In-Vehicle Air Pollutant Exposures from Daily Commute in the San Francisco Bay Area, California

Abstract

With urbanization and increased vehicle usage, understanding the exposure to air pollutants inside the vehicles is vital for developing strategies to mitigate associated health risks. In-vehicle air quality influences the comfort of the driver during long commutes and has gained significant interest. This study focuses on studying in-vehicle air quality in the San Francisco Bay Area in California, an urban setting with significant traffic congestion and varied emission sources and road conditions. Each trip is about 80.5 km (50 miles) in length, with commute times of approximately one hour. Two low-cost portable sensors were employed to simultaneously measure in-vehicle pollutants (PM_2.5, PM₁₀, and CO₂) during morning and evening rush hours from May 2023 to December 2023. Seasonally averaged PM_2.5 varied from 5.07 µg/m³ to 6.55 µg/m³ during morning rush hours and from 4.38 µg/m³ to 4.47 µg/m³ during evening rush hours. In addition, the impacts of local PM_2.5, vehicle ventilation settings, and speed of the vehicle on in-vehicle PM concentrations were also analyzed. CO₂ buildup in vehicles was studied for two scenarios: one with inside recirculation enabled (RC on) and the other with circulation from outside (RC off). With RC off, CO₂ concentrations are largely within the 1100 ppm range recommended by many organizations, while the average CO₂ concentrations can be three times high under recirculation mode. This research suggests that low-cost sensors can provide valuable insights into the dynamics of air pollution in the in-vehicle microenvironment, which can better help commuters reduce health risks.

1. Introduction

The major sources that contribute to air pollutant emissions in urban air are area sources and, in particular, vehicle exhaust [1]. Research has shown that most of the toxic pollutants present in the air are emitted from mobile sources rather than industrial sources. In-vehicle air quality is crucial to the health and well-being of millions of commuters globally. Most of the urban residents are exposed to various air pollutants, especially the particulate matter (PM) daily during their commute [2]. PM can come from exhaust emissions of internal combustion engines as well as non-exhaust emissions from tires, brakes, and resuspension. Thus, the measurement of particle mass concentrations and particle size are significant parameters that determine the ambient air quality of many countries [3]. According to the US EPA, transportation accounted for 28% of the total US GHG emissions in 2022, with on-road vehicles making 80% of all the transportation emissions [4]. Studies indicate that in 2010 in the US, exposure to PM_2.5 caused 107,000 premature deaths, and of these, 28% were ascribed to transportation [5]. Long-term exposure to PM_2.5 is associated with the development of ischemic heart disease, nasal inflammation, damage to lung function and the respiratory system [6]. In urban areas, driving to and from work is part of life, and the time spent in vehicles tends to get longer. The American Driving Survey 2014-17 indicated that the average daily commute in the US is approximately 51 min and 50.69 km (31.5 miles) [7]. The air quality inside the vehicle microenvironment can significantly impact the health and well-being of commuters. In-vehicle air pollutant exposure has been studied in many cities around the world, especially after the availability of portable and/or low-cost monitoring devices since the late 1990s. Select studies are summarized in

Table 1. Select studies on in-vehicle PM_2.5 exposure.

Study Area	PM2.5 Concentrations	Devices Used	Year of Publication
Barcelona, Spain	On average, 30 µg/m3 during morning rush hour and 25 µg/m3 during evening rush hour	DustTrak (TSI, 8520)	2012 [11]
Sacramento, CA, USA	7.1 ± 3.3 µg/m3	DustTrak (TSI 8520)	2017 [8]
Taipei, Taiwan	Range: 8–80 µg/m3 Average: 15 µg/m3	Temtop P600 Air Quality Laser Particle Detector	2021 [12]
10 cities in Asia, Africa, and South America	16–65 µg/m3 on average	Dylos Laser Particle Counter	2021 [9]
Birmingham, UK	6.4 ± 2.7 µg/m3	OPS Spectrometer, TSI Model 3330	2023 [10]
This Study	5.07–6.55 µg/m3, and 4.38–4.47 µg/m3 during morning and evening rush hours, respectively.	Temtop M2000 and Flow 2	2024

, with PM_2.5 concentrations reported during windows closed and RC on, similar to this study. The exposure can vary depending on the vehicle types, traffic conditions, ambient air quality, as well as the driving behaviors of the commuters, etc. These studies pointed out the effectiveness of ventilation settings in reducing in-vehicle PM exposure. PM concentrations were generally the lowest under the RC on mode with windows closed [8,9,10]. However, these studies did not simultaneously investigate the impact of ventilation on in-vehicle CO₂ concentrations.

Meanwhile, electric vehicles (EVs), which include battery electric vehicles (BEV), plugin hybrids (PHEV), and non-plugin gasoline-electric hybrids, are increasing in the commute fleet around the world. The state of California has the highest share of electric vehicle sales in the USA, at 25% in the first half of 2023 [13]. EV’s share is expected to increase in the future due to policy incentives, such as California’s progressive goal that requires all new cars and light trucks sold should be zero-emission vehicles by 2035, such as BEVs and PHEVs [14]. President Biden sets a goal of 50 percent electric vehicle sales by 2030 [15] in the USA.

Concerning on-road emissions, EV increases in fleets can result in lower tailpipe emissions. However, EVs are generally 20% heavier than vehicles with internal combustion engines and tend to generate more non-exhaust emissions from tires, brakes, dust resuspension, etc. A study indicated that EVs resulted in only 1–3% reduction of PM_2.5 with almost no reductions in PM₁₀ on the road [16]. The transition from petroleum-powered vehicles to EVs will change on-road emissions and in-vehicle exposure, which warrants constant monitoring to better understand the impact on commuters. As vehicles are striving to become more energy efficient, the recirculation mode is often automatically activated, especially when air conditioning is turned on. Needless to say, energy efficiency is even more critical for EVs, and the recirculation on mode is generally preferred, which can save up to 6% of energy [17]. While this can help with PM reductions, the resultant high carbon dioxide (CO₂) levels have been increasingly reported.

According to ANSI/ASHRAE (the American Society of Heating, Refrigerating, and Air-Conditioning Engineers) Standard 62.1 [18], the CO₂ concentration in indoor environments should not exceed 700 ppm above outdoor air levels. Considering that the outdoor CO₂ concentration is ~400 ppm, a practical limit for indoor environments is approximately 1100 ppm.

From the literature, limits for indoor CO₂ concentrations for non-industrial environments typically range from 1000 to 1500 ppm_v (parts per million per volume), primarily aimed at managing general indoor air quality concerns and sickness caused by CO₂ exposure [19]. However, the specifics behind these limits are often not explicitly stated. From the data analysis, across different seasons, at any given point the average CO₂ concentration does not exceed the OSHA and NIOSH standards of 5000 ppm_v over an 8-h workday [19].

Although CO₂ is not an air pollutant in the USA, excess levels in enclosed environments can result in various health issues. Studies indicated that exposure to CO₂ of 2500 ppm for 2.5 h caused relatively large decrements in decision-making performance [20], and CO₂ exposure at 3000 ppm can cause reduced mental performance, increased blood pressure, and stress [21].

CO₂ levels can build up rapidly inside vehicle cabins, especially when the windows are closed and the vehicle’s cabin recirculation mode is selected (RC on). This buildup is primarily due to CO₂ from human respiration and poses health and safety risks to the vehicle occupants. Studies have shown that in confined vehicle cabins, CO₂ levels can rise rapidly due to occupant exhalation and metabolism. A study in California, USA reported in-cabin CO₂ concentrations ranging from 2500–4000 ppm in the recirculation mode, while it went down to 620 ppm to 930 ppm with RC off [22]. Researchers have observed CO₂ concentrations reaching as high as 2000 parts per million (ppm) within just 12 min [23,24] in an idling vehicle. The situation intensifies with multiple occupants.

In-vehicle CO₂ levels can be influenced by factors such as the number of occupants, cabin volume, vehicle ventilation settings, etc. The confined nature of vehicle cabins exacerbates the problem, trapping exhaled CO₂ and leading to rapid concentration increases, especially in the absence of proper ventilation.

Vehicle speed serves as an indication of road congestion, and a study showed a negative correlation between PM_2.5 and speed, indicating PM_2.5 increased with speed decrease [9]. However, a weak positive correlation between vehicle speed and PM_2.5 was observed in another study [10]. Since congestion often occurs during commutes, its impact on air quality warrants further study.

Accurate and precise assessment of pollutant concentrations is necessary as they are related to human health and the environment [25]. The EPA sensors workshop held in 2013 emphasized the advantages of the development of portable air quality monitoring sensors that provide real-time measurements [26]. Since the 2013 EPA workshop, portable air quality sensor technology has evolved with sensors available to monitor ozone, nitrogen dioxide (NO₂), particulate matter (PM), and volatile organic compounds (VOCs) [27].

The air quality inside vehicles has become a growing concern as commuters spend substantial time in traffic, particularly in densely populated cities with heavy congestion. The confined space of a vehicle can result in higher concentrations of PM and other pollutants compared to outdoor air due to factors such as poor ventilation and air recirculation within the cabin. Despite the well-documented health impacts of PM exposure and the significant time spent by urban commuters in vehicles, there is a need for a comprehensive understanding of the dynamics of in-vehicle PM concentrations and the factors influencing them. This study leverages low-cost, portable sensors to quantify in-vehicle air pollutants, focusing on PM_2.5, PM₁₀, and CO₂ concentrations inside vehicles during peak commute hours in the San Francisco Bay Area, with data collection over several weeks to capture the wide range of commute conditions. This study analyzed the impact of various factors such as ventilation settings, traffic conditions, and the impact of ambient local air quality on in-vehicle pollutant levels. This research aims to enhance knowledge about personal exposure to traffic-related pollutants and offer actionable insights for improving public health in urban environments.

2. Materials and Methods

Two low-cost portable air quality monitoring sensors, namely the Elitech Temtop M2000 and Plume Labs Flow 2, were employed to continuously monitor real-time concentrations of PM_2.5, PM₁₀, and CO₂. The Flow 2 sensor tracked the geographical coordinates of the commute and generated a map of the commute route with AQI (Air Quality Index) concentrations. The Temtop M2000 uses a non-dispersive infrared (NDIR) sensor for CO₂ measurement, laser PM sensors for PM_2.5 and PM₁₀ measurement, and a Dart electrochemical HCHO sensor for formaldehyde measurement [28]. Flow 2 uses a laser particle counter sensor to measure PM concentrations and a thermally activated metal oxide sensor to measure gases, i.e., NO₂ and VOCs [29]. Flow 2 is controlled by an app on the cell phone, and it also records coordinate data, all of which is stored in the cloud. Additionally, the Flow 2 sensor was integrated with a personal mobile to record latitudes and longitudes along the commute. This allowed computation of the speed of the vehicle during commute and mapping the pollutant concentration levels. Both the instruments record air quality data every minute, while the GPS coordinate data is recorded every second. The times of the devices were synchronized before sampling to facilitate subsequent data analysis.

The study area (Figure 1) covered an 80.5 km (50-mile) stretch along I-680 between the cities of Sunnyvale and Walnut Creek in the San Francisco Bay Area of California. This route was selected as it is representative of the traffic conditions and commuting patterns in the region. The study area passes through three counties—Santa Clara County, Alameda County, and Contra Costa County— and covers multiple cities that include Sunnyvale, Pleasanton, Livermore, and Walnut Creek. The commute route takes about an hour to traverse under the morning and evening rush hour traffic conditions. In the study, pollutant monitoring was conducted simultaneously using the Temtop M2000 and Flow 2 sensors during the morning peak hours from 7:00–9:00 a.m. and evening peak hours 5:00–7:00 p.m. [30] for the months from May to December 2023, capturing seasonal variations. The instruments were started before entering the vehicle the actual time when the trip began was noted, and the data was extracted. The windows were closed for all the trips. The commute times ranged from 50 min to 1 h 17 min, and two riders were in the vehicle during the study period.

The vehicle used in this study is a Nissan Rogue S, a 2023 model compact SUV that is 100% gas and has a cabin air filter, with a passenger volume of 2.98 m³ (105.4 cu. ft.) [31]. This study aimed to investigate the relationship between in-vehicle PM_2.5 concentrations and ambient (local) PM_2.5 concentrations obtained from the local monitoring stations. Advanced statistical analyses were performed to study the impact of various parameters on PM, and data visualization techniques were utilized to represent the pollutant concentration data from both sensors. Furthermore, the study explored the buildup of carbon dioxide (CO₂) inside the vehicle under different ventilation conditions, as well as the correlation between vehicle speed and PM concentrations.

For data collection during the commute, the sensors were positioned inside the vehicle, on the dashboard. This placement was chosen since this is a stable location around the breathing zone. Two ventilation conditions were tested to investigate the CO₂ buildup in vehicles. RC on is when the vehicle’s cabin recirculation mode is activated. This mode circulates the cabin air internally, preventing fresh air intake from outside. For newer vehicles, RC on is oftentimes automatically set when using air conditioning, to save energy. RC off is when the vehicle’s cabin recirculation mode is deselected, which allows fresh air intake into the cabin from outside through air filters.

Based on sensor performance studies from the South Coast Air Quality Management District (AQMD) [32,33,34,35]. PM_2.5, PM₁₀, and CO₂ data from the Temtop M2000 were used for analysis, while the GPS coordinate data were taken from the Flow 2 sensor.

A trip was considered a one-way commute from Sunnyvale to Walnut Creek or the other way around. PM_2.5, PM₁₀, CO₂, and GPS were simultaneously measured for each trip, and data was recorded every minute by the sensors and statistically analyzed as daily, monthly, and seasonally to capture temporal variations. Trip data were taken from May 2023 to December 2023, which spanned almost four seasons: spring included May, summer included June to August, fall included September to November, and winter included December.

A study indicated that the on-road air quality can contribute 30% of in-vehicle PM_2.5 [10] exposure. Therefore, it is worthwhile to study the correlation between PM_2.5 levels inside and outside of vehicles. In this study, the ambient PM_2.5 concentrations were used to represent on-road PM_2.5 concentrations. The local (ambient air) PM_2.5 concentrations were obtained from the EPA database. The four local PM_2.5 monitoring stations chosen are San Jose-Jackson, Pleasanton-Owens Ct, Livermore, and Concord (Figure 1), all owned and operated by the Bay Area Air Quality Management District (https://www.baaqmd.gov/, accessed 15 September 2024).

Statistical analysis (such as the average and standard deviation of trips) and regression of the data were mainly performed with Microsoft Excel. For data visualization, Tableau was used to develop box plots and commute maps with PM concentrations.

3. Results

3.1. Temporal Variation of Particulate Matter

The average in-vehicle PM concentrations for morning and evening rush hours are summarized in

Table 2. Summary of in-vehicle particulate matter for morning rush hours.

Type	Season	Temtop M2000 PM2.5 (µg/m3) (Mean ± SD)	Temtop M2000 PM10 (µg/m3) (Mean ± SD)	Average Relative Humidity (%)	Average Local PM2.5 (µg/m3)
In-vehicle	Spring	5.96 ± 5.56	7.76 ± 6.13	79.63	7.53
In-vehicle	Summer	6.55 ± 3.75	8.99 ± 5.19	78.13	6.89
In-vehicle	Fall	5.07 ± 2.33	6.95 ± 3.85	81.08	6.46
In-vehicle	Winter *	6.31 ± 3.85	8.95 ± 5.90	95.50	7.72

* Since only 6 trips were taken during winter, the statistical results of daily and monthly were presented but were not used to draw conclusions on seasonal trends.

and

Table 3. Summary of in-vehicle particulate matter for evening rush hours.

Type	Season	Temtop M2000 PM2.5 (µg/m3) (Mean ± SD)	Temtop M2000 PM10 (µg/m3) (Mean ± SD)	Average Relative Humidity (%)	Average Local PM2.5 (µg/m3)
In-vehicle	Spring	4.38 ± 3.24	5.42 ± 3.57	57.17	7.47
In-vehicle	Summer	4.47 ± 1.89	5.65 ± 2.70	52.90	6.24
In-vehicle	Fall	4.46 ± 1.64	5.95 ± 2.80	57.23	4.95
In-vehicle	Winter *	7.07 ± 4.91	9.55 ± 7.49	80.00	10.31

* Since only 6 trips were taken during winter, the statistical results of daily and monthly were presented but were not used to draw conclusions on seasonal trends.

respectively. These values are all lower than the US EPA’s National Ambient Air Quality Standards (NAAQS) of 35 µg/m³ (24-h average). Frequently, the morning PM concentrations were higher compared to evenings, and this is consistent with studies of [9,11]. This may be attributed to the higher humidity in the mornings and subsequent absorption of moisture around the particles, which affects portable laser-based monitoring devices the most [36,37]. In a study by Kim et al. [36], the PM increased with increasing RH up to 70% RH. When RH > 70%, steam particles combined, forming water drops, decreasing PM concentrations. Such a trend of higher morning in-vehicle PM concentrations than evening warrants more studies in the future.

Analyzing the morning peak data from Temtop M2000, the PM_2.5 and PM₁₀ were higher for summer, followed by spring and fall. The measured PM concentrations are comparable with those from Sacramento, CA [8], which is in the same state but are lower than those reported in Taipei, which used a similar Temtop device [12].

Figure 2 shows monthly in-vehicle PM concentrations in the form of average and 95% confidence interval (CI) for morning and evening rush hour data. The higher in-vehicle PM concentrations observed in July and August are likely related to higher temperatures and congestion from lane closure due to tree removal along I-680 [38]. Figure 3 displays maps of PM concentrations to the coordinates of the commute. These trip-based concentration maps can identify hotspots of higher in-vehicle exposure for specific trips and can also be helpful for traffic management or source reduction.

Color-coded PM concentrations along a trip were plotted in Figure 3. The trip in Figure 3 is a morning trip with RC on; the ambient local PM_2.5 for that trip was 4.33 µg/m³, and ambient humidity was 79%. For this particular trip, the average in-vehicle PM_2.5 recorded was 5.19 µg/m³ and PM₁₀ was 6.02 µg/m³. As seen from the figures, higher PM concentrations were observed while crossing cities such as Fremont, Pleasanton, Dublin, and San Ramon, which are predominantly residential communities with local businesses, and higher traffic was observed in these areas due to people commuting to and from work.

3.2. Relationship between In-Vehicle PM_2.5 and Local PM_2.5 Concentrations

To obtain the PM_2.5 local concentrations, the concentrations measured at the four monitoring stations—namely San Jose-Jackson, Pleasanton-Owens Ct, Livermore, and Concord—were averaged according to the commute time. Of these stations, Pleasanton-Owens Ct is a near-road site and is a Special Purpose Monitor (SPM) [39], while the other three are State or Local Air Monitoring Stations (SLAMS) [39].

For example, for a trip from 7:30 a.m. to 8:30 a.m. on a given day, the local PM_2.5 at the four stations was selected and averaged from 7:00 a.m. to 9:00 a.m. as the local PM_2.5 concentration for that trip.

The correlation between in-vehicle and ambient PM_2.5 was also plotted in Figure S1, with fitting parameters in Table S1 in the Supplementary Material. Non-linear correlation showed a better fit, while the exponential fitting seemed to be the closest.

To further analyze in-vehicle exposure to outside PM_2.5, the ratio of inside/outside (Y′) was computed using average in-vehicle PM_2.5 and average local PM_2.5 concentrations of each trip [40]. The fraction (Y′ − 1) quantifies the fraction by which in-vehicle PM_2.5 exceeds (or is less than) PM_2.5 local [40]. In this study, 68% of in-vehicle PM_2.5 values are lower than the corresponding ambient values. This may be a result of the air filters in the vehicle. The ratio Y′ for in-vehicle PM_2.5 from Temtop M2000 ranged from 0.29–2.09, with an average of 0.92.

For the whole data set, scatter plots were created, and logarithmic, exponential, and linear trendline equations were fitted (Figure 4) to the data using an 80/20 training-testing approach [41].

Y′ = In-vehicle PM_2.5/Local PM_2.5

(1)

Y′ − 1 = (In-vehicle PM_2.5 − Local PM_2.5)/Local PM_2.5

(2)

The logarithmic, exponential, and linear curves were plotted with 80% training data and tested on the remaining 20% of the data (local PM_2.5 vs. Y′). Evaluation metrics such as root mean square error (RMSE) and mean absolute percentage error (MAPE) were calculated for all the curve fittings (

Table 4. RMSE and MAPE for logarithmic, exponential, and linear fittings of PM_2.5.

Sensor	Equation	R2	RMSE	MAPE
Temtop M2000	Logarithmic	0.52	0.33	27.51%
	Exponential	0.38	0.35	29.72%
	Linear	0.29	0.35	33.26%

). For Temtop M2000, though the logarithmic curve showed lower RMSE and MAPE compared to the exponential, the exponential curve was selected to have unit compatibility and have a better fit even for higher local PM_2.5 concentrations.

The exponential fitting suggests that the ratio of in-vehicle to ambient PM_2.5 does not increase or decrease linearly with changing ambient levels. Instead, the relationship follows a non-linear pattern, which could arise from factors such as vehicle ventilation dynamics, traffic influences, or nonlinear atmospheric processes affecting the relationship between indoor and outdoor particulate levels as ambient concentrations vary. The non-linear correlation is also reported by Goel et al. [40]. However, their study reported a logarithmic correlation between the ratio of in-vehicle to ambient PM_2.5 concentrations and ambient PM_2.5 concentrations.

3.3. CO₂ Buildup In-Vehicles

Figure 5 presents the CO₂ concentrations with RC on and off for both morning and evening trips for four seasons (See also Table S2 in supplementary material). Across all the seasons, whether morning (AM) or evening (PM) rush hour, the CO₂ concentrations were significantly higher when the RC was on compared to RC off. Spring season recorded the highest CO₂ (exceeding 3000 ppm), particularly during evening hours with RC on. Summer, Fall, and Winter recorded similar CO₂ levels during morning and evenings with slight variations. When RC was on, CO₂ levels were 2.11–2.86 times higher than RC off. This difference highlights the substantial buildup of CO₂ that can occur in the confined vehicle cabin environment when the recirculation mode is selected. Maintaining adequate fresh air ventilation is crucial for keeping in-vehicle CO₂ concentrations at acceptable levels and reducing their negative impacts on driver comfort and cognitive performance during long commutes. With fresh air intake, the CO₂ levels are largely within the 1100 ppm guidelines set by ASHRAE [18].

Furthermore, as shown in Table S2, when RC was on, higher CO₂ concentrations were observed in the evening than in the morning for three seasons measured: spring, summer, and fall. This is likely due to the higher temperatures in the evening rush hours (than mornings) and the accumulation of CO₂ throughout the day inside the vehicle cabin. In comparison, when RC was off, higher CO₂ concentrations were observed during the morning rush hours for summer, fall, and winter. This is consistent with the diurnal cycle of CO₂ concentrations, which peak in the morning and are at their lowest in the afternoon [42]. By not using the recirculation mode, the average CO₂ concentrations are mostly within the ASHRAE guideline. Letting fresh air in is especially helpful for keeping healthy levels of CO₂ during long or congested commutes. The control of CO₂ concentrations is largely the choice of vehicle occupants, while the choice of ventilation settings is associated with energy use. For BEVs, this can be a challenge due to the competition between personal comfort and vehicle range.

Figure 6 below displays the aggerated data of in-vehicle CO₂ together with PM_2.5 for both RC on and RC off conditions. In Figure 6a with RC off, the CO₂ concentrations range from 784 ppm to 1328 ppm, with an average of 1007 ppm. In-vehicle PM_2.5 levels range from 3.09 μg/m³ to 16.16 μg/m³, with an average of 6.33 μg/m³.

In comparison, in Figure 6b with RC on, the CO₂ concentrations range from 1656 ppm to 4718 ppm, with an average of 2851 ppm. In-vehicle PM_2.5 levels range from 3.14 μg/m³ to 12.47 μg/m³, with an average of 5.16 μg/m³.

In this study, RC on mode resulted in rapid CO₂ build-up to 5000 ppm at max in some of the trips. Under RC off conditions, with fresh air coming into the vehicle, CO₂ concentrations were greatly reduced. The average CO₂ levels were 2.11–2.86 times higher at RC on compared to RC off. The RC off mode is recommended to keep the levels of CO₂ within the ASHRAE guidelines. In our study, in-vehicle PM_2.5 levels with RC on are mostly lower than that of RC off, which is in agreement with studies [8,9,10]. However, higher PM_2.5 levels were observed in a few of our trips during RC on. One such is included in Figure S2. This is likely due to the high ambient PM_2.5 concentrations.

3.4. Speed vs. In-Vehicle Particulate Matter Concentrations

Some studies indicated that higher in-vehicle PM concentrations can be associated with congestion, which can be indicated by reduced speed [9,10]. This seems to be also reflected by the trip-based pollution graphs, such as Figure 3. Therefore, the vehicle speed during the commute was calculated from distance using the Haversine Formula [43], with the latitudes and longitudes from the Flow 2 sensor.

Distance (in miles) = a cos (cos (radians(90 − lat1)) × cos (radians(90 − lat2)) + sin(radians(90 − lat1)) × sin(radians(90 − lat2)) × cos (radians(long1 − long2))) × 3958.756

(3)

where, lat 1, lat 2—latitude 1 and latitude 2, long 1, long 2—longitude 1 and longitude 2.

The vehicle speed can then be averaged for the whole trip. The correlation of vehicle speed and PM was studied for select trips in each month and during both AM and PM rush hours, as shown in

Table 5. The correlation between vehicle speed and in-vehicle PM.

Date and Trip	Average Speed (km/h)	Correlation R between PM2.5 and Speed	Correlation R between PM10 and Speed
19 May 2023 a.m.	90.93	0.076	0.076
25 May 2023 p.m.	76.28	0.01	0.008
5 June 2023 p.m.	77.38	0.15	0.28
22 June 2023 p.m.	70.58	0.09	0.04
20 July 2023 a.m.	96.56	0.15	0.10
10 August 2023 a.m.	96.67	0.22	0.32
25 September 2023 a.m.	104.60	0.16	0.20
10 October 2023 p.m.	86.87	−0.07	−0.01
8 December 2023 p.m.	78.86	−0.12	0.14
14 December 2023 p.m.	64.90	0.21	0.17

. The correlation coefficient (R) between vehicle speed and in-vehicle PM (both PM_2.5 and PM₁₀) is largely weakly positive, consistent with that of the Birmingham, UK, study [10].

Figure 7 below displays the speed vs. PM concentrations on 8 December 2023. Throughout the commute, the average speed maintained by the vehicle was 78.90 km/h (49 mph) on the highway, (where the speed limit is 104.60 km/h (65 mph)) without air conditioning and with circulation from outside (RC off). During the beginning of the trip for the first 10 min, the vehicle maintained a relatively low average speed of 65.98 km/h (41 mph). During this period, the average PM_2.5 concentration was 4.60 μg/m³. Subsequently, at 5:20 p.m., the congestion eased, and the vehicle’s speed increased, averaging 109.44 km/h (68 mph) for the next 23 min until 5:42 p.m. Throughout this higher-speed segment, the average PM_2.5 observed was at 3.90 μg/m³. Following this, the vehicle’s speed was significantly reduced due to congestion, with an average of 37.01 km/h (23 mph) for the remainder of the trip. However, the average PM_2.5 concentration was 4.00 μg/m³. It was during this lower-speed segment that an elevated PM_2.5 concentration of 10 μg/m³ was recorded at 5:50 p.m.

The correlation coefficient (R) between vehicle speed and in-vehicle PM_2.5 was −0.12, indicating a weak negative correlation. This suggested that slow speed has some correlation with PM_2.5 increase. The correlation coefficient (R) for vehicle speed and PM₁₀ was 0.14, indicating a weak positive correlation. A study in the same state but a larger city (Los Angeles) indicated that brake and dust particles were in the PM₁₀ range [44]. Speed reduction is expected to be negatively correlated to PM₁₀ due to increased brake and tire abrasion, but this was not observed in this study, or from another study [10]. This warrants further investigation.

In the mornings, traffic jams tend to occur on the local roads of Walnut Creek after I-680 is exited. During the evening rush hour, the lower vehicle speed was due to higher traffic in evenings compared to mornings, and the areas with lower vehicle speeds are Pleasanton-Livermore areas; exit from I-680 to enter I-880. Minutes after the low speed, when the vehicle starts moving, high PM concentrations were observed likely due to the movement of vehicles ahead generating more PM.

Similar analyses were also performed for multiple trips. Overall, the correlation between vehicle speed and in-vehicle PM_2.5 and PM₁₀ concentrations was generally weak, with correlation coefficients (R) ranging from −0.12 to 0.32 across different trips. While some instances of PM spikes coincided with reductions in vehicle speed, there were also cases where elevated PM levels occurred despite constant speeds. Spikes in PM concentrations were sometimes observed after sudden reductions in vehicle speed, potentially due to the resuspension of settled particles or outside air intake. Lower vehicle speeds may be associated with higher in-vehicle PM levels in some cases, potentially due to reduced dispersion and dilution, as well as the presence of vehicular traffic. The results from this study suggest that vehicle speed may be one of many contributors to in-vehicle PM concentrations [9,10].

4. Conclusions

In Vehicle exposure to air pollutants during rush hour commutes in the San Francisco Bay area, CA, was studied with two low-cost sensors. Each trip length is about 80.5 km (50 miles), mostly on the highway with commute times of about an hour, with windows closed. The seasonal in-vehicle PM_2.5 during morning rush hours varied from 5.07 µg/m³ to 6.55 µg/m³, and PM_2.5 concentrations during evening rush hours were in the range of 4.38 µg/m³ to 4.47 µg/m³. Overall, in-vehicle PM_2.5 concentrations are much lower than EPA’s NAAQS’ 24-h average concentration of 35 µg/m³, and 68% of these are lower than the corresponding ambient values (values observed at the state monitoring stations) [45]. Seasonal variations of in-vehicle PM_2.5 were evident during the morning rush hour, with the highest in summer and lowest in the fall, while those during the evening rush hour did not exhibit strong variations. Higher PM_2.5 levels during the morning rush hour than evening rush hour were consistently observed for all four seasons studied. PM₁₀ largely follows the trend of PM_2.5. In this study, an exponential relationship was found between the PM_2.5 concentration ratios of in-vehicle vs. ambient/local (Y′) with local PM_2.5 concentrations. The average CO₂ level with the cabin recirculation mode engaged (RC on) was 2851 ppm for two occupants, while it decreased to 1007 ppm under the fresh air mode. In contrast, the average PM_2.5 concentration was 5.16 µg/m³ with RC on and 6.33 µg/m³ at RC off. This research demonstrates the effectiveness of using low-cost sensors for real-time monitoring of in-vehicle air quality. The findings provide valuable insights into the dynamics of air pollution within vehicles, emphasizing the importance of maintaining adequate ventilation and monitoring air quality to reduce health risks.

5. Limitations

The portable devices used in this study may be affected by temperature, relative humidity, and other environmental conditions since these devices do not have the sophisticated conditioning mechanisms of the FEM/FRM grade devices [27]. Although the Temtop M2000 has been regarded as fairly accurate [32,33,34], we did not do side-by-side calibration with FEM devices in this study. Low-cost devices, though very convenient in operation and usage, have limitations and uncertainties that should be estimated and compared with standard instruments [46]. The uncertainty of the low-cost devices may affect observations with small differences. This study only investigated windows closed conditions, which lacks dynamic range compared with other studies. This study did not simultaneously measure ambient air and cabin air at the same location due to the limited number of instruments and used the average ambient PM_2.5 concentrations from 4 local monitoring stations. This approximation can introduce inaccuracies in the correlation estimates. Using one sensor outside and one sensor inside the vehicle should be considered in future studies.