic Variables Exogenous Variables: Import/Export FAF OD Road/Rail Network Characteristics Air/Water Impedances Commodity Module Structural Direct Demand Output: Domestic Commodity FAZ OD by Commodity Groups Import/Export Output: Import/Export Commodity FAZ OD by Linked Mode IMPLAN FAF Port Data (POLB & POLA) Waterborne Transborder PMA (Pacific Maritime Association) Mode Split Module Output: Commodity (Groups) OD by Linked Mode Rail Waybill Road/Rail Network Characteristics Air/Water Impedances Transshipment Module (FAZ + Gateways + TLN) Output: Commodity OD (FAZ + Gateways + TLN) by Specific Mode Conversion of Tonnage to Truck/Rail Trip Rail Waybill Waterborne T-100 Vehicle Inventory and Use Survey (VIUS) Weight-In-Motion (WIM) Network Module Output: Truck Network Flows by Commodity, by Truck Class, by Time of Day, and by Season : input variables for the model : input data for calibration 8
Continued.. How does CSFFM compare with a conventional commodity-based model? Commodity Generation (Regression Equation) Commodity Distribution (Gravity Model) Commodity Module Structural Direct Demand Domestic Commodity FAZ OD by Commodity Groups Total Demand Import/Export Commodity O/D Mode Choice (Logit Model - truck, rail, water, air, Intermodal) Freight Truck Trip Table (Payload factors) Mode Split Module Commodity (Groups) OD by Linked Mode Path-based commodity OD Transshipment Module (FAZ + Gateways + TLN) Commodity OD (FAZ + Gateways + TLN) by Specific Mode Conversion of Tonnage to Truck/Rail Trip Assignment Network Module Truck Network Flows by Commodity, by Truck Class, by Time of Day, and by Season < Conventional "Model> < CSFFM > 9
SUMMARY OF PROGRESS 10
Overall Progress Total funding: $1,400,000 Start: 10/11 (was to be 1/11); end: 6/13 Projected expenditure in Q1 FY13: $263,322 Projected total expenditure thru Q1 FY13: $632,050 % budget expended thru Q1 FY13 (projected): 45% (Q4 FY12: 26%) % of work complete thru Q1 FY13 (projected): approx 52% (Q4 FY12: 30%) On schedule: yes Within budget: yes 11
Task Schedule Scenario Analysis/Report Update Model 2010 Validate Model Develop/Calibrate Vehicle Network Module System Architecture and CUBE program Design Develop /Calibrate Transshipment Module Develop /Calibrate Mode Split Module Develop /Calibrate Commodity Module Seasonal Factor Enhancement Inventory Temporal Assignment Enhancement Truck Tour Enhancement 100% Completed 80-90% Completed Oct 2011 Data and System Prep Jan 2012 Jun 2012 Sep 2012 Jan 2013 On schedule Jun 2013 12
Completed Tasks thru Q4 FY12 Data Preparation (Task A.1) Collected necessary socioeconomic and demographic data for model calibration Collected necessary traffic counts for model validation Model Construction (Task A.2) Purchased/Prepared software platform and hardware Finalized the CSFFM design and structure for the module development. Continued to code each module Model Calibration (Task A.3) Coordinated with Statewide Rail Plan Network Development Continued to refine road and rail networks Built a framework for network quality control 13
Completed Tasks thru Q4 2011-12, cont. Demand Module B.1 Completed literature review Obtained IMPLAN data and extract data for testing model Formulated and coded IO-constrained direct demand model estimator Tested Model using initial network skims Transshipment Module B.2 Completed literature review Developed transshipment network module Tested Model with simple network Developed calibration algorithm for California network 14
Completed Tasks thru Q4 FY12, cont. Enhancement Module B.3 (Truck Tour Module) Completed literature review Formulated truck tour model Prepared truck data for evaluating model Tested model Enhancement Module B.4 (Inventory Temporal Assignment) Completed literature review Completed model formulation Developed test scenario Developed the parameter calibration algorithm Implemented at California FAF level (5 zones) Enhancement Module B.5 (Policy Sensitive Seasonality Factors) Continued literature review 15
Progress in Q1 FY13 Data and Model System Preparation Evaluated ATRI sample data Designed data flows and data structure connecting modules Continued to code each module Demand Module Continued to Implement/calibrate commodity model Formulated mode split model Analyzed travel costs for transportation modes Investigated seasonal factors for agricultural products 16
Progress in Q1 FY13, cont. Transshipment Model Began testing the models via freight data in California Tested Model with Air Mode at California FAF level Generated payload factors for California using California portion of VIUS 2002 Network Model Completed editing road and rail networks Started network QA/QC on CSFFM network CSFFM/CSTDM Integration and Peer Advisory Committee Meeting in July 2012 Identified action items from peer advisory committee s comments Responded to peer advisory committee s comments with tech memos 17
KEY DECISION FOR CSFFM MODULES 18
Data Preparation for commodity model Considerations for commodity Group: Available data; Homogeneity of goods; Mode share; Trip length distribution Magnitude of flow (considering weight but NOT value of shipment); Further discussion on each group commodity groups tech memo Commodity group SCTG Codes Commodity group SCTG Codes G1 Agriculture products 1-4 G9 Chemical/ pharmaceutical products 20-23 G2 Wood, printed products 26-29 G10 Nonmetal mineral products 31 G3 Crude petroleum 16 G11 Metal manufactured products 32-34 G4 Fuel and oil products 17,18,19 G12 Waste material 41 G5 Gravel/ sand and non metallic minerals 10-13 G13 Electronics 35,38 G6 Coal / metallic minerals 14-15 G14 Transportation equipment 36-37 G7 Food, beverage, tobacco products 5-9 G15 Logs 25 G8 Manufactured products 24,30,39,40,42,43 Disaggregation from FAF to FAZ Agricultural data: Farm and related information Gravel data (CA Department of Conservation) Employment and Establishment information: establishments at Zip code level 19 Manufacturing GDP: Manufacturing employment at FAZ level
Commodity Flow Model Framework Two models were estimated: 1) Total Flow Model 2) Domestic Flow Model 2) Translog Structure Equation System California FAZs Domestic flows Between FAF regions Domestic flow by destination Domestic flow by origin Export(exogenous) Total flow by origin Calibration for FAZs 1) Linear Regression models Import (exogenous) Total flow by destination 20
Import and Export Total Import/Export = Total flows domestic flows Assume that the share of foreign origin/destination by mode, exit/entry, and commodity group is the same as the base year for every FAF region. 21
Continue Shares (%) of US-Mexico freight movements by gateway will be used for disaggregating commodity OD flows from FAF -> FAZ -> Gateways. Transborder tonnage data from the BTS is used. Mexico/truck/Otay Mesa 22 Mexico/truck/San Ysidro 0 East Asia/Water/PoLA/LB 20 Total 82 23 11 Mexico/ truck/ Otay Mesa Mexico/ truck/ San Ysidro East Asia/ Water/ POLA/LB... Total 10 5 75 130 12 320 Domestic flows Between FAF regions 480 Domestic flow by destination Domestic flow by origin 101 Export(exogenous) 421 Total flow by origin 90 570 Import (exogenous) Total flow by destination Matrix O- Observe (FAF)
Seasonality Analysis for Agricultural Products Data Sources CropScape from the United States Department of Agriculture (USDA) USDA QuickStats USDA s California report Assumption Transportation seasonality is identical to production seasonality because agricultural products are transported whenever harvesting happens. Methodology Divide the year into analysis periods (days, months and/or seasons) Compute the total production by crop for each one of the proposed FAZs Calculate the given period s percentage share among total annual production Figure. Production Concentration in each FAZ 23
Percentage of Annual Production 24
Magnitude of Annual Production 25
Mode Split Module A fractional split model approach used to estimate mode shares The multinomial logit functional form at the core of the models For each of the commodity groups a mode split model is considered 12 commodity specific mode split models are estimated For commodity groups 12 (Waste) and 15 (Logs) fixed proportions are used No model for commodity group 6 (Coal/Metallic Minerals) given that there are no significant flows in CA Models estimated with FAF3 data 26
Continue 27
Input Data Issues and Assumptions Total pipeline flows are assumed to be fixed Only affects groups 3 (Crude Petroleum) and 4 (Fuels) No modal attributes are assumed for FAF s Multiple Modes & Mail mode category There are cases where some modal attribute is missing for a particular OD The missing values are imputed using equations estimated with the observed data. 28
Transshipment Module Assign the commodity flows to transshipment logistic nodes. FAZ-TLN, TLN-TLN, or TLN-FAZ OD flows for each mode will be obtained from FAZ-FAZ OD flows Applied to domestic freight movement Modes for transshipment model Air, Water and Intermodal mode(truck-rail, rail-water, watertruck) Truck only and Rail only : simply travel to FAZ to FAZ via the truck or rail network without transferring at TLNs. Commodity Module Mode Split Module Air, Water, Intermodal Truck and Rail Network Module Transshipment Module 29
Continue Input Data OD flow by commodity and mode Rate of conversion of 1 unit of commodity to 1 unit of a mode Capacity at TLNs (Transshipment Logistic Nodes) from Inverse Problem Free flow transfer time at TLNs Free flow travel time on each mode OD path (from network by mode) Obtain relative capacity changes from base year Process Run a path-based entropy-maximization model for OD estimation (i.e. Origin Destination (input) >> Origin TLN Destination (output)) Run Inverse Problem for Transshipment module for model calibration Run Forward Problem for Transshipment module for assigning demand Run the assignment problem with the updated capacity changes and OD flows 30
Implementation for Transshipment Module (Air Mode with FAF level OD flows) 31
Implementation: Air mode Calibration Results for 2007 FAF OD Capacities Transfer time 32
Continue Assumption All centroids are accessible to all airports. Assume that Transfer time is 4hours and loading/unloading time is 1hour. The change of capacity: there has not been too much change based on size of runways from NTAD data. Summary of Results 33
UPCOMING TASKS 34
Upcoming Tasks for Q2 FY13 Model Construction Start System architecture and CUBE program design Continue to code each module Model Calibration Complete network QA/QC for CSFFM network Complete coding the locations for WIM stations and observed traffic counts on CSFFM network Model Validation Develop future and freight policy scenarios further Begin to analyze the ATRI GPS data for model validation 35
Upcoming Tasks for Q2 FY13, cont. Demand Module B.1 Implemented/calibrated commodity model Implemented/calibrated mode split model Transshipment Module B.2 Continue to implement transshipment model for the other modes (water, intermodal modes) Expand FAF level to FAZ level Enhancement Module B.4 (Inventory temporal assignment) Conclude the model feasibility Prepare the model documentation Peer Advisory Committee Meeting Tentatively scheduled in October 2012 36
QUESTIONS OR COMMENTS? 37