Metrics of access-to-destinations have enjoyed increasing use in transportation planning practice over the past decade. By calculating, for example, the number of jobs the average transit rider can reach in 45 minutes under various scenarios, analysts have been able to support decision-making in a variety of contexts. In Houston, most famously, a nearly cost-neutral bus network redesign reversed a decade-long decline in ridership. In Virginia, access to jobs is one of six criteria by which the state DOT evaluates projects. Measuring access-to-destinations is especially useful in situations where it is not practical to run a full travel demand model or where an easily understood metric is needed for communication with stakeholders.

In current practice, measurements of access-to-destinations are “modally-specific”: they measure access to jobs by public transport, or by bicycle, or by car, but they do not consider how people might use one mode or another to access a given destination. This modal specificity is not a problem when, say, considering two alternative bus networks. But modal specificity makes it impossible to assess the overall impacts of a project that would affect multiple modes, such as road pricing or the conversion of a general-purpose lane to a bus-only lane. 

Some academics have begun to consider the potential of modally-general access-to-destinations metrics in theory, and forms of modally-general access calculations have long been used as a component of larger travel demand models. But to our knowledge, modally-general access-to-destination metrics have not yet been used in practice.

We are building an open-source Python tool (https://github.com/Ives-Street/connectome) to measure modally-general access. This prototype can be used in any area where GTFS, OpenStreetMap, and population income data is available, and it can be used at any spatial scale and resolution to test scenarios. We calculate origin-destination impedances for each demographic subgroup on the basis of car ownership, willingness-to-cycle, and income, and determine the most convenient mode for each O-D pair and subgroup. Together with data on the locations of destinations such as job locations and a valuation function (cumulative or gravity/decay), we can sum the value of modally-general access-to-destinations across the entire population. Our preliminary findings indicate that this approach may be predictive of modal split.

There are many open questions. In what cases might this approach be useful? How best to compare between modes for a given trip (strict minimum-impedance, or logsum)? How to balance clear communication and theoretical completeness? We look forward to dialogue and collaboration around how these methods can best serve the needs of cities and agencies.

Background:

Travel demand models typically have a few transportation supply options (e.g., drive, transit, walk/bike) whose characteristics are largely exogenous.  Today, there are more supply options (e.g., shared mobility services, micromobility), whose services respond to market cues.  In some cases, they compete for shared resources (e.g., drivers and/or automated vehicles for shared mobility services).  Travel demand models struggle to account for the behavior of these modes.

Description:

System dynamics (SD) offers a rigorous approach to dealing with time-lags and feedback effects in complex systems and is ideal for gaining insight into potentially large changes to the transportation system.  SD models are aggregate models that run very fast (typically, a few seconds), and are thus a useful addition to the strategic modeling toolbox.  The travel modeling community was introduced to the qualitative side of SD in a May 2020 Zephyr webinar (which replaced a workshop at the cancelled 2020 Innovations in Travel Modeling conference). 

The presentation for MoMo focuses on the quantitative side of SD, with a stock-flow model that represents the offerings of several providers of shared mobility services, and how they might respond to market cues.  The supply side of the model includes a common pool of regional shared mobility resources (e.g., drivers and vehicles); and several service vendors with their own characteristics (fare, cost structure, target utilization). The demand side of the model includes traveler wait time for each vendor (which is a function of both vendor characteristics and the number of travelers), and a simple nested logit model that first considers traveler response to the several shared mobility vendors, and then, the competition between shared mobility, other modes, and whether or not trips are made (induced travel).

The model was calibrated using Transportation Network Company data (geographic and temporal distribution of trips, wait time, trip time and fare) from New York City.  The response to differing cost structures (e.g., those for vehicles with automated driving systems) was then tested. 

Why Application is Noteworthy: 

This endogenous model of transportation supply with multiple service providers includes (1) competition for upstream resources,  (2) competition for customers, and (3) traveler responses to the offered services.  This stock-flow model runs by itself, or it can be integrated with Department of Energy’s POLARIS agent-based model of travel supply and demand.  It provides insights to policymakers regarding traveler mobility choices, road congestion, and energy consumption, enabling testing of various policy options. 

Project Status: 

The first phase will be complete by April 2025, with the development of a quantitative system dynamics model, and an initial integration with POLARIS. 

External-internal (EI) and internal-external (IE) trips play a critical role in regional travel demand modeling, yet accurately capturing their distribution remains challenging due to limited data availability for insightful analysis. This study develops separate custom destination choice (DC) models for external stations, leveraging passive data to improve the estimation of EI/IE trip distributions. Each external station is treated as a unique travel choice agent, influenced by exogenous factors that are largely unknown but specific to that station. The use of passive data for external travel studies has become state-of-the-practice in recent years and provides significant benefits over prior methods such as roadside interviews and license plate matching techniques that produce safety and privacy concerns. However, passive data collection functions as a sample survey that is subject to sampling bias and has sparsity issues when presented in a trip table format. As a result, OD flows derived directly from passive data may be incomplete and should not be used without further processing. As an alternative, a gravity distribution model often developed using the OD flows from passive data may be used to estimate the EI/IE trip distribution. However, this approach can be questionable as the travel impedance used by the gravity model is not based on real trip origins or destinations.

To address these challenges, a DC model is estimated for each external station, with all internal traffic analysis zones (TAZs) in the study area serving as the destination choice set. Using the Houston-Galveston Area Council (H-GAC) travel demand model (TDM) as a case study, the model incorporates multiple data sources, including observed EI/IE vehicle travel patterns represented by the OD trip tables from AirSage for autos and INRIX  for medium and heavy-duty trucks, highway network impedance metrics, and zonal activity data  (i.e., Internal productions and attractions by trip purpose) from the H-GAC TDM.

Analysis of data collected from the H-GAC region confirms that EI/IE trip distributions differ significantly across external stations, reinforcing the need for station-specific models. 47 external stations were analyzed, and distinct DC models were developed for each. The model estimation results demonstrate that destination choice behavior is well explained by zonal productions and attractions, as well as highway impedance between external stations and internal zones, as indicated by the high R-squared and correlation statistics. The distinct coefficients for each DC model further highlight the importance of differentiating external stations for analyzing their unique travel patterns.

The study findings further proves that traditional methods relying on universal assumptions about external trip distributions may not fully capture the diverse context of the external stations that is evidenced by the passive data; on the other hand, the DC modeling approach improvs the accuracy and sensitivity of external travel analysis. Overall, this study contributes to the growing body of work on passive data applications in transportation planning, providing a framework for enhancing regional models and better understanding external travel behavior.

Background to Problem

Transportation planning requires tools to quantify the benefits of transit in reducing congestion, improving accessibility, and achieving sustainability goals. Travel Demand Models (TDMs) can evaluate how transit systems contribute to the economic, environmental, and social benefits of a region. 

Description of Application

This presentation examines how TDMs can be used to evaluate transit benefits through two case studies: Bay Area Rapid Transit’s (BART) Role in the Region study and Washington Metropolitan Area Transit Authority’s (WMATA) 2024 Benefits of Transit study. TDMs demonstrate that transit improves the economic viability of a region by providing access to major destinations. Combining TDM accessibility data summaries and socioeconomic data sources, the studies were able to highlight the benefits of connecting people and their desired destinations. In the Bay Area, census tracts within a half-mile of a BART station have a 13% higher job access score and generate 16% less vehicle miles traveled (VMT). The WMATA study shows that more than 70% of the DC’s jobs and businesses are within walking distance of a Metro station or bus stop. 

TDMs can model a future without transit. Without BART, regional traffic on the Bay Bridge would surge by 73% during peak periods, and weekday delays for all drivers could increase by 10-19 hours between major destinations. WMATA study shows that without transit, travel time on major corridors like I-66, I-95, or New York Avenue would double, adding 20-30 minutes per trip.

TDM and emissions models can quantify the increase in Vehicle Miles Traveled (VMT) and Greenhouse Gas (GHG) emissions without transit. Without BART, VMT will increase by 1.6 million miles and gasoline burned will increase by 35,000-70,000 gallons daily. Without WMATA, an additional 1.2 million metric tons of greenhouse gases will be generated annually, equivalent to adding 1.2 million vehicles per day. Increased VMT would also raise safety concerns. In 2023, WMATA transit was 20 times safer than driving, helping prevent nearly 30 fatalities and over 2,500 injuries from car crashes annually.

Further processing of TDM data using economic analysis software provides a deeper understanding of transit’s economic impact. Without BART, the region could lose $1.2 billion and 5,000 jobs. WMATA study estimates that annually, transit adds over $9 billion in economic growth and saves $30 million in freight costs, helping keep consumer prices lower.

TDMs can help estimate how transit can reduce regional travel and housing costs, connect people to destinations and experiences, reduce traffic congestion, improve public health, and reduce emissions.

Statement on Why Application is Noteworthy

This application demonstrates how TDMs, combined with emissions and economic models, can quantify and communicate transit’s wide-ranging benefits of the current transit system and the potential benefits of improving transit performance. They provide valuable insights to guide policy, planning, and investment in transit infrastructure.

Project Status

Both projects were completed in 2024

Availability and cost of parking are key factors influencing travel patterns, particularly mode choice between private vehicles and alternative modes in urban areas with constrained parking supply. However, incorporating parking constraints into travel demand models presents several challenges. Parking data is highly localized, varies between public and private lots, and lacks consistent cost data across all areas. Aggregating parking data at the Transportation Analysis Zone (TAZ) level can introduce errors, complicating model integration. Forecasting future parking costs also remains a challenge, as there are no widely established methodologies for projecting how prices will evolve. Asserting static parking prices effectively assumes that new developments maintain the existing ratio of available parking, but there is growing interest in reducing minimum parking requirements ratio to encourage mode shifts away from private vehicle use.

The Boston Region Metropolitan Planning Organization (MPO) maintains TDM23, a trip-based travel demand model for the Boston region, where parking supply is regulated through policies such as the City of Boston and Massport’s parking freeze. To explore the parking cost modeling and forecasting, a clustering analysis was conducted using parking costs derived from lot-specific prices and then averaged across nearby TAZs. The algorithm was trained on multiple variables, including job density, employment access by transit, industry-sector job density, access density, parking freeze designation, distance from downtown, intersection density, and land use. The strongest correlations were observed with job density and employment access by transit, leading to the grouping of TAZs into cost-based clusters. However, comparison among clusters revealed some limitations, with systematic misestimations in some neighborhoods. Misclassifications were particularly evident in the Seaport and UMass Boston areas, while TAZs along the Emerald Necklace Conservancy remained largely consistent with expectations. Aggregating parking costs at the neighborhood level helped smooth out some of these inconsistencies but lost useful granularity of dense development near transit stations.

To better understand how the assertion to maintain observed parking costs in forecast scenarios might cause problems in the model, a parking demand estimation utility was developed that post-processes the estimated vehicle arrivals and departures by trip purpose and time of day. Parking demand estimates were tested with parking freeze supply constraints and used in a model application for the Allston Multimodal Project, a major infrastructure and development project, to compare additional vehicle demand with planned parking development. 

This research explores a potential methodology for developing an adjustable parking cost mechanism for forecast years, incorporating demographic, network, and development patterns. While the clustering algorithm effectively reassigns TAZs to parking cost clusters, finding a set of variables that produces distinct, non-overlapping clusters remains a challenge. A parking demand utility provides a valuable tool for validating assumptions about how adjusted costs influence travel behavior. Further refinement is needed to improve this process and enhance the integration of adjusted parking costs into travel demand models. 

FTA’s Simplified Trips on Project Software (STOPS) is a key tool within the United States for modelling and forecasting transit ridership. At a minimum, STOPS is used for submitting applications to the FTA’s Capital Improvement Grant (CIG) program. Given that requirement, many regions have invested in STOPS becoming part of their modelling toolbox, and have developed custom tools and workflows for exchanging data with their STOPS models. In some regions, this might look like a full integration with a regional travel demand model; in others, it might simply be a set of scripts for reading STOPS’s large text file results. I’d like to briefly share what I have been working on in terms of scripting tools and foster collaboration between regions on the topic of STOPS.

How does your region use STOPS? Do you have ideas about how to generalize and share something you have been working on for STOPS? Please join us!

In December 2021, the Colorado Transportation Commission approved the Greenhous Gas (GHG) Transportation Planning Standard. Under the Standard, CDOT and the five Colorado Metropolitan Planning Organizations (MPOs) are required to achieve individually set GHG reductions for three different planning horizons - 2030, 2040, and 2050. To determine compliance with the mandated reduction levels, agencies must model their existing networks and all future regionally significant capacity projects included in their planning documents (10-Year Plan, Four-Year Prioritized Plan in Non-MPO areas, and the MPO's TIPs and RTPs) using their travel demand models to generate transportation data sets for emissions analysis using EPA's Motor Vehicle Emission Simulator (MOVES). 

The Pueblo Area Council of Governments (PACOG) responded to the new requirements by first enhancing their existing travel models's mode choice and truck models, and second, by integrating a GHG mitigation scenario testing module into their travel model's existing scenario builder. With this new tool set, PACOG can rapidly evaluate performance of "what if" travel model scenarios to met mandated GHG reduction targets. Based on the MPO's context, "toggles" for four GHG reduction strategies that were deemed to be feasible and implementatble with the MPO region were included in the GHG scenario testing module:

• Increasing work from home percentages by income category
• Increasing frequency of fixed-route bus service
• Increasing transit bus travel speeds relative to auto travel speeds
• Increasing non-motorized mode share for short trips

As a "proof of concept" of the new GHG Policy Conformity Toolkit, the MPO created a 2020 baseline scenario and three initial future year scenarios, consistent with the current, adopted MPO plans for the required 2030, 2040 and 2050 planning horizons. Scenario sets for nine GHG mitigation scenarios were then created using the GHG scenario builder module, and the GHG reductions associated with each of the nine scenarios were evaluated. Through this process, the MPO was able to expedite alternatives processing and evalulation, and to quickly establish which strategies, at what application levels, and in which combinations of strategies would be most effective in achieveing the mandated GHG reduction targets.