Florida Multimodal Statewide Freight Model Model inputs and structure, and integration with regional truck-touring models Presented by Colin Smith, RSG

Freight Issues in Florida Florida has a large, complex, multimodal freight transportation system: 15 deep water seaports (international trade moving through Florida’s ports was valued at $56.9 billion in 2009) 2,786 miles of rail lines carried 1.6 million carloads and 83 million tons of freight in 2008 Miami International Airport first in the nation in total international freight tonnage Strategic Intermodal Systems Highways cover 4,300 of roadway and carry the majority of the state highway system’s truck traffic Issues include how to manage and invest in the system to meet statewide objectives: For example – SIS objectives: economic competitiveness, interregional connectivity; efficiency; choices; intermodal connectivity; energy, air quality, and climate; and emergency management and safety Florida does have a number of specific freight issues that are important, related to the complexity and size of its transportation system, its coastal location with a large number of ports, and its proximity to central and south America which make Miami Airport such an important air freight handling airport

Motivation for the Model Enhance Florida’s statewide freight forecasting capabilities AND Provide a framework to support enhancements to regional freight forecasting New model is multi-modal Includes updated commodity flow information More detailed framework: additional policy sensitivity Represent characteristics of firms and shipments Represent supply chains and distribution channels Capture trip-chaining that occurs New model is multi-modal Models all shipments to, from, and within Florida And movements by all modes: truck as existing model, but also now models freight movement by rail, water, and air, and connections between those all of the modes so that we can understand multi-modal trips, e.g. imports by water to truck, truck-rail-truck shipments, etc. Includes updated commodity flow information Currently using FHWA’s freight analysis framework data, updated in 2007 FDOT is purchasing updated Transearch data More detailed framework: additional policy sensitivity Represent characteristics of firms and shipments Represent supply chains and distribution channels Capture trip-chaining that occurs Model framework is design to allow us to evaluate and answer more policy questions Freight demand is based on the business needs of firms in Florida, and so we represent the locations, sizes, and industries of firms in Florida is detail Freight is moved in individual shipments and not annual commodity flows, so we model shipments specifically How shipments are moved through intermodal and distribution centers is important and so those supply chains and distribution channels are explicitly modeled Framework is also designed to connect to more detailed regional truck touring models that capture the patterns of truck deliveries in urban areas. 3

Model Structure Overview Florida Statewide Model Conversion to modal trip tables Passenger Model National Supply Chain Model (Firms, Shipments, Modes) National/Statewide Networks Regional Model Pax Model Regional Truck Touring Model Regional Networks The new statewide model is being integrated into the existing statewide model framework, alongside the passenger model. The freight demand model produces shipments by mode and origin/destination that are (for the statewide model) converted to modal trip tables and assigned to national and statewide networks. The detailed shipment information from the statewide model can also be transferred to regional models to support a regional truck touring model. The overall framework that we are transferring to Florida is based on a combined national and regional model framework implemented in Chicago for CMAP, the Chicago MPO, and FHWa, so it is explicitly designed to support regional modeling.

Model Integration: CUBE The freight model is implemented in R and Cube. Cube provides the overall structure, skimming and assignment, and gives control to R which runs the demand model components via a command line window. Cube Window This screenshot shows the model running, with Cube starting a command window to run the R script. Status messages are printed to the command window and logged in a file and R also uses a progress bar to show how far through the current model component it is. Command Line Window Progress Bar Window

Statewide Model Framework Firm Synthesis Supplier Selection Goods Demand Distribution Channel Shipment Size Mode and Transfers Network Assignment Synthesizes a list of businesses in Florida, the rest of the US, and an international sample Connects suppliers to buyers based on the commodities produced by the supplier and consumed by the buyer Distributes commodity flows amongst the paired suppliers and buyers For each buyer/supplier pair, selects whether shipments are direct or involve intermediate handling (intermodal, distribution center) For each buyer/supplier pair, converts an annual commodity flow to shipments by size and frequency Introduce how the model works: The first three components deal with firms and the interactions between them Firm synthesis creates a list of business locations in Fl, the US, and internationally. These represent the origins and destinations of freight shipments Supplier selection models the process of firms who need specific commodities as inputs to their business (buyers) select a supplier to buy that commodity from. That connection from supplier to buyer then represents a link over which commodities will flow In goods demand, we have an input of observed aggregate freight demand that is then allocated to those connections between suppliers and buyers 2. The next three steps deal with how those commodity flows between each supplier and buyer are turned into shipments and moved. - Distribution channel: models whether shipments are moved through an intermodal center or distribution center - Shipment size model divides the annual flows between each supplier and buyer into individual shipments - The mode and transfer models then estimates some key characteristics of the shipment movement: the mode for each leg of trip and the transfer locations 3. Finally, the model converts the modal shipment movements into a demand for vehicle movements (e.g. trucks, rail cars) and assigns those to the transportation networks to estimate truck volumes on roads, rail car volumes, and demand at intermodal connections. Identifies the mode for each leg of the trip from supplier to buyer and the transfer locations Assign the trips to the multi-modal networks based on the mode(s) and transfer locations 6

What is “firm synthesis”? A model that creates a complete set of “agents” that represent firms The model embodies each of those firms with individual characteristics that affect the behavior that we are interested in modeling: Where are they located? How large is the firm? What industry do they operate in? Which commodities do they consume? Which commodities do they produce? Concept is similar to population synthesis used in activity-based models In this case, households are the agents that are created They are described in terms of location, size, etc. An introduction to firm synthesis, which is the first component of the model 7

What is “supplier selection”? Model to establish business relationships between producers (suppliers) of goods and consumers (buyers) of goods Data FAF industry to commodity crosswalk, to determine type and percentage of produced commodity by each industry type Make and use tables in the 2007 Benchmark Input-Output Account, to determine type and percentage of commodities used by each industry Firms’ characteristics and amount of incoming and outgoing commodities are from the first module Rule-based system: Uses “Distance” and “Establishment size of both supplier and buyer” as the input criteria for assessing the suitability of candidate suppliers The output is the probability of partnership for each pair of supplier-buyer The firms with the highest probability of partnership will be selected as the best suppliers An introduction to Supplier Selection, which is the second component of the model

Model Geography: Firms and Commodity Flow Data Individual firms are synthesized in each traffic analysis zone (TAZ) based on employment data for each industry (County Business Patterns, InfoUSA, QCEW, and other local data) TAZs provide detailed spatial resolution, particularly in metropolitan areas, for firm locations and shipment origins and destinations Firm locations. Each point represents one business TAZ system provides detailed spatial resolution for firm locations, shipment origins and destinations, and connections to the transportation networks TAZ systems covers Florida, plus additional buffers into AL and GA (not shown on map) Commodity flow data is relatively aggregate but the model incorporates the much more detailed firm location data and this is used to disaggregate the commodity flows Across the whole of Florida, the TAZs provide a significant level of detail FAF commodity flow data, a model input, uses large FAF zones. This is disaggregated down to TAZs based on the firm allocations and economic (input/output) data 9

Freight in Florida: Commodity Flows to and from Florida If we look at the geographical distribution of domestic movements from Florida, adjoining states in the South East dominate, with important flows to the northeast and California Going back to our example of flows to the Houston area: Cereal grains is the largest movement from Florida to Houston, at 800,000 tons per year Following this there are several commodities that are actually largely imported from international origins to Florida, transhipped and then moved on to Houston: that accounts for much of the Fuel Oils, Gasoline, and Coal products Looking at the overall mode shares (for the domestic movements), truck dominates, with some multimodal and rail trips. But, once we look at the combination of commodity and mode, we see cereal grains, the top commodity, is almost entirely moved by rail or a combination of modes (e.g. truck to rail) The model is designed to capture those commodity specific mode choices 10

Freight in Florida: Freight flows to the state Now looking at flows to Florida, we see a different distribution, with energy rich states becoming important alongside our neighboring states. Looking again at our Houston example, other coal and petroleum products account for most of the flows from Houston to Florida, as would be expected given the number of refineries in that region. In this case, pipelines are important, along with trucks and rail. Again, when we look at the connection between commodities and modes: pipelines are used for most of the other coal and petroleum products, with rail being used for some of the bulk commodities and trucks for the higher value goods 11

Model Geography: Transportation Networks Model covers all of Florida and includes transportation networks across the USA and internationally Uses newest multi-modal transportation networks: highway, rail, seaports and waterways, airports, and intermodal connections Uses network information to understand transportation costs (including storage costs during transshipment), capacities, and resulting travel times Model outputs vehicle and commodity flows on networks and through intermodal/distribution centers The new model is multi-modal and as such incorporates new networks for each mode and for the connections between modes Key freight locations such as ports, airports, and truck terminals are represented A rail network connects to those locations. Movements of freight by rail is modeled based on observed data from the USDOTs Waybill data FDOT has just developed a new highway network using up to date information from NAVTEQ 12

Mode and Transfers Model Shipments are Assigned to the Following Modes: Truck (Full truck load, less than truck load) Rail (Carload, intermodal) Water (international, non-international) Air Paths for shipments with complex distribution channels, i.e. that are not direct from shipper to buyer, can use a single mode or be multimodal, e.g. Truck – Truck Truck – Rail – Truck Truck – Air – Truck Truck – Water The mode and transfer model includes all freight modes. And also combinations of modes – as a shipment may use more than one mode to get from supplier to buyer, for example, truck to rail to truck. The choice of mode is made based on the full logistics cost of the path from supplier to buyer and not just the transportation costs. Seller Buyer Truck Air Rail Water Annual logistic costs calculated for all available mode-paths for each firm pair The least cost mode-path selected for each firm pair

Sample Model Sequence #1 sequence that illustrate how the model represents shipments. Start with the a business connection between a wholesaler in Tampa and a Pharmaceutical manufacturer in Houston, TX. We don’t model spatial detail in Houston, but we do in Florida: in this case, the wholesaler is located in Southwest Tampa, close to MacDill air force base The total commodity flow between Houston and Tampa for Pharmaceuticals is 1110 tons, of which a proportion is allocated to this buyer-supplier pair based on the size of the firms: in this case 60 tons In this case we identify that the freight will be moved through one distribution center en route from supplier to buyer, located in the Tampa region In the shipment size model, we estimate a shipment size of 20,000lbs, which means 6 shipments over the course of the year to deliver the total annual flow between the supplier and buyer. This means that on any given day there is an approximately 2% chance of that shipment taking place. The shipment is moved entirely by truck (20,000lbs equates to approximately a full truck local of pharmaceuticals) 14

Policy Sensitivity Changes in land use and economy: spatial distribution of employment and mix of industries Changes in transportation supply Major highway network changes Changes to rail capacity Intermodal facility capacity changes, including deep water ports and airports Changes in distribution center network New facilities (with regional significance) Changes in commodity flow origins and destinations outside Florida Explain the policy sensitivity built in to the model and how FDOT plans to use this new tool - This slide describes the broad types of policies that the model is sensitive to 15

Connections to the Statewide Model Statewide model is designed to be integrated with regional models Statewide model develops shipment paths including transfer locations and modes, and intercity truck movements Statewide model outputs include a list of regional shipment pick-ups and deliveries that must be met in a given day: that can then be microsimulated in the regional model Possible modes of operation Integrate complete statewide model with regional model and run models sequentially Provide an extract from the statewide model of regional shipment demand and run the regional model separately The statewide model is specifically designed to be connected to the regional models, either in a remote way – via an output produced by the statewide model that is transferred and used as an input to the regional model – or in a more directly integrated manner in the same software application

Regional Model Framework Shipment Flows (from Statewide Model) Vehicle and Tour Pattern Choice Number of Tours and Stops in each Tour Stop Sequence and Stop Duration Tour Start Time and output of Trip List The regional model framework which was developed as part of the project that RSG did with FHWA and CMAP connects directly to the national scale supply chain model that is being implements in Florida. The model is somewhat analagous to the tour building part of the activity based model. Instead we are trying to connect together delivery and pick up activities at businesses as opposed to facilitating work and shopping and other personal activiities.

Truck Touring Models Chicago Region Warehouse 120 Implemented in Chicago region for CMAP as part of FHWA project Model links shipment deliveries and pick-ups together into truck tours Tours built for different truck types and for different patterns: single stop, single loops, several returns to warehouse Output is a trip schedule similar to that from an activity-based model of personal travel Delivery tour An example of how a truck tour is constructed.

Regional Policy Sensitivity Regional model input is the output from the statewide model, so statewide model’s sensitivity feeds through to regional model Local (more detailed) distribution of employment and land use affects local truck travel patterns Local changes in transportation supply, non-truck traffic growth causing congestion, and resulting truck travel times Local policies such as truck routes, truck prohibitions, delivery windows, and size limits can be modeled An explanation of regional model policy sensitivity

Questions? Colin Smith, RSG colin.smith@rsginc.com 802-295-4999 www.rsginc.com Questions?