Modeling the resilience of urban mobility when exposed to the COVID-19 pandemic: A qualitative system dynamics approach

Abstract

In December 2019, coronavirus disease (COVID-19) was detected in Wuhan, China. Due to the rapid spread of the disease, containment measures were adopted, which caused unprecedent shifts in individual mobility. Although some studies explored the impacts of the COVID-19 pandemic on travel patterns and resilience of transport systems based on different analysis techniques, there is a lack of studies addressing the impacts of the pandemic on the sustainability and resilience of urban mobility systems using in-depth and holistic methods, such as system dynamics. This research aims to characterize the dynamics present in urban mobility systems when exposed to pandemics and analyze the changes needed for systems to increase their resilience to pandemics using qualitative system dynamics modeling. The framework comprises the characterization of cause-and-effect relationships and the creation of systems’ causal loop diagrams (CLD) in their basic state of functionality, when affected by pandemics, and still operating owing to its resilience. Our findings indicated that the CLD of a resilient system is driven by strategic preparedness and response plans, as well as research and development, which balance the spread of the pandemic and increase support on technological strengths and the activities performed from home.

1. Introduction

Urban mobility systems comprise different elements that interact strongly with each other. These elements (which themselves can be subsystems) when exposed to threats can cause not only unexpected effects on certain parts of the system, but also endanger the overall functioning of the system. An urban system that is resilient should be able to counteract major negative threats and still preserve the basic functionality.

Health-related threats to urban mobility systems, such as pandemics, have been increasingly addressed due to the unprecedented impacts caused by the COVID-19 pandemic. All over the world, to control the rapid spread of the COVID-19 disease, a series of containment measures (such as partial and total lockdown) have been adopted, causing economic, social, and environmental tensions that impacted the sustainability and resilience of urban systems.

Promptly, the pandemic led to the collapse of health systems, the closure of commercial establishments, unemployment, wage reduction and impairment of population income (Higginson et al., 2020; Sharifi & Khavarian-Garmsir, 2020), a ban of long-distance trips due to border closures between cities and countries, and restrictions (or prohibitions) to short distance trips (Narayan, Phan, & Liu, 2021). On the other hand, the levels of atmospheric NO₂ and CO (pollutants directly associated with the transportation sector) have diminished (Gupta, 2020; Kakderi, Komninos, Panori, & Oikonomaki, 2021).

Due to the imbalance in the systems, response measures were needed. Examples included providing financial assistance to employees, business, and unemployed people, adapting existing infrastructure for pedestrians and cyclists (Kakderi et al., 2021), adapting and creating smart technologies to properly inform procedures to prevent infection and the spread of the disease, minimizing face-to-face contact, identifying infected individuals, and facilitating quarantine measures (Sharifi & Khavarian-Garmsir, 2020). Studies indicated that the use of digital services, e-tools, teleworking, and smart city solutions have significantly complemented policies to control the pandemic and help to develop more resilient and environmental-friendly behaviors (Kakderi et al., 2021). After implementing the measures, the systems gradually resumed activities considered essential.

To strengthen the resilience of cities in the face of pandemics, Sharifi and Khavarian-Garmsir (2020) suggest allocating more space dedicated to active modes and public spaces, implementing smart city and sustainable planning, and Transit-Oriented Development, as well as a health system prepared to quickly act and respond to pathogens with pandemic potential.

In this context, mobility systems were strongly affected by the pandemic due to the need to restrict crowds, inner and outer travel to control the spread of the disease and, consequently, the number of infected people (Chinazzi et al., 2020; Mahmoudi & Xiong, 2022; Wells et al., 2020). Thus, several containment measures were aimed at restricting short- and long-distance trips, preventing crowds in public transport, limiting non-essential activities, etc. Therefore, a framework of analysis focusing on the resilience of urban mobility systems while considering the effects of its different subsystems is crucial to avoid a collapse (partial or total) in the medium and long term.

Specifically, regarding urban mobility systems, the containment measures caused unparalleled shifts in the purpose of daily trips, travel demand, and travel mode choices. For instance, the total number of trips for education and work purposes reduced almost to zero in Italy (Moslem et al., 2020). Overall daily trips have drastically declined (Aloi et al., 2020; Hadjidemetriou, Sasidharan, Kouyialis, & Parlikad, 2020; Hasselwander et al., 2021; Sahraei, Kuşkapan, & Çodur, 2021; Sharifi & Khavarian-Garmsir, 2020), in which public transit ridership registered the largest decrease (Aloi et al., 2020; Hasselwander et al., 2021; Munawar, Khan, Qadir, Kouzani, & Mahmud, 2021; Shakibaei, de Jong, Alpkökin, & Rashidi, 2021). In contrast, several studies found evidence of increased use in active modes (Fenu, 2021; Kamga & Eickemeyer, 2021; Sharifi & Khavarian-Garmsir, 2020; Zhang, Hayashi, & Frank, 2021) and in individual travel modes (Campisi et al., 2020; Sharifi & Khavarian-Garmsir, 2020; Zhang et al., 2021).

Our model covers the pre-pandemic, pandemic and post-pandemic periods, but we believe that singular changes in travel patterns may affect the sustainability and resilience of urban mobility and they are expected to be in the long-term, as mentioned by Manzira, Charly, and Caulfield (2022). Studies argue that a portion of users who shifted to private cars instead of public transport are unlikely to return to using public transport (de Haas, Faber, & Hamersma, 2020; Simić, Ivanović, Đorić, & Torkayesh, 2022). According to Shakibaei et al. (2021), private cars may become a predominant mode of transport during the post-pandemic period in Istanbul causing major traffic congestion and air pollution. Reductions in concentration levels of NO₂ (Aloi et al., 2020; Sharifi & Khavarian-Garmsir, 2020) and CO were registered due to changes in travel patterns (Baldasano, 2020; Dantas, Siciliano, França, da Silva, & Arbilla, 2020; Saadat, Rawtani, & Hussain, 2020). However, Ozone (O₃) concentrations significantly increased, which indicates a clear need to implement holistic measures and policies for reducing pollutants, allowing the identification of effects that may be counterintuitive, such as the increase in secondary pollutants that cause health problems (Sharifi & Khavarian-Garmsir, 2020).

In addition, the pandemic has intensified disparities between social groups (Gusheva & Gooyert, 2021; Munawar et al., 2021) as detected between those able to work remotely and those who must commute daily (Gutiérrez, Miravet, & Domènech, 2020). Thus, to achieve a sustainable and resilient urban mobility system to the pandemic, policies implemented must ensure the system's ability to maintain its current conditions, reduce the impacts adopting alternative mobility solutions; or adapt to change and transform subsystems; while maintaining an acceptable level of economic, environmental, and social sustainability.

Given that urban mobility (sub)systems are strongly interrelated and fundamentally dynamic in time and space, an analytical approach that encompasses different stakeholders and/or agents, such as system dynamics, is advantageous. System Dynamics (SD) modeling is a valuable tool to comprehend decision-making processes in challenging conditions (for instance, considering the effect specific actions have on one part of the system that affect other parts of the system (or subsystems) and eventually modify the behavior of the entire system (Nabavi, Daniell, & Najafi, 2017)). Moreover, in-depth qualitative models provide the conception of the system's hypothesized behavior, and the proper formulation of quantitative models (Tenza, Perez, Martinez-Fernandez, & Gimenez, 2017).

In the literature, the influence of different government policies on urban mobility sustainability has been investigated using comprehensive and holistic system dynamics models (e.g., Melkonyan et al., 2020). On the other hand, the impacts of the COVID-19 pandemic on travel patterns have been addressed using different methodologies, such as an artificial neural network model (e.g., Ghanim, Muley, & Kharbeche, 2022) and exploratory analyses (e.g., Aloi et al., 2020; Chan, Chen, Ma, Sze, & Liu, 2021; Fenu, 2021; Hasselwander et al., 2021; Manzira et al., 2022; Munawar et al., 2021; Orro, Novales, Monteagudo, Pérez-López, & Bugarín, 2020; Sahraei et al., 2021; Shakibaei et al., 2021). Besides that, the resilience and vulnerability of transport systems to the COVID-19 pandemic was introduced using a three-stage integrated Fermatean fuzzy model by Simić et al. (2022). Although some studies explored urban mobility sustainability using system dynamics models and the impacts of the COVID-19 pandemic on travel patterns and the resilience of transport systems based on different analysis techniques, there is a lack of studies addressing the impacts of the COVID-19 pandemic on the sustainability and resilience of urban mobility systems using system dynamics.

The study conducted by Lara and Rodrigues da Silva (2021) characterized the resilience of urban mobility systems in the face of environmental threats (such as climate change, flooding, and natural hazards). The present study extends the research carried out by Lara and Rodrigues da Silva (2021), however, considering the COVID-19 pandemic as a threat. Therefore, considering the study problem and the research gap, we intend to answer the following research questions:

• RQ1: Which components of urban mobility systems are most affected by the coronavirus disease (COVID-19) pandemic?

• RQ2: How does the COVID-19 pandemic affect travel behavior and public transport demand?

• RQ3: How do these changes in travel behavior affect sustainable urban mobility systems?

• RQ4: What changes are needed for urban mobility systems to increase their resilience in the face of the COVID-19 pandemic?

Overall, RQ1 is important to help urban and transport planners to understand how components of urban mobility interact, respond, and react when affected by a pandemic. RQ2 aims to focus on changes in travel behavior and identify critical impacts on public transport demand. RQ3 discusses the impacts of these changes in travel demand and mode choices on the sustainability of urban mobility. RQ4 provides new insights regarding the resilience of urban mobility systems. Therefore, the novelty of this approach comprises the integrated analysis of social and institutional aspects, health security, behavioral changes in travel patterns and impacts on the sustainability and resilience of urban mobility into a single and simple model. Moreover, it can help the development of effective and integrated response plans, the decision-making process and building resilient cities.

This section introduced the problem of study and the research questions we intend to answer. Section 2 presents an overview describing the concept of sustainability and urban mobility resilience, system dynamics modeling, and system dynamics approaches in urban mobility. In Section 3, the materials and method are described. In Section 4, we present the qualitative system dynamics models and discuss the results. Finally, the conclusions are drawn and suggestions are made for future work in Section 5.

2. Theoretical framework

In this section, we introduce a concise literature review on sustainability and resilience of urban mobility, in which we present the concepts and the definition of resilience of urban mobility adopted for the present study. Moreover, we describe a theoretical background regarding the system dynamics modeling and system dynamics approaches in urban mobility.

2.1. Sustainability and resilience of urban mobility

The concepts of sustainability and resilience have many similarities; however, their relationship is still not clear. The study conducted by Bocchini, Frangopol, Ummenhofer, and Zinke (2014) showed that resilience and sustainability are concepts that complement each other and should be addressed together. Additionally, most definitions and frameworks presented in the literature consider resilience as a short-term phenomenon, while sustainability is mainly related to long-term phenomena that consider the effects on ‘future generations’ (Saunders & Becker, 2015). Lizarralde, Chmutina, Bosher, and Dainty (2015) discusses tensions that occur when stakeholders create their own representations and concepts, which can create different interpretations of what is considered ‘sustainable’ and ‘resilient’, which can even contradict each other, requiring unexpected trade-offs in the decision-making process. On the other hand, Saunders and Becker (2015) argue that to ensure that the needs (economically, socially, culturally, and environmentally) of future generations are met, a resilient community should also be a sustainable community.

Marchese et al. (2018) identified three generalized frameworks that relate sustainability and resilience. Framework 1 describes resilience as a component of sustainability, operating with the notion that “increasing the resilience of a system makes that system more sustainable, but increasing the sustainability of a system does not necessarily make it more resilient” (Marchese et al., 2018, p. 1276). Framework 2 considers sustainability as a component of resilience, operating with the assertion that “increasing the sustainability of a system makes that system more resilient, but increasing the resilience of a system does not necessarily make that system more sustainable” (Marchese et al., 2018, p. 1276). Framework 3 describes resilience and sustainability as “concepts with separate objectives that lack hierarchical structure, and that can complement or compete with each other” (Marchese et al., 2018, p. 1278). On the other hand, Rogov and Rozenblat (2018) argue that resilience is related to the adaptation process while sustainability concerns the transformation process of the system.

In this study, we argue that by increasing the economic, social, and environmental sustainability of urban mobility systems, they become more likely to resist, adapt or transform in the face of health-related threats, such as epidemics and pandemics. For example, reducing the number of trips (due to the increase in remote working, distance education, and online shopping) and increasing the coverage of active modes and e-tools, leads to a reduction in CO₂ and NOx emission levels (environmental sustainability), face-to-face interactions, crowds in public transport, and the portion of income spent on transport fares (social sustainability). These measures also ensure greater health security for making essential trips, and reaffirm confidence in public transport (social sustainability). Additionally, policies supporting subsidies per transported passenger, as well as financial assistance programs for public transport companies, help public transport companies to maintain revenue at acceptable levels (economic sustainability), as well as supply meeting demand while respecting social distancing rules (social sustainability). Therefore, the concept of resilience considers sustainability as a component of resilience, in accordance with Framework 2 described by Marchese et al. (2018) and other studies in the field (e.g., Fior, Galuzzi, & Vitillo, 2022; Manzira et al., 2022; Melkonyan, Gruchmann, Lohmar, & Bleischwitz, 2022).

In the urban mobility context, a sustainable urban mobility system should encourage residents to use public transport (Shakibaei et al., 2021) and active modes, and reduce the affinity towards private cars (Manzira et al., 2022), minimize transport-related emissions and promote equitable access to services and transport infrastructure. Additionally, according to Lara and Rodrigues da Silva (2021) and based on the principles of persistence, adaptability, and transformability proposed by Folke et al. (2010) and Fernandes et al. (2019), we define the urban mobility resilience as a system that can:

• i)maintain the existing conditions or quickly return to desired functions of mobility maintaining an acceptable level of economic, environmental, and social sustainability (persistence).
• ii)absorb the first damage and minimize the impacts from a disturbance providing alternatives to return to the basic mobility functionalities, without compromising economic, environmental, and social sustainability (adaptability).
• iii)adapt to change, and rapidly transform subsystems that restrain current or future adaptive capacity (transformability) throughout temporal and spatial scales.

2.2. System dynamics modeling

The System Dynamics Society defines system dynamics (SD) as a computer-aided approach for strategy and policy design (System Dynamics Society, n.d.). The main objective of SD is to support people to make better decisions when they are confronted with complex, dynamic systems. SD complements system thinking and uses simulation modeling based on the feedback systems theory. The core concept of SD is that the structure of a system represented by its primary feedback loops captures the interactions between system parts and how these lead to patterns of behavior over time. In general, SD models apply to problem solving in complex systems (Richardson & Pugh, 1981; Sterman, 2002). Developing an SD simulation model usually involves casting the problem under review, first in terms of the behavior of the critical components of the system, and second in a formal replicable mathematical structure (Rich et al., 2009). A commonly used approach in SD modeling is to first create qualitative diagrams - Causal Loop Diagrams (CLD) - of the primary feedback loops of a system, which then form the basis for developing quantitative models. Besides being the first step of developing quantitative models, qualitative models enable a comprehensive understanding of the system behavior, which is important for identifying policies or strategies responses that may be counterintuitive, impairing the performance, or malfunction of the system (Lara & Rodrigues da Silva, 2021). This study focuses on the aspect of the development of a qualitative model in the form of CLDs.

CLD comprises a graphical depiction of cause-and-effect relationships between the components of a system (Sterman, 2002). CLDs consist of four components: the elements of the system (entities), the interrelations between the elements (links), time delays and feedback loops. Elements used in a CLD need to fulfill only one criterion: they must be non-specified quantities. This means that it must be possible to state whether the element increases or decreases. The interrelations between elements have a direction and a polarity. The direction indicates which of the two connected elements is the influencing entity (cause) and which is the connected entity (effect). The polarity indicates how an element changes when the influencing element changes (Ford, 2019). The polarity “same” indicates that if the cause increases, the effect increases more than expected, and if the cause decreases, the effect decreases below what it would otherwise have been. The polarity “opposite” indicates that if the cause increases, the effect decreases below what it would otherwise have been, and if the cause decreases, the effect increases more than expected. In a CLD, the interrelations are represented by arrows. The polarities are marked near the arrowhead by a “+” for the polarity “same” and a “-” for the polarity “opposite”. A closed circuit of interrelations between different elements forms a feedback loop. Feedback loops are assigned with a polarity which can be either reinforcing or balancing. A reinforcing feedback loop is characterized by zero or an even number of links with the polarity “opposite”. A balancing feedback loop is characterized by an odd number of links with the polarity “opposite”. Reinforcing loops lead to exponential growth while balancing loops lead to stabilization and dynamic equilibrium. A more detailed description of CLD can be found in Barbrook-Johnson and Penn (2022).

2.3. System dynamics approaches in urban mobility

System dynamics is a method widely used to characterize and analyze systems in different areas of knowledge. The most relevant areas for the study presented here are transport and mobility, on the one hand, and health and medicine on the other hand. Shepherd (2014) identified 54 references on system dynamics models in transportation in the following domains: modeling the uptake of alternate fuel vehicles (22%), supply chain management with transportation (11%), highway maintenance/construction (9%), strategic policy at urban, regional and national levels (24%), airlines and airports (19%) and emerging areas (15%). A more recent systematic literature review of system dynamics models in relation to strategies for freight transport decarbonization found 50 relevant studies (Ghisolfi et al., 2022). Darabi and Hosseinichimeh (2020) identified 301 relevant references on system dynamics applications in health and medicine. The studies were classified into the categories: regional health modeling (38%), disease-related modeling (35%) and organizational modeling (27%). Forty-two studies covered the topic of infectious disease modeling. System dynamics has also been used to model the transmission of Covid-19 under different policy scenarios (Feng & Lu, 2020; Jia, Li, & Fang, 2022; Struben, 2020; Sy et al., 2020). However, relatively few studies using system dynamics to assess the resilience of urban mobility systems in the face of threats, such as Covid-19, were found.

One of the most used models to evaluate urban mobility resilience is equivalent to the indicator-based one. Nevertheless, it has the drawback of excluding the system's complexity and its systemic interaction regarding feedback effects among the system's components (Feofilovs et al., 2020). Hence, assessing the resilience of urban mobility as a complex system using SD modeling can bring substantial advantages and contributions to the area. That is because SD does not only describe component interdependencies and their relational dynamics but also captures effects in the short and the long term. Another key aspect is the ability to show the root cause (or causes) of a problem (Mallick, Radzicki, Daniel, & Jacobs, 2014).

Joakim et al. (2016) developed a conceptual model of social vulnerability and resilience using an SD approach to simulate the impact of potential adaptation policies in the Vancouver Metro. The method involved a multi-step approach, incorporating literature reviews, discussions with local decision-makers and stakeholders, and ground-truthing by consulting local emergency managers, community planners, engineers, decision-makers, and other relevant stakeholders.

Macmillan et al. (2018) used qualitative SD models to analyze the relationships between active travel and walking and cycling infrastructure. The models were carefully outlined using evidence from the literature and knowledge from policy makers, professionals in the field, and community to ensure contextual and cultural fit. Moradi and Vagnoni (2018) explored passenger urban mobility using SD to identify the current path of low carbon mobility transition and to study the most probable pathways under the 2030 targets. For this, the main regimes of urban mobility and the dynamics of the transition process were studied at all levels. Furthermore, Suryani et al., 2020, Suryani et al., 2019 addressed the urban mobility and traffic congestion problem using SD simulation and scenario analysis.

3. Materials and method

In this section, we present the method and database used for assessing the resilience of urban mobility. The framework applied in this study consists of four steps:

• a)Characterization of cause-and-effect relationships: detecting cause-and-effect relationships (and feedbacks) among the system's elements using a literature review. Moreover, defining the system's boundaries and time horizons.

The literature review was conducted from the Web of Science (WOS) database in August 2021. The WOS database was chosen as it has a wide repository of different academic disciplines, contributing to broader output data. A search for terms was carried out using the Boolean operator AND in the field tag Topic (TS), which includes the title, abstract, and author keywords. Combinations were performed in pairs, then in trios, and so on until, in the last combination, all terms were used. The group of terms included: “resilience”, “mobility”, “threats”, “vulnerability”, “transport*”, “system dynamics”, “COVID-19″, “Coronavirus” and “SARS-CoV-2″. The search was limited to articles, books, book chapters, notes, and proceedings papers, all published in English.

After filtering the research area, 770 publications were found. A content analysis was conducted, in which terms were searched and counted using the R package. After obtaining the content analysis, the database was reduced to 607 publications by selecting those whose count was higher than the average. In addition, the compliance of the titles and abstracts was evaluated through reading, which resulted in a final list of 292 publications. Most of the publications selected were journal articles (89.7%), while 10.0% were proceedings papers, and 0.3% were book chapters, all published between 2016 and 2021. Most originated from the United States of America (27.1%), Italy (11.6%), and China (10.6%).

Based on the literature review, subsystems that compound the urban mobility system can be identified. The variables used to describe the behavior of each subsystem and the urban mobility as a whole can be investigated, as well as their associations.

• b)Qualitative model formulation of the system in its basic state of functionality: identification of the main causal links, structure, feedback processes, and creation of CLD.

Based on the variables compiled in step a), similar variables can be aggregated, and recurrent variables in the literature review can be selected. Furthermore, an analysis of the causal relationships for each variable selected can be carried out, as well as their feedback processes. To keep the model as simple as possible while still providing a satisfactory range of accuracy, the main causal relationships and feedback processes regarding the basic state of functionality of urban mobility systems need to be selected. To this end, discussions with specialists in the field can be carried out to identify the most important links to describe the system behavior. The resulting cause-and-effect relationships and feedback processes are summarized in Appendix A (Table A1) and are described in detail in Section 4.

• c)Qualitative model formulation of the system affected by a pandemic (RQ1, RQ2 and RQ3): analysis of structure, feedback processes, creation of CLD in the face of a pandemic.

Analogously to step b), similar variables (compiled in step a)) can be aggregated, and recurrent variables in the literature review specific to the system affected by pandemics can be selected. Furthermore, discussions with specialists can be carried out to identify the most relevant links to describe the system behavior under threat. Validation of the system behavior can be done by using observed data (for example, the Community Mobility Reports of Austria (Google, 2021)). The resulting cause-and-effect relationships and feedback processes are summarized in Appendix A (Table A2) and are described in detail in Section 4.

• d)Qualitative modeling of a resilient system (RQ4): analysis of structure, feedback processes, creation of CLD describing the attributes of persistence, adaptability, and transformability of an urban mobility system considering a pandemic.

In this step, the consistency of the links and feedback processes from step c) can be verified so that the attributes of persistence, adaptability, and transformability are strengthened, and consequently, the resilience of the system. The analysis is performed considering the literature specific to the system resilience, sustainability, and vulnerability to pandemics. Furthermore, discussions can be held with experts to verify links and feedback processes that describe the system's ability to resist, adapt or transform in the face of pandemics. The resulting cause-and-effect relationships and feedback processes are summarized in Appendix A (Table A3) and are described in detail in Section 4.

Fig. 1 shows the whole process of the development of a quantitative model of the resilience of urban mobility when exposed to pandemics in the form of a data flowchart. The whole process starts by defining the model boundaries and identifying the key elements. Logical considerations and results of a literature review form the basis of this step. Intermediate results are discussed and refined in an iterative process whereby peers participate. The same procedure applies to the subsequent steps of the development of the CLDs of the basic state of functionality, the pandemic affected system and the resilient system. The pandemic affected CLD can be validated by comparing the resulting behavior over time with observed data (in our case we used the Community Mobility Reports from Austria (Google, 2021) for infection numbers and for mitigation policies we used Open Data Österreich (2021)). The black elements in Fig. 1 highlight the work which is presented in this paper. The gray elements mark ongoing work concerning the development of the quantitative model. A quantitative stock-flow model is built on the basis of the CLD representing the resilient system. Once more, the model is refined in an iterative process whereby peers participate. Community Report Data and data about infection numbers and mitigation policies are used to estimate the parameters of models for Brazil and Austria. The resulting calibrated models are used to carry out simulations and to evaluate different policy strategies and scenarios. The scenarios are defined in an iterative stakeholder process.

Fig. 1.

Data flowchart to develop a quantitative model of the resilience of urban mobility when exposed to pandemics.

4. Results

The method starts by detecting interrelations and interconnections among the elements of the system in three different conditions: i) basic state of functionality, ii) affected by health-related threats and iii) a resilient system.

4.1. Characterization of cause-and-effect relationships

We identified, through a literature survey, a significant number of variables and their associations through a literature review, which were further selected and aggregated to describe causal relationships and to build the qualitative models. The present study comprises an extension of another study (Lara & Rodrigues da Silva, 2021), which described and analyzed the dynamics of urban mobility systems when affected by environmental threats (such as climate change, flooding, and natural hazards).

The CLD described by Lara and Rodrigues da Silva (2021) was simplified in this study to specifically address our problem of study (which is related to the effects of the pandemic on travel behavior, public transport demand, and the effects of changes in travel behavior on urban mobility sustainability and resilience). We do not address ‘fuel consumption’ and ‘trip costs’ as they would require a more complex and detailed investigation, which is not the focus of this paper. Although we found studies indicating that vehicle fuel consumption declined in many cities around the world (e.g., China (Rahman et al., 2021), USA (Du, Rakha, Filali, & Eldardiry, 2021), Canada (Tian, An, Chen, & Tian, 2021)) during the pandemic, which caused reductions in CO₂ emissions, we found no evidence that fuel consumption is associated with changes in transport demand and travel behavior. Therefore, to respond to the research questions (RQ1 to RQ4, introduced in Section 1), we decided not to address ‘fuel consumption’ and the Material and energy flow subsystem. Instead, we focused on the other seven subsystems described by Lara and Rodrigues da Silva (2021), namely: Institutional, Social, Economic, Infrastructure, Natural, Demand, and Transport mode. A complete analysis incorporating the Material and energy flow subsystem is recommended for quantitative analysis, which we intend to carry out in future studies. The CLD of the urban mobility system in its basic state of functionality is represented in Fig. 2 , in which the subsystems are shown in different colors.

Fig. 2.

Causal loop diagram of the urban mobility system in its basic state of functionality (adapted from Lara & Rodrigues da Silva, 2021).

4.2. CLD: System in its basic state of functionality

In the second step, considering the variables identified and selected, we created a qualitative model of the system in its basic state of functionality. The basic state of functionality is an archetype as it is equivalent to common system structures that simulate the system's performance in everyday use (Lara & Rodrigues da Silva, 2021). There were nine major feedback loops in the CLD. Five of them are reinforcing (R1 to R5) and three are balancing (B1 to B3), as presented in Fig. 2.

Due to the complexity of the model indicated in Fig. 2, a schematic representation showing the feedback loops and a list of references related to the causal relationships found in the literature review are summarized in Appendix A (Table A1). The interpretation of the causal loops is exemplified next. Reinforcing loop R1 shows that an increase in public travel demand increases active and public transport policies, which leads to an increase in investments in sustainable transport infrastructures, decreasing private motorized travel demand, which in turn decreases traffic congestion and travel time, which leads to increases in public travel demand. As the explanation of the other loops is analogous, they will not be detailed.

Overall, the Transport mode and Demand subsystems have a significant effect on the CLD as shown in Fig. 2, as they account for 46.2% of the variables. Additionally, Fig. 2 shows that the priorities of transport policymakers (active, public, or private motorized transport policies) are bound to economic interests (loops R4 and R5). Balancing loop B1 indicates that an increase in economic performance decreases vulnerable social groups as more jobs are available, however this can cause a decrease in public travel demand, increasing private motorized demand and traffic congestion, which in turn decreases the economic performance. Loop B1 is counterintuitive suggesting that policies (such as awareness campaigns) are needed to prevent an increase in income leading to a drastic increase in private motorized trips.

4.3. CLD: System affected by health-related threats

The third step comprises the qualitative model formulation of the system affected by the COVID-19 pandemic. The CLD representing the system behavior created eleven major feedback loops. Four of them are reinforcing (Rcv1 to Rcv4) and seven are balancing (Bcv1 to Bcv7) loops. A schematic representation of the feedback loops shown in Fig. 3 and a list of references related to the causal relationships found in the literature survey are summarized in Appendix A (Table A2). The explanation of reinforcing (Rcv) and balancing (Bcv) feedback loops is similar to the CLD concerning the system in its basic state of functionality.

Fig. 3.

Causal loop diagram of the urban mobility system affected by the COVID-19 pandemic.

Overall, the model resulted in an increase in the contribution of the variables that compound the Social (from 15.4% to 27.3%) and Institutional (from 15.4% to 31.8%) subsystems in relation to the basic state (RQ1). Social and Institutional subsystems significantly affect the CLD as they represent 59.1% of the variables (all represented in italics in Fig. 3).

Fig. 3 is also useful to address RQ2. It indicates that an increase in contagion increases the number of infection and death cases, which in turn increases containment measures (loop Bcv1). An increase in containment measures increases the activities done from home, which decreases crowding and contagion (loop Bcv2). An increase in activities from home causes a decrease in the distances travelled, which decreases public travel demand, crowding and contagion (loop Bcv3). Besides that, an increase in containment measures increases government financial assistance, decreasing vulnerable social groups, which increases activities from home (loop Bcv5). However, a decrease in vulnerable social groups causes a decrease in public travel demand (loop Bcv6). An increase in public travel demand decreases private motorized demand, which increases economic performance, technological strengths and activities from home, decreasing public travel demand (loop Bcv8). An increase in economic performance, increases private motorized transport policies, decreasing investments in sustainable transport infrastructure, which increases crowding and contagion (loops Bcv7 and Rcv1). An increase in investments in sustainable infrastructure investments increases traffic safety and economic performance (loop Rcv2), which in turn increases technological strengths and activities from home (loop Rcv3).

To validate the system behavior, we analyzed two major loops (Bcv2 and Bcv3) using data from the Community Mobility Reports from Austria (Google, 2021) from March 16 to December 26, 2020 (Fig. 4 ). According to the literature review presented in Section 3 - item a), no articles published in Austria were found. Austria is a developed country with 8,978,929 inhabitants, 49.3% male and 50.7% female (Statistics Austria, 2022). In 2019, the modal split on average corresponded to 48.1% for pedestrians, 2.9% for cyclists, 36.8% car users and 12.2% public transport passengers, while in 2020, these proportions changed to 47.4%, 4.4%, 40.3%, and 7.8%, respectively (ETH Zürich & Universität Basel, 2022). In general, changes in Austrian travel patterns were similar to other European countries already pointed out in the introduction (such as Spain (Aloi et al., 2020; Fenu, 2021; Sahraei et al., 2021), Italy (Fenu, 2021; Moslem et al., 2020; Sahraei et al., 2021), France (Fenu, 2021; Sahraei et al., 2021), Germany, Turkey, and Sweden (Sahraei et al., 2021), Belgium (Fenu, 2021)). Thus, Austria can be considered a good example to validate the models formulated, as this database is open, reliable, widely accessible, and available, in addition to contributing to studies in the area, as we did not find studies for this country. Furthermore, in future studies we intend to calibrate and validate models using data observed from 2019 to 2022 in Austria and Brazil. These countries have markedly distinct characteristics from each other, and therefore the models can be analyzed in different contexts.

Fig. 4.

Covid-19 infections, containment measures and mobility in Austria week 8–52/2020; Source: (Google, 2021), (Open Data Österreich, 2021).

According to loop Bcv31 , growing infection numbers will trigger containment measures which reduce out of home activities such as going to work and public travel demand. Due to exponentially growing infection numbers (bold line in Fig. 4), the Austrian government implemented a lockdown as a containment measure against the pandemic in week 12. This led to a drastic decrease in out of home-based activities. The number of visits at workplaces dropped by 60% (dotted line with squares in Fig. 4). The need for travel and public transport use decreased sharply. The number of visits at transit stations dropped by about 70% (dotted line with diamonds in Fig. 4). Less out of home activities and public travel demand reduced crowding and contagion. With a time delay, less contagion reduced the number of infections. The reduced crowding and contagion resulted in a time delayed drop in the infection numbers with a peak in week 13. Lower infection numbers reduced the need of containment measures. Therefore, Austrian containment measures were relaxed with a time delay: reopening of shops in week 16, reopening of food establishments in week 20, reopening of hotels, fitness studios, etc. in week 22 and the end of most restrictions in week 25. The gradual easing of restrictions led to an increase in out of home activities, as well as longer journeys and public transport use. Thus, crowding and contagion were increased again. During the warm climate in summer, less indoor activities and school holidays helped to keep infections at a low level. In the middle of the holidays when people started to trickle back to work, infection numbers started to rise again. By the end of the school holidays, infection numbers reached a threshold and containment measures were put into force again culminating in partial and a full lockdown in weeks 45 and 47. This time, the effect of the containment measures was less strong than in the first lockdown. It is worth mentioning that the drop in out of home activities and public travel demand started already in the week before the partial lock. The cause was the Austrian national holiday which fell on a Tuesday and was used by many employees for a break. The number of visits at workplaces in this period dropped only by about 24% and the number of visits of transit stations only by about 50%. The containment measures were relaxed for Christmas but tightened again directly afterwards. This example demonstrates that our balancing feedback loops Bcv2 and Bcv3 are able to replicate the system behavior as seen in the observed data.

In a nutshell, considering the database from Austria, due to the growing infection numbers, measures to contain the spread of the disease were implemented causing impacts on travel behavior and public transport demand. The first lockdown caused a drastic decrease in out of home-based activities, in which the number of visits at workplaces dropped by 60% and the number of visits at transit stations dropped by about 70%. Considering the reduction in infection numbers, containment measures were relaxed (partial lockdown) causing an increase in out of home activities, as well as longer journeys and public transport use. However, infection numbers increased again, and another lockdown was needed but its effect in travel behavior was less strong than in the first lockdown. The number of visits at workplaces in that period dropped only by about 24% and the number of visits of transit stations decreased by about 50%. As can be seen from the database, the system that was disturbed by the pandemic makes urban mobility no longer sustainable (RQ3). Considering this, institutional interventions to control contagion, changes in travel patterns, as well as social and economic impacts are needed so that the system can return to its basic state of functionality.

4.4. CLD: Resilient system

In the fourth step, we created a qualitative model of a resilient system, in which the principles of persistence, adaptability, and transformability were applied (see Fernandes et al., 2019; Folke et al., 2010). The CLD indicated in Fig. 5 resulted in eight major feedback loops, in which five are reinforcing (Rr1 to Rr5) and three are balancing (Br1 to Br3) loops. A schematic representation of the feedback loops in Fig. 5 and a list of references related to the causal relationships are summarized in Appendix A (Table A3). The interpretation of the loops is analogous to the CLD indicated in Fig. 2.

Fig. 5.

Causal loop diagram of a resilient urban mobility system. Schematic diagrams at the bottom of the figure highlight in red the attributes of (a) persistence, (b) adaptability, and (c) transformability.

Based on the analysis of the system under threat (Figs. 3 and 4), we identified variables, links, and feedback loops that play a central role to characterize a resilient system. In other words, by observing how the system reacts to threats to maintain its basic state of functionality, rearrangements between the system's components, and measures to increase sustainability can be predicted, and thus increase the system resilience to pandemics. Thus, the CLD from Fig. 5 shows that the Social and Institutional subsystems have a greater impact on the diagram, as they express 60.0% of the variables. The model resulted in an increase in the contribution of the variables that compound the Social (from 15.4% to 32.0%) and Institutional (from 15.4% to 28.0%) subsystems, compared to the basic state.

The dynamics indicated in the CLD of a resilient system have a preventive pattern, while in the CLD of the system exposed to threats they have a reactive pattern. Additionally, Fig. 5 indicates that an increase in public awareness increases public travel demand and active travel demand, increasing active and public transport policies (loops Rr1 and Rr2). An increase in strategic preparedness and response plans increases research and development, increasing technological strengths, i.e., measures and policies regarding innovation, Intelligent Transportation Systems, digitalization, etc., which positively affect activities from home (loops Rr4 and Rr5). An increase in activities from home decreases distances travelled, which increases active travel demand, decreasing private motorized travel demand (loop Rr4). A decrease in private motorized travel demand causes an increase in economic performance, which increases support in strategic preparedness and the response plans (loops Rr4 and Rr5).

In addition, an increase in activities from home decreases crowding, contagion, and spread of the pandemic, consequently decreasing containment measures and government financial assistance, which increases vulnerable social groups (loop Rr5). The contagion and spread of the pandemic are balanced by containment measures and activities from home, which decreases crowding (loop Br1). An increase in strategic preparedness and response plans and research and development causes a decrease in the spread of the pandemic, decreasing containment measures, activities from home and government financial assistance (loops Br2 and Br3).

Hence, in response to RQ4, for urban mobility systems to increase their resilience to the COVID-19 pandemic, they must be persistent, adaptable, or transformable. In a persistent system, research and development and strategic preparedness and response plans are predicted to balance the spread of the pandemic, strengthening the support in technological strengths and the activities carried out from home. Support for activities from home and public awareness are important to reduce the need for long trips, favoring the use of active and public travels, and decreasing private motorized travel demand. Considering this, the urban mobility system can maintain its basic functionality while maintaining an acceptable level of economic, environmental, and social sustainability. Alternatively, the system must be adaptable or transformable, in which government policies and measures are rapidly adapted or modified, such as government financial assistance and technological strengths, causing the strengthening of activities from home. In addition, policies to increase public awareness towards the use of sustainable transport modes are crucial to increase public and active travel demand, also increasing active and public transport policies. Taking this into account, the system can absorb the first damage, or adapt to change, and rapidly transform subsystems that restrain current or future adaptive capacity.

5. Conclusions

It can be observed that urban mobility systems are complex and often entail participation from many stakeholders, leading to feedback with distinct time lags (Shepherd, 2014). There is a body of literature that has explored urban mobility sustainability using system dynamics models (Melkonyan et al., 2020) and the impacts of COVID-19 pandemic on travel patterns and the resilience of transport systems based on different analysis techniques (Aloi et al., 2020; Chan et al., 2021; Fenu, 2021; Ghanim et al., 2022; Hasselwander et al., 2021; Manzira et al., 2022; Munawar et al., 2021; Orro et al., 2020; Sahraei et al., 2021; Shakibaei et al., 2021). However, studies are scarce concerning the impacts of the COVID-19 pandemic on the sustainability and resilience of urban mobility systems using in-depth and holistic methods of analysis, such as system dynamics. Thus, the study characterized the dynamics found in urban mobility systems in their basic state of functionality, affected by disruptive events and under resilient functioning through qualitative system dynamics modeling. Considering this, effects of the COVID-19 could be identified on components of urban mobility systems, travel behavior and public transport demand, sustainability, and resilience of urban mobility systems.

The results indicated that the Social and Institutional subsystems were most affected by the coronavirus disease pandemic, in which there was an increase of 59.1% in the contribution of the variables of those subsystems (RQ1). The CLD of the urban mobility system in its basic state of functionality pointed out that economic interests significantly help shape transport policies, which is in line with the findings in Malvestio, Fischer, and Montaño (2018) study. The results of the CLD throughout the pandemic threat showed that containment measures, and activities from home are important to balance the contagion. These results are consistent with Manzira et al. (2022) study, which indicated that public transport trips have a negative correlation with reported cases of COVID-19 infections. In addition, considering the system operating in the lockdown period, we found a drastic decrease in out of home activities and public transport usage (RQ2), which agrees with the study carried out by Aloi et al. (2020) and Hadjidemetriou et al. (2020). On the other hand, measures need to be implemented to control the impacts on public transport usage in the long term, such as policies to increase public awareness towards sustainable transport modes usage to ensure the urban mobility sustainability (RQ3). The CLD of a resilient system has a preventive pattern and is driven by strategic preparedness and response plans and research and development, which balance the spread of the pandemic and increase support on technological strengths and the activities carried out from home, which is consistent with the results found by Simić et al. (2022). In addition, public awareness is critical to reinforce the use of sustainable transport modes and increase urban mobility resilience (RQ4), which is in accordance with Manzira et al. (2022) and Chan et al. (2021) studies. According to Manzira et al. (2022), to achieve a resilient transport system it is necessary to rebuild trust in public transport (due to people's fear of contagion), reduce incentive for private motorized trips and encourage the use of active modes. Besides that, the success of response to and recovery from public health threats depends on mutual cooperation between the government and community (Chan et al., 2021).

We found that sustainability can be considered a component of resilience, in which it is recommended to adopt measures that increase the sustainability of elements of the system to strengthen the resilience of urban mobility systems in the face of pandemic threats. For example, policies aimed at encouraging the use of active modes and e-tools, implementing low-carbon transport (environmental sustainability), surveilling the compliance and application of social distancing rules in public transport, implementing financial assistance programs for unemployed people (social sustainability), encouraging remote work, providing subsidies on transport fares, implementing financial assistance programs for public transport companies, developing and improving scientific research, innovations and smart technologies (economic sustainability).

Our findings extend Lara and Rodrigues da Silva's (2021) study by addressing the COVID-19 pandemic as a threat to urban mobility systems. The study has the potential to help develop comprehensive quantitative models to evaluate urban mobility resilience when threatened by a pandemic. Hence, this study not only answers the questions raised in the introductory section, but also integrates health security into transport planning.

This article provided a broad understanding of the dynamics in urban mobility systems using qualitative models (CLD). The CLDs can assist in the integration of urban and transport planning, in the decision-making process and in building resilient cities. However, the models were based on evidence from the literature presented for several cities around the world, that is, they are generic models that do not portray the regional, cultural, and social behaviors of a specific population group. Although the model considering the system affected by the pandemic was validated using observed data, we analyzed the most cited dynamics in the literature. In addition, the study addresses the initial phase of the pandemic (March 16 to December 26, 2020), in which COVID-19 prevention vaccines had not yet been widely distributed and applied to the population.

For future investigations, we suggest analyzing the effect of vaccines on the system affected and on the resilience of the system. Moreover, the qualitative models using CLDs presented in this paper form the basis for the development of quantitative models using stock and flow diagrams. Thus, we recommend incorporating the Material and energy flow subsystem to the SFDs. We intend to calibrate and validate models using data observed from 2019 to 2022 in Austria and Brazil. The models will be used to analyze policies of both countries, in which different policy scenarios will be simulated to improve the preparedness and resilience of urban mobility to future pandemic events.

Declaration of Competing Interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Appendix A

Table A1.

Schematic representation of feedback loops in Fig. 2 and list of references related to causal links.

Reinforcing (R) or balancing (B)	Feedback loop	References
R1	Public travel demand $\stackrel{+}{\to }$ Active and public transport policies $\stackrel{+}{\to }$ Transport infrastructure investments (sustainable) $\overline{\to }$ Private motorized travel demand $\stackrel{+}{\to }$ Traffic congestion $\stackrel{+}{\to }$ Travel time $\overline{\to }$ Public travel demand	Arellana, Márquez, and Cantillo (2020), Chaves, de, Fontoura, and Ribeiro (2019); Hasselwander et al. (2021); Moradi and Vagnoni (2018); Munawar et al. (2021); Orro et al. (2020); Shakibaei et al. (2021); Sharifi and Khavarian-Garmsir (2020); Spadaro and Pirlone (2021); Suryani et al. (2019)
R2	Private motorized transport policies $\overline{\to }$ Transport infrastructure investments $\overline{\to }$ Private motorized travel demand $\stackrel{+}{\to }$ Private motorized transport policies	Arellana et al. (2020), Hasselwander et al. (2021); Kamga and Eickemeyer (2021); Munawar et al. (2021); Pfaffenbichler, Emberger, and Shepherd (2010); Sharifi and Khavarian-Garmsir (2020); Suryani et al. (2019)
R3	Public travel demand $\overline{\to }$ Private motorized travel demand $\stackrel{+}{\to }$ Traffic congestion $\stackrel{+}{\to }$ Travel time $\overline{\to }$ Public travel demand	Chaves et al. (2019), Munawar et al. (2021); Orro et al. (2020); Sharifi and Khavarian-Garmsir (2020); Spadaro and Pirlone (2021); Suryani et al. (2019)
R4	Economic performance $\stackrel{+}{\to }$ Private motorized transport policies $\overline{\to }$ Transport infrastructure investments $\overline{\to }$ Private motorized travel demand $\stackrel{+}{\to }$ Traffic congestion $\overline{\to }$ Economic performance	Arellana et al. (2020), Chaves et al. (2019); Gusheva and Gooyert (2021); Hasselwander et al. (2021); Mostafavi, Abraham, and DeLaurentis (2014); Munawar et al. (2021); Shakibaei et al. (2021); Sharifi and Khavarian-Garmsir (2020); Spadaro and Pirlone (2021)
R5	Economic performance $\stackrel{+}{\to }$ Active and public transport policies $\stackrel{+}{\to }$ Transport infrastructure investments $\overline{\to }$ Private motorized travel demand $\stackrel{+}{\to }$ Traffic congestion $\overline{\to }$ Economic performance	Arellana et al. (2020), Gusheva and Gooyert (2021); Hasselwander et al. (2021); Mostafavi et al. (2014); Munawar et al. (2021); Shakibaei et al. (2021); Sharifi and Khavarian-Garmsir (2020); Spadaro and Pirlone (2021)
B1	Vulnerable social groups $\stackrel{+}{\to }$ Public travel demand $\stackrel{+}{\to }$ Active and public transport policies $\stackrel{+}{\to }$ Transport infrastructure investments $\overline{\to }$ Private motorized travel demand $\stackrel{+}{\to }$ Traffic congestion $\overline{\to }$ Economic performance $\stackrel{+}{\to }$ Employment $\overline{\to }$ Vulnerable social groups	Arellana et al. (2020), Chaves et al. (2019); Gusheva and Gooyert (2021); Hasselwander et al. (2021); Kamga and Eickemeyer (2021); Mostafavi et al. (2014); Munawar et al. (2021); Shakibaei et al. (2021); Sharifi and Khavarian-Garmsir (2020); Spadaro and Pirlone (2021); Suryani et al. (2019)
B2	Economic performance $\stackrel{+}{\to }$ Employment $\overline{\to }$ Vulnerable social groups $\stackrel{+}{\to }$ Public travel demand $\overline{\to }$ Private motorized travel demand $\stackrel{+}{\to }$ Traffic congestion $\overline{\to }$ Economic performance	Arellana et al. (2020), Chaves et al. (2019); Gusheva and Gooyert (2021); Kamga and Eickemeyer (2021); Moradi and Vagnoni (2018); Munawar et al. (2021); Shakibaei et al. (2021); Spadaro and Pirlone (2021); Suryani et al. (2019), Xie and Wang (2018)
B3	Private motorized travel demand $\stackrel{+}{\to }$ Traffic congestion $\stackrel{+}{\to }$ Travel time $\overline{\to }$ Private motorized travel demand	Arellana et al. (2020), Munawar et al. (2021); Orro et al. (2020); Shakibaei et al. (2021); Sharifi and Khavarian-Garmsir (2020); Spadaro and Pirlone (2021); Suryani et al. (2019)

Table A2.

Schematic representation of feedback loops in Fig. 3 and list of references related to causal links.

Reinforcing (Rcv) or balancing (Bcv)	Feedback loop	References
Rcv1	Transport infrastructure investments (sustainable) $\overline{\to }$ Crowding $\stackrel{+}{\to }$ Contagion $\stackrel{+}{\to }$ Number of infection and death cases $\stackrel{+}{\to }$ Containment measures $\stackrel{+}{\to }$ Activities from home (adaptation to measures) $\overline{\to }$ Travel distance $\stackrel{+}{\to }$ Public travel demand $\overline{\to }$ Private motorized travel demand $\stackrel{+}{\to }$ Private motorized transport policies $\overline{\to }$ Transport infrastructure investments	Aloi et al. (2020), Arellana et al. (2020); Dodson (2014); Hasselwander et al. (2021); Kamga and Eickemeyer (2021); Moslem et al. (2020); Mouratidis, Peters, and van Wee (2021); Munawar et al. (2021); Orro et al. (2020); Sahraei et al. (2021); Shakibaei et al. (2021); Sharifi and Khavarian-Garmsir (2020); Suryani et al. (2019)
Rcv2	Transport infrastructure investments $\stackrel{+}{\to }$ Traffic safety $\overline{\to }$ Traffic congestion $\overline{\to }$ Economic performance $\stackrel{+}{\to }$ Active and public transport policies $\stackrel{+}{\to }$ Transport infrastructure investments	Arellana et al. (2020), Chaves et al. (2019); Gusheva and Gooyert (2021); Mostafavi et al. (2014); Munawar et al. (2021); Suryani et al. (2019))
Rcv3	Activities from home $\overline{\to }$ Travel distance $\overline{\to }$ Active travel demand $\overline{\to }$ Private motorized travel demand $\stackrel{+}{\to }$ Traffic congestion $\overline{\to }$ Economic performance $\stackrel{+}{\to }$ Technological strengths $\stackrel{+}{\to }$ Activities from home	Arellana et al. (2020), Chaves et al. (2019); Dodson (2014); Kakderi et al. (2021); Leung, Burke, and Cui (2018); Mouratidis et al. (2021); Munawar et al. (2021); Shakibaei et al. (2021); Spadaro and Pirlone (2021); Suryani et al. (2019)
Bcv1	Contagion $\stackrel{+}{\to }$ Number of infection and death cases $\stackrel{+}{\to }$ Containment measures $\overline{\to }$ Contagion	Arellana et al. (2020), Hasselwander et al. (2021); Kakderi et al. (2021); Moslem et al. (2020); Sharifi and Khavarian-Garmsir (2020)
Bcv2	Crowding $\stackrel{+}{\to }$ Contagion $\stackrel{+}{\to }$ Number of infection and death cases $\stackrel{+}{\to }$ Containment measures $\stackrel{+}{\to }$ Activities from home $\overline{\to }$ Crowding	Arellana et al. (2020), Hasselwander et al. (2021); Kakderi et al. (2021); Kamga and Eickemeyer (2021); Moslem et al. (2020); Pase, Chiariotti, Zanella, and Zorzi (2020); Shakibaei et al. (2021); Sharifi and Khavarian-Garmsir (2020)
Bcv3	Crowding $\stackrel{+}{\to }$ Contagion $\stackrel{+}{\to }$ Number of infection and death cases $\stackrel{+}{\to }$ Containment measures $\stackrel{+}{\to }$ Activities from home $\overline{\to }$ Travel distance $\stackrel{+}{\to }$ Public travel demand $\stackrel{+}{\to }$Crowding	Aloi et al. (2020), Arellana et al. (2020); Hasselwander et al. (2021); Kamga and Eickemeyer (2021); Moslem et al. (2020); Mouratidis et al. (2021); Munawar et al. (2021); Orro et al. (2020); Pase et al. (2020); Sahraei et al. (2021); Shakibaei et al. (2021)
Bcv4	Transport infrastructure investments $\stackrel{+}{\to }$ Traffic safety $\overline{\to }$ Traffic congestion $\overline{\to }$ Economic performance $\stackrel{+}{\to }$ Private motorized policies $\overline{\to }$ Transport infrastructure investments	Arellana et al. (2020), Chaves et al. (2019); Gusheva and Gooyert (2021); Hasselwander et al. (2021); Mostafavi et al. (2014); Munawar et al. (2021); Sharifi and Khavarian-Garmsir (2020); Suryani et al. (2019)
Bcv5	Government financial assistance $\overline{\to }$ Vulnerable social groups $\overline{\to }$ Activities from home $\overline{\to }$ Crowding $\stackrel{+}{\to }$ Contagion $\stackrel{+}{\to }$ Number of infection and death cases $\stackrel{+}{\to }$ Containment measures $\stackrel{+}{\to }$ Government financial assistance	Arellana et al. (2020), Hasselwander et al. (2021); Moslem et al. (2020); Pase et al. (2020); Shakibaei et al. (2021)
Bcv6	Contagion $\stackrel{+}{\to }$ Number of infection and death cases $\stackrel{+}{\to }$ Containment measures $\stackrel{+}{\to }$ Government financial assistance $\overline{\to }$ Vulnerable social groups $\stackrel{+}{\to }$ Public travel demand $\overline{\to }$ Crowding $\stackrel{+}{\to }$ Contagion	Aloi et al. (2020), Arellana et al. (2020); Hasselwander et al. (2021); Kamga and Eickemeyer (2021); Moslem et al. (2020); Munawar et al. (2021); Orro et al. (2020); Sahraei et al. (2021); Shakibaei et al. (2021)
Bcv7	Transport infrastructure investments $\overline{\to }$ Crowding $\stackrel{+}{\to }$ Contagion $\stackrel{+}{\to }$ Number of infection and death cases $\stackrel{+}{\to }$ Containment measures $\stackrel{+}{\to }$ Activities from home $\overline{\to }$ Travel distance $\overline{\to }$ Active travel demand $\overline{\to }$ Private motorized travel demand $\stackrel{+}{\to }$ Private motorized transport policies $\overline{\to }$ Transport infrastructure investments	Aloi et al. (2020), Arellana et al. (2020); Dodson (2014); Hasselwander et al. (2021); Kamga and Eickemeyer (2021); Leung et al. (2018); Moslem et al. (2020); Mouratidis et al. (2021); Shakibaei et al. (2021); Sharifi and Khavarian-Garmsir (2020); Suryani et al. (2019)
Bcv8	Activities from home $\overline{\to }$ Travel distance $\stackrel{+}{\to }$ Public travel demand $\overline{\to }$ Private motorized travel demand $\stackrel{+}{\to }$ Traffic congestion $\overline{\to }$ Economic performance $\stackrel{+}{\to }$ Technological strengths $\stackrel{+}{\to }$ Activities from home	Arellana et al. (2020), Chaves et al. (2019); Dodson (2014); Kakderi et al. (2021); Mouratidis et al. (2021); Munawar et al. (2021); Shakibaei et al. (2021); Spadaro and Pirlone (2021); Suryani et al. (2019), Xie and Wang (2018)

Table A3.

Schematic representation of feedback loops in Fig. 5 and list of references related to causal links.

Reinforcing (Rcv) or balancing (Bcv)	Feedback loop	References
Rr1	Public awareness $\stackrel{+}{\to }$ Public travel demand $\stackrel{+}{\to }$ Active and public transport policies $\stackrel{+}{\to }$ Public awareness	Gusheva and Gooyert (2021), Moradi and Vagnoni (2018); Spadaro and Pirlone (2021)
Rr2	Public awareness $\stackrel{+}{\to }$ Active travel demand $\stackrel{+}{\to }$ Active and public transport policies $\stackrel{+}{\to }$ Public awareness	Gusheva and Gooyert (2021), Moradi and Vagnoni (2018); Spadaro and Pirlone (2021)
Rr3	Research and development $\overline{\to }$ Spread of pandemic $\stackrel{+}{\to }$ Number of infection and death cases $\stackrel{+}{\to }$ Containment measures $\stackrel{+}{\to }$ Activities from home $\overline{\to }$ Travel distance $\stackrel{+}{\to }$ Public travel demand $\overline{\to }$ Private motorized travel demand $\stackrel{+}{\to }$ Traffic congestion $\overline{\to }$ Economic performance $\stackrel{+}{\to }$ Strategic preparedness and response plan $\stackrel{+}{\to }$ Research and development	Aloi et al. (2020), Arellana et al. (2020); Chan et al. (2021); Chaves et al. (2019); Dodson (2014); Hasselwander et al. (2021); Kakderi et al. (2021); Moslem et al. (2020); Mouratidis et al. (2021); Munawar et al. (2021); Orro et al. (2020); Sahraei et al. (2021); Shakibaei et al. (2021); Sharifi and Khavarian-Garmsir (2020); Spadaro and Pirlone (2021); Suryani et al. (2019), Xie and Wang (2018); Zhang et al. (2021)
Rr4	Strategic preparedness and response plan $\stackrel{+}{\to }$ Research and development $\stackrel{+}{\to }$ Technological strengths $\stackrel{+}{\to }$ Activities from home $\overline{\to }$ Travel distance $\overline{\to }$ Active travel demand $\overline{\to }$ Private motorized travel demand $\stackrel{+}{\to }$ Traffic congestion $\overline{\to }$ Economic performance $\stackrel{+}{\to }$ Strategic preparedness and response plan	Arellana et al. (2020), Chaves et al. (2019); Dodson (2014); Kakderi et al. (2021); Leung et al. (2018); Moradi and Vagnoni (2018); Mouratidis et al. (2021); Munawar et al. (2021); Shakibaei et al. (2021); Sharifi and Khavarian-Garmsir (2020); Spadaro and Pirlone (2021); Suryani et al. (2019)
Rr5	Strategic preparedness and response plan $\stackrel{+}{\to }$ Research and development $\stackrel{+}{\to }$ Technological strengths $\stackrel{+}{\to }$ Activities from home $\overline{\to }$ Crowding $\stackrel{+}{\to }$ Contagion $\stackrel{+}{\to }$ Spread of pandemic $\stackrel{+}{\to }$ Number of infection and death cases $\stackrel{+}{\to }$ Containment measures $\stackrel{+}{\to }$ Government financial assistance $\overline{\to }$ Vulnerable groups $\stackrel{+}{\to }$ Public travel demand $\stackrel{+}{\to }$ Active and public transport policies $\stackrel{+}{\to }$ Transport infrastructure investments $\overline{\to }$ Private motorized travel demand $\stackrel{+}{\to }$ Traffic congestion $\overline{\to }$ Economic performance $\stackrel{+}{\to }$ Strategic preparedness and response plan	Aloi et al. (2020), Arellana et al. (2020); Chan et al. (2021); Chaves et al. (2019); Gusheva and Gooyert (2021); Hasselwander et al. (2021); Kakderi et al. (2021); Kamga and Eickemeyer (2021); Moslem et al. (2020); Mostafavi et al. (2014); Munawar et al. (2021); Orro et al. (2020); Pase et al. (2020); Sahraei et al. (2021); Shakibaei et al. (2021); Sharifi and Khavarian-Garmsir (2020); Spadaro and Pirlone (2021); Suryani et al. (2019), Zhang et al. (2021)
Br1	Spread of pandemic $\stackrel{+}{\to }$ Number of infection and death cases $\stackrel{+}{\to }$ Containment measures $\stackrel{+}{\to }$ Activities from home $\overline{\to }$ Crowding $\stackrel{+}{\to }$ Contagion $\stackrel{+}{\to }$ Spread of pandemic	Arellana et al. (2020), Hasselwander et al. (2021); Kakderi et al. (2021); Moslem et al. (2020); Munawar et al. (2021); Pase et al. (2020); Shakibaei et al. (2021); Sharifi and Khavarian-Garmsir (2020)
Br2	Research and development $\overline{\to }$ Spread of pandemic $\stackrel{+}{\to }$ Number of infection and death cases $\stackrel{+}{\to }$ Containment measures $\stackrel{+}{\to }$ Activities from home $\overline{\to }$ Travel distance $\overline{\to }$ Active travel demand $\overline{\to }$ Private motorized travel demand $\stackrel{+}{\to }$ Traffic congestion $\overline{\to }$ Economic performance $\stackrel{+}{\to }$ Strategic preparedness and response plan $\stackrel{+}{\to }$ Research and development	Aloi et al. (2020), Arellana et al. (2020); Chan et al. (2021); Chaves et al. (2019); Dodson (2014); Hasselwander et al. (2021); Kakderi et al. (2021); Leung et al. (2018); Moslem et al. (2020); Mouratidis et al. (2021); Munawar et al. (2021); Sharifi and Khavarian-Garmsir (2020); Shakibaei et al. (2021); Spadaro and Pirlone (2021); Suryani et al. (2019), Zhang et al. (2021)
Br3	Strategic preparedness and response plan $\stackrel{+}{\to }$ Research and development $\overline{\to }$ Spread of pandemic $\stackrel{+}{\to }$ Number of infection and death cases $\stackrel{+}{\to }$ Containment measures $\stackrel{+}{\to }$ Government financial assistance $\stackrel{+}{\to }$ Economic performance $\stackrel{+}{\to }$ Strategic preparedness and response plan	Arellana et al. (2020), Chan et al. (2021); Hasselwander et al. (2021); Kakderi et al. (2021); Sharifi and Khavarian-Garmsir (2020); Zhang et al. (2021)

Data availability

• Data will be made available on request.