Pollution hot spots and the impact of drive-through COVID-19 testing sites on urban air quality

To be successful, commitments to climate change and environmental policy will require critical changes in human behavior and one important example is driving and idling. Idling is defined as running a vehicle’s motor while not in motion. Idling studies have repeatedly demonstrated that this behavior is costly, harmful to human health, and highly polluting. However, with the onset of COVID-19, the use of drive-through services to provide food, pharmaceuticals, and medical testing has increased. To understand this phenomenon further, we worked cooperatively with our government partners to compare the concentrations of PM2.5 at three regulatory sensor locations with nearby drive-through COVID-19 testing sites during average to elevated pollution days. Salt Lake City, UT (USA), where this study was undertaken, has seen a dramatic rise in drive-through services since the onset of the pandemic and community concern is also high due to poor local air quality. More importantly, the Salt Lake Valley is home to one of the largest research grade air quality networks in the world. Fine particulate matter sensors were installed or already in place at or adjacent to COVID-19 testing sites in the area, and we used data from nearby Utah Division of Air Quality monitors to provide comparative PM2.5 concentrations. Due to their placement (e.g., further distance from large roads and other emitting sources), we found that testing sites showed lower PM2.5 concentrations during average air quality days despite increased idling rates. However, when urban pollution rates were elevated due to atmospheric inversions, extensive idling around testing sites led to hyper local PM2.5 concentrations or pollution hot spots. This suggests that idling has serious compounding effects in highly polluted urban areas and policies minimizing vehicle emissions from idling and congestion could conceivably curtail pollutant exposure in a range of settings.


Introduction
Commitments to climate-change and other environmental problems will require critical changes in human behavior (Pearson et al 2016) to assure long term human health and well-being. One important example of behavior modification, which may contribute to the efforts to improve human health and the environment, is vehicle idling reduction. Idling is defined as running a vehicle's motor while not in motion (Carrico et al 2009). Idling studies have repeatedly demonstrated that such a habit is inefficient, damaging to the vehicle, and highly polluting (Kinsey et al 2007, Mendoza et al 2022b. Most importantly, it is harmful to human health (Vishnevetsky et al 2015, Breton et al 2016.
In addition, a range of fields have tackled this interdisciplinary problem. As a clear consensus has emerged, there appears to be a strong relationship between pollution generated by idling (Kinsey et al 2007, Richmond-Bryant et al 2009, Kim et al 2014, Lee et al 2018, interventions to reduce idling (Eghbalnia et al 2013, Meleady et al 2017, Mahmood et al 2019, Rumchev et al 2021, and the resulting impacts of idling on human health , Lynn et al 2014. With the onset of COVID-19, however, the use of drive-through services for food, pharmaceuticals, and medical testing has increased (Cohen et al 2022). Since the scientific understanding of the dispersion of pollutants in urban areas due to idling is extremely complex and difficult to measure with precision, gaps remain in this area. As a result, few scientific studies of this phenomena exist, even with the increased use of drive-through conveniences since 2020.
In Salt Lake City, Utah, where this study was undertaken, such services have become increasingly common. At the same time, idling has been identified by local stakeholders as an area of concern (Benney et al 2021, Mendoza et al 2022a) due to the interactive effects of idling and other forms of air pollution. Since the Salt Lake Valley is home to one of the densest research grade air quality networks in the world (Mendoza et al 2019), air quality sensors were installed or already available at or adjacent to COVID-19 testing sites in the area. To learn more about vehicle congestion and idling at COVID-19 testing sites during the 2020-21 pandemic, a government-academic partnership was developed to consider the human health and policy implications of this phenomena. The goal of this partnership was to quantitatively evaluate if traffic and vehicle idling impacts local PM 2.5 particulate formation.
To advance this understanding, we compared the concentrations of PM 2.5 at three regulatory sensor locations with nearby drive-through COVID-19 testing sites in an urban area during wintertime average to elevated pollution days. We used data from nearby Utah Division of Air Quality (UDAQ) monitors to provide comparative PM 2.5 concentrations (Utah Division of Air Quality 2021). Due to their placement (e.g., further distance from large roads and other emitting sources), we found that testing sites showed lower PM 2.5 concentrations during average air quality days despite increased idling rates. However, when urban pollution rates were elevated due to atmospheric inversions, extensive idling around testing sites led to hyper local PM 2.5 concentrations or pollution hot spots. This suggests that idling has serious compounding effects in highly polluted urban areas and policy may need to account for these factors to protect human health and well-being. In addition, best management practices and policies minimizing vehicle emissions from idling and congestion could conceivably curtail pollutant exposure in a range of settings.

Methodology
This study compares the concentrations of PM 2.5 at three regulatory sensor locations with nearby drive-through COVID-19 testing sites in Salt Lake City, UT. This urban area was studied because of the ability to use high quality, regulatory and research grade sensors to enable direct measurement of the phenomena of interest. Additionally, the region's network was substantially enhanced in 2015 which allows our team to understand pre-existing conditions in an urban area during average to elevated pollution days. We used data from nearby UDAQ monitors to provide comparative PM 2.5 concentrations.
As shown in figure 1, Salt Lake County (SLCo), Utah, sits within a valley, or basin, bracketed by the Great Salt Lake (Northwest), and the Wasatch (East), Oquirrh (West), and Traverse (South) mountains (figure 1). Given the right meteorological conditions, the SLCo basin traps airborne pollutants emitted from sources, including vehicles, under an inversion layer resulting in stagnation and preventing the normal vertical mixing of warm and cool air (Whiteman et al 2014). Strong and weak inversions lasting over multiple days typically result in secondary particulate matter formation (Kroll and Seinfeld 2008). This results in air pollution concentrations that exceed the national ambient air quality standards (United States Environmental Protection Agency 2022) and raise the risk of aggravating certain health conditions especially in susceptible populations (Pope III et al 2015, Pirozzi et al 2018, Johnson et al 2021.

Study sites
The testing sites were strategically placed by the Utah Department of Health and Human Services to provide quick and efficient drive-up and drive-through access to COVID-19 testing. This arrangement provided for readily available testing coverage across the central part of SLCo and the east side of the Valley. Three COVID-19 testing sites were chosen because they specifically catered to vehicles (drive-through), primarily served three geographically diverse areas of the Salt Lake Valley [i.e., North (Urban/Industrial), East (Urban), and South (Urban/Commercial)], were accessible, had non-diesel/gas power sources, and exhibited high traffic and vehicle congestion where idling could occur. Site locations were characterized as urban, mixed with nearby industrial and commercial area sources. Main contributors to PM 2.5 emissions affecting the COVID-19 testing sites were the I-15 interstate corridor connecting the North-South ends of the valley (State Fairgrounds and Workforce), I-80 interstate (Highland High), industrial facilities (State Fairgrounds and Workforce), Salt Lake City International Airport and railroads (State Fairgrounds).

Study timeline
The vaccination site measurement duration with the study sensors are listed in table 1.
We used data from three UDAQ sites to compare PM 2.5 readings for the entire study period and across two time periods encompassing atmospheric inversion events (Bares et al 2018). While the captured event was a complex inversion with 'partial mix-outs' (hence the zig zags of PM 2.5 up and back down), northerly winds were present for much of the first event from 2-14 February. The short but rapid increase in pollution between 25-28 February was associated with lighter winds and a more classic inversion. By this time, the temperatures increase due to additional sunlight, and it becomes more difficult to maintain a strong inversion. This was especially true in this instance because it was so late in the inversion season, which typically ends at the end of February.

Instrument description
The COVID-19 testing site locations had ES-642 remote dust monitors (Met One Instruments Inc., Grants Pass, OR 97526) measuring PM 2.5 with a manufacturers uncertainty of 1 µg m −3 (Met One Instruments 2013) and PM 2.5 inlet sharp cut cyclones installed. The sensors used in this study are frequently used and have been rigorously evaluated against environmental protection agency regulatory sensors and cited in monitoring and research activities with a precision and accuracy similar to regulatory grade instrumentation (Mendoza et al 2019). The sensors were all well-maintained with proper flow rates and filter changes throughout the study. The humidity was well below the threshold (<40% humidity in this study) where particle hygroscopicity (>90% humidity) starts to have a bias effect on the laser sensors. Biases between the different ES-642 may account for part of the signal differences, however these should be less than 5% per manufacturer specifications (Met One Instruments, Inc. 2013).

COVID-19 testing site activity
The total daily number of COVID-19 tests administered at the testing sites is shown in figure 3. Detailed information on the testing sites, including operating days and hours, and tests by type is found in appendix.

Case studies
Due to the testing schedule, only the Highland High site can be studied to compare weekday air pollution on testing (Mondays and Wednesdays) and non-testing days. Both the Copperview and State Fairgrounds sites tested from Monday to Saturday which made a comparison between testing and non-testing days intractable. Figure 4 shows the timeseries comparison of the Highland High COVID-19 Testing Site air quality data with the Hawthorne Elementary School UDAQ regulatory sensor data. Two inversion periods are noted spanning from early to mid-February, and late February to early March. A statistical comparison of these data across all testing and non-testing days is shown in figures 5(a) and (b). Although Highland High PM 2.5 readings are lower than Hawthorne, on testing days (Mondays and Wednesdays) the PM 2.5 readings are proportionately higher, with a slope of 0.704, than on non-testing days (Tuesdays, Thursdays, and Fridays), which have a slope of 0.543.

Highland high testing site
The comparison of non-inversion testing days and non-inversion, non-testing days is shown in figures 6(a) and (b). During non-inversion periods, PM 2.5 concentrations were proportionately higher for the COVID-19 testing site locations during testing days (slope of 0.64) compared to non-testing days (slope of 0.366).
The analysis of inversion testing days and inversion, non-testing days is shown in figures 7(a) and (b). Although the values at Hawthorne (UDAQ) are still higher than at Highland High during inversions, the testing day PM 2.5 showed relatively higher values (slope of 0.69) than non-testing days (slope of 0.49).
The diurnal cycles during inversion testing days and inversion, non-testing days are shown in figures 8(a) and (b). During inversions the testing day Highland High PM 2.5 readings increased in the afternoon (during and after the 4-7 pm testing hours) at a greater rate than Hawthorne's (UDAQ) readings. This leads to more similar evening concentrations at both sites. On the other hand, during non-testing days, there is a larger divergence in PM 2.5 readings.    Figure 9 provides a timeseries comparison of the Workforce COVID-19 Testing Site PM 2.5 data with the Copperview UDAQ regulatory sensor data.

Workforce testing site
The analysis of inversion testing days and inversion, non-testing days is shown in figures 10(a) and (b). During inversions the testing day PM 2.5 (slope of 0.972) showed substantially higher values than the non-testing days (slope of 0.645) at Workforce when compared to UDAQ data (Copperview).    Figure 11 displays a timeseries comparison of the State Fairgrounds COVID-19 Testing Site PM 2.5 data with the Rose Park (UDAQ) regulatory sensor data.

State fairgrounds testing site
The analysis of inversion testing days and inversion, non-testing days is shown in figures 12(a) and (b). Similarly, to the Copperview site, during inversions the testing day PM 2.5 (slope of 0.988) showed relatively higher values than non-testing days (slope of 0.661) at State Fairgrounds compared to Rose Park (UDAQ).

Key findings
On days when the COVID-19 testing sites were closed, we found consistently lower PM 2.5 concentrations compared to their corresponding UDAQ sites. This is primarily due to the proximity of UDAQ sites to high-emitting sources, including interstate highways and industrial facilities. However, during testing days, this difference was reduced, due to the addition of local pollution emissions from vehicles traveling to the testing sites and idling while waiting in line. It is well-known that high-traffic areas in cities are subject to increased pollution levels compared to areas away from roads (Chen et al 2022). During inversion stagnation events, when the local emissions from vehicles traveling and idling were able to build up more, relative to non-inversion days, the concentration difference between testing and UDAQ sites was further narrowed. Looking at the diurnal cycles of PM 2.5 concentrations, the inversion-driven pollution build-up is particularly apparent in the late afternoon and evening hours, when testing lines and, therefore, idling levels, were highest. This suggests that the evening nocturnal surface inversion works to maximize the relative signal of hyper-local emissions from added car traffic and idling cars at the testing sites.
Staff working at the testing sites mentioned a final rush during the last 30 min of testing hours. Hourly business patterns of COVID testing sites on Google.com appears to support this pattern in some locations as the busiest times were found to be at the end of the testing site's hours of operation. In other locations, post 9am-5pm business hours were found to also be high traffic periods, and this further supports the patterns found in our study. Therefore, the compounded effects of multi-day and nocturnal surface inversions in the meteorology and the late afternoon maximum in vehicular emissions from the testing site are hypothesized to both be important factors to explain this phenomenon. For the evening surface inversion, the most notable example is at Highland High as testing hours were from 4-7pm.
Because the number of daily tests administered at the testing sites is relatively small and stable over time, (figure 1, appendix tables A1-A3), it is apparent that vehicular idling while waiting in line for testing was a substantial contributor to PM 2.5 concentrations. As a result, it is likely that the main source of emissions may be attributed to the additional local traffic associated with the vehicles arriving at the testing site as well as both idling and the associated behavior of stopping and restarting while waiting in line. During inversion events, when pollutants easily accumulate and do not ventilate well, even a small amount of additional contaminants can cause a sustained effect.
The Highland High site is sufficiently removed from large emission sources that even during inversions the testing site does not approach the Hawthorne UDAQ pollution levels. However, the Workforce and State Fairgrounds sites show similar PM 2.5 concentrations as their corresponding UDAQ sites (Copperview and Rose Park) during inversion testing days. During non-inversion testing days, the testing site PM 2.5 levels are far below UDAQ site levels. This underscores the strong local impact of inversions on pollution levels. As the atmospheric boundary layer lowers, the PM 2.5 levels increase and remain elevated overnight into the early morning hours.

Policy implications
Our findings suggest that drive-through facilities may have serious compounding effects in highly polluted urban areas and policy may need to account for these factors in an effort to protect human health and well-being. Hyperlocal pollution may also unnecessarily increase the health risks for health care workers  in such conditions as well as staff in drive-through facilities. This study was conceived after receiving reports of long wait lines of up to four hours at COVID-19 testing sites in early January 2022. Although this effect had largely subsided by the time this study took place, there could be some improvements to reduce the impact of these sites on local air quality if another surge occurs. Since the benefits of turning the engine off vary by emission type (Gaines et al 2012), instead of drive-through settings for testing, for example, when testing lines extend beyond the ten minute mark (when net harm is maximized), large parking lots could be provided and patients could be required to walk up to the protected testing sites. This would not only protect workers and patients, but also maximize the net benefits of efficiency and pollution control. Additionally, this would improve the local air quality, be more cost effective, and would be considered more humane by reducing worker exposure overall.
While this study focused on drive-through COVID-19 testing sites, the amount of drive-through facilities, including food establishments, banks, and pharmacies, among others, is non-trivial and this number has increased as a result of the COVID-19 pandemic. As important as it is to focus on ambient air quality impacts, the health of the workers staffing these facilities must be taken into consideration as well. Many of these businesses have windows that open to facilitate customer interaction and exchanges (e.g. food pick-up) and expose the staff to elevated pollution levels stemming from vehicles, especially during inversion periods.

Conclusions
This study quantified PM 2.5 concentrations at three COVID-19 testing sites and compared them against regulatory air quality sensors nearby. Although, the testing sites were generally farther away from large pollution sources than the regulatory sensors and generally read lower PM 2.5 concentrations, on testing days during atmospheric inversion periods, the PM 2.5 readings became comparable. Our findings suggest that traffic-related emissions, from both idling and added traffic to and within the COVID-19 testing sites are the cause of the observed increased PM 2.5 concentrations. The impact of stagnating pollution from hyperlocal sources is a concern, particularly for staff working at high traffic facilities, including drive-throughs.

Data availability statement
No new data were created or analysed in this study.

Appendix . Testing Site Information
The testing site operating days and hours, as well as the number of tests administered by type are shown for Highland High (table A1), Workforce (table A2), and State Fairgrounds (table A3).