Simulations

Main previous achievements

We focus in this section on the most critical tools in HARMONIC: traffic simulator and land use model.

Traffic simulator and interface

Over the past decades, André de Palma and his co-authors have developed and extensively used the mesoscopic transport simulator METROPOLIS (de Palma et al., 1997; de Palma and Marchal, 2002; Saifuzzaman et al. 2016). A study of pricing and vehicle emission policies was explored in Vosough et al. (2022). In recent years, the simulator has undergone significant revisions to incorporate features such as trip chaining, different vehicle types and explicit modelling of bottleneck queues (Javaudin and de Palma, 2024). The simulator METROPOLIS was used to study ride sharing in Ile-de France (de Palma, et al. 2022), and to evaluate the impact of thermic vehicle emissions on air quality in the French island of La Réunion (Le Frioux et al., 2023).

LUTI model

EUrbanSim is a land use model designed to simulate the long-term effects of various policies on urban development, residential locations of households, job locations of workers, real estate trends, and local real estate and labor markets. It considers the dynamic interactions between the decisions of individuals, households, and firms, as well as the decisions made by stakeholders and public policies. The preliminary version, UrbanSim, initially developed by P. Waddell, was adapted to the deregulated American context. We developed the European version, EUrbanSim, to integrate the imperfections in real estate and labor markets induced by rigidities, regulation and other public policies specific to the French context (see de Palma, Picard, Motamedi, 2015, Antoniou and Picard, 2015, and Picard and de Palma, 2019).

Interaction between the traffic simulator and the land use model

Integrating a dynamic traffic model with a land use model is challenging yet essential. The integration of EUrbanSim and METROPOLIS is documented in de Palma, Picard, and Motamedi (2015), de Palma, Motamedi, and Saifuzzaman (2015), and de Palma and Picard (2019). Currently, data transfer (e.g., exporting origin-destination matrices from EUrbanSim to METROPOLIS2, and travel time data from METROPOLIS2 to EUrbanSim) must be done manually. This time-consuming process requires significant data preparation before the models can interact (see 2.2.4 and 2.2.5).

Air quality

The integrated chain of models, METROTRACE suite (Le Frioux et al., 2023), computes the economic costs of population exposure to road air pollution. Using a Gaussian dispersion and exposure model, it evaluates population air pollution exposure and related costs and investigates the impact of two policies: replacing old thermic vehicles with electric vehicles, and allowing flexible departure times for commuting trips. An application is currently underway in Île-de-France.

Activity models

IPR collected data in 2023 on trips in Ile-de-France on weekdays and weekends. It provides information on the variability across the 7 days of the week, rather than relying solely on an "average day" analysis, and will serve as the foundation for modelling weekly activity patterns in 2.2.2.1.

Policy analysis

HARMONIC partners have analyzed various transport and land use policies, and conducted impact studies over the past years. The most representative studies are described in Bierlaire et al. (2015), Prager (2019), and Small et al. (2024): radial pricing, modular and cordon pricing, exploratory studies of low-emission zones in Île-de-France, introduction of electric vehicles in Île-de-France and La Réunion, impact of flexible and staggered work hours, short- and long-term consequences of public transport pricing, impacts of teleworking, and studies on the variants of Line 18 in Île-de-France. Land artificialization is currently under consideration for the Société des Grands Projets.

Family decision and mobility

The partners have extensively collaborated on family models across various domains, including residential location (Picard et al., 2015), carpooling (de Palma et al., 2022b), mode choice (Picard et al., 2018), and departure time decisions (de Palma et al., 2015a). A recent survey discussing this growing literature can be found in de Palma et al. (2024c).

State of the art

We focus here on the topics most important for HARMONIC, and refer to Small et al (2024) for more.

Synthetic populations

Most agent-based transport simulators rely on a synthetic population of individuals or households (disregarding freight), with their sociodemographic characteristics and daily mobility patterns. This approach enables more precise and disaggregated results compared to the traditional origin-destination matrices used in conventional transport simulators. Hörl and Balac (2021) introduced a methodology to generate a detailed synthetic population for any region in France, which will serve as input for our transport and land use simulators.

Car ownership - Electric and automated vehicles

Kwon and Kim (2020) examine advances in household vehicle ownership modelling over the past two decades. They focus on methodological improvements, data availability, and the integration of new factors like environmental concerns and shared mobility. Zhao and Chen (2020) emphasize the recent shift towards incorporating big data and machine learning techniques.

(Family) Mode choice and activity models

For a survey on activity patterns at the individual level, see Miller (2013). Activity-based travel demand models take an agent-based approach. More sophisticated models simulate the travel and activity schedules of household members, while considering their interactions (Roorda et al., 2009). Knowledge of the schedules and activities of each household member is required to maximize a collective utility function. In a similar vein, Arentze and Timmermans (2009) developed models for household activity-travel behavior, focusing on how households allocate time and resources across different members' activities. The literature is still rapidly developing.

Freight

Relatively few attempts have been made to include freight in traffic simulators. A notable exception is Ben-Akiva’s (2020) traffic simulator SimMobility. The current lack of urban B2B freight data in France stems from the fact that the existing standard to collect such data is highly sophisticated, costly and complex to administer. As a result, only 5 surveys have been conducted in France over the past 30 years, whereas reliable, reproducible, and affordable household surveys are routinely conducted.

Traffic simulators

With recent advances in computing power, transportation simulators have become essential tools for analysing specific scenarios or policies, taking into account the complex interactions of various effects (Nguyen et al., 2021). These simulators range from microscopic models, which examine traffic infrastructure impacts at the neighbourhood level, to macroscopic models that explore aggregate effects at regional level. Mesoscopic models, such as METROPOLIS (de Palma et al., 1997) and MATSim (Axhausen et al., 2016) adopt an intermediate approach between microscopic and macroscopic simulations. Using an agent-based methodology, they capture heterogeneous effects while simplifying congestion modelling to enhance scalability for large-scale scenarios.

LUTI models

LUTI (Land Use/Transport Interaction) models traditionally assume general equilibrium, but do not include a detailed traffic model (Anas, 2013; Anas and Chang, 2023). More disaggregated models such as UrbanSim, originally developed by Waddell (2002), either assume partial equilibrium (in some markets), or recognize that public interventions prevent partial equilibrium. UrbanSim was further developed and adapted to France, to be used by operators, such as Société du Grand Paris.

Indicators: welfare, equity, pollution
Equitable transport systems enable more people to commute easily and affordably (see, e.g., Nicolas et al., 2003). Equity and pollution play crucial roles (see Kitchin at al., 2015). Welfare indicators such as access to quality healthcare and education influence where individuals choose to live and work; areas offering higher living standards attract more residents. Air quality and other pollution indicators significantly affect residential choices, with people often preferring to live and work in areas with lower pollution levels to ensure better health and quality of life. For example, cities with better air quality and access to green spaces are more attractive.

Air quality
Pollutants like PM, O3 and NO₂ impose significant economic burdens and health costs in urban areas (Walton et al., 2015; Vlachokostas et al., 2012; Martinez et al., 2018, and de Bruyne and de Vries, 2020). In 2023, the European Environment Agency released a report detailing the burden of disease from air pollution in Europe, highlighting the significant health impacts and the substantial economic costs associated with pollutants like PM₂.₅ and NO₂ across European cities (see also European Environment Agency's home page, Health Effects Institute and World Health Organization (WHO).

Family mobility decisions
Traditionally, households were modeled as single decision-making units in economics and transportation research. However, the development of non-unitary models, particularly in the Economics of the Family, recognized the diverse preferences of household members and the complexity of joint decision-making processes. These models have been increasingly applied to transportation decisions, incorporating concepts like bargaining, altruism, and Pareto optimality to better reflect intra-household interactions (see Bhat and Pendyala, 2005, Timmermans and Zhang, 2009, and the recent survey of de Palma, Picard and Lindsey, 2024). These interactions have also been formulated as optimal transport models by Galichon (2008); see also the seminal book of Browning, Chiappori and Weiss (2014).