COVID-19 transmission in U.S. transit buses: A scenario-based approach with agent-based simulation modeling (ABSM)

Abstract

The transit bus environment is considered one of the primary sources of transmission of the COVID-19 (SARS-CoV-2) virus. Modeling disease transmission in public buses remains a challenge, especially with uncertainties in passenger boarding, alighting, and onboard movements. Although there are initial findings on the effectiveness of some of the mitigation policies (such as face-covering and ventilation), evidence is scarce on how these policies could affect the onboard transmission risk under a realistic bus setting considering different headways, boarding and alighting patterns, and seating capacity control. This study examines the specific policy regimes that transit agencies implemented during early phases of the COVID-19 pandemic in USA, in which it brings crucial insights on combating current and future epidemics. We use an agent-based simulation model (ABSM) based on standard design characteristics for urban buses in USA and two different service frequency settings (10-min and 20-min headways). We find that wearing face-coverings (surgical masks) significantly reduces onboard transmission rates, from no mitigation rates of 85% in higher-frequency buses and 75% in lower-frequency buses to 12.5%. The most effective prevention outcome is the combination of KN-95 masks, open window policies, and half-capacity seating control during higher-frequency bus services, with an outcome of nearly 0% onboard infection rate. Our results advance understanding of COVID-19 risks in the urban bus environment and contribute to effective mitigation policy design, which is crucial to ensuring passenger safety. The findings of this study provide important policy implications for operational adjustment and safety protocols as transit agencies seek to plan for future emergencies.

1. Introduction

Transit bus use was considered a major vector for community disease transmission during the COVID-19 global pandemic (Shen et al., 2020). The limited space and frequent on-board/discharge of riders across an urban space make the bus particularly vulnerable to COVID-19 transmission. To reduce the risk of indoor transmission in public transit, the Center for Disease Control and Prevention (CDC) along with other science policy bodies responded with guidance on social distancing and other safety protocols, which many transit agencies rapidly implemented (Hymon, 2020; Meyer, 2020; WMATA, 2020). Managing respiratory disease spread in the bus transit industry, among other public transportation sectors, is critical to ensure that captive populations, which were reliant on buses throughout the pandemic (Lou et al., 2020), have a low transmission risk. While the understanding of indoor transmission has advanced compared to the early phase of the COVID-19 pandemic, the risks associated with buses remain largely unknown despite their significance in transmission risks. Buses, while enclosed, have other characteristics that influence the spread of disease (Zhang et al., 2021). Understanding risks in the context of these factors is an important first step to reassuring riders.

Bus transit is particularly crucial for those having limited mobility options in accessing well-being determining destinations. In comparison with choice riders, who have the flexibility to choose their means of transportation, this group of riders relies on transit as their critical lifeline to fulfill their daily needs such as job, school, grocery, and medical care (Taylor and Morris, 2015). When the stay-home order was issued in March 2020 at the beginning of the COVID-19 pandemic, transit ridership across USA dropped by nearly 90% (Ahangari et al., 2020; DeWeese et al., 2020; Jenelius and Cebecauer, 2020) across subways, light rails, and commuter rails, and the bus ridership dropped 60% in the first quarter of 2020 (APTA, 2021). The drastic drop in ridership cascaded into a financial crisis that many transit agencies have yet to resolve, even with the funding support from the federal recovery package in mid-2021 (Blumgart, 2022; EBP US, Inc, 2021; Verma, 2020). The budget deficit has led to reduced transit service in service coverage, hours, and frequency, which further exacerbates the mobility challenge among essential transit riders (Babalık, 2021; He et al., 2022; Liu et al., 2020; Palm et al., 2021). During the pandemic, the available bus service and the headway schedule were increasingly dissonant with ridership needs (González-Hermoso and Freemark, 2021). Despite the transmission risk of COVID-19 disease, it is clear that riders lacking access to a private vehicle and people from lower-income households were still more likely to continue using transit (He et al., 2022).

The ridership data indicated that there were an average of 5.365 million riders boarding buses on daily basis during the pandemic (APTA, 2021). Although the public transit industry responded swiftly by adopting CDC's guidance to lower the transmission risk, such as instituting mask mandates, opening windows, and other protocols, scientific evidence is scarce in evaluating the outcome of these policies. Without clear evidence on the effectiveness of these mitigation policies, uncertainty remains in the efficacy of transit industry management strategies deployed during COVID-19 to reduce transmission risk.

In this study, we develop an agent-based simulation modeling (ABSM) to evaluate the effectiveness of individual and combinations of policies in a transit bus setting. We examine the specific policy regimes that transit agencies adopted in the early stages of the COVID-19 pandemic in USA. Our study helps to characterize the efficacy of mitigation measures on COVID-19 transmission. Based on the standard design of buses used in USA, we model the on-boarding and off-boarding activities and examine how masks, ventilation, seating capacity control, and passenger positivity rate affect disease transmission. We establish base scenarios and expand them with additional policy settings to explore the effectiveness of commonly applied guidelines adopted by the U.S. transit industry during the pandemic. Our findings show that with no mitigation, the highest infection rate reaches as high as 85% for higher- and 75% for lower-frequency bus service. We find that wearing face-covering (or surgical masks) significantly reduces onboard transmission rates to as little as 12.5% in both service frequency scenarios. The most effective prevention outcome results from the combination of KN-95 masks, open window policies, and half-capacity seating control during higher-frequency bus service, with nearly 0% onboard infection rate. We also find that seating capacity control (the social-distancing policy) provides little improvement in effectiveness for lower-frequency bus services when KN-95 masks and open-window policies are already in place. Our results advance the understanding of COVID-19 risks in the urban bus environment and contribute to effective mitigation policy design, which is crucial to ensuring passenger safety. Our findings can help to refine operational adjustments and boarding protocols to service providers as transit agencies seek to resume pre-pandemic service capacity and facilitate ridership increase across the country.

In Section 2, we begin with a literature review on closed space COVID-19 transmission, buses as constrained spaces and the risks of airborne transmission disease, and we present our reference method for transit bus operations and design features. In Section 3, we present details of the ABSM specification and review the simulation settings represented in each scenario. Finally, we present our modeling results and discuss findings within the context of mitigation measures based on the commonly used suggestions among public health and transit agencies.

2. Literature review

The management and control of respiratory illness in public transportation plays an important role in minimizing community outbreaks. During the COVID-19 (SARS-CoV-2) outbreaks, transit buses were identified as a major source of respiratory disease spread in public space (Yasri and Wiwanitkit, 2020). The frequent on- and off-boarding activities in a constrained space can facilitate the transmission of particles, which can be suspended in the indoor air from minutes to hours (Peng and Jimenez, 2021) without effective mitigation. One estimate based on a single tourist bus event in Thailand early in the pandemic resulted in uncertainties in infection rates ranging from 4% to 24% (Yasri and Wiwanitkit, 2020). Using daily transportation data leaving Wuhan City and the infection data in destination cities from January 2020, daily frequencies of using bus transport were highly correlated with the number of daily COVID-19 infection cases (Zheng et al., 2020). A cohort study at the beginning phase of the pandemic (January 2020) showed that COVID-19 transmission on buses was very high: 24 of 68 passengers became infected after riding the bus with one previously infected passenger (Shen et al., 2020). During a community outbreak incident on January 19, 2020, with two buses making a 100-min round trip for a 150-min worship event, bus passengers riding with the source patient had a 34.4% (or 11.4 times) higher risk of infection than those riding the bus with no source patient identified. All of these studies were very early in the pandemic and no mitigation was in place.

Although the public transit industry responded to CDC's guidance in issuing and implementing mitigation policies to lower the transmission risk, we lack quantitative evidence on the efficacy of the implemented protocols on buses, which included mask mandates, seating capacity control, ventilation adjustments, and routing restrictions, among other policies. In a comprehensive report released by American Public Transportation Association, initial (largely anecdotal) evidence suggested that mask-wearing was effective in reducing person-to-person transmission within a public transit setting (APTA, 2020). This is consistent with what is now well established: that airborne transmission was a primary cause of COVID-19 population spread (Bazant and Bush, 2021). The risk of transmission has been estimated as 18.7 times greater (95% CI: 6.0 to 57.9) when comparing a closed environment to an open-air environment (Nishiura et al., 2020). This evidence is consistent with other studies indicating that close physical proximity and longer time durations in a shared space can increase the probability of transmission, regardless of the space configuration. This has been shown as applicable for commercial flights (Olsen et al., 2003), an exercise facility (Lendacki et al., 2021), a university classroom (Peng and Jimenez, 2021), a health care facility (Fisher et al., 2020), a restaurant (Chang et al., 2021; Lu et al., 2020), and a school bus (Abulhassan and Davis, 2021a). Transmission risk factors associated with public transit use specifically include proximity in passenger seating (Hu et al., 2021); turbulent airflow within the vehicle (Edwards et al., 2021); time shared in-vehicle (Hu et al., 2021); mask type (Dzisi and Dei, 2020; Edwards et al., 2021) and passenger compliance (Dzisi and Dei, 2020).

Transit service providers adjusted real-time operational policies to lower the potential transmission risk among transit riders. These adjustments primarily aimed to limit service capacity, including lowering the service frequencies, changing timetables and vehicle schedules, and reducing the time duration of daily service (Calderón Peralvo et al., 2022). Additional regulations regarding crowding were also issued; these included policies like limiting station-level boarding and stop-skipping when onboard capacity reaches the limit (Gkiotsalitis, 2021). The primary passenger-oriented mitigation was a mask-wearing policy, alongside social-distancing compliance. Well-fitted surgical masks have been shown to reduce transmission risk on public transit (Zhang et al., 2021). The prevention effectiveness has been assumed from 90% to 30%, where surgical masks are shown to have the highest effectiveness and home-made masks the lowest (Zhang et al., 2020). Studies using computational fluid dynamics find evidence that wearing a mask can reduce the particle dispersion distances as well as the particle count being released into the vehicle environment (Edwards et al., 2021).

A recent literature review conducted by Calderón Peralvo et al. (2022) summarized the preventative policies and operational adjustments that transit agencies operationalized to minimize the COVID-19 transmission risk. The most commonly adopted mitigations included social distancing (6 ​ft), ventilation improvement, mask mandate, and a few other sanitation protocols. Social distancing and ventilation effectiveness are considered among the most important factors for confined space transmission. The physical distancing of 1.6–3.0 ​m helps to avoid large droplet transmission (e.g., from talking) and up to 8.2 ​m for smaller aerosol particles (Sun and Zhai, 2020). Given the difficulty of maintaining social-distancing measures, hygiene and cleaning services were also recommended as a way to reduce transmission risk (DemiRbiLek et al., 2020).

Modeling COVID-19 transmission within a highly uncertain public bus environment is challenging for several reasons, with the primary one being testing the effectiveness of simultaneously enacted mitigation policies. Although social distancing reduces infection rate by 20%–40% during the first 30 ​min, the effect of ventilation requirement is uncertain and dependent on exposure time, air distribution systems, and social-distancing conditions. In addition to ventilation, virus movement is affected by the direction and intensity of airflow (Fisher et al., 2020). There are only a few studies examining the characteristics of aerosol dispersion and simulating the infection rate in a bus environment (Edwards et al., 2021; Zhang et al., 2021); most of them focus on one or two particular factors without considering the full environmental setting. Initial passenger infection rates, which are critical in reducing onboard passenger transmission, have not been included in modeling strategies to date. These are also significant gaps in our understanding of how concomitant mitigation policies work together to reduce passenger transmission. The relationships between the transmission rate and the mitigation policies during bus service (such as the face-covering policy, seating capacity control, and open window policy) have not quantified. Our study begins to address these gaps by examining transmission risk in the context of vehicle ventilation setting, passenger seating control, masking policy, and realistic bus boarding and alighting simulation.

3. Methodology

We develop an agent-based simulation modeling (ABSM) to simulate the behavior of transit riders on a bus and the spread of COVID-19 infections. An ABSM allows us to account for individual-level behavior (Crooks et al., 2018; Macal and North, 2010), integrate both social and spatial features (Crooks and Heppenstall, 2012), introduce randomness (Bonabeau, 2002; Crooks and Heppenstall, 2012), and account for different time scales (Batty and Longley, 2003), which are important in a setting where other environmental factors change over time (Crooks et al., 2018). ABSM's have been used in other examinations of COVID-19 transmission including the influence of social distancing policies on epidemiologic and economic outcomes (Silva et al., 2020) and the use of genomic surveillance to estimate containment outcomes (Rockett et al., 2020). The models perform well when data are constrained or fine-grained estimation is appropriate, especially within closed-space facilities (Cuevas, 2020). Research has also shown that ABSM's can successfully capture transmission related to exposure time (D'Orazio et al., 2020), incubation period (Chang et al., 2020; Silva et al., 2020), social distancing (Chang et al., 2020; Gharakhanlou and Hooshangi, 2020), infection probability (Cuevas, 2020), and severity and death rates (Silva et al., 2020).

Our model uses a conventional urban bus design, and we developed a standard bus schematic after reviewing a number of different bus designs in USA (Table 1 ). The classic transit bus has a capacity that ranges from 39 to 52 seats, and wheelchair lifts are often required for American transit operators. Seating maps for buses can vary extensively, for example, the Hagley Coach shows three different designs for the same-sized buses (Hagey Coach Inc., 2021). According to descriptions from Federal Transit Authority (FTA), (FTA, 2021), the Colorado DOT (Colorado DOT, 2021), and other sources (Table 1), the bus environment can vary in terms of both passenger seating capacity and vehicle dimensions. Based on these standards, we established a bus design with 40 seats (capacity), 40 feet in length (480 inches), 102 inches in width, and the leg-to-seat space at 40 inches. This design complies with the commonly applied standards and is realistic for our modeling approach.

Table 1.

Bus design standards.

Company	Capacity	Length (ft)	Width (in)	Leg-seat space (in)	Source
ABSM	40	40	102	40	Our ABSM Design
CL DOT	42	40	NA	NA	Colorado DOT (2021)
CL DOT	30	35	NA	NA	Colorado DOT (2021)
Grech Motors	37–43	26–35	72	28–32.5	Grech Motors (2021)
Hagey Coach	54	NA	NA	NA	Hagey Coach Inc. (2021)
Krystal Int'l	31–37	31/35/38	102	NA	Absolute Bus Sales (2021)
General Motors	48	40	102	NA	Market Street Railway (2020)
Nova Bus	41	40	102	NA	Nova Bus (2017)
Tesco Bus	16	33	96	31	Tesco Bus (2019)

We perform 500 simulations for each of our proposed scenarios using two types of agents: passenger and virus. We discuss each of the scenarios in the next section. We assign passenger agents with entering and departure times (stops). The virus spreads when a passenger is infected while seated on the transit bus. We simplify the on- and off-board movements for passengers by placing them directly in their seats. Additionally, we assume that a passenger agent stays in their seat between entering and exiting times. We use a uniform distribution for the random sampling of infection probability to generate new passenger-agents, both in the beginning of each 12-h simulation and at every bus stop, and to identify exiting passenger-agents during on- and off-board simulations.

Modeling the virus agent is challenging because of external factors (i.e., wind speed, wind direction, temperature, humidity, virus size, virus traveling speed and distance) that can affect the transmission of COVID-19. To model the spread of the virus through the air, we first assign a random floating movement to our virus agent once an infected passenger has boarded the bus. We also assign each virus agent a “life-span” metric (in minutes) according to the literature, ranging from 8 to 14 ​min (Stadnytskyi et al., 2020). When this life-span is reached, a virus will disappear (or die) in the modeling environment. To capture infection, if there is direct contact between a passenger and a virus agent (e.g., sitting next to each other) the uninfected passenger can be infected based on their mask type (e.g., 20% infectious probability if wearing surgical masks, 5% infectious probability if wearing KN-95 masks). The infection probability adopted in our model assumes that all riders are required to wear the same type of masks uniformly for each scenario.

The probability of being infected can also change depending on outside air exchange. When virus agents are close to an open window, for instance, there can be exportation of some of the transmittable virus droplets outdoors or dilution of the virus. If windows (shown in green, Fig. 1 ) are open during the bus ride, the virus movement will change depending on the airflow, which is a function of the number of opened windows as well as their locations within the bus. We also position doors (shown in blue, Fig. 1) in the front and back of our bus design. We use the right front door for passenger on- and off-boarding and the back door for emergency use.

Fig. 1.

Illustration of transit bus design applied in our ABSM. Note: The bus we deploy has a capacity of 40 seats, length of 40 feet, width of 102 inches, and leg-seat space of 40 inches. Agents in this figure are illustrative purposes only.

To start the simulation (Fig. 2 ), we assume that each passenger will board the bus within 60 ​s of time (A1). The initial number of boarding passengers that are infected is set at 10% of the total initial bus riders (A2). Each simulation starts with at least one infected passenger; infected passengers produce virus-agents that circulate based on their seated locations. These virus-agents move within the bus geometry and depending on the specific masking protection can infect other non-infected passenger-agents. Newly infected agents create more virus-agents around their seated locations, and this pattern continues until the end of the 12-h simulation. Further details of passenger infection, virus movement and interactions between the agents are included in the next section.

Fig. 2.

ABSM simulation flowchart.

3.1. Modeling scenarios

For our modeling scenarios, we vary the probability of infection, the transmission type, the ventilation (windows opening), the average trip time between two stops (headways), overall bus trip duration, and passenger capacity limit. Our variables and their respective ranges are taken from the current literature (Table 2 ). For example, it is clear that ventilation plays a critical role in reducing indoor transmission (Dai and Zhao, 2020). An infection probability of 2% is widely used in COVID-19 studies (Dai and Zhao, 2020; Sun and Zhai, 2020b) with some research adopting between 10% and 20% (Cuevas, 2020; Dorra Louati et al., 2020). We vary our infection rate between 5% and 100% based on the various masking possibilities and include a pre-pandemic non-masking setting (100%). We use a typical bus design with 20 windows on two sides of the bus (10 on each side) and randomly select opening and closing windows in our simulation rules. To realistically model the randomness of windows opening between simulations, and to select different sets of windows as a part of the mitigation strategies, we keep 10 out of 20 windows randomly open (50% of all windows) in each simulation for scenarios that implement window-opening policies as a mitigation strategy.

Table 2.

Model metrics and scenarios.

Metric	Parameter	Source
Probability of Infection	5% (N95/KN-95 mask)	Andrejko et al. (2022)
	20% (surgical mask)	Cuevas (2020)
	50% (cloth mask)	Dai and Zhao (2020)
	100% (no mask)	Dorra Louati et al. (2020)
		Ganapathy et al. (2021)
		Shen et al. (2020)
		Sun and Zhai (2020)
Passenger Capacity	25%	Abulhassan and Davis (2021)
	50%	Colorado DOT (2021)
	75%	Melnick and Darling-Hammond (2020)
	100%	Shen et al. (2020)
		Sun and Zhai (2020)
		Wang et al. (2020)
Ventilation (Simplified)	Ventilation on	Abulhassan and Davis (2021)
	Ventilation off	Dai and Zhao (2020)
		Ganapathy et al. (2021)
		Thiboumery (2021)
Average Travel Duration Between Bus Stops	5 ​min	Abulhassan and Davis (2021)
	10 ​min	Shen et al. (2020)
	20 ​min
	30 ​min
Overall Simulation Duration (Cleaning)	8 ​h	Chicago Transit Authority (2021)
	12 ​h	Montgomery County DOT (2021)
	16 ​h	SFMTA (2021); The MTA (2021)
	20 ​h	WMATA (2020)
	24 ​h	Washington Metropolitan Area Transit Authority (2021)
Transmission type	Droplet and airborne	Lu et al. (2020)

We develop scenarios for two types of bus service, 10-min and 20-min headways. Ten-min headways are consistent with higher-frequency bus service and 20-min headways proxy lower-frequency bus service. We develop three scenarios for the higher-frequency service: (1) a base scenario, which reflects pre-pandemic conditions when no masks were required, windows stay closed, and full capacity service was permitted; (2) a midway-mitigation scenario, surgical masks are required; and (3) the most stringent scenario, where KN-95 masks are required, 50% of windows stay open, and the bus operates at half seating capacity (20 seats). For the lower frequency service, we set the first three scenarios to reflect the same levels of mitigation as the higher-frequency service (described above), with an additional scenario to evaluate the effectiveness of the passenger capacity control policy that was deployed by many transit agencies. This additional scenario (LF_2) allows us to examine what happens if a lower-frequency bus is occupied at full passenger capacity while imposing KN-95 and having windows open (50% of all windows) as mitigation measures. The detailed descriptions of each modeling scenario for higher- and lower-frequency services are shown in Table 3 .

Table 3.

ABSM simulation scenarios.

Bus Service Type	Higher Frequency Bus Service			Lower Frequency Bus Service
	(10-min headways)			(20-min headways)
Modeling Parameters	HF_Base	HF_1	HF_2	LF_Base	LF_1	LF_2	LF_3
Bus Design
Allowable Capacity	40	40	20	40	40	40	20
Windows	Closed	Closed	Open	Closed	Closed	Open	Open
Passenger Levers
Infection Probabilitya	100%	20%	5%	100%	20%	5%	5%
Embark/Debark Social Distance	—	—	6 ​ft	—	—	—	6 ​ft

^aThe type of masking establishes the probability of infection: 100% reflects no masking (early pandemic); 20% denotes surgical masks; 5% reflects KN-95 masks.

We use a 12-h simulation period (720 ​min) across the seven scenarios. We randomly generate the initial number of passengers (with a minimum of one passenger) and the initial-infected passengers is 10% of the total initial passengers. The maximum number of passengers is a function of the allowable capacity (40 or 20 passengers for full or half capacity) for any given simulation. We adopt a 2-min stop time to allow passengers to exit in the first min, and new passengers to enter in the second min at each stop. For example, in higher-frequency services, the bus stops at the 10th min, allows passengers to exit at the 11th min, and has new passengers enter at the 12th min. The next time the bus stops is at the 22nd min (takes 10 ​min to the next stop), passengers exit at the 23rd min and new ones enter at the 24th min, and so on for 720-min simulation period, at which point the bus service ends for the day. Since we are modeling passengers on urban transit buses, we set a maximum onboard time of 1h. We randomly assign the entering and exiting times for each passenger to achieve a realistic onboard time duration. We quantify passengers who enter, depart and are infected, the exiting passenger, exiting-and-infected passenger, as well as passengers who are entering, seating, and seating-and-infected at every time step (from 0 to 720 ​min).

4. Results and discussion

Fig. 3 shows our results for the average onboard infection rates across each of our scenarios. As we might expect, the highest average on-board infection rates, 38% and 23%, are observed in the higher- (left) and lower-frequency (right) base scenarios (HF_Base and LF_Base, green curves), respectively. These scenarios reflect pre-masking requirements. Mitigation scenarios with surgical/face-coverings only (HF_1 and LF_1, brown curves) show a significant decrease in the overall infection rate. This finding is consistent with the demonstrated importance of even limited masking in transit buses in reducing COVID-19 transmission during the onboard time (Dzisi and Dei, 2020; Edwards et al., 2021; Zhang et al., 2021). The average infection risks in both higher- and lower-frequency buses decline with additional mitigation policies (HF_2, LF_2, and LF_3), including masking with better prevention rates (i.e., KN-95 masks), opening windows, and the implementation of seating control policies.

Fig. 3.

Average on-board infection rate for 7 scenarios in higher-frequency bus services (left) and lower-frequency bus services (right) with the standard error (SE).

There is a high level of variability in the high frequency (no mask) base scenario (top, Fig. 4 ), where onboard infection can reach as high as 85%. Results also show that the rate of infection is relatively steady for the first 200 ​min, increasing significantly after about 200-min. This is likely due to the effect of circulating virus agents, which are generally low at the start of a run and gradually increase as passengers’ board and exit the bus. The surgical mask scenario (HF_1) results in significantly lower infection rates with the highest around 20% (middle, Fig. 4). In the HF_2 scenario (bottom, Fig. 4), under a more stringent mitigation setting that includes KN-95 masking, open windows, and a reduction in seating capacity reduction (or social-distancing) to 50% of original bus capacity (from 40 to 20 seats). Both scenarios show significant reductions in transmission risk, where the highest infection rates are less than 12.5% and 3%, respectively. The benefit of the HF_2 policies is a reduction in the variability of infection risk.

Fig. 4.

Infection rate for higher-frequency (10-min) bus services (3 scenarios). Note: Redline denotes the scaled average infection rate among all simulations; HF_Base refers to pre-pandemic situation; HF_1 refers to mitigation strategy of surgical mask mandate (20% transmission rate); HF_2 refers to KN-95 mask mandate (5% transmission rate), half-capacity seating, and open-window policies.

The highest onboard passenger infection rates for our lower-frequency bus service scenarios reach 75% without other mitigation policy interventions (LF_Base, Fig. 5 ). We also find that the lower-frequency buses have relatively steady infection rates for the first 400 ​min of service duration, with a maximum of nearly 25% infection rate. This initial steady onboard infection rate indicates that, absent other mitigation measures, reducing bus service time as a proposed transmission-prevention method may not be as effective as some transit agencies perceived. The results from our second scenario (LF_1), where surgical masking is required, are consistent with the simulation results from the higher-frequency service model. When comparing the results from the third scenario, with KN-95 masking and windows opening (LF_2), and the fourth scenario, with KN-95 masks, open windows, and seating capacity limit (LF_3), we do not find a large variation in the simulation outcomes for infection risk. This indicates that the combination of KN-95 and open windows is sufficiently effective to limit infection risk among onboard passengers on buses with lower-frequency headways. Opening windows to dilute viruses while wearing KN-95 seems to reduce infection risk regardless of changes in allowable bus capacity.

Fig. 5.

Infection rate for lower-frequency (20-min) bus services (4 scenarios). Note: Redline denotes the scaled average infection rate among all simulations; LF_Base refers to no mitigation implementation; LF_Base refers to pre-pandemic situation; LF_1 refers to mitigation strategy of surgical mask mandate (20% transmission rate); LF_2 refers to KN-95 mask mandate (5% transmission rate) and open-window policies; and LF_3 refers to KN-95 mask mandate (5% transmission rate), half-capacity seating, and open-window policies.

5. Conclusions and future work

We use agent-based modeling to simulate the effects of COVID-19 in a standard U.S. transit bus environment. We model base (pre-pandemic) and mitigation scenarios for both higher-frequency (20-min headways) and lower-frequency (10-min headways) bus services. Despite initially low, infection rates for the base scenario for the high frequency and low frequency base scenarios reaches as high as 85% and 75%, respectively. Our simulation results provide evidence for the high effectiveness of masking policies as well as the efficacy of opening windows (or ventilation mitigation) in both higher- and lower-frequency bus services. We find that wearing surgical masks (or any comparable face covering) alone significantly reduces modeled onboard transmission rates, from the highest rate of 80% in higher-frequency buses and 75% in lower-frequency buses to 12.5% in both service types. The most effective prevention outcome is reflected when using the combination of KN-95 masks, open window policies, and half-capacity seating control during higher-frequency bus service, with an outcome of a near zero percent, on average, onboard infection rate. We also find that seating capacity control (namely social-distancing policy) offers little improvement in effectiveness among lower-frequency bus services when KN-95 masks and open window policies are both implemented.

Policy Implications We examined policy regimes adopted by transit authorities in the early phases of COVID-19 pandemic. Among these policies, the most widely used protocols for transit buses included 6-ft social distancing, masking restrictions, ventilation improvement as well as seat capacity controls. Although these transmission risk mitigation practices were deployed by many transit authorities in the U.S. (Federal Transit Administration, 2022), evidence is scarce as to the effectiveness of each protocol individually, and the combined efficacy of the policies. As the transit industry experiences recovery in bus ridership, modeling the COVID-19 transmission risk under the commonly adopted policies will provide an integral policy reference going forward. From an equity perspective, our findings on the high effectiveness of KN-95 masks as well as its combined effect with open window policies and seating capacity control, even during higher-frequency bus services, provides important evidence to support the safety protocols adopted during the pandemic by many transit agencies. Moreover, it suggests that under any time of high infectious disease spread, masking is an optimal policy. Our evidence on the effectiveness of these mitigation measures also suggests that captive riders can be protected, and that incentives should include offering masks at time of on boarding. From the transit agency's perspective, understanding the effectiveness of the mitigation policies can help with planning for future bus operations under both normal and emergency conditions. Our finding on the marginal difference made by social distancing in addition to the protection of KN-95 and the open window policy points to the potential limiting of the application of transit social distancing policies. The COVID-19 pandemic will have long-term effects on the travel behavior in both passengers' needs for transit services and their perception of safety towards public transit systems (de Palma et al., 2022). As we slowly recover from the ongoing pandemic and prepare for future activities, research evidence shows that essential transit riders and socially vulnerable groups are more likely to use public transit continuously (He et al., 2022; Palm et al., 2022; Paul and Taylor, 2022). This paper contributes to a better understanding of the uncertainties of COVID-19 transmission risk among transit riders and provides a quantitative evaluation of policy regimes adopted by transit agencies in USA, which are crucial to addressing potential health disparities under changing transit operational policies (de Palma et al., 2022).

Future work should build on a number of aspects of our research. Disease transmission is determined by various characteristics of viruses, for example, virus agent size, traveling speed and distance of disease agent, prevention effectiveness of masks, as well as particle counts and respiration rates based on different passenger types (e.g., adults versus children). Future studies can improve understanding of passenger transmission of infectious airborne diseases by incorporating respiratory and local environmental characteristics. Another factor to consider is the efficacy of ventilation systems in comparison to window opening mitigation as well as their impact on virus-agent's movement. Additional scenarios reflecting both measures would shed further insight on virus movement. Our study also simplifies the boarding and alighting pattern through random passenger assignment within a realistic time range; future research could improve this by incorporating context-based simulation with a greater range of ridership patterns.

Replication and data sharing

The data models and software code from this study are available on http://github.com/sachraa1/BusABM.

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.

Biographies

Sachraa G. Borjigin is a Postdoctoral Research Fellow at the Department of Environmental Health Sciences, University of Michigan, Ann Arbor. He obtained his Ph.D. degree in Civil Engineering (Systems) from the Department of Civil and Environmental Engineering, University of Maryland, College Park in May 2022. His main research focus is at the intersected area of transportation systems, urban development, human mobility, environmental health sciences, public policy, and with a goal of advancing equity in communities. He is also interested in evaluating urban and societal problems with policy penetrations, integrated through various computational models and methods (e.g., AI & Machine Learning.)

Qian He is a Postdoctoral Research Associate at the University of Maryland, College Park. Her research focuses on the risks and resilience of vulnerable communities at the intersection of the urban built environment, sustainability, and community development. Her publications and ongoing projects use advanced modeling and urban informatics techniques to examine how planning and public policies affect the well-being of historically disadvantaged communities such as the Black, Indigenous, and People of Color (BIPOC) communities facing natural and societal hazards.

Deb A.Niemeier is the Clark Distinguished Chair Professor and Director of the Center for Disaster Resilience. At the University of Maryland, Deb Niemeier's research targets aspects of the built environment that give rise to structural inequality, particularly within the context of climate change. She holds the Clark Distinguished Chair and is a Professor in Civil Engineering and the director of the Center for Disaster and Resilience, established to help foster community solutions to interdisciplinary societal problems. Her work is justice-centered and focused on structural inequities and questions of access. She is interested in understanding how formal and informal governance processes in urban planning shape community resilience in the face of disasters.