New Mexico Statewide Travel Demand Model Evaluation Final Report

New Mexico State University Department of Civil Engineering New Mexico Statewide Travel Demand Model Evaluation Final Report Prepared for the New Mexico Department of Transportation, Santa Fe, NM Peter T Martin PhD PE Dusan Jolovic PhD July 13th, 2017 Disclaimer The views and opinions expressed in this document are solely those of the authors. The views expressed in this document do not necessarily represent the opinions of the New Mexico Department of Transportation, the State of New Mexico, or New Mexico State University. 2 Summary Project Title: New Mexico Statewide Travel Demand Model Evaluation Contracting Institution: New Mexico State University, 1780 E University Ave, Las Cruces, NM 88003, (575)-405-1127 Person Submitting Report: Peter T. Martin Report Written By: Peter T. Martin, Dusan Jolovic Report Date: June 2017 Principal Investigator: Peter T. Martin, PhD, PE, 575-646-2267, wales@nmsu.edu Administrative Officer: Alisha Giron, Director of Grants and Contracts, 575-646-1590, ogc@nmsu.edu Contract Period: November 1 2016 to August 15, 2017 Facilities and Administration Rate: 20% NMDOT Contract Manager: Jessica Griffin, AICP Statewide Planning Bureau Chief New Mexico Department of Transportation P.O. Box 1149, SB-1N Santa Fe, NM 87504 Cell: 505 231-7769 Email: Jessica.Griffin@state.nm.us 3 Table of Contents Table of Contents .................................................................................................................... 4 1. The Scope of the Report ................................................................................................... 6 2. Modeling Platforms and Data Sources .............................................................................. 7 2.1. The traditional 4-stage transportation planning model ............................................................ 7 2.2. Detailed description of modeling platforms .............................................................................. 9 2.3. Commercial and publically available data sources .................................................................. 17 3. Review of National Practices in the Statewide Modeling ................................................ 21 4. The Current New Mexico Statewide Travel Demand Model ............................................ 38 5. Discussions with Modeling Planners from similar State DOTs ......................................... 40 6. An Evaluation of the Current NMSTDM and MPO Models .............................................. 43 6.1. MPO Models........................................................................................................................... 44 6.2. Usage of the NMDOT model and services provided ................................................................ 48 6.3. Triggers for model updates ...................................................................................................... 49 6.4. Known gaps and deficiencies ................................................................................................... 50 6.5. NMSTDM Budget ..................................................................................................................... 50 6.6. Inter-State Cooperation ........................................................................................................... 51 6.7. Model Functions ...................................................................................................................... 51 6.8. Platform Choice ....................................................................................................................... 52 6.9. Data Requirements .................................................................................................................. 53 7. The Highway Performance Monitoring System ............................................................... 53 8. Final Recommendations ................................................................................................. 55 8.1. General..................................................................................................................................... 55 8.2. Platform Choice ....................................................................................................................... 56 8.3. Freight ...................................................................................................................................... 56 8.4. Potential Future Improvements .............................................................................................. 57 8.5. Costs and expenditure ............................................................................................................. 57 4 8.6. 9. Data Requirements .................................................................................................................. 58 References...................................................................................................................... 59 APPENDIX .............................................................................................................................. 64 5 1. The Scope of the Report The New Mexico Statewide Travel Demand Model (NMSTDM) has served a number of purposes since its initial development, providing support for projects around the state such as traffic projections for engineering design projects and Intermodal/Inland Port Facilities, and scenario testing project analysis for New Mexico Department of Transportation (NMDOT) Districts. The statewide model matters. A national research group forecasts that in 2017, $109 billion in goods will have been shipped to and from New Mexico, predominantly by trucks (TRIP - A National Transportation Research Group, 2017). The average driver in the Albuquerque area loses 36 hours per year, and the Santa Fe driver about 19 hours due to congestion. New Mexico’s population reached 2.1 million in 2015, which is a 15% increase since 2000. Vehicle Miles Traveled (VMT) has increased by 16% from 2000 to 2015. For the same period, New Mexico’s gross domestic product increased by 24%, compared to 27% for the US. We expect New Mexico VMT will increase by 20% by 2030 (TRIP - A National Transportation Research Group, 2017). If the NMSTDM is to continue to serve the state’s needs and be able to provide support for planned infrastructure projects, the model must be updated, expanded and enhanced. This report addresses the way we should improve the NMSTDM. Consultants have supported modeling development and operations at NMDOT since 2008, but most recent support has been by one individual. So it is time to evaluate the state model and its platform. This report evaluates the current model and recommends improvements. This report reviews national practices for State DOT travel demand models, different modeling platforms, various data sources available, the most common modeling concepts, and evaluates the NMSTDM. It concludes with specific recommendations on how the NMSTDM should be improved. 6 2. Modeling Platforms and Data Sources 2.1. The traditional 4-stage transportation planning model The traditional transportation planning model dates back to the pioneer definition of the classical 4-stage transportation planning model of the 1960s. The stages are distinct: 1. Trip Generation – the number of people and vehicles that want to move. 2. Trip Distribution – where do they come from, and where do they want to go. 3. Trip Assignment – which routes do they want to take. 4. Modal Split – how many walk, ride transit, or private car. These models operate at a regional level, representing large numbers of travelers over large areas. Today, planners refer to these traditional 4-stage models as “macro”. They remain the core of statewide transportation models. The first step “trip generation” establishes the demand for travel. There are three different approaches to demand estimation for trip generation: 1. Traditional trip based models. 2. Activity based models. 3. Tour based models. Traditional trip based models apply the individual person trip as the unit of analysis. Trip-based models are widely used in practice to support regional, sub-regional, and project-level transportation analysis and decision-making. These models can be four-step or three-step when the modal choice step is omitted. The trip generation step estimates the numbers of trips produced by and attracted to each geographical traffic analysis zone (TAZ). These zones are geographical areas, which identify land use and the attendant transportation characteristics. For example, a TAZ with several large retail “box stores” will attract shopping travel following time and network constraints. The TAZ data describes socioeconomic and demographic characteristics. The TAZ is identified by its centroid that is linked to highways, roads, and streets as connectors. A TAZ is identified as a geographically defined area where the traffic characteristics are consistent. When traffic changes, there is a need for the definition of a new TAZ. The term “granularity” means the degree of detail associated with the zones. Low granularity is associated with large area TAZs. 7 High granularity is associated with small detailed TAZs, often urban. The higher the granularity, the better the quality of the model prediction. The lower the granularity, the lower the cost of building and maintaining the model. The 2014 model has 919 TAZs, which are based on Census block groups, a geographical unit used by the US Census Bureau. It is the smallest geographical unit for which the bureau publishes sample data. The trip distribution step links production and attraction of trips. The traffic assignment step predicts the specific network facilities or routes used for each trip. The mode choice step determines the travel mode for every trip (e.g., car or transit). Activity-based models share similarities to traditional trip based models. Activities are generated, destinations for the activities are identified, travel modes are determined, and the specific network facilities or routes used for each trip are predicted. Activity-based models are more advanced than the trip based models. They represent realistic constraints of time and space and the linkages among activities and travel. They can model both individuals as well as multiperson households. These linkages enable them to show the effect of travel conditions on activity and travel choices more realistically. These models can incorporate the influence of detailed person-level and household-level attributes and the ability to produce detailed information across a broader set of performance metrics. These capabilities are possible because activity-based models work at the individual person-level rather than on zones. (Castiglione, Bradley, & Gliebe, 2014). Tour based models arrange travel into “tours.” These tours are travel events that start at one location and return to that same location. For example, a person traveling to work and returning home, is a “home-based work tour.” A tour can have two or more trips. Activity based models are tour-based, but a model does not necessarily have to be activity based to be tour based (Travel Forecasting Resource, 2017). The three approaches to demand prediction are Trip based models (“traditional”), Activity based and Tour based models. The data demands for the traditional models are less than activity models which in turn, need less data than tour based models. The state of New Mexico has a low population with few urban travel areas. Although activity and tour based models offer 8 attractive sophistications, they are unlikely to prove affordable in New Mexico because of the prohibitive cost of data acquisition. 2.2. Detailed description of modeling platforms Transportation models emerged with computers. Large main-frame computers drove the earliest models in the 1960s. Model innovation followed intense research in both Europe and North America. They evolved sometimes independently, sometimes cohesively. Today, mergers and buy-outs have brought us just four companies or vendors – two European in origin, and two North American. The vendors pedigrees reflect their development. The emergence of faster microchips led to the development of the micro simulation model (often referred to as “micro”). Here, individual vehicles and even pedestrians are modeled. Instead of the 4-stage broad approach of travel demands among vast groups, micro models simulate individual units as vehicles or platoons. Traffic Engineers simulate traffic behavior to test their designs. Recently a third model type has emerged. The mesoscopic (“meso”) model sits between the large granularity of macro models and the fine granularity of the micro models. Sometimes, micro models share data with meso models or macro models. “VISUM”, for example is a macro model that can function with its micro simulation cousin, “VISSIM.” When these macro, meso, or micro models combine, we append the suffix “hybrid”. So a “mesoscopicmicroscopic hybrid simulation” is a combination of a mid level meso model interacting with a fine-level micro simulation model. The traditional 4-step model, however, remains the basis of all of today’s tools. As granularity increased, mesoscopic and later microscopic models emerged. Often, the different models developed synchronously. Mesoscopic planning models emerged as high granularity planning tools to supplement traditional 4-step models. Microscopic traffic simulation tools (confusingly called both simulators and models) followed the need for traffic engineers to test new traffic management designs. The micro tools were merged with meso, and macro tools. So today’s transportation planning model is so much more than one big computer program. They have many working parts and a complex array of connections passing or processing data. A transportation model amounts to a system. Commonly, transportation planners and 9 transportation engineers refer to these complexities as “platforms”. A platform, therefore, can encompass a regional travel demand model, a large network traffic prediction tool, a traffic management microsimulation tool, and special routines for special features such as crash analysis, adaptive traffic control, or Port of Entry assessment. In contrast to the multiplicity of components, there are only four primary platform vendors for the US: 1. PTV’s VISUM 2. Citilabs’ Cube Suite 3. Caliper’s TransCAD 4. TSS’s Aimsun These vendors have platforms that encompass macro, meso, and micro tools. Each vendor offers a complex variety of options. The technical effectiveness of these platforms is moot. Each has been proven to be effective. So comparisons based solely on technical merits are academic. Each platform is technically sound. This means that selecting the best platform for NMDOT does not rest on the technical competence of the platforms. Instead, other factors emerge as keys to making a selection. Legacy is about institutional memory. There is a natural inertia that opposes change. Platforms vary in user friendliness. Each platform is capable of taking another platform’s data to enable cross platform data transfer. Some platforms such as Cube, TransCAD, and Aimsun focus on planning while the PTV platform has functions for other transportation professionals such as transit modeling, safety analysis, and traffic signal optimization. This evaluation seeks to address the costs of procuring, licensing, maintaining, updating, and expanding a Statewide Transportation Planning model. Licenses can be annual, or other fixed times. Maintenance costs can be fixed price, triggered by requests, based on agreed access limits, or they can be a combination of several of these factors. Sometimes, license costs are integrated into maintenance costs. Updates may be complementary or priced by the hour for custom updates. Sometimes, software vendors develop special features for one client, and then charge for others. Similarly, training costs can be hard to define. Some vendors include introductory training in their licensing and maintenance contracts. Others offer standard feepaying training programs. 10 One more key element further compounds the complexity of vendor pricing. The small group of transportation software vendors competes aggressively for business. These leads to different marketing strategies and different pricing policies. For example TSS share partial pricing information online; while PTV, Citilabs, and Caliper do not. Vendors are reluctant to quote when there is no contract to purchase under discussion. Citilabs and PTV told the authors that no pricing is possible without working out a detailed model specification with parameters such as:   zones, vehicle miles traveled (VMT), base Origin/Destination (OD) Matrices.  detailed description of existing data sets  stakeholders  consultant relationships   existing platform components already in state specification of model functions based on organizational structure, partners, and staff complement and expertise budget for data collection Further, each of the vendors , when approached, were eager to tell the authors that pricing was “highly negotiable”. Even TSS, who provided more detail than the other three vendors emphasized that there was considerable variability associated with price negotiation. In conclusion, transportation planning models are intricate software systems, or platforms. With so many components, licensing alternatives, supplementary features, and maintenance arrangements, no detailed priced tags can be presented. So although the following offers pricing and costs, the numbers will only serve as a general guide. PTV VISUM is developed by PTV, a company located in Karlsruhe, Germany with US offices in Oregon. PTV VISUM brochure and the official PTV website describes the software. VISUM is software for traffic analyses, forecasts, and GIS-based data management. It consistently models all road users and their interactions and has become a recognized standard in the field of transport planning. Transportation experts use PTV VISUM to model transport networks and travel demand, to analyze expected traffic flows, to plan public transport services, and to develop advanced transport strategies and solutions (PTV America, 2017). VISUM modeling functions: 11   four-step modeling,  demand modeling with simultaneous distribution and mode choice calculation,  tour-based demand modeling (VISEM), nested demand modeling, a computational tool to select activity based assignments VISUM has interfaces to: TransCAD and Cube. It is integrated with its microsimulation counterpart, VISSIM and can analyze environmental metrics. VISUM has sophisticated tools for modeling traffic signal control systems, for accident data (VISUM Safety), for optimization (VISUM Optima), and post processing (VISUM Data Analytics). The latest version is VISUM 16 that allows modeling simulation-based dynamic assignment (SBA). This means that individual vehicles and their interaction are simulated more realistically, allowing better representation of congestion effects and delays over time. SBA is the ideal stepping stone between strategic modeling in PTV VISUM, and the hybrid simulation in PTV VISSIM microsimulation (PTV America, 2017). PTV VISUM Safety visualizes crash data collected by the police and identifies black spots and high-risk sections. Detailed information about each individual crash allows users to find similarities and contributing factors to draw conclusions about causes; and to develop, plan and optimize effective and cost-efficient mitigation measures. PTV Optima is a model-based solution which offers precise, real-time traffic information for the entire traffic network and produces reliable forecasts for the next 60 minutes. The process is a quasi real-time VISUM model which represents each typical day for a particular traffic area. PTV VISUM applies dynamic traffic assignment to calculate the time-related traffic flows and turn rates for networks based on travel demand. It transfers this information to PTV Optima, where data is combined with PTV VISUM's data to adjust capacity, speed and volume from a base model to the current local flow and road conditions. PTV VISUM Data Analytics enable graphical data analyses without prior modeling work, through powerful visualization tools. The cost of PTV VISUM depends on the number of zones, number of time profiles, and the intersection modeling needs (conversation with Joseph Lubliner, a PTV Sales Manager). 12 Table 1 shows rudimentary pricing regime. It is presented to show the initial stage of a costing process that would entail detailed specification. The maintenance is 15% of the license cost, per year. Table 1 PTV VISUM Pricing Size 1 2 3 4 400 1,000 3,000 5,000 Number of Time Profiles 10,000 30,000 50,000 80,000 Price $5,600 $8,100 $18,500 $32,500 Number of Zones PTV VISUM Add-on Modules Size Junction Modeling 1 2 3 4 $1,600 $2,700 $3,100 $3,500 Cube is a suite of products for transportation planning, developed by Citilabs (Citilabs, 2017). According to Citilabs brochure and a discussion with Katie Brinson (Global Account Director), Cube is a modular, tightly integrated, fully featured product line for the transportation planning process. It addresses passenger demand, freight demand, microsimulation, air quality and reporting. With a strong integration with ArcGIS, the fundamental Cube modules that cover passenger demand modeling or strategic planning are Cube Base, Cube Voyager, and Cube Analyst. Cube Base is the user interface for the entire Cube platform. It provides interactive data input and analysis, GIS functionality in ArcGIS, model building and documentation, and scenario development and comparison. Users can create and test different scenarios, while the Application Manager makes it easy to present the model in a clear way. Cube Voyager is the extension for personal travel forecasting. It can model four-step, discrete choice, tour based, and activity-based models. With the right quantity and quality of data, users can model highway performance and public transportation operations with high precision. Cube Analyst is the extension developed for both static and dynamic estimating and updating of base year automobile, truck, and public transit trip tables. In a set of iterative calculations, Cube Analyst automatically determines the statistically most likely matrix for the set of input data values. Users have to supply observed travel demand data. 13 Other extensions of Cube software are Cube Avenue, Cube Cargo, Cube Land, and Cube Cluster. Cube pricing is flexible and negotiable with new clients. The price also depends on the number of licenses. The basic package is Cube Voyager, which includes Cube Base and Cube Voyager. Cube Cluster can be added for a 10% premium, which allows the model to run over multiple cores, without any limits. The standard rate for Cube Base, Voyager, and Cluster is $17,050. Due to the existing relationship between Citilabs and New Mexico’s Mid Region Council of Governments, the price could be discounted by 25% for any desired license. This is valid only for NMDOT purchases. One year of complimentary maintenance would also be provided with purchase. Additional information on Citilabs’ license structure is shown in Table 2. The stated cost relates to product cost to own the software, where NMDOT will never lose access. However, Citilabs recommends annual maintenance which provides access to software updated and technical support. The annual maintenance for Seat 1 at NMDOT would be $2,790. Thus, the estimated price for year one for a single seat of Cube Voyager is equal to $12,787.50. The license will have full capability (e.g., no technical limits applied), and can float between workstations. The license does not require a hardware key (i.e., dongle). To add Cube Cargo module, which models truck traffic, would cost additional $14,250. The cost of onsite training is $2,000 per day. An introductory course lasts for 3-5 days. These sums are preliminary and subject to large variation when a model structure is specified in detail. Table 2 License Structure of Citilabs Products Stand Alone Products SUGAR CUBE Citilabs Packages Base Voyager Analyst Avenue Land Cargo Dynasim Compact Dynasim Pro Dynasim Pro + Parking Network Editor Dynasim Dynasim Compact Pro Dynasim Pro + Parking Sugar Net Editor Packaged Products Sugar Access Cube Base x Voyager Analyst Avenue x x x x x x x x x Land Cargo Traffic x x x x x x x x x x x x Modelin g MMM Node x x x x x x x x x x x x x x x x x Compact x x x x x x x x x x Access x Virtualization/Server (+25%) Distributed Cluster (+10%) o o o o o o o o o o o o o o o o o o Discounts: 2-5 seats of same package 10%; 6+ seats of same package 25%; Annual Maintenance (tech support and updates) is 18%; Sugar Network Editor includes updates for current version TransCAD is the GIS designed platform for use by transportation professionals to store, manage, and analyze transportation data (Caliper Corporation, 2017). The software combines 14 GIS and transportation modeling capabilities in a single integrated platform. TransCAD extends the traditional GIS data model incorporating transportation data objects: transportation networks, matrices, routes systems, and linear-referenced data. It offers a complete solution for modeling commodity flows and truck movements. Freight traffic can be readily assigned to the transportation network and there are specialized assignment procedures available for rail waybill assignment. TransCAD provides:    a powerful GIS engine with special extensions for transportation, mapping, visualization, and analysis tools designed for transportation applications, application modules for routing, travel demand forecasting, public transit, logistics, site location, and territory management. A standard license costs $12,000: a single annual workstation license with technical support by phone and email is free. When an annual support contract expires, Caliper charges $1,200 for a one-year support extension. A 10-hour of prepaid consulting by phone, email, or any other webbased meeting tool costs $2,500 (Caliper Corporation, 2017). Again, these costs are preliminary. Transport Simulation Systems (TSS) Aimsun 8.2 is the latest update of Aimsun traffic modeling software, released in April 2017. TSS is the software development company based in Barcelona, Spain. Aimsun fuses travel demand modeling macroscopic functionalities and mesoscopic-microscopic hybrid simulation. The Aimsun hybrid simulator provides simultaneous micro and meso simulation. Users can model large areas while zooming into specific areas. Fourstep demand modeling was added in 2015. Transportation projects can be initiated by entering raw geographical and socioeconomic data (TSS Aimsun, 2017). If building the hybrid meso-micro model, users have to input data at a microscopic level. This may not suitable for large areas which have to be modeled at the mesoscopic level. It would be time consuming or even not viable for big meso areas. Aimsun can exchange data with the most popular CAD and GIS platforms. Aimsun can import data from VISUM, Cube and TransCAD. Aimsun has interfaces to adaptive traffic signal control simulation. (TSS Aimsun, 2017). Aimsun features Aimsun Online, which is a quasi real-time traffic management solution. 15 Table 3 shows the basic structure of TSS pricing for Aimsun in 2017. The company offers two license types: stand-alone and network. A stand-alone license uses a dedicated dongle. The software can be installed on several computers but only functions on one computer at a time. A network license is installed on a server and then shared between all users connected to the server, depending on the number of licenses bought. TSS pledges to fix bugs in all supported versions free of charge. Membership to the Aimsun forum is free for all licensed users. TSS offers the following optional services:  Software Update Subscription is free for the first year and then optionally renewable for an annual fee of 10%. It guarantees access to all major updates of Aimsun with no need   to pay update fees when a new version is released. Technical Support is available in 8-hour "packs". A single pack costs $950. Training - TSS organizes frequent courses at their offices in New York City and Portland, OR. It offers in-company training anywhere in North America. Table 3 Aimsun Pricing Aimsun has five “Editions”, each providing special features. Aimsun Small Edition: Microsimulation multi-modal intersections or small areas with total precision. This edition is limited to 20 intersections and 25 miles of lanes. Aimsun Standard Edition: A fast and accurate microsimulation for small corridors or urban centers. The Standard edition is limited to 100 intersections and 150 miles of lanes. Aimsun Professional Edition: It comes in three editions: Professional for Microscopic Simulation, Professional for Mesoscopic Simulation and Professional for Travel Demand Modeling. There is no network size limitation. 16 Aimsun Advanced Edition: This edition accesses both static (deterministic) and dynamic (stochastic) traffic assignment, mesoscopic, microscopic and hybrid mesoscopic-microscopic simulation along with OD matrix manipulation tools for large sized applications. Aimsun Expert Edition: It has all the benefits of the Advanced Edition, and travel demand modeling. The Professional, Advanced and Expert editions have no limitations on the number of intersections and miles of modeled lanes. To conclude, the four modeling platforms, their tools and routines are bewilderingly multifaceted, which accounts for the complexity of their pricing structures. The only way to define costs and tools is to begin detailed procurement discussions with vendors. However, some clear principles emerge in terms of the needs of New Mexico: 1. NMs low population and tax base constrains options. 2. Modeling sophistication entails detailed, and costly data collection. 3. Synergy with MPO and other partners is critical to NMs modeling future. 2.3. Commercial and publically available data sources The collection, management, and processing of data is a substantial factor in the cost and effectiveness of a transportation model. Travel data is available freely and for a price. In general, the cost associated with data increases with the amount of pre-processing from the provider. So publically available free data will be noisier than commercially produced data. Free data will have more extraneous elements that will have to be screened and managed. Commercial data that’s sold will follow the format based on prior user requests. This may or may not suit NMDOT needs. And these needs emanate from the specification of the Statewide Model with purpose, users, stakeholders, and TAZ’s defined. This section addresses the availability of publically available sources. The Freight Analysis Framework (FAF), produced through a partnership between the Bureau of Transportation Statistics (BTS) and the Federal Highway Administration (FHWA), integrates data from a variety of sources to create a comprehensive picture of freight movement among states and major metropolitan areas by all modes of transportation (Federal Highway Administration, 2017). With data from the 2012 Commodity Flow Survey (CFS) and 17 international trade data from the Census Bureau, FAF incorporates data from agriculture, extraction, utility, construction, service, and other sectors. FAF4 is the latest edition. The FAF4 baseline edition provides estimates for tonnage and value by regions of origin and destination, commodity type, and mode for 2012, the most recent CFS year. Data are available from FHWA. Releases of FAF4 products provided forecasts through 2045: state-to-state flows for 1997, 2002, and 2007; truck flows assigned to the highway network for 2012 and 2045; and domestic ton-miles and distance bands (Hwang, et al., 2016). However, FAF does not:   estimate flows accurately for local regions and individual routes,  address the effects of capacity limitation or forecast future capacity expansion,  estimate temporal variations in freight flows, adjust for changes in costs of transportation (Katsikides, 2016). The American Community Survey (ACS) is an ongoing survey by the U.S. Census Bureau that began in 2005 and provides data every year (US Census Bureau, 2017). It gives communities the current information they need to plan investments and services. ACS is renewed every year to provide up-to-date information about the social and economic needs of American communities between the decennial censuses. Through the ACS, agencies and educators know more about jobs and occupations, educational attainment, veterans, and whether people own or rent their homes. The National Household Travel Survey (NHTS) is the primary source of information about how people across the Nation travel. The U.S. Department of Transportation collects this information for Congress, national, state and local policymakers and transportation planners to study daily travel in the United States. It is repeated every 5 to 7 years. Questions in the survey are about household composition, travel experiences, and places travelled on a specific day. Results help us to understand how the Nation’s transportation systems work – roads, bike paths, and public transport. The survey is sponsored by FHWA (Federal Highway Administration, 2017). Transearch by Global Insights database is a source for US county-level freight movement data by commodity group and mode of transportation. A state-of-the-art US Intermodal Freight Flow database, it projects freight flows by truck, air, rail, and water for all 3,000+ US counties. Twice 18 a year, Global Insight provides detailed annual data for 38 different commodity groups. This database is the only system that links freight forecasts with detailed economic activity from Global Insight's Business Demographic database (IHS Global Insight, 2017). There are several service features:   freight flows for all 3,000+ US counties,  truck, air, rail, and water freight information,  annual data on 38 commodity groups, semi-annual forecasts to 2030. AirSage data is the largest provider of consumer locations and population movement intelligence in the U.S. Every day, AirSage captures and analyzes more than 15 billion anonymous, realtime, cellular-signal data points to identify travel patterns and transportation trends. AirSage data enables transit and air quality studies, and our understanding of commuter patterns. It enables logistics improvement and predicting cut-through traffic, and special event mitigation planning. It enables us to determine whether proposed roadways can offer improvements or create new problems. The cost of AirSage data begins at $10,000 (AirSage, 2017). Inrix is real-time, predictive and historical data from more than 300 million sources, such as commercial fleets, GPS, cell towers, mobile devices and cameras. It calculates traffic metrics at a 100 meter granularity. The traffic flow information is complemented with data for incidents, weather, and shockwaves. Inrix traffic data across the US help public agencies and consultants to monitor, measure and mange road network performance. (Inrix, 2017) American Transportation Research Institute (ATRI) has been engaged in transportation studies and operational tests since 1954. ATRI’s mission is transportation research focusing on the trucking industry’s role in a safe, efficient, and viable transportation system. (American Transportation Research Institute, 2017). ATRI is part of the American Trucking Association and is a nonprofit research organization. Its research addresses all aspects of trucking. ATRI has access to and maintains a number of unique and anonymous trucking industry data sets:    operational and economic, finance, GPS, 19   travel times and performance, safety data. ATRI collects about 1 billion truck position readings every 1 to 2 weeks. It collects, stores, processes, and analyzes data, turning it into useful information (Transportation Research Board, 2014). Surface Transportation Board (STB) Waybill sample is a stratified sample of carload waybills for all US rail traffic with more than 4,500 revenue carloads annually. The STB is an independent adjudicatory and economic-regulatory agency charged by Congress with resolving railroad rate and service disputes and reviewing proposed railroad mergers. The States use STB Waybill as a major source of information for developing state transportation plans. (Surface Transportation Board, 2017). This section has shown that vast sources of data are publically available at low cost or free. While this remains a valuable source of data for transportation modeling, even free data carries a cost. Data has to be selected, downloaded, processed, and input. 20 3. Review of National Practices in the Statewide Modeling This section reviews the modeling process of all states with statewide transportation models. By 2006, 19 states had operational travel demand models, according to the NCHRP Synthesis 358 (Horowitz, 2006). The review serves to provide a realistic picture of the state-of-the-practice in the US today. The review ranges from California to Vermont, it represents states of all sizes, densities, rural and urban. California uses a tour based travel demand model that can forecast all types of travel and interregional trips. It incorporates statewide networks for roads, rail, bus, water, and air travel. The model, developed in Cube (Citilabs, 2017) modeling software, has been updated to provide future year datasets for 2020, 2035, 2040, and 2050. Cambridge Systematics and HBA Specto Inc. updated the model in 2014 (Cambridge Systematics, Inc. & HBA Specto, 2014). Connecticut has a four-step operational model with a base year of 2010 built in the Tranplan platform. Cambridge Systematics built the model. The Tranplan model has 1,858 TAZs (1807 internal and 51 external). Horizon years for the current model are: 2018, 2020, 2025, 2030, 2035, and 2040. Connecticut Department of Transportation (CTDOT) is working on developing new population and employment forecasts and will expand the modeling years to 2045 and 2050. The model is being continuously updated, so the latest update year is 2017. CTDOT is working with Cambridge Systematics to develop a new travel demand model in Cube. The new model will have 4,029 TAZs with 1,810 TAZs in Connecticut, 1,613 TAZs in New York, 499 TAZs in Massachusetts and 107 TAZs in Rhode Island. According to Judy Raymond, Transportation Supervising Planner, CTDOT’s travel demand model provides VMT, bus and rail ridership estimates, rail station boarding, trip tables by mode (Drive Alone, Share Ride, Bus and Rail) and by purpose: Home-based-school, home-basedshopping, home-based-other, non-home-based-work and non-home-based non-work. The VMT by county and by functional classification is used as an input file into EPA’s mobile vehicle emission software, MOVES2014a for regional and project level transportation conformity purposes. The model compares changes in ridership and/or VMT and traffic patterns for corridor alternative analysis. The new model estimates rail ridership and boarding by station for a new rail line by opening year, as well as future input into the MOVES air quality emission model. The triggers for model update are: 21  Transportation regional conformity for either the regions’ long range transportation plan or Statewide Transportation Improvement Program (STIP). Connecticut acts as the MPO  for regional conformity.  New demographic data availability.  five years based upon the ACS 5-year data. Updating the model the every 10 years based on Census data, the goal is to update every Calibrating new rail station boarding – an interim update. This is done every three years. CTDOT has approximately $2 million set aside for updating the travel demand model in three phases: 1. Phase One is a four-step trip based model (in Cube). It will expand the model area west to the Hudson River in New York, north to near the Massachusetts Turnpike, and east to just west of Providence, Rhode Island. These TAZs will be geographically larger than the internal Connecticut TAZs. 2. Phase Two will have a population synthesizer. 3. Phase Three will be an activity based travel model. Model updates are part of the annual work program. However, due to limited staffing and other related tasks (air quality conformity modeling), CTDOT has been working with the University of Connecticut (UCONN) in proposing a long term data collection program. UCONN developed the statewide household travel survey which was completed in 2016. The household travel survey collected and analyzed demographic and travel behavior information from 7,500 households in Connecticut to better understand travel and mobility patterns (Connecticut Department of Transportation, 2016). No funding is in place for this program, but it is hoped that funding will be found so more surveys can be completed (rail and bus O/D surveys; Bradley Airport special generator, etc.) in a relatively timely manner in order to keep the model as updated as possible. Consultants are developing the new travel demand model with some input from CTDOT staff. Staff will test the new model when delivered and will create all future model years (2020, 2025, 2030, 2035, 2040 and soon 2045 and 2050). There are five people who also work on Air Quality modeling for project and regional transportation conformity purposes and the Congestion Mitigation Air Quality (CMAQ) Program analysis for federal funding. 22 Connecticut does not have a freight model. It is expected that after Phase Three of the new model, CTDOT will be able to add a freight component. Delaware (and Maryland East shore) models 1.2 million people and 5,300 square miles. The model can produce both Average Annual Daily Travel (AADT) forecasts and peak tourist season travel forecasts. The model has unique features:   a separate home based regional shopping trip type,  and maintenance, built-in select link analysis procedures,  the ability to code several future year scenarios into a single network file for ease of use an evacuation model, air quality post processor, a project benefit estimator. The following on the latest Delaware Department of Transportation (Del DOT) model were obtained from Mr. Michael DuRoss, an Assistant Director in the Planning Office. The model contains 2,100 regional TAZs. Del DOT applies a subarea approach by integrating tax parcel TAZs within the study areas. Routinely, Del DOT goes up to 32,000 TAZs which is the maximum limit that the Windows Operating System can accommodate. Built by WRA, the model is a five-step travel demand model, developed in Cube. The fifth step is assignment feedback through a process known as “re-skimming.” Del DOT combines trip generation and mode choice within subarea approaches, to eliminate long run times for transit cost skims. Del DOT uses trip and mode generation rates to get auto, transit, bike, and pedestrian trips for individual synthetically populated housing units. Bicycle and pedestrian modeling creates data needed for congestion mitigation and air quality funded project documentation, and scenario evaluation for bicycle and pedestrian project alignments. The base year for the model is 2015, the year when the latest update. Horizon years are 2020, 2030, 2040, and 2050. Delaware uses Cube Cargo based on FAF2, FAF3, Surface Transportation Board Waybill sample (Surface Transportation Board, 2017), and Transearch data. Cube Cargo allows modeling at county and TAZ level freight flows by commodity tons and value for truck, rail, water, and air modes. Del DOT is yet to implement a pipeline mode. The model was used to develop the statewide freight plan under Moving Ahead for Progress in the 21st Century Act (MAP-21) and is now undergoing an update to prepare a Fixing America’s Surface Transportation (FAST) Act compliant freight plan by August 2017. 23 The Del DOT performs travel demand modeling for the three MPO’s in the state. Each of three counties is an MPO with long-range plan and conformity responsibilities. To accomplish modeling workloads, the work is distributed between one FTE in Del DOT, three people at a consultant and five graduate students from the University of Delaware. Two students work on model applications and three on the data collection. The triggers for model update are long-range plan or conformity determinations. However, the MPOs update TAZ data annually, so the assignments in the model are recalibrated annually. Trip generation and trip distribution are generally recalibrated every three years. Del DOT spends about $450,000 annually: $250,000 to a consultant for model development, $100,000 to the University of Delaware for monthly multimodal travel surveys. Del DOT has a list of improvements for the next five years focused on developing a human health sub-model to better support the bicycle and pedestrian travel demand model. Del DOT generates health impact assessment performance data such as calories, health incident reduction, and VMT to person miles travelled (PMT). Florida has a four-step demand model developed in Cube with a base year of 2010. The traditional gravity model is divided into short trips under 50 miles and long distance trips over 50 miles. It provides travel forecasting over the entire state reflecting long-range demographic and socioeconomic growth. The model has 8,519 sub-county sized internal TAZs; 594 TAZs in Georgia and Alabama, that support the freight part of the model. The 185 TAZs representing the rest of the US were developed for the freight model only. Population, household, and employment data are based on 2010 parcel level data for each county from property appraisals. The model population totals by county match Bureau of Economic and Business Research (BBER) 2010 published data. The passenger and the freight models are shown in Figure 1 (Florida Department of Transportation, 2016). The freight model is built to address the following: (i) freight demand at the traffic analysis zone level, (ii) trip chaining in the supply chain, (iii) commodity representation produced and consumed by various industries, (iv) estimates of shifts in long and short haul demand resulting from transportation investments, 24 (v) provision of direct connections to pick-up and delivery trips (RSG, 2016). Figure 1 Florida Statewide Travel Demand Model (Florida Department of Transportation, 2016) Georgia has a model built in Cube and calibrated to 2010 conditions. Identified as GSTDM, it has two major components: a freight and a passenger model. Both use the four-step modeling process. It has 3,505 TAZs representing 48 states and Washington D.C. (FHWA - Travel Model Improvement Program (TMIP) , 2012). The GSTDM is designed for a variety of transportation planning studies and projects:   testing of project alternatives, preparing updates to Statewide long-range transportation plans. The model is used to assess the impact of large scale corridor improvements such as:   interstate widening,  construction of new facilities,  corridor toll system analysis, improvements to existing facilities. The model’s policy-level applications are:   freight diversion analysis between truck and rail, intercity passenger rail ridership forecast development, 25  high speed rail alternative analysis (Atkins, Atlanta, GA, 2013). Indiana predicts changes in travel patterns with its model. It responds to changes in regional developments such as:   new or wider roads,  shopping malls,  new buildings, large industries. The model addresses changes in both land use (such as more residential developments or new industrial sites); and demographics (more or less people in a specific area or more residents driving), and transportation supply (Indiana DOT, 2013). TransCAD is an Indiana Department of Transportation (INDOT) transportation modeling software package that integrates the modeling algorithm with GIS. This brings productivity and accuracy benefits. The model is used for:   systems level performance analysis,  input into other planning/modeling tools,  initial OD project studies,  new corridor alternative analysis,  long range planning analysis, project scoring and evaluation exercises (Indiana DOT, 2013). Kentucky has a model developed in TransCAD. Kentucky is rural so the modal split is not a model step, yielding a 3-step model. The base year is 2010 and is able to generate intermediate years as needed. Kentucky is using a 2016 scenario to validate the model against 2016 count and class information. The model is available upon request for use on state projects. Model categories are small urban area models, MPO urban area models, the Kentucky Statewide Traffic Model, county level models, sub area models (Kentucky Transportation Cabinet, 2014). The statewide model provides traffic growth rates, diversion analysis, and average VMT and VHT 26 savings. The main trigger for model update is US Census information and road network updates. The initial development (in 2017 dollars) cost about $2 million. Updates every three years cost $500,000. The updates are TAZs and zonal data update, trip matrices update, updates on the TAZs numbering system, network update, changes in the way the model is organized, correction to the model Geographic Information System Developer’s Kit (GISDK) script, and improvements to the modeling methodology (The Corradino Group, Inc., Parsons Brinkerhoff, Inc.; Stantec, 2012). According to Scott Thomson, Model Team Leader in Kentucky DOT, Kentucky funds the model by Federal Highway Administration appropriations and the available funding for transportation planning. The consultants perform most of the work on the model. Kentucky DOT provides network data, observed speeds, and volume counts from their Highway Information System. Today, Kentucky DOT is updating the model network. Kentucky DOT is seeking to improve freight movement within the statewide model. Kentucky’s statewide model has a base and future OD matrix that is the source of freight movement. It was created using Transearch and ATRI data. Kentucky canvas their major freight generators and modal hubs to estimate volumes and distribute trips to match (The Corradino Group, Inc., Parsons Brinkerhoff, Inc.; Stantec, 2012). Louisiana has a model that estimates auto and truck volumes on both state and non-state roadways. The model is developed in TransCAD. It is a four-step model and it consists of 2,287 TAZs. The base year is 2010 and the last update was in 2015. The model is designed to produce reliable estimates of regional and inter-regional travel within, from, and to Louisiana, while complementing the intra-regional focus of the travel demand of MPO models. The model informed the needs analysis and is used to assess the impacts of major capacity improvements and future demographic trends. (CDM Smith, 2014). The trigger for the update is the need for the statewide transportation plan update. The last update of the plan was in 2001. According to John Fu, Transportation Planning Engineer in Louisiana DOT, there is no information on the total or current cost of the model. Louisiana DOT does not have annual funding for keeping the model up to date. In the future, they plan to recalibrate the model using the latest AADT data. The model does not have a freight part. Massachusetts has a model built for estimating travel speeds, vehicle emissions, and VMT. It is based on the latest planning assumptions, demographic data, surveys, and professional judgment 27 about travel characteristics and growth. With over 60,000 network links and 3,500 TAZs, it covers Massachusetts, Rhode Island and part of New Hampshire. The network reflects Massachusetts Department of Transportation (MassDOT) Planning’s official Road Inventory files, as well as existing transit lines. The model has a base year of 2010, which has been calibrated to replicate traffic and travel patterns based on recorded traffic counts and the 2010 U.S. Census data. Future model year scenarios (up to the year 2035) have been developed from the base year model. They contain the latest socio-economic planning assumptions. The model has updates of roadway and transit networks. It incorporates regionally significant transportation projects that are reasonably expected to be completed by specific milestone years. (Massachusetts DOT, 2011). Maryland has a multi-layer model operating at the regional, statewide, and urban levels. It covers local trips, long distance trips, and both person and freight trips. The model covers Washington DC, Delaware and parts of New Jersey, Pennsylvania, Virginia and West Virginia. The statewide level has 1,588 zones, of which 1,151 zones are located within the state of Maryland. The collaboration with the state’s Metropolitan Planning Organizations (MPOs) helped model development (FHWA - Travel Model Improvement Program, 2014). The model is an essential tool meeting a broad range of transportation planning and policy analysis needs such as forecasting future levels of congestion, analyzing inter-regional traffic flow patterns, evaluating tolling and pricing strategies, assessing highway corridor improvements, and estimating green house gas emissions. (Zhang, Cirillo, Xiong, & Hetrakul, 2011). Michigan has a model that is an analytical framework for assessing transportation system performance and deficiency analyses, long range plan development, and system level project analyses. Built in TransCAD (Caliper Corporation, 2017) and GIS, it has 2,307 in-state zones and 85 outstate zones with 13,000 highway links and 8,000 nodes (Nellett, Banninga, Johnson, Witherspoon, & Whiteside, 1999). It is a three-step model (no mode split). The current model has a 2010 base year and it was last calibrated in 2014. The model forecasts 24-hour passenger vehicle traffic in 5-year increments to a horizon year of 2040 (Frankovich & Karen Faussett, 2014). A separate model for commercial trucks is under development. The most important model uses are: 28  forecasting growth rates for Highway Performance Management System (HPMS)  reporting,  and economic benefit analysis,  providing external trip information for MPO models,  calculating regional VMT and Vehicle Hours Travelled (VHT) for air quality conformity work zone diversion analysis in rural areas, corridor studies, alternatives analysis, and select link analysis in rural areas. The model was developed by a consultant in the mid-1990s and updated in-house following the 2005 Michigan travel counts household survey. It was re-calibrated in 2012 and 2014. According to Jesse Frankovich, a Statewide Model Specialist, MDOT has a two-year contract (to conclude near the end of 2018) with a consultant to completely redevelop the statewide model. There is no set schedule for model updates. The redevelopment underway is to take advantage of 2015 household survey data and other data sources. The updates will incorporate many of the recent advancements in travel demand modeling and will be able to address more types of policy questions. The new model will have a more detailed network and zone system, utilize big data to calibrate OD patterns, estimate mode shares, model four time periods, and have components for estimating summer and weekend travel as well as the impacts of automated vehicles. The new model should have about 4,500 TAZs, 2015 base year, and should be an enhanced four-step model, though only passenger vehicles and trucks will be assigned to a network. The 2005 Michigan Travel Counts household survey cost about $2.1 million and the 2015 Michigan Travel Counts III survey cost about $2.4 million. The survey data supports MPO models. The statewide model redevelopment contract has a budget of $1.23 million. Michigan Department of Transportation (MDOT) spends approximately $800,000 every four years on socioeconomic forecasts and freight data that supports the statewide model. This data supports MPO models beyond the statewide model. MDOT asks for larger funding amounts as needed for surveys and model development. MDOT has 4 full-time employees (FTE) supporting the statewide model, and all maintenance and analysis is done in-house. The new model is being entirely developed by consultants, but once completed, the model will be run and maintained in-house. 29 MDOT has a commodity flow model that disaggregates FAF data to the TAZ level. The base year for freight model is 2007. This model estimates truck flows between the statewide TAZs for 21 commodities, based on industry employment. MDOT does not use this model extensively, and does not have a working model for service vehicles. Documentation on the freight model is unavailable. The new statewide model will have a fully integrated freight model developed with Transearch and the ATRI data. Nevada uses TransCAD. According to the Mark Wooster, a Traffic Information Systems Assistant Chief in Nevada DOT, the current model is a three-step model, with a base year of 2007. It uses 2,891 TAZs with an average area of 0.65 square miles per TAZ. The forecast years are 2020, 2025, 2030, 2035, and 2040. The model covers eleven western states: Washington, Oregon, Montana, Idaho, Wyoming, Utah, Colorado, California, Arizona, New Mexico, and Nevada. As expected, the model is most detailed within Nevada. It frames travel as short distance resident trips, short distance visitor trips, long distance person trips, and short and long distance truck trips. The purpose of the model is to test new major corridors, such as the potential new Interstate I-11 corridor, and identify deficiencies on state highways and interstate facilities outside the coverage area of the current urban transportation models. The updates are based on available funding, working with other DOT projects, as changes to the state highway network occur. The cost of the model update underway is $90,000. There is no information on the total model cost. There is no annual funding for the model. The funding is obtained whenever possible and is based on need requests. A consultant team developed the whole model. The plan for the future is to update the model to TransCAD version 7. The model will update with the highway network as new highway improvements are built in the future. The model estimates truck traffic, for short or long trips. It estimates short distance truck trips by TAZ using trip generation rates based on households and employment by industry from the FHWA Quick Response Freight Manual II (QRFM2). Long distance truck trips are estimated using the FAF commodity flow data. Ohio has an integrated land-use and transportation model that was created with several components: an exogenous economic model, several spatial models, and several transportation 30 models covering residents, visitors and cargo. (Parsons Brinckerhoff, 2010). The model evaluates the impact of:    alternative roadway locations, other modes of travel such as bus and light rail, policy decisions on the transportation system. The model supports MPOs in urban areas. Ohio modelers work in coordination with MPOs. The work combines travel demand modeling with the Motor Vehicle Emissions Simulator (MOVES) transportation pollution emission model developed by the EPA, thereby delivering regional emissions analyses of long-range transportation plans (Ohio DOT, 2012). Further, the modeling provides project level air toxic emissions estimates for federal requirements and funding eligibility in support of other departmental planning activities. The following information was obtained from Mr. Gregory T. Giaimo, P.E. in the Ohio DOT Office of Statewide Planning and Research. The Ohio’s model is Cube and has 5,116 TAZs. The model has Java components with a base year of 2010. Consultants developed the model, while the data preparation (e.g., network development, data sorting) was done in house. The most important modeling services provided by the model are: 1. traffic forecasts for projects 2. project prioritization 3. long-range planning Major review and model revalidation are scheduled every 10 years. Ohio DOT has four FTEs which maintain the model – two people to work on the data preparation and two to update. The yearly cost is about $250,000 which is equivalent to two consultant staff. The annual funding stream is established as part of biennial budgeting Federal State Planning and Research process. Future plans for model improvements are:    increase number of TAZs to 20,000, replace FAF3 with FAF4, replace the short distance commercial vehicle model with a more explicit distribution and a warehouse layer, 31  replace the short distance passenger transport model with the Coordinated Travel – Regional Activity Modeling Platform (CT-RAMP2) model, used in major Ohio’s MPO areas. The CTRAMP models are characterized by a number of features:   a full simulation of travel decisions for discrete households and persons,  generation of travel, explicit tracking of time in half-hourly increments and use of time constraints on the explicitly modeled intra-household interactions across a range of activity and travel dimensions. These important features allow for greater behavioral realism in representing the response to numerous transportation policies. The freight component consists of three main parts:    the land use model, developed in Cube, a long distance freight model, developed in Cube, a short distance freight and non-freight commercial vehicle model, developed in Java. The land use model (SEAM/SLUM) is a simple allocation. It is driven by “Impact Analysis for Planning” (IMPLAN) which is a county level econometric data and forecast tool. (IMPLAN, 2017). The passenger model uses the outputs of this model. The long distance freight model (ACOM) takes information from the land use model along with FAF data and forecasts at the FAF zone level and has TAZ level employment. SEAM/SLUM and ACOM are packaged with the assignment steps in a standalone freight model in Cube. It is easier to distribute since it does not have the Java components. The short distance freight and non-freight commercial vehicle model is a tour based microsimulation model based on establishment surveys the Ohio DOT collects. Texas Statewide Analysis Model Version 3.0 (SAM-V3) is a state-of-the-practice multimodal travel model that provides:    highway traffic forecasts for both highway passenger travel and freight transport, intercity and high speed passenger rail ridership, freight rail tonnage and train forecasts, 32  forecasts of air passenger travel to and from Texas airports. Developed in 2013, it runs in TransCAD. The SAM-V3 provides travel forecasts at a level of detail suitable for use in comparative analyses of large-scale transportation corridor projects and other large-scale investments. The model can analyze transportation outcomes and economic impacts of state level transportation, land use, and economic policy decisions and strategies. The demographics in the model have been updated for the year of 2010. The latest additions to the model are:   enhanced multimodal passenger analysis capabilities,  (Transportation Research Board, 2010),  trip assignment by time of day,  details in the commodity based freight models, estimated roadway capacities based upon the Highway Capacity Manual  incorporation of intersection delay,  feedback of congested travel time from assignment to distribution, increased flexibility and expandability. (Alliance Transportation Group, 2013). Texas followed three major principles during the development of this model, to more easily and effectively distribute the model to users: (i) only data sources that can be freely distributed are incorporated in the model; (ii) model years, key field names and the file locations are not “hard coded” in the model, to enable easy modification; (iii) the model is designed with future updates in mind and with a simple user accessible interface. The model can deliver forecasts of years 2020, 2030, and 2040, by not “hard coding” the year in the model scripts. The model will enable assessing:   high speed intercity passenger rail,  non-traditional freight modes and truck policies,  toll diversion, port improvements, 33    new international bridges, changes in land use and development patterns, rail emissions, and freight rail realignment changes in international trade patterns (Texas Department of Transportation, 2013). Oregon has an integrated transport-land use model. Oregon has been operating its second generation of the model in PTV VISUM since 2008. It represents the elements of the land use, economy, and transport system in the State of Oregon using a set of connected modules that cover different components of the full system. The current model is a disaggregated and complex customized framework that combines a PECAS spatial allocation model with activity-based micro-simulation transport models (Parsons Brinckerhoff & HBA Specto Incorporated & EcoNorthwest, 2010). PECAS is production, exchange and consumption allocation system for urban and regional modeling to support transportation and economic planning (Waddell, 2011). Vermont built their four-step model in TransCAD. It has 966 TAZs. The base year for the model is 2010. The original model was developed in the 1990s. At that time, the model processes were run in the SAS Model Manager 2000 platform, and the network was in the Tranplan software format. According to the James Sullivan, the Research Projects Director from the University of Vermont Transportation Research Center, the model has many functions, it:   analyzes construction effects,  estimates travel demand to and from New England census urban areas,  estimates travel demand between towns,  models emergency scenarios, estimates transportation energy. The model calculates performance measures for resiliency modeling. Validation is a test of a model’s faithful reproduction. Known travel metrics are tested and compared to model outputs. When a model projects metrics that are close to actual, validation is good. The usual practice is for modelers to inspect validation findings and make a qualitative assessment. Resiliency modeling is a process where model validation is automatically quantified. Here, validation generates a set of performance measures that deliver a score. 34 The triggers for model update are data availability, which is usually the release of National Household Travel Survey data. There is no information on the overall cost of the model. Vermont has a constant of $80,000 per year to fund the model. The model has been hosted at the University of Vermont Transportation Research Center for the last 10 years. Vermont DOT models trucks as a separate trip type. There is no specific freight model that uses commodity flows. According to the report from the University of Vermont Transportation Research Center, the new external-travel sub-module was built with the support of the GIS of 2010 US Census Urban Areas within 100 miles of Vermont (Sullivan & Sentoff, 2016). The Census Urban Areas Boundary files are simplified representations from the TIGER geographic database. The Census Urban Areas and Urban Census boundaries are associated with demographic data from the ACS. The geographic representation of a single-year ACS for a rural state (e.g., Vermont) is typically poor. However, ACS pooled-data can be used to obtain improved demographic, social, economic, and housing characteristics data. Although single-year ACS estimates are typically only valid for areas with populations over 65,000, the pooled 5-year data is valid for populations of almost any size. Figure 2 shows the architecture of the Vermont Statewide Model. Figure 2 Description of Vermont Statewide Model (Sullivan & Sentoff, 2016) 35 Virginia uses Cube Voyager for its statewide model, but TransCAD for its regional models. There are one statewide model and ten regional models. Regional models are maintained by MPOs. The statewide model has about 300,000 links and 100,000 nodes. It is a four-step model with 5,000 TAZs. The base year for the model is 2012, while for the latest validation the base year was 2015. Two years ago, Virginia Department of Transportation (VDOT) ran Cube Voyager. After Cube updated to version 6.4, VDOT had problems with the GIS component. Every time Cube was updated, VDOT had to update its GIS, which required time and effort for 10 different regional models. Sometimes VDOT or MPOs had to purchase more GIS licenses to be able to run Cube models. Moreover, some of the models developed in Cube had inconsistencies when transferred to the newer Cube versions. VDOT was not satisfied with Cube. Two years ago, all Virginia’s regional models were transferred to TransCAD. According to Peng Xiao, Travel Demand Modeling Manager at VDOT, TransCAD has better GIS functionality and more options for coding and network development. In the future, VDOT plans to move its statewide model to TransCAD. The uses are traffic forecasting, congestion mitigation, safety studies, and the impact of construction of big facilities on traffic. VDOT has a program called SMART SCALE, which prioritizes future projects on a 6 year timeframe. VDOT’s SMART SCALE (§33.2-214.1) is about picking the right transportation projects for funding and ensuring the best use of limited tax dollars. Transportation projects are scored based on an objective, outcome-based process that is transparent to the public and allows decision-makers to be held accountable to taxpayers. Once projects are scored and prioritized, the Commonwealth Transportation Board (CTB) selects projects for funding. Today, the model is under revision. The latest update was in September 2014. VDOT found inconsistencies in the models, so consultants are investigating. Regular model updates are every 2 years, depending on staff availability. VDOT gets $5 million for 4 years for all 11 models (one statewide and ten regional models). Approximately $2 million is spent on the statewide model. MPOs are required to revise their models every 10 years. VDOT has 4 full time modelers, and 10 modelers from two big consultants to support work on the model. Freight is part of the statewide model with two components: 36   Long distance freight, Short distance. The input to the freight model depends on data availability. VDOT purchased data from Transearch and FAF data. VDOT does not use the freight model extensively, and cannot confirm whether purchasing Transearch data proved worthwhile. Wisconsin DOT (WisDOT) uses Cube 6.4. According to the Chris Chritton, AICP, an Urban and Regional Planner, the model has 14,500 TAZs, including external TAZs. WisDOT is satisfied with its Cube platform. The consultant developed the whole model. The statewide model has been just completed, so there are no recent updates. The model’s base year is 2010 and is based on 2010 US Census data and 2009 NHTS data. Major updates are every 10 years. The updates align with the release of the US Census data. The next major update is scheduled after the release of 2020 US Census Data. When there is an urgent need for a minor update, WisDOT contracts consultants. The model provides the growth rates for non-MPO areas in the state. MPO areas have their regional models. Since the model has been completed only recently, it is not yet in service. There was a preliminary testing of the model where WisDOT conducted a feasibility tolling study. WisDOT used a beta version of the model in the tolling study. The total cost of the model since its inception is estimated at $1 million, including the cost of Transearch data. WisDOT did not survey households because all survey data were obtained through NHTS. There are 8 FTEs in the traffic forecasting group. The most important model outputs are growth rates, future AADTs, OD patterns, link analyses, and VMT. The model has an embedded freight component and is a four-step trip based freight demand model. It is based solely on Transearch data. Wisconsin does not use the FAF framework. The freight model is not used, but will be in the future. WisDOT uses macroscopic traffic models to develop travel demand model (TDM) forecasts and to conduct deterministic Highway Capacity Manual (HCM)-based traffic analyses. WisDOT will add a dynamic traffic assignment (DTA) software to its traffic analysis toolbox. DTA modeling provides a method to consider the spatial and temporal effects of congestion and roadway capacity on drivers’ route choice, mode choice and potentially the time of departure choice. Initially, WisDOT plans to utilize the DTA software to better identify diversion traffic 37 patterns and routes associated with a work zone and/or capacity expansion project. Ultimately, WisDOT is planning to DTA for several other applications: emergency response planning, incident management, capacity reduction analysis, road or lane closure assessment and alternative scenario planning. The review reveals that each state with a statewide transportation model has a different approach toward modeling. The approach depends on needs and future projections, reliability of datasets, available funding to collect data and develop the model, and the required data granularity. Some states such as Ohio have emissions analysis as one of their statewide model outputs while others like Florida focus on freight modeling. In short, a state’s population, resources, and needs influence the degree of statewide model sophistication. The review reveals one consistent theme. States strive to synergize their models with their MPO and consulting partners. 4. The Current New Mexico Statewide Travel Demand Model The NMDOT statewide model is the second generation of the State’s traffic demand model. The first generation model was completed by contractor services in 2010. The model is integrated with the five MPO models for Santa Fe, Mid Region, Farmington, Las Cruces, and El Paso, as well as small city models for Hobbs, Roswell, Gallup, Carlsbad, Deming and Española. The City of El Paso is modeled as one external zone with data from 2011-2012. El Paso MPO is updating their model. Some of these local and MPO models are five to eight years old. Most of the MPO and city models are built in PTV VISUM. The exception is the City of Albuquerque, which runs its model in Cube. MPOs are designated forums for cooperative decision making in metropolitan areas with populations greater than 50,000. NMDOT contracts with El Paso MPO for transportation planning in southern Dona Ana and Otero counties (NMDOT, 2015). The El Paso and MidRegion MPOs are designated as Transportation Management Areas. This means they receive direct federal allocations of Surface Transportation Program and in the case of El Paso MPO, Congestion Mitigation and Air Quality Improvement Program funds. The NM statewide model has 919 traffic analysis zones (TAZs) including external zones. These are built for all highways entering NM from bordering states (e.g., US62 in Texas). NMDOT has some of their ports of entry (POE) modeled. The Santa Teresa facility is partially modeled and 38 needs more modeling. The NMSTDM models the operation of the Interstates, US routes and State routes. Local roads, which are the majority of lane-miles in most roadway networks for New Mexico, are not part of the model. The rail corridors are not part of the model with the exception of the Burlington Northern Santa Fe Railroad Refueling Station (BNSF), which is a freight generation point. NMDOT’s Statewide Planning Bureau oversees the development of the statewide model, which is completed by consultants, and utilizes the model in-house to provide forecasts, data or documentation for project design, corridor studies, and long range planning. For example, NMDOT staff and consultants use forecast data provided by NMDOT from the NMSTDM for 20-year design horizons for pavement design. 39 5. Discussions with Modeling Planners from similar State DOTs A select group of states are of special interest to NMDOT. Colorado, Arizona, and Texas are bordering states. Utah and Wyoming are geographically large western states with small urban centers. These two states are similar to New Mexico because they are expansive with small populations. The state of Wyoming has no statewide model and is not discussed. The state of Utah, in contrast, has a small but energetic team of statewide planners. Colorado Department of Transportation (CDOT): The discussion with Eric Sabina, the manager of the Information Management Branch at CDOT, revealed several ideas. Instead of a traditional four-step demand model, CDOT decided to develop an Activity Based Model. CDOT uses TransCAD developed by the Caliper Corporation. The software is under construction and will be updated through in-house resources. CDOT has its own programmers who developed parts of the software for their model. Their model is expected to be operational sometime in 2017. Once the model is completed, CDOT will provide modeling services as generated by stakeholders. Once operational, the triggers for the updates will be new releases of the US Census data, and the collection of new socioeconomic data in Colorado. The cost for the whole project is estimated at about $1.2 million dollars. Consultants did 50% of the work. The rest was managed by CDOT planning officers. Future modeling plans will follow completion of the model. Once the model is operational, future improvements will flow from requests and suggestions from users. Arizona DOT (AzDOT): Keith Killough, Director of Transportation Analysis at AzDOT revealed that AzDOT and Arizona MPOs run their models in TransCAD. It is their third generation, and is a three-step model. The fourth step will address transit. Consultants used Java predominantly, to build the original modeling software. Subsequently, AzDOTs in-house planners and engineers used GIST, which is similar to C++. GIST is a data structure that can be used to build a variety of disk-based search trees. GIST is an example of software extensibility in the context of database systems. Today AzDOT modifies the code in GIST and is incorporating the FAF4 standards and public transit. AzDOT will use Level Of Service to estimate transit performance. The model provides the following services:  toll revenue assessments, 40   performance measures,  corridor assessment,  long-range & short-range planning, support for consultants and eight MPOs. AzDOT shares the model with agencies and consultants. AzDOT shares its network and zonal files, but never the core code. When consultants and MPOs run their own scenarios, they report errors and inconsistencies to AzDOT. This helps the state to make continuous model improvements. AzDOT cooperates with FHWA’s Travel Model Improvement Program (TMIP). In 2011 and 2014, TMIP provided useful feedback on the state of their model. There are no specific triggers for model updates – rather, the model is subject to continuous improvement. The total cost to develop the four generations model was about $1 million. Much of the development work was through in-house staff, that proved cost effective. The first generation of the model was little more than a sketch. The second generation of the model cost about $250,000, and the third about $300,000. The estimate for the fourth generation model is around $150,000. The fourth generation model will cover the FAF4 update and the cost of the NHTS data. AzDOT, had 4 in-house staff to develop their model. The process took several years. Beginning with 1,000 TAZs the team brought the number to 6,000. AzDOT estimate the cost would have been $75,000 had consultants been hired. The model works at the level of state roads, but sometimes drills down to collectors when integrating with MPO models. AzDOT applies the FAF procedures for freight modeling. AzDOT bought Transearch for $500,000 but abandoned it because the data could not be applied to validate the model. AzDOT uses commodity based modeling because it is a more realistic representation of freight flows, although it models only trucks carrying no loads. In the view of AzDOT, modeling freight and long distance trips (50+ miles) is specialized and is not easy. AzDOT collects all data necessary for model development, which is sometimes supplemented by USDOT data. ACS supplies demographic data (US Census Bureau, 2017). AzDOT would like to embrace upcoming technologies such as, driverless cars and self-driving pods, in their model. The challenge to including these in a model is in identifying and collecting meaningful future attributes such as capacity, speed, economics, and reliability. 41 There is no linking with bordering states. The greatest potential for linking is with large urban areas such as Las Vegas and El Paso. AzDOT recommendations for NMDOT are:    NMDOT should “keep it simple,” consultants should be highly experienced, NMDOT must have in-house staff to supervise their consultants. Utah DOT (UDOT) began building its statewide model in 2007. Walt Steinvorth, the Planning Manager at UDOT revealed that the initial model cost was about $300,000-$400,000, and $150,000 every year for annual updates. UDOT prefers continuous annually funded updates. UDOT uses Cube software and are happy with the tool. UDOT does some simple in-house modeling, but leave the complex parts to their consultant: programming, calibration, and validation. The model is used for all types of planning, environmental impacts, and forecasts for state agencies whenever required. The triggers for the updates are new US Census information and MPO updates. When an MPO updates its model, UDOT incorporates the updates into the statewide model. In most of the cases, MPOs and UDOT work together to decide on necessary model updates. The last major update was in 2014 (mostly new socioeconomic data for long range planning), but minor updates are on a yearly basis. Since the inception of UDOT’s statewide modeling efforts, the estimated total cost of the model is over a million dollars. UDOT has three in-house staff that maintains the model and call on their consultant for fixes and updates. The consultants do all the development and major updates. Utah DOT does not have any major updates planned in the near future. UDOT stated that it would be hard to obtain substantially more funding to do some major changes in the model. Texas DOT (TXDOT): Geena Maskey is a planner at TxDOT, and via email she provided the following information about the Texas statewide model. The Texas statewide model SAM-V3 is developed in the TransCAD platform and has 4,667 TAZs. The developer is the Alliance Transportation Group. The model supports passenger and freight modeling, statewide planning, the Texas Freight Mobility Plan and feasibility studies. There is no specific trigger for model 42 updates, but the model is typically updated every five years. SAM-V3’s base year is 2010. TxDOT is working on the new SAM-V4 model. The base year for the new model is 2015. The estimated cost of the Texas model is approximately $600,000 for SAM-V3 and $450,000 for the upcoming SAM-V4 model. TxDOT has one full time planner that manages the statewide model maintenance service contract with a consultant. The new model, SAM-V4 is scheduled to be ready approximately mid -2018. 6. An Evaluation of the Current NMSTDM and MPO Models The evaluation following reviews current model documentation, operating uses and procedures, calibration and validation requirements, and any gaps or deficiencies. The analysis is based on discussions with Brian Degani, Engineering Coordinator, Statewide Planning Bureau of the Asset Management and Planning Division and from documents (NMDOT, 2009), (Eco Resource Management Systems Inc., 2014), (NMDOT, 2015). The analysis gives a model description, the current purpose and use of the model, and its gaps and deficiencies. The core purpose of the NMDOT model is to:   support policy and planning for freight, tolling and long range planning,  support development and application of performance measures,  assess intercity transportation projects,  explore the potential for congestion pricing, augment external trip making estimates for MPO models. Mode choice is not modeled, which makes the current model a three-step travel demand model. With a base year of 2006, the model applies:    socioeconomic data (housing, employment, etc.), population and urban area census data, existing and proposed geometric configurations of roadways and the other trip producers and attractors. The model was developed in PTV VISUM software. The core model is the simple model for passenger vehicles and freight. The last minor updates (including the current population forecasts) were implemented in 2014, with VISUM 14 as the latest software version. These 43 updates are in coordination with NMDOT’s consultants Cambridge Systematics and the Alliance Transportation Group. NMDOT’s model consists of TAZs from five MPOs: Farmington, Mid Region, Santa Fe, Mesilla Valley and El Paso. In total, the model has 33 counties in NM and El Paso County, Texas NMDOT proposes expanding the model to accommodate rapidly growing areas, such as the active oil and gas regions in the northwest and southeast, with the number of TAZs growing from 919 to 4,000 (Brian Degani). Further study could refine the assessment of the expansion of TAZs in terms of costs and benefits. Expanding the number of TAZs could accommodate the impacts from extractive industries in the three NW and eight SE counties. The FAF is part of the NMDOT model disaggregated into several zones with one zone for each county. It is based on FAF3 procedures and local trip generation. The latest available procedures are from the FAF 4th generation. NMDOT has an intercity transit component in its model - Rail Runner Express. The NM 2040 Plan (NMDOT, 2015) proposes incorporating city transit services provided in Albuquerque and Las Cruces into the NMSTDM. The model incorporates POEs (e.g., Santa Teresa, Columbus, and Antelope Wells), Santa Teresa Sorting Facility, and several intermodal nodes (e.g., truck to rail, rail to truck). POEs need additional modeling details. 6.1. MPO Models According to Sonia Perez, Regional Transportation Planner/Travel Forecasting Lead, El Paso MPO (EPMPO) developed their model in TransCAD. It is a four-step model with 869 TAZs. The base year is 2012. The model is a result of 90% in-house development. For congestion performance, the model outputs volume to capacity ratio. For emission analyses, the model outputs VMT. Triggers for model updates are conformity related projects and new funding availability. EPMPO spent the large sum of $600,000 to calibrate and validate its model. Travel surveys and field data (e.g., traffic counts) cost a further $500,000. The next travel demand model is planned in 5 years, with a similar cost estimate. The annual funding for maintaining and operating the model is $300,000. 44 The consultant develops the base year model and calibrates the model. The consultant develops a dynamic traffic assignment sub model. MPO staff code the forecast years and run alternative scenarios. The MPO is developing a model update for the 2045 long range plan. EPMPO is satisfied with TransCAD for the purpose of regional modeling. They were looking into using PTV VISUM/VISSIM for micro-simulation but were not able to import the network satisfactorily. They are now looking into using TransModeler (i.e., Caliper’s microsimulation platform) for micro-simulation. EPMPO shares the model with TxDOT. According to Derrick Garcia from Farmington MPO (FMPO), their model is built in PTV VISUM. FMPO contracted a consultant to build and update the model. The updates are performed every five years. The next update is scheduled for 2018. Mr. Garcia maintains the model. Horizon years are 2025 and 2040. The most important services the model provides are build/no build scenario comparisons. The annual software maintenance is about $900. The cost to build a model was $25,000. The forthcoming update will cost substantially less. FMPO has one full time consultant employee to support the model update when necessary. The model has 295 TAZs. It is a three-step model with a 2010 base year. The last update was in 2013. Mr. Garcia is completely satisfied with the VISUM platform. The model assists in the development and evaluation of future transportation improvement projects. It forecasts traffic volumes for roadways within the Farmington area. The latest update and modification takes advantage of external data from the NMSTDM, and refinements of the data along with improved modeling techniques (Eco Resource Management Systems Inc., 2014). Mesilla Valley MPO’s model consists of 235 TAZs, and is a four-step model. The base year is 2015. The model was calibrated in March 2017 and will be updated by the end of June 2017. Mesilla Valley MPO (MVMPO) measures VMT, volume to capacity ratio, and level of service for Mesilla Valley roadways using the VISUM model. The parameters for the VISUM model were developed in coordination with NMDOT, other MPOs in New Mexico, and the EPMPO. The travel behavior parameters in the model are based on the 2001 Las Cruces Household Travel Survey. The VISUM model uses a schematic of major roadways and land uses to predict travel. The network has generalized local roadways to offer access points into the system. The land uses are generalized and located in TAZs. Each TAZ in VISUM is populated with housing and jobs. 45 The model is calibrated to historic traffic counts from the MPO (Transport 2040 Metropolitan Transportation Plan Update Mesilla Valley Metropolitan Planning Organization, 2015). According to Tom Murphy, MVMPO Officer, the model is for general traffic forecasting for the horizon year of 2040. It is one element used to update the MPO’s Metropolitan Transportation Plan (MTP), which is updated every five years. MPO staff initiates the update two years before its adoption by the MPO’s Policy Committee. The model update is triggered by the requirement of the update of the MTP and by the significant change in demographics. There is no total cost estimate. MVMPO pay $3,000 for their annual license. Recently, MVMPO contracted consultants for $62,000 for model updates. The MPO devotes a certain portion of time for model updates in its Unified Planning Work Program. Tom Murphy and Michael McAdams are involved in the model updates and special modeling efforts. Tom Murphy is the primary modeler. MVMPO has employed a consultant for the calibration of the model. The MPO staff is responsible for the updating of demographics and for monitoring of the process. The staff has been involved in all aspects of model’s update and calibration. MVMPO will provide modeling for any upcoming projects. The MPO is presently providing modeling assistance for a major NMDOT project. In the future, the model could be used in conjunction with traffic impact analysis. Mid Region MPO (MRMPO) model is built in Cube. It is a four-step model with 907 TAZs, 14,000 links and 5,700 nodes. The base year is 2012 (AECOM and ASU, 2015). Forthcoming base year will be 2016, with 2040 projecting horizon. Currently, MRMPO has three seats with Cube Base, Voyager, Analyst, and Avenue licenses. Cube Analyst and Avenue modules are not fully implemented yet. There are Python components that complement the model. For example, the destination choice and the mode choice are both independent Python implementations for the model. Currently, the model is undergoing significant update. The MPO’s Unified Planning Working Plan states that the model runs are conducted upon request from various agencies, for development of the Metropolitan Transportation Plan, and the Transportation Improvement Program. Updates are done periodically, to the model's socioeconomic and demographic data (e.g., decennial census), the roadway network and transit network. Additionally, coordination with NMDOT is included, to ensure alignment of inputs and outputs between MRMPO’s model and the statewide model. Beyond this, the model serves as 46 project support both for local agency projects pursuing project development, and for the development of various scenarios as part of Metropolitan Transportation Plan. The model also provides travel time inputs for the MPO’s land use model. The total cost since the model inception is unknown. However, the current validation of the model FY17 is $71,000. MRMPO has both annual funding and one-time funding to support the model. Larger one-time amounts are required when significant model updates are underway. On the top of that, for maintenance purposes (e.g., updates and refinements, excluding licensing) MPO has $25,000/year. MRMPO has two FTE and relies on consultants supports. MPO has an on-call from a local consultant for various modelling support needs, and larger consultant team for the current model validation effort. Project support is consistent with the MTP/TIP, and larger scenarios are developed by staff in conjunction with the modeling team. Smaller-scale project support is conducted by consultants, with raw-model output provided to them with minimal project-level network modification. Turning movements are no longer included in model delivery, and postprocessing is typically handled by the consultants. Future plans include migration of the portions of the model to a scripting language, such as Python. The MPO does not want to rely on a proprietary platform such is Cube. The current model is too complex. The MPO staff is working on simplification of procedures, and better accessibility and usability. Emme platform is another possible option to replace Cube. Emme comes with a Python API integration that would allow transition to a Python model, which is the goal of the MPO. Keith Wilson, Senior Planner and Mark Tibbetts, MPO Officer and Program Manager provided the information on Santa Fe’s MPO model. Santa Fe MPO uses VISUM for travel demand modeling. The VISUM software has been used for over a decade, and the MPO is satisfied with the platform. Current model has 325 TAZs, 3,950 links and 1,573 nodes. It is a four-step model, with the base year of 2015. The current model has been used for the following major infrastructure projects: - NE/SE Connector Location Study, - I-25/Cerrillos Road Interchange Study, 47 - Santa Fe County Sustainable Land Development Plan, - NM599 Interchange Prioritization Study, - St Francis Drive Corridor Study, - Interstate 25 Corridor Study, - Richards Avenue Extension White Paper, and - Several Large private development Master Plans. Furthermore, the MPO have shared the model with NMDOT and obtained input information from NMDOT, such as external TAZ structure information. Santa Fe MPO does not have the resources to do the modeling work in house for the projects. The MPO provides modeling files and require the users to do the modeling work themselves. Santa Fe MPO do help on interpretation of the inputs and outputs, if requested. There are no specific triggers for the model updates. Santa Fe MPO is currently in process of creating updated model to reflect more detailed network coding and to update existing and future demographics. The update has been in process for the last 2 or 3 years. Delays in finalizing updated model have been due to MPO Staff availability, and lack of traffic count data for validation. If no further delays occur, model is expected to be fully updated by end of September 2017. Maintenance of the model is supported by consultant assistance. Since 2014, the MPO spent $79,000 for services and the model updates. The consultant has done the bulk of the work on the model. The initial cost of the software was around $13,000. The MPO pays annual maintenance fee to PTV of $1,632. The funding for the modeling contracts is planned within MPO’s Unified Planning Work Plan. The MPO programmed small amount of staff time and costs to manage the model. If the specific task is large, the MPO asks for larger amount of funds. Staffing is sporadic and it is based on workload on other tasks. When working consistently on modeling activities with consultant, workload was about 0.25FTE. MPO Staff has provided input on network checks, coding, on the demographics, and TAZ structure information. 6.2. Usage of the NMDOT model and services provided The model: 48   helps MPOs and NMDOT Districts with planning efforts (short and long range),  forecasts traffic demand volumes for passenger cars and freight travel demand volumes,  forecasts external and through traffic for the MPOs and other city models,  evaluates the impact of changes of population and employment,  enhance safety,   tests scenarios, such as evaluation of possible additional lanes to relieve congestion or tests different growth assumptions and network and facility improvements, determines freight forecasts as well as exporting model run results for GIS and presentation purposes, forecasts traffic impacts. The model provides travel demand modeling of passenger cars and freight volumes on a 2006 calibrated and validated base model year. It forecasts future networks for 2020, 2030 and 2040. 6.3. Triggers for model updates The following triggers for model updates are based on the practices of other states:   Decennial Census (every 10 years),  New socioeconomic (population and employment) datasets,  Freight Allocation Framework changes (e.g., FAF3 to FAF4),   Outdated base-year traffic counts – 10+ years Substantial changes to land use and modeling networks (interchanges and bypasses), New nodes, turns and connectors. The NMDOT model was updated in 2014 in terms of the model procedures, network and demographic datasets, but traffic counts are from 2006. The model was recalibrated in 2014 by Cambridge Systematics to improve forecast reliability. The frequency of the updates is every 5 years, but could be based upon adding new modeling forecast networks, updated BBER data, updating MPO boundaries, adding intercity transit and intermodal and inland port facilities, and updating new assignment procedures to calibrate and validate the 2006 base-year network. 49 6.4. Known gaps and deficiencies    NMDOT discovers gaps and deficiencies through observations and modeling experience as well as discussions with their modeling consultant. NMDOT considers the current model out-of-date because it is based on 2006 traffic counts. Trip generation needs to be refined to have better long distance analysis. Trip generation relying only on county-to-county flows is a weakness in the model because trip generation should be based on wider scale origin and destination matrices i.e. regionally   based instead of county based. Inconsistencies found in the 2006 FAF network required external trip approach revisions. The model needs better OD long distance truck data due to the need for freight planning and projections, freight intermodal facilities, POEs and facilities requiring a re-  evaluation/update of the FAF Freight Process.  Mexico. The freight component is based upon only one FAF TAZ which is the entire State of New Interregional transit for rail, bus and air should be updated since explicit modal share data is limited or unavailable. For this effort, it would be necessary to collect more data on  modal share in NM. There is no 2050 future year forecasting network for long range planning and MPO consistency. 6.5. NMSTDM Budget Update costs to the NMSTDM are part of the Federal Fiscal Year (FFY) 2017/2018 Planning Work Program (PWP) from federal State Planning and Research (SPR) funds. The budget for both state and federal funds for FFY2018 is $425,000. There is no budget information for 2019 because the FFY2019/2020 PWP is not yet programmed. Traditionally, NMDOT has taken a phased approach to updating the model, where Phase I provides a functioning operational model which can be used for estimates and planning applications. And Phase II incorporates more advanced options. However, limited funding raises five fundamental questions for NMDOT: 1. Should NMDOT retain use of the PTV VISUM platform, or switch to another? 50 2. What should NMDOT do about the first generation model? 3. Should NMDOT re-model though consultants, or in-house? 4. The POEs are changing travel throughout the state, so how should their impact be addressed? 5. Should NMDOT invest in activity-based models and tour-based models? 6.6. Inter-State Cooperation NMDOT uses VISUM, not TransCAD, which is used by several bordering states (Texas, Arizona, and Colorado). NMDOT can read the outputs from El Paso’s TransCAD model, but can neither use the El Paso MPO model nor supplement its own NMDOT statewide model. While Texas, Arizona, and Colorado all use TransCAD, there are no interstate connections. While planners see model sharing on their borders as worthwhile, there are no proposals to share. The three MPOs across NM have their models in VISUM. If NMDOT were to switch to TransCAD, there would be new operational challenges between New Mexico MPOs and NMDOT. It would be more complicated to transfer the data from one model to another. 6.7. Model Functions The current model has 919 TAZs. Many of these zones are large. FAF3 is in place with FAF4 proposed for the freight update. Other freight data collection options:    Airsage (GPS based), Transearch, Inrix data. Current model outputs are:    vehicle miles travelled (VMT), vehicle hours travelled (VHT),  average speeds on links,  capacity range groups,  volume to capacity (v/c) ratio, VMT/(v/c ratio). 51 NMDOT has new needs especially in terms of POEs. The current TAZs are too big. This smaller scale modeling need is in the direction of the modification of macro models toward more mesoscopic modeling. Microsimulation is not of high importance. The NM 2040 plan identifies multimodal access and connectivity (NMDOT, 2015). This suggests interchangeability between transportation modes (e.g., rail, bus, air, transit); however, there is no data available from NMDOT’s Transit Bureau or the Aviation Division to track connectivity. 6.8. Platform Choice Technically, all four platforms are strong systems, well proven, and capable of serving any state across the US. Further, platform choice cannot rest on cost comparisons. It is not possible to propose a range of scenarios and expect vendors to respond positively. Phone conversations, emails, and years of working with vendors tells us that pricing is so loose, it is simply not possible. It would be like asking a Realtor to price houses with only “a nice home suitable for a family,” without identifying age of house, neighborhood, number of bed rooms, land, etc. etc. The only way to derive comparisons of the platform costs is for NM DOT to specify the detail of the model, the tools, and the training requirements. These needs are further influenced by NM DOTs model staffing plans. Therefore, this means that the next steps for NM DOT is to: 1. define the needs of the model, its degree of granularity, and its users. 2. identify future staffing complement – number of employees, skill level, and consulting budget. 3. approach vendors with detailed plans and negotiate preliminary costs. Given that these three steps are yet to be taken, it is still possible to offer here some preliminary recommendations for platform choice. We recommend the PTV platform, as the most favorable, for the following six reasons: a) NMDOT’s MPO partners use VISUM which means there is institutional knowledge throughout the state. Intra-State cooperation overwhelms the opportunity for Inter-State cooperation, because NM has a small population, small urban areas, and limited future budgets. b) The only major state border region worthy of cooperation is the El Paso/Las Cruces area. Collaboration is already in place. 52 c) Sharing sub models with MPOs and other agencies could be powerfully synergistic, were NMDOT to follow the Arizona DOT approach already described. d) PTV products are supported by an excellent customer service organization. Unlike other vendors, PTV are particularly well equipped to help small teams with limited staff. Further, PTV have one of the better training programs. Although originating from Germany, the expansion of the US presence has been marked. e) The PTV platform is one of the more user friendly of the four platforms. f) Traffic Engineers favor PTVs microsimulation VISSIM. Leadership of NMDOT Planners would encourage future collaboration with Traffic Engineers. 6.9. Data Requirements For the updated model, the base year could be 2014 or 2016, depending upon core data from sources such as the BBER, and the US Census. If data were available for different years, NMDOT could interpolate. The existing model of 900 TAZs took a year to build. Expansion to 4,000 TAZs would likely take at least two years. The model incorporates socio-economic data: employment, personal income, households and total population. Like most states, NMDOT is interested in looking at new data sources and upcoming technologies, such as: activity based data, UBER pods, connected vehicles, connected freight, and taxi bots. These innovations, which are related to travel behavior, could be applied to corridors and sub-area analyses to test growth assumptions and facility improvements. Brian Degani from NMDOT recalled that the travel behavior surveys for only one MPO were costly: between $100,000 to $200,000. 7. The Highway Performance Monitoring System The Highway Performance Monitoring System (HPMS) was developed as a systematic measure encompassing the scope, condition, performance, use and operating characteristics of the Nation's highways. It is a national transportation information system and is reflective of all public roads. HPMS serves the needs of the States, MPOs, local agencies, and other customers in assessing highway condition, system performance, air quality trends, and future investment requirements (Federal Highway Administration, 2015). HPMS data can be incorporated into 53 statewide transportation planning models. So this section discusses the way agencies are making use of this linkage. Specifically, the section addresses how agencies apply HPMS date to calibrate and validate their statewide planning models. NMDOT has a Traffic Data Collection Section within the Data Management Bureau, that maintains 120 automatic traffic recorders (ATRs) statewide. The recorders are Automatic Vehicle Classification devices that operate every day of the year. ATRs are distributed in a way to achieve statistically valid classifications. They cover roads with various functional classifications, volume groups, and urban and rural area types. The data are augmented with short-term portable counters. These data are used in the NM HPMS, which is a federally FHWA mandated program to provide information to Congress on the nation’s roadways. The state transportation agencies are required to report VMT measures to FHWA through HPMS. In Texas, many agencies use travel demand models and their statewide model to forecast VMT. This approach combines both traffic count data with the socioeconomic data. There are limitations with this approach:   roadway coverage in urban areas is low,  limited detail of existing statewide model,  varying demographic estimates used as inputs,  low-volume roads are not well represented,  variability of seasonal patterns, uncertainty of mobility and capacity of improvement projects. The approach is suitable for forecasting VMT by functional class and area type, for different jurisdictions. But the major limitation of using travel models to estimate VMT is the time and effort required to produce estimates (Williams, Chigoy, Borowiec , & Glover , 2016). Rhode Island, among others, populates its statewide model using HPMS data. Rhode Island runs the model and compares VMT output to their original HPMS dataset (Rhode Island Division of Planning, 2016). Kentucky calibrated and validated their model based on HPMS data (Bostrom, 2002). Maryland simulates VMT with their statewide model and compares the outputs to the HPMS data. HPMS data validated the statewide model (Erdogan, et al., 2014). 54 These examples show that agencies use HPMS data to calibrate and validate their statewide models. While a statewide model can estimate VMT and the agency can report these findings to FHWA, there is still concern about the ability of the model to predict VMT accurately, especially for the smaller local roads. This means that for this approach, the data input for a statewide model has to be highly accurate and the model has to be calibrated and validated with high precision. 8. Final Recommendations This section summarizes the many recommendations and suggestions throughout the report. The general section addresses the entire modeling approach for New Mexico. The platform choice advocates for PTV software. The freight aspects are summarized with a list of recommendations. The future improvements section makes detailed suggestions on immediate model improvements and ideas for future benefits. The costs and expenditure section recommends future staffing requirements for NMDOT. The recommendations conclude with some data requirements. 8.1. General 8.1.1. The current model should be retained and updated. The current model’s base year is over a decade old. 8.1.2. The current model should incorporate the 2016 socioeconomic data. 8.1.3. Update the number of TAZs from 919 TAZs, to 4,000 TAZs to redefine the granularity of the zones to a better level, based on recommendations from Brian Degani. 8.1.4. Further TAZs will be needed in Artesia, Carlsbad, and Roswell in the southeast and the Farmington area in the northwest parts of the state, depending on the boom and bust cycles of the extractive industries located in those counties. NMDOT or a consultant should collect O-D data on roads affected by the extractive industries, and use it to further validate the statewide model. 8.1.5. The POEs are changing travel patterns throughout the state, so they should be modeled. POEs are modeled superficially and need to have 55 O-Ds to enable trip generation – trips need to be distributed and assigned. 8.1.6. Activity-based and tour-based models are an unnecessary sophistication, given NMDOTs resources and needs. 8.1.7. While state-of-the-art modeling such as “Uber” and “ConnectedCars” is attractive, the state-of-the-art in modeling needs to advance beyond its primitive stage. No such modeling is recommended given NMDOTs limited resources and needs. 8.2. Platform Choice NMDOT should keep the PTV VISUM/VISSIM platform because three of five MPOs in NM have their models in VISUM. So if NMDOT were to change its platform, the MPO models would lose compatibility. VISUM can import models from other platforms (e.g., TransCAD, EMME, Cube). VISUM can analyze crash data through its PTV VISUM Safety feature, which would draw support from NMDOT safety analysts, thereby using and contributing to model development. Intra-State collaboration (with MPOs) is more important to NMDOT than Inter-State collaboration. While neighboring states such as Texas, Arizona and Nevada operate the same TransCAD platform, there is no evidence of sharing models or even data. Budget limitations for NMDOT suggest that future Interstate collaborations would be a low priority. The need is to build a coherent NM Statewide Model, which is a substantial task. Wherever possible, Intra-State collaboration would contribute directly to the NM Statewide Model, across the state. 8.3. Freight The current model should be updated with freight data applying FAF4 because: 8.3.1. previous NMDOT freight models were based on FAF2, and FAF3, so there is NMDOT institutional experience, 8.3.2. FAF data is free, while Transearch data is not, 56 8.3.3. Transearch data may offer more details, but FAF presents more opportunities for analysis throughout the US, 8.3.4. FAF covers more commodities than Transearch (e.g., crude petroleum). (RS&H, Inc. , 2016). 8.4. Potential Future Improvements 8.4.1. Scenario development allows various scenarios to be assessed. 8.4.2. NMDOT should work with a more modern version of VISUM. 8.4.3. Were toll roads introduced to NM, they should be modeled thereby providing assessment for operations. 8.4.4. The statewide 2040 plan (NMDOT, 2015) recommends NMDOT have a statewide model with transit options. The model should have transit routes updated to test potential transit projects in the state. Then, the model could evaluate potential transit projects throughout the state such as modeling of bike and pedestrian routes and facilities. 8.4.5. As the volume of freight increases, the model should incorporate detailed assessment of the POE and intermodal facilities at Santa Teresa. 8.4.6. Highway traffic generated by air transportation should be modeled, including rail, bus, and air facilities at the Albuquerque International Airport. 8.4.7. NMDOT should consider emissions analysis capabilities, as there is a trend across DOTs to add emission components to their statewide models. 8.5. Costs and expenditure NMDOT needs two full time VISUM modelers to:    supervise external data collection, liaise with MPOs to incorporate their models into the statewide model, maintain the statewide model. NMDOT should hire consultants to collect data and support the modeling effort. 57 8.6. Data Requirements NMDOT should: 8.6.1. apply the Highway Performance Monitoring System VMT data to calibrate and validate the model. If NMDOT were to develop a highly detailed statewide model, it could estimate VMT and report to HPMS, 8.6.2. obtain travel household activities data from the National Household Travel Survey Database, but should also consider doing some travel behavior surveys, 8.6.3. hire consultants to investigate all possible data sources and their costs: AirSage, Streetlight, HERE, Inrix, TomTom, Nokia (NEXT), Strava, and Transearch, 8.6.4. look into Inrix data because it could require time intensive processing for advanced applications and analyses (WSP Parsons Brinkerhoff & David Evans and Associates & DKS, 2016), 8.6.5. consider Commodity Flow Survey 2012 data from US Census Bureau (USDOT - Department of Transportation Statistics, 2015). 58 9. References AECOM and ASU. (2015). MRCOG Travel Demand Model Update. Albuquerque, NM: Mid Region Council of Governments . AirSage. (2017, 4). AirSage Transportation Data. (AirSage, Producer) Retrieved 4 2017, from http://www.airsage.com/Industries/Transportation/ Alliance Transportation Group. (2013). Texas Multimodal Statewide Analysis Model Version 2.0. Retrieved 2016 л 12 from http://www.alliance- transportation.com/projects/texasstatewidemodel.asp American Transportation Research Institute. (2017, 4). About ATRI. (ATRI, Producer) Retrieved 4 2017, from ATRI: http://atri-online.org/ Atkins, Atlanta, GA. (2013). Development of Statewide Model - Report . Georgia Department of Transportation. Bostrom, R. (2002). County-Level Travel Demand Modeling Using TransCAD. 2002 Conference on Transportation Planning for Small and Medium Size communities, (p. 34). Caliper Corporation. (2017). TransCAD. (Caliper, Producer, & Caliper) Retrieved March 2017, from TransCAD Transportation Planning Software: http://www.caliper.com/tcovu.htm Cambridge Systematics, Inc. & HBA Specto. (2014). California Statewide Travel Demand Model, Version 2.0 - Short Distance Personal Travel Model. California Department of Transportation. Castiglione, J., Bradley, M., & Gliebe, J. (2014). Activity-Based Travel Demand Models: A Primer. Washington DC: TRB. CDM Smith. (2014). Louisiana Statewide Transportation Plan. Louisiana DOT. Citilabs. (2017). Citilabs . (Citilabs, Producer, & Citilabs) Retrieved March 2017, from CUBE Transportation and Land Use Modeling: http://www.citilabs.com/software/cube/ Connecticut Department of Transportation. (2016, March 10). CTDOT and UConn Launch Household Travel Survey. Retrieved March 05, 2017, from Connecticut Department of Transportation: http://www.ct.gov/dot/cwp/view.asp?A=1373&Q=578000 59 Eco Resource Management Systems Inc. (2014). New Mexico State Travel Demand Model Forecast Documentation 2014 Update. Santa Fe: Transportation Planning Division. Erdogan, S., Ducca, F. W., Patnam, K., Deng, Z., Yang, D., & Wang, X. (2014). Inclusion of time dependent networks in Maryland statewide transportation model. The National Center for Smart Growth Research. Federal Highway Administration. (2015). HPMS - Division Review Guidelines - 2015. Washington DC: FHWA. Federal Highway Administration. (2017). Freight Analysis Framework. (FHWA, Producer) Retrieved 3 2017, from Freight Management and Operations: https://ops.fhwa.dot.gov/freight/freight_analysis/faf/ Federal Highway Administration. (2017, 4). National Household Travel Survey. (FHWA, Producer) Retrieved 4 2017, from USDOT: https://www.nationalhouseholdtravelsurvey.com/#faq FHWA - Travel Model Improvement Program (TMIP) . (2012). Georgia Department of Transportation Statewide Travel Model - Peer Review Report. FHWA. FHWA - Travel Model Improvement Program. (2014). Maryland State Highway Administration (SHA) Maryland Statewide Travel Model Peer Review Report. Washington DC: FHWA. Florida Department of Transportation. (2016). 2010 Florida Statewide Model 6.0 - User Guide. FDOT. Frankovich, J., & Karen Faussett. (2014). MDOT Statewide Passenger Travel Demand Model. Horowitz, A. (2006). A Synthesis of Highway Practice. University of Wisconsin-MIlwaukee, Center for Urban Transportation Studies. National Cooperative Highway Research Program 358. Hwang, H.-L., Hargrove, S., Chin, S.-M., Wilson, D., Lim, H., Chen, J., . . . Davidson, D. (2016). The Freight Analysis Framework Version 4 (FAF4). Oak Rigde National Laboratory. FHWA. 60 IHS Global Insight. (2017, 4). Transearch Insight. (IHS Global Insight) Retrieved 4 2017, from Intermodal Freigh Flow Database: http://www.ihsglobalinsight.com/ProductsServices/ProductDetail838.htm IMPLAN. (2017, 4). IMPLAN. (IMPLAN) Retrieved 4 2017, from http://www.implan.com Indiana DOT. (2013). Transportation Modeling. Retrieved 2016 л 12 from http://www.in.gov/indot/3133.htm Inrix. (2017, 4). Inrix Traffic. Retrieved 4 2017, from Inrix: http://inrix.com/products/traffic/ Katsikides, N. (2016). Federal Highway Freight Data. Washington DC, US. Kentucky Transportation Cabinet. (2014). Retrieved 2016 л 12 from Traffic Demand Modeling: http://transportation.ky.gov/planning/pages/traffic-demand-modeling.aspx Massachusetts DOT. (2011). Statewide Travel Demand Model. Retrieved 2016 л 12 from https://www.massdot.state.ma.us/planning/Main/MapsDataandReports/Data/StatewideM odel.aspx Nellett, R., Banninga, G., Johnson, C., Witherspoon, L., & Whiteside, L. (1999). Michigan's Statewide Travel Demand Model. Statewide Travel Demand Forecasting (pp. 100-115). TRB. NMDOT. (2009). New Mexico Statewide Model poweroiunt. GIS Department. NMDOT. NMDOT. (2015). Planning Procedures Manual. Asset Management and Planning Division. Santa Fe, NM: New Mexico DOT. NMDOT. (2015). The New Mexico 2040 Plan . Santa Fe: NMDOT. Ohio DOT. (2012). Modeling & Forecasting. Retrieved 2016 л 12 from Statewide Planning & Research : https://www.dot.state.oh.us/Divisions/Planning/SPR/ModelForecastingUnit/Pages/default .aspx Parsons Brinckerhoff & HBA Specto Incorporated & EcoNorthwest. (2010). Oregon Statewide Integrated Model (SWIM2). Oregon Department of Transportation. Parsons Brinckerhoff. (2010). Ohio Statewide Model. Ohio DOT. 61 PTV America. (2017, 4). PTV Group. (P. Group, Producer) Retrieved 4 2017, from PTV Visum: http://vision-traffic.ptvgroup.com/en-us/products/ptv-visum/ Rhode Island Division of Planning. (2016). Rhode Island Statewide Model Update. Providence, RI: RI Statewide Planning Program. RS&H, Inc. . (2016). SWOT Analysis of TRANSEARCH and FAF Data. Florida Department of Transportation. RSG. (2016). Florida Freight Supply-Chain Intermodal Model (FreightSIM). Florida Department of Transportation. Sullivan, J., & Sentoff, K. (2016). Vermont Travel Model 2015-2016 . University of Vermont Transportation Research Center . Burlington, VT: Transportation Research Center . Surface Transportation Board. (2017, 4). Industry data - Economic Data: Waybill. Retrieved 4 2017, from Surface Transportation Board: https://www.stb.gov/stb/industry/econ_waybill.html Surface Transportation Board. (2017, 4). Industry Data - Economic Data: Waybill. Retrieved 4 2017, from Surface Transportation Board: https://www.stb.gov/stb/industry/econ_waybill.html Texas Department of Transportation. (2013). Statewide Analysis Model - Third Version (SAMV3). TxDOT. The Corradino Group, Inc., Parsons Brinkerhoff, Inc.; Stantec. (2012). Model Update - Kentucky Statewide Traffic Model. Final Report. (2015). Transport 2040 Metropolitan Transportation Plan Update Mesilla Valley Metropolitan Planning Organization. Las Cruces: Mesilla MPO. Transportation Research Board. (2010). Highway Capacity Manual 2010. Washington DC: TRB. Transportation Research Board. (2014). Developing Freight Fluidity Performance Measures. Washington DC: Transportation Research Board. Travel Forecasting Resource. (2017, 5). Tour Based Models. Retrieved 5 2017, from Travel Forecasting Resource: http://tfresource.org/Tour-based_models 62 TRIP - A National Transportation Research Group. (2017). New Mexico Transportation - By the Numbers. Washington DC: TRIP. TSS Aimsun. (2017, 4). Reasons to choose Aimsun Traffic Modelling Software. (TSS, Producer) Retrieved 5 2017, from TSS Aimsun: https://www.aimsun.com/aimsun/top-features/ US Census Bureau. (2017). American Communit Survey (ACS). (U. C. Bureau, Producer) Retrieved 4 2017, from US Census Bureau: https://www.census.gov/programssurveys/acs/about.html US Census Bureau. (2017). American Community Survey (ACS). (U. C. Bureau, Producer) Retrieved March 2017, from https://www.census.gov/programs-surveys/acs/ USDOT - Department of Transportation Statistics. (2015). 2012 Commodity Flow Survey. USDOT. Waddell, P. (2011). Spatial Input-Output Models: PECAS. University of California Berkeley, Department of City and Regional Planning. Williams, T. A., Chigoy, B., Borowiec , J., & Glover , B. (2016). Methodologies Used to Estimate and Forecast Vehicle Miles Traveled (VMT) . Texas A&M Transportation Institute . WSP Parsons Brinkerhoff & David Evans and Associates & DKS. (2016). Data Assessment and Plan - ODOT Freight Highway Bottlenecks Project. Oregon DOT. Zhang, L., Cirillo, C., Xiong, C., & Hetrakul, P. (2011). Feasibility and Benefit of Advanced Four-Step and Activity-Based Travel Demand Models for Maryland . University of Maryland. Maryland State Highway Administration. 63 APPENDIX Questions sent to MPOs: 1. Is your current modeling software a commercial product or custom designed. 2. What is it? 3. If custom designed, what proportion of the work was in-house? 4. What modeling services and activities does your model currently provide? 5. What are the most important services and activities the model provides? 6. What are the triggers for model updates? What is the frequency of updates, or are they based on necessities and available funding? 7. Can you estimate the cost of your MPO model? Do you have any estimates for planned future model extensions and updates? 8. Do you have annual funding for keeping the model up to date? 9. What is the staffing Full Time Equivalent (FTE) and consultant support required for modeling? Do consultants do all the work or is it a shared effort? 10. Do you have any plans for model expansion or model improvements? 11. What is the total number of TAZs in your model? 12. Is your model a 4-step model? 13. What are the base year for your model, and the year of the last update? 14. Are you satisfied with your modeling software? Are you considering changing? 15. Do you share your MPO model with NMDOT? Does it matter if NMDOT used software? 16. If you have any documents in the public domain, could you please send them to me? 64 Questions sent to DOTs: 1. Is your current modeling software a commercial product or custom designed. 2. What is it? 3. If custom designed, what proportion of the work was in-house? 4. What modeling services and activities does your model currently provide? 5. What are the most important services and activities your model provides for DOT and the State? 6. What are the triggers for model updates? What is the frequency of updates, or are they based on necessities and available funding? 7. Can you estimate the cost of your statewide model? Do you have any estimates for planned future model extensions and updates? 8. Do you have annual funding for keeping the model up to date? 9. What is the staffing Full Time Equivalent (FTE) and consultant support required for modeling? Do consultants do all the work or is it a shared effort? 10. Do you have any plans for model expansion or model improvements? 11. What is the total number of TAZs in your model? 12. Is your model a four-step model? 13. What are the base year for your model, and the year the last update? 14. What is your freight model? Can you elaborate on inputs and outputs and the purpose of your freight model or provide any document? New Mexico DOT is interested in freight modeling. 15. If you have any documents in the public domain, could you please send them to me? 65