Introduction

Air pollution is recognized as a public health emergency for humans. Multiple epidemiological studies have been conducted globally, with evidence showing adverse health effects and premature death linked to short-term and long-term exposure to particulate matter and gas pollutants (Chen and Kan 2008, Nuvolone et al. 2018; Samoli et al. 2001, World Health Organization 2004, 2013, 2016). Consistent research findings were shown from studies that were conducted in Ireland. Since the implementation of the Air Pollution Act in the 1990s, Ireland has experienced a decline in ambient concentrations of particulate matter, nitrogen dioxide (NO2), and carbon monoxide (CO) (Carthy et al. 2020; Clancy et al. 2002), and the health benefits of air pollution controls showing reduced respiratory and cardiovascular deaths in Dublin. However, more recent research evidence highlighted there was no safe level of air pollutants to cause adverse effects. Quintyne et al. (2020) found the linkage of poor air quality with a rise in hospital admissions for asthma, chronic obstructive airway disease, and heart failure in Dublin city. Specifically, associations of outdoor particulate matter exposure with premature death from type 2 diabetes and cardio-respiratory diseases have been reported (McVicar et al. 2023; White et al. 2020). For example, Carthy et al. (2020) highlighted the associations between NO2 pollution and asthma in older adults in Ireland. All stoke admission risks in Dublin were associated with increased PM2.5 and NO2 pollution during the winter season (Byrne et al. 2020), suggesting the exceedance of these pollutants in the short term could lead to acute health problems. The persistent efforts to mitigate air pollution, even at relatively lower levels, will continue to yield health benefits for the public.

In September 2021, the World Health Organization (WHO) launched stringent Air Quality Guidelines (AQGs) for major air pollutants, including inhalable particulate matter (PM10, with particle size equal to or less than 10 μm), fine particulate matter (PM2.5, with particle size equal to or less than 2.5 μm), NO2, ozone (O3), sulfur dioxide, and CO to provide better protection for the general public (WHO 2021). Nearly, the whole global population (99%) breathes fine particles (PM2.5) that exceed the new WHO AQGs annual level (5 µg/m3) (Yu et al. 2023). European Environment Agency (2022) estimated that 96% of the urban population in Europe was exposed to PM2.5 above the AQGs in 2020. As part of the European Union (EU) Green Deal, the EU Commission proposed to revise the Directive on Ambient Air Quality to align with the recently updated WHO AQGs (Tankosić 2023). To adhere to the legislation and EU directives, it is critical to maintain engagement in evidence-based policymaking, effectively reducing air pollution for the benefit of public health in Ireland.

Air quality characterization at a global or regional scale was mostly assessed using remote sensing (Brown et al. 2021). These methods have also been employed at the national or local level, for example, Kumari et al. (2022) employed satellites and ground-based data to characterize the air quality during the COVID lockdown in Dublin. Additionally, statistical models (e.g., linear or non-linear) using a small number of fixed site measurements in combination with geographic and temporal parameters have been developed to predict air pollutants at the city level (Alam and McNabola 2015, Basu et al. 2019; Donnelly et al. 2019). These studies are limited by a need for high-resolution data. Notably, most epidemiological studies regarding air pollution and health effects are generally based on ambient concentrations measured from limited representative regulatory stations or derived from modeling results (McDuffie et al. 2021). In complement to conventional methods, mobile monitoring proves to be an effective approach to assessing the temporal and spatial variabilities of air pollutants at a fine scale.

Google Street View vehicles equipped with air sensors have been employed for comprehensive traffic air pollution investigation in North America (e.g., Denver, San Diego, Oakland) and European cities (e.g., Amsterdam, Copenhagen, London) (Alexeeff et al. 2018; Apte et al. 2017; Kerckhoffs et al. 2022; Whitehill et al. 2020). Along with the rapid economic growth, urbanization, and increasing number of private cars (Tubridy et al. 2022), people in Dublin experience bad traffic congestion similar to that of other European citizens. Road traffic emission represents a major source of PM2.5, CO, and NO2, with nitrogen oxides (NOx) serving as one of the main precursors of O3 formation. Severe traffic congestion could lead to toxic levels of traffic-related air pollutants (Jin and Jin 2023), and there was evidence indicating that health damage could be attributed to diesel vehicles in Dublin, Ireland (Dey et al. 2018). Ireland’s national and local air quality monitoring networks are essential to ensure compliance with air quality standards. However, limited studies provided a microscale level assessment of traffic-related air pollutants. Responding to this data gap, high-resolution monitoring of gas (CO, NO2, O3, nitrogen monoxide (NO), carbon dioxide (CO2)) and particulate pollutants (e.g., PM2.5) was performed using Google’s electric Street View car at street level in Dublin city, Ireland.

Using high-resolution criteria pollutants mapping allows a comprehensive understanding of the spatiotemporal variation of air pollution within Dublin city, thereby enabling the formulation of mitigation policy targeting localized air pollution phenomena. The objectives of this paper are to (1) assess the daytime trend, weekday variations, and seasonal patterns of criteria air pollutants; (2) assess the temporal variations of local and background contribution to time serial air pollution; and (3) address the spatial complexities of hyperlocal air quality. This meta-data enables real-time information about the exposure level and hotspots for further air quality management and public awareness. Furthermore, it is critical for a growing urgency for mitigation strategies and future environmental policy to meet the zero-pollution action plan for 2030 and the zero-pollution version for 2050.

Methodology

Study area

Dublin is the capital of the Republic of Ireland, with a population of 1.45 million in 2022 (Central Statistics Office 2022) and an approximate area of 118 km2. The Dublin city center is characterized by a mix of residential and commercial buildings. A study indicated that about 10.5% and 9.8% of Ireland’s national CO2 and PM2.5 emissions from traffic sources occurred in Dublin city (Alam et al. 2018). It has a transport network with national roads (N) and regional roads (R), tram (i.e., LUAS), train (i.e., DART), and cycle lanes. Dublin has a maritime climate without significant temperature extremes and is characterized by mild-warm summer (12–16 °C) and cool winter (4–8 °C) (McHugh et al. 2023). The prevailing wind direction in Ireland is from the southwest (Broderick et al. 2006).

Measurement protocol and mobile air sensing platform

A hyperlocal air pollutants monitoring campaign using an all-electric Street View vehicle (Jaguar I-PACE) equipped with Aclima’s mobile air measurement and data acquisition platform was set up in Dublin, Ireland, over 16 months. This Google Street View car was characterized by zero tailpipe emission. The Aclima’s mobile air sensing platform was driven through roads and streets in Dublin, following fixed routes and time slots to capture air pollutants at a high spatiotemporal resolution. In general, mobile monitoring was repeatedly conducted in Dublin city on weekdays from Monday to Friday during daytime hours, between 7:00 and 17:00, with a limited number of monitoring sessions extending until 18:00 and 19:00. A Global Positioning System (GPS) was used to record the geographical location, which was synthesized with the air monitoring platform for date and time. Figure 1 shows the mobile monitoring routes.

Fig. 1
figure 1

Overview of the “Google Air View” mobile monitoring routes (orange dots) in Dublin city from May 2021 to May 2022. The blue dots indicated the location of the urban background, suburban background, and urban traffic sites

Full size image

Aclima’s mobile measurement platform is equipped with laboratory-grade instruments, which are characterized by fast time-response (up to 1 Hz) air monitors that measure size-resolved particle number (PN) counts and NO2, NO, CO, and CO2, except O3 (3 s). PM2.5 concentrations were measured using the light scattering optical particle counter (Whitehill et al. 2020), which was reported along with PN in five size ranges of 0.3–0.5 µm, 0.5–0.7 µm, 0.7–1.0 µm, 1.0–1.5 µm, and 1.5–2.5 µm. Additionally, the real-time equipment measures NO (chemiluminescence), NO2 (cavity-attenuated phase shift spectroscopy), O3 (UV absorption), CO2 (non-dispersive infrared absorption), and CO (UV fluorescence).

This study focused on mobile monitoring data collected from May 2021 to May 2022. The metadata platform provides information about data collection, data management, data processing methods, measurement calibration, quality control and quality assurance (QA/QC) protocol, and uncertainty quantification. QA/QC procedures for the Google Air View project in Dublin and data processing protocols/algorithms in this study followed the reports of Apte et al. (2017) and Whitehill et al. (2020). Instrument performance including precision, bias, and drift for the monitored pollutants (NO, NO2, O3, CO, CO2, PM2.5) is shown in Table S1.

Ambient monitoring data

Hourly PM2.5, NO2, and O3 concentrations measured from urban traffic, urban background, and suburban across Dublin were obtained from the national and local air quality networks (https://airquality.ie/). The ambient air quality data collected during the mobile monitoring period were incorporated for comparison. Information about the ambient sites can be referred to Table S2 of the Supporting Information (SI).

Data cleaning and data aggregation

We performed a series of data processing and data aggregation steps to convert data into different time resolutions. Initially, concentrations from one-second (1 s) measurements (except O3) that fall below the limit of detection (LOD) were substituted with half the LOD of the monitored air quality features. The 8 s rolling median was aggregated from 1 s monitoring during the “drive pass” time window (Apte et al. 2017). Any 8 s rolling median below the LOD will be substituted with the corresponding 1/2 LOD value. The LODs for the monitored pollutants at 1 s and 8 s resolutions are listed in Table S1 of the SI. The statistical summary of the raw data for the 1 s and 8 s rolling median can be seen in Table S3 of the SI. The final dataset consists of 5.03 million 1 s measurements of PM2.5, NO2, NO, CO, and CO2, and 1.45 million of O3. Furthermore, we excluded any missing values for the key variables (e.g., 23.8–38.6%) (Table S3b) and retained 377,113 8 s measurements. The data was subsequently down-sampled to 1-h and daytime resolution to reduce the size of the dataset.

Statistical analysis

In this study, PM2.5 concentrations are reported in the unit of µg/m3. NO2 and NO recorded in parts per billion volumes are converted to NOx, which is expressed as NO2 in the unit of µg/m3 in this study. We employed the Kolmogorov–Smirnov test to check central tendency and data distribution, and the median values were reported as core estimates. One-hour and daytime median values were derived from 8 s median aggregation. Instead of the daytime average, a daily 1-h maximum of O3 concentration was utilized in air quality evaluation. Correspondingly, subsampled data (e.g., high pollution days) was produced based on the 1-h and daytime resolution that showed concentrations exceeded the WHO AQGs and EU legal standards to explore the pollution episodes and hotspots. Statistical results were expressed in terms of interquartile range, including 1st, 5th, 25th, 75th, 95th, and 99th percentile. Data processing and statistical analysis were performed using the R statistical computing environment (https://www.r-project.org/), and ggplot2 R package was used for data visualization.

Baseline extraction

Atmospheric background and accumulated local emissions contribute to time series variation of air pollutant concentrations in urban area (Li et al. 2019b). In mobile monitoring, the background signal (e.g., low-frequency variations) is characterized by smooth variation at an hourly scale, while the rapid concentration spikes occur from seconds to minutes (Tan et al. 2014). In this study, we employed a baseline extraction method to divide the monitored concentrations \(({C}_{t}(k,j)\)) into the background and local emission concentrations using the following equations:

$${C}_{t}(k,j)={C}_{bg}(k,j)+{C}_{lc}(k,j)$$
(1)
$${C}_{bg}\left(k,j\right)={C}_{t, i}(k,j)$$
(2)
$${P}_{lc}(k,j)={C}_{lc}(k,j)/{C}_{t}(k,j)$$
(3)

where Clc(k, j) refers to local emission concentrations for pollutant k at time j, and Cbg (k, j) represents the background concentration for pollutant k at time j. The contribution from local emissions pollution \(({P}_{lc}(k,j)\)) is the ratio of \({C}_{lc}(k,j)/{C}_{t}(k,j)\).

The hypothesis proposes that the lowest percentile (\({C}_{t, i}(k,j)\)) signifies a background concentration characterized by a stable variation, which is minimally influenced by peaks originating from local sources (e.g., on-road pollution). The same method has been employed in previous studies to divide on-road background and accumulated vehicle emissions of the mobile monitoring air concentrations (Bukowiecki et al. 2002; Li et al. 2019b; Wei et al. 2021). In this study, we employed the 5th percentile for a “drive pass” time window (i.e., 8 s) to reconstruct a complete baseline to account for the background contribution to PM2.5 and NO2. Additionally, minima of 1-h median and daytime median were employed for the estimation of short-lived and longer-lived local pollution events.

Optimized hotspot analysis

The identified highly polluted days that exceeded the WHO AQGs for 24-h levels of NO2 and PM2.5 were employed as input (Table 3) for optimized hotspot analysis. Both pooled data and single-day monitoring were included for analysis. In this study, the significant hot (i.e., significant spatial clusters of high values) and cold spots (i.e., low values) were calculated using the Getis-Ord Gi* statistics in Arc GIS Pro. We employed “count incidents within fishnet grid” method for aggregating datasets, while PM2.5 and NO2 concentrations were spatially joined in the analysis field. This approach has been utilized for hotspot analysis in prior research (Cummings et al. 2021). The tool aggregates the measured pollutants (e.g., points) and performs scale of analysis, and the outputs were corrected for both spatial dependence and multiple calculations. The outputs included a new feature class with a z-score, p-value, and significance level categories (Gi_Bin: 99% hot, 95% hot, 90% hot, not significant, 90% cold, 95% cold, and 99% cold).

Results and discussion

Summary statistics of monitored air pollutants

The current meta-data covers a total of 377,113 single-street segment (8 s) monitoring across Dublin. Table 1 shows descriptive statistics of mobile monitoring for PM2.5, NO2, O3, CO, and CO2. Specifically, the median NO2 concentration was 27.0 µg/m3 across single-segment measurements, and 95th of NO2 concentration reached 280 µg/m3. The median PM2.5 concentration was 6.6 µg/m3, with the 95th reaching 19.3 µg/m3. The median and 95th of measured CO2 were 442.3 ppm and 495.9 ppm. Median O3 and CO concentrations across the segments were 46.2 µg/m3 and 0.316 ppm, respectively, with 95th reaching 77.8 µg/m3 and 0.697 ppm. The mobile platform enables measurements with a high spatial coverage of traffic-related air pollutants at the hyperlocal level. The findings revealed that particle and gas pollutant concentrations exhibit significant dynamics across road segments. Previous studies indicated that spatiotemporal variability of air pollutants through mobile monitoring would be affected by various factors, including background matrix, meteorology, road traffic (e.g., speed limits, fleet composition), street topology, or pollution of spikes originating from nearby sources (Hofman et al. 2022a; Tang et al. 2020).

Table 1 Summary descriptive statistics of air pollutants from a single-street segment (8 s) from multiple measurements of mobile monitoring
Full size table

As shown in Table 2, this mobile monitoring campaign is characterized by a high coverage density and measuring frequency for 1491 h on 233 sampling days. The 1-h median for NO2 varies from 14.2 to 470.4 µg/m3. The results in Table 2 showed that the EU legislation for 1-h NO2 limited values (i.e., 200 µg/m3) exceeded 6 h/year (0.4%). Some evidence suggests short-term exposure to a 1-h mean value of NO2 below 200 µg/m3 is associated with adverse health effects for patients with severe asthma (Latza et al. 2009). The daytime median of NO2 was 24.8 µg/m3, and 115 (49.4%) of the sampling days exceeded the 24-h WHO AQGs for NO2 (25 µg/m3) during the monitoring period. During daytime monitoring, PM2.5 concentrations vary from 1.4 to 29.5 µg/m3, with a median of 6.3 µg/m3. The results showed that the targeted WHO AQGs for 24 h PM2.5 (15 µg/m3) were exceeded in 21 days/year (9%). The WHO recommended to implement corrective actions to mitigate against PM2.5 pollution if the 24-h mean exceeded 3–4 days/year. The dominance of NO2 is reflected in the numbers of hours and days that pollutant concentrations exceeded the hourly and daily recommended concentrations. Tubridy et al. (2022) highlighted a breach of the EU limit value for NO2 and PM2.5 was evident in Dublin. These results were consistent with the findings from the Environmental Protection Agency (EPA), Ireland, which reported that NO2 is the primary pollutant from the transport sector in Dublin (Spohn et al. 2021), followed by PM2.5 and PM10 (Tang et al. 2019).

Table 2 Summary descriptive statistics of hourly and daytime concentrations for PM2.5 (µg/m3), NO2 (µg/m3), O3 (µg/m3), CO (ppm), and CO2 (ppm) from mobile monitoring
Full size table

As shown in Table 2, the AQG levels (range from 100 to 120 µg/m3) did not exceed for 8-h mean O3 concentrations. The findings from the mobile monitoring, which recorded the average 8-h O3 concentration (48.0 µg/m3), align with previous results observed in Dublin at the urban site. McHugh et al. (2023) indicated that O3 concentrations in Dublin at urban sites were consistently lower than those in suburban or coastal areas. The WHO has no AQGs set for CO2 and concentrations of CO remained below the AQGs throughout the monitoring period.

Time series of air pollutants

Figure 2 shows the temporal variability of PM2.5, CO, NO2, CO2, and O3 between May 2021 and May 2022. In addition to hourly and daytime variations, the 5-day moving average and data smoothing across the monitoring period are illustrated in Fig. 2. The smoothing pattern shows the PM2.5 concentration exceeded the WHO AQGs for an annual value of 5 µg/m3. Furthermore, the smoothing trend highlights a rise in concentrations observed during the winter season. The monthly variations of the monitored pollutants are presented in Fig. S1. PM2.5 concentrations were lower from May to August, experiencing a steady increase in December and January. These findings align with results obtained from fixed-site monitoring at community sites in Dublin. For example, Lee et al. (2022) found that the ambient PM2.5 concentrations (9–13 µg/m3) in Dublin surpassed the WHO AQGs with significant seasonal variability showing higher PM2.5 levels in December, which could be attributed to solid fuel burning for domestic heating in the winter season.

Fig. 2
figure 2

Time series of hourly a PM2.5, b CO, c NO2, d CO2, and e O3 from Google Air View vehicle during May 2021 and May 2022 in Dublin city. Hourly median concentrations on each day were used for calculation, except the O3 hourly maximum was utilized for calculation. Notes: The red dash lines show the 5-day moving average variations. The solid lines were drawn using the smoothing (gam, loess) trend function and the band indicates the confidence interval (0.95 by default). The black dashed lines indicate to the 24 h and annual of WHO AQGs for the specified pollutants

Full size image

As can be seen in Fig. 2, the measured NO2 is highly dynamic during the on-road mobile monitoring. The 5-day moving average shows a higher frequency of elevated NO2 pollution events compared to PM2.5. The smoothing pattern demonstrated the NO2 concentrations were above the WHO AQGs for an annual value of 10 µg/m3. Much greater monthly variation is evident for NO2 than PM2.5 (Fig. S1). There was moderate evidence showing that short-term exposure to a 24-h mean NO2 value of ~ 50 µg/m3 increased respiratory hospital admission and mortality (Quintyne et al. 2021). The CO and CO2 concentrations fluctuate slightly across seasons. In general, the 5-day moving average shows similar peaks and valleys occurred across measured pollutants except O3. For the 8-h average O3, an inverse trend was shown compared to NO2, with the maximum concentrations in April and May, while minimum values were shown in November and December.

The potential relationships between the mobile monitored pollutants were investigated using Spearman’s correlation coefficients, and results are shown at temporal variations from 1-h and daytime levels in Fig. S2. Weak correlations were found for 1-h and daytime median PM2.5 with CO2 (rs = 0.26), CO (rs = 0.25), and NO2 (rs = 0.27–0.30), most likely due to the varied sources of PM2.5 emission in Dublin. A report from EPA Ireland indicated major sources of ultra-fine particles in Dublin were secondary pollutants and oil, traffic, solid fuels, and cooking (Ovadnevaite et al. 2021). Moderate to strong correlations were found for NO2 with CO (rs = 0.48–0.67) and CO2 (rs = 0.47–0.74), suggesting that they share similar emissions from the transport sector (Ghahramani and Pilla 2021). In contrast, negative correlations were shown for O3 with CO2, CO, and NO2. This is expected since O3 is a secondary pollutant, with NO2 serving as its precursor. There was no notable weekday variability observed for the monitored pollutants (Fig. S3). In contrast, other studies have reported varying results, such as Martínez Torres et al. (2020) found that NO2 concentrations were higher from Wednesday to Friday compared to other weekdays in Dublin. Weekday-weekend variation is not the primary focus of this study and has been discussed in previous articles. For example, Lee et al. (2022) suggested that weekday PM2.5 concentrations were higher than weekends in Dublin.

Diurnal variation of air pollutants

Figure 3 illustrates the boxplots of hourly variation of the mobile monitored PM2.5, NO2, O3, CO, and CO2, respectively, during daytime (mainly from 7:00 to 17:00) between May 2021 to May 2022. One-hour medians are utilized to characterize the diurnal variation. In terms of temporal evolution, although mobile monitoring was conducted from different postal areas and across various seasons in Dublin, the basic trend illustrates elevated levels of PM2.5 in the early morning (8:00–9:00) and in the early evening around 17:00, respectively, during weekdays (Fig. 3). Figure S4 shows the diurnal variation in average PM2.5 from ambient locations characterized by urban background, suburban background, and urban traffic. Peak ambient PM2.5 values occurred at 20:00–21:00, showing higher night-time concentrations than daytime. These elevated levels are most likely attributed to regional transportation emission from residential solid combustion. These findings align with the results from air pollution monitoring conducted outside three maternity hospitals in Dublin. The study reported a similar daytime pattern for PM2.5 and showed higher night-time PM2.5 concentrations than daytime levels (Lee et al. 2022). Additionally, the results indicated these high concentration events for PM2.5 occurred mainly during working hours on weekdays in December.

Fig. 3
figure 3

Diurnal trends (left) and heatmap (right) of a PM2.5, b NO2, c O3, d CO, and e CO2 by Google Air View vehicle in Dublin city (n = 1486 h). Box plots represent the 25th and 75th percentiles, with the median (50th percentile) represented by the horizontal black line. Whiskers represent the lowest and largest values of 1.5*IQR

Full size image

Hourly averages of NO2 exhibited a distinct daytime variation trend, with peaks observed at 7:00–8:00 in the morning and 17:00–18:00 in the early evening. Martínez Torres et al. (2020) found a similar NO2 diurnal pattern in Dublin, showing the highest NO2 concentration at 8:00–9:00 and 17:00–18:00. These findings indicate that attribution of the direct impact of traffic emission on NO2 is more evident than for PM2.5, suggesting NO2 serves a better indicator for characterizing traffic-related air pollution in both short-term and long-term. These patterns coincided with the morning rush hour and traffic hour after work, and the valley values were observed at noon. A previous study revealed that the average hourly traffic achieved the peak value during the morning (8:00 am) and afternoon traffic (4:00–5:00 pm) in Dublin city (Tang et al. 2019). Moreover, the time and distribution pattern of the peaks shifts with the seasons, with the morning peaks in June occurring at 7:00–8:00. Evening peaks for NO2 concentrations mostly occurred at 16:00 in December and January. Previous findings in the UK suggested a distinct shift in peak timing to earlier evenings in the winter season (Wyche et al. 2020). Figure S3 shows that ambient NO2 levels gradually increased from 5:00 am, and the morning peaks occurred at about 8:00 in the morning at the urban traffic site and 9:00 at the urban and suburban background sites. The evening peaks occurred at 18:00 at the traffic site and shifted to 19:00–21:00 at the urban and suburban background sites. These results highlight the impact of non-traffic emissions.

O3 exhibited a contrasting diurnal trend compared to PM2.5 and NO2, and the highest O3 concentrations were observed in the afternoon (13:00–15:00). This could be due to photochemical effects that increase O3 concentrations during the daytime. The diurnal pattern for O3 is consistent with previous findings from ambient sites in Dublin (Tripathi et al. 2012) and modeling results (Donnelly et al. 2019). The O3 concentrations from mobile monitoring exhibited consistently lower concentrations than other suburban or background areas. A previous study was conducted in Dublin to investigate the spatial–temporal variation of traffic and urban monitoring sites (Perillo et al. 2022), and the results demonstrated reductions in PM2.5 and NO2 concentrations and an increase in O3 concentrations during COVID restrictions. Diesel vehicles are prominent sources of NO2 and PM2.5 (Tang et al. 2019). The CO diurnal profiles showed a similar but less clear pattern than NO2, which reflects the diesel-powered vehicle fleet. As for CO2, higher concentrations were mostly shown in the early morning.

Partitioning pollution into the background and local contribution

The total pollutants time series from mobile monitoring were divided into background and locally emitted contributions. In this study, we chose the 5th percentile of 1 h (Table S5), minima over 1-min, 1-h, and daytime data as background at different time resolutions. Compared to the 5th percentile of the whole dataset, the corresponding averages of minima for 1-h NO2 deviated by about 5%, while they doubled for PM2.5. The development of the 5th percentile curve for PM2.5 and NO2 shows a consistent background trend compared to the minima during the entire monitoring period (Fig. S5). Similarly, strong correlations were shown for PM2.5 from the estimated background (1-h minimum) and those measured at the urban background site (rs = 0.76, p < 0.01). Based on these results, the uses of minima are considered valid to obtain an estimate of the local contribution at different time resolutions where mobile monitoring was conducted. Furthermore, as suggested by Zimmerman et al. (2020) and Hofman et al. (2022b), a change on the order of 1 h is considered a short-lived event, and a longer-lived event is defined as occurring on a time scale of daytime basis (e.g., 2–8 h).

Figure 4 shows the division of total pollution into background contributions alongside short-lived and longer-lived events for PM2.5 and NO2. Of the total enhancement above background, the short-lived event enhancement (1 min to 1 h) accounts for 20.8–42.2% and 36.0–40.6% of the total concentration for PM2.5 (Fig. 4a) and NO2 (Fig. 4b). For short-lived events, NO2 exhibits sharper peaks than PM2.5. These results indicate the general public is at potentially higher risk during evening rush hours when background and local traffic emissions are tremendous. As can be seen in Fig. 4c and d, the longer-lived events (i.e., daytime) account for about 25% of the total concentration for both pollutants. Furthermore, previous findings suggested there are currently severe issues associated with air pollution in Dublin arising from both the transport sector and the use of solid fuel for domestic heating (Ovadnevaite et al. 2021). As for NO2, vehicle traffic is the main source along with electricity generation stations and industry. NO2 levels fluctuate more than PM2.5 during daylight hours at the local level.

Fig. 4
figure 4

Time series decomposition of background and local contribution for PM2.5 and NO2. Notes: a and b refer to short-lived events for PM2.5 and NO2 (i.e., 1 h), and c and d refer to longer-lived events for PM2.5 and NO2 (i.e., ~ 8 h)

Full size image

The decomposition results for the highly polluted days with the background and local contributions are illustrated in Fig. S6. Of particular interest are the pollution episodes that occurred on 17th January and 29 March 2022, revealing decomposed background concentration of NO2 and PM2.5 reaching 53.6–77.8 µg/m3 and 13.0–20.4 µg/m3, which were mainly due to background contribution, could be regional transported sources. During these regional pollution events, high concentrations of NO2 and PM2.5 were observed at urban and suburban background ambient sites. Subsequently, we compared the diurnal variability of PM2.5 and NO2 to illustrate the temporal trend and potential contributions from local and background sources during highly polluted days. As shown in Fig. 5, for short-lived events, traffic emissions contribute to both PM2.5 and NO2. The results show the diurnal trends are evident in NO2 concentration in the short term. The mean NO2 concentration during traffic peak time is a factor of 2 to the background and a factor of 2–3 to the off-peak. These results are consistent with previous findings from Zimmerman et al. (2020), which suggested near road concentrations of PM2.5 were only moderately elevated, while NO2 has much higher near roadway concentration enhanced (> 3–5 times). The probability of experiencing adverse health responses could be greatest when the concentration of NO2 is elevated. During polluted days, local NO2 has two peaks associated with morning (8:00) and early evening (16:00–18:00) rush hour, contributing to 54.4–68.0% of total NO2 concentrations. This means that the monitored NO2 is significantly enriched mainly from local traffic emissions. Additionally, there were peaks around 17:00–18:00 for background PM2.5 and NO2; this may suggest that their temporal variations are influenced by non-traffic sources. Overall, NO2 had the larger temporal variations (33.2%), whereas PM2.5 was less variable (22.3%). This is consistent with the traffic site monitoring results as shown in Fig. S3. Song et al. (2021) suggested diesel vehicles have larger NOx emission factors than gasoline vehicles. These findings highlighted the influence of diesel traffic on elevated NO2 levels in Dublin.

Fig. 5
figure 5

Diurnal variation in background and short-period emission of PM2.5 and NO2 during the entire study period. Notes: Lines represent short-lived events and areas represent background concentrations; the golden color refers to PM2.5 and the blue color refers to NO2

Full size image

Figure S7 enables a comparison of the time series variability of PM2.5 and NO2 from mobile monitoring and ambient air quality monitoring stations (AQMS) on the same timescale. The time series variability of ambient and mobile results exhibited similarities but showed significant differences in concentrations. For example, mobile results were significantly higher than the traffic monitoring site for PM2.5 (7.2 µg/m3, p < 0.01), but less for NO2 (41.1 µg/m3) (Table S3). A total of 20.2% of the monitoring days showed higher mobile NO2 concentration than the corresponding traffic site measurement. As expected, PM2.5 is spatially homogeneously distributed across the city, and higher associations were found for mobile monitoring with AQMS measurement of PM2.5 (rs = 0.72–0.80) than for NO2 (rs = 0.33–0.37) (Table S4). Carthy et al. (2020) found that the Dublin traffic zone had the highest NO2 levels compared to other areas. A study by Li et al. (2019a) highlights that mobile measurement represents the urban stationary monitors within a 1 km2 area. These findings indicated that mobile monitoring exhibited a high variability compared to air quality monitoring stations, emphasizing the need for extensive spatial coverage to capture the variability of NO2 from traffic emission.

Hotspots of traffic-related air pollutants

By ranking the most polluted days along with sampling routes at the community level from high to low, Table 3 shows the details of several cases in which daytime NO2 and PM2.5 exceeded the WHO guidelines of 25 µg/m3 and 15 µg/m3 throughout the monitoring period. Mapping of the elevated NO2 and PM2.5 concentrations for these days is displayed in Fig. 6, and pollution hotspots were identified for PM2.5 and NO2. Table 3 outlines the hotspots description for PM2.5 corresponding to optimal hotspots. For example, high values of PM2.5 cluster are observed at highly trafficked areas in Dublin, notably along National Roads such as N1 at O’Connell Street Upper, N2 near Grangegorman Upper, N3 at Dalymount, and N4 at Saint John’s Road West. In addition, Regional Roads like R101 at Cork Street and R109 from Parkgate to Chesterfield venue were PM2.5 hotspots with 90 to 99% confidence. High concentrations of PM2.5 clusters are also linked to industrial activities, for example, vehicle traffic at the landside of Dublin Port (Djordjević et al. 2023), and emissions from nearby restaurants. Our findings in Fig. 6 further indicated that hotspots of NO2 are more dispersedly distributed than those of PM2.5. To compare the results for pooled data, Fig. S8 shows consolidated information derived from single-day output. The location and timing of hotspots varied from day to day, and the recurrent hotspots in specific locations suggest that there are areas in Dublin where traffic-related pollutant concentrations are consistently elevated relative to the surrounding area. A recent publication indicates a signification association between urban green space and decreased NO2 concentrations in Dublin (Sabedotti et al. 2023). Identification of hotspots for traffic-related pollution provides potential policy instruments from an urban planning perspective to prioritize urban public transit, walking, and cycling networks within the city. Establishing motorized traffic-free zones and expanding public green spaces would also contribute to reducing the PM2.5 and NO2 levels.

Table 3 Description of identified highly polluted daytime median of PM2.5 and NO2 and elevated concentrations
Full size table
Fig. 6
figure 6

Spatial distribution of highly polluted a PM2.5 and b NO2 along with optimized hotspots analysis results for c PM2.5 and d NO2 in Dublin

Full size image

Limitation of the study

While our study provided comprehensive insights into Google Air metadata, it is limited by the lack of weekend and nighttime monitoring of traffic-related air pollutants, which hinders a thorough assessment of diurnal variations and the analysis of meteorological effects on traffic-related pollution. The varied monitoring time of each segment could affect the spatial and temporal assessment of the measured pollutants. Our assessment of the decomposition of short-lived events did not consider the traffic fleet composition or street topology. There were no supporting data to explore the origin of pollution sources about the contribution from longer-lived events. Further research could explore the integration of regional pollution sources to develop targeted PM2.5 and NO2 mitigation policies aimed at addressing traffic pollution in Dublin.

Conclusions

In this study, we demonstrate the potential for a mobile monitoring approach to investigate the hyperlocal distribution of air pollutants in Dublin, Ireland. This investigation characterizes traffic-related gas and particle pollutants after the COVID pandemic and provides an illustration of temporal variability, hotspot characterization, and local and background concentration contributions for mobile monitored PM2.5 and NO2. Median concentrations of daytime NO2 and PM2.5 exceeded the WHO 24-h AQGs for 49.4% and 9% of the sampling days throughout the monitoring period. The diurnal pattern was more evident for NO2 than PM2.5, with peak values occurring at 8:00 and 17:00–18:00. Seasonally, PM2.5 and NO2 concentrations are higher in winter than in the summer season, while no clear weekday variations were shown. Our findings demonstrate that elevated traffic-related NO2 in Dublin exhibited significant diurnal and spatial variations. Detailed knowledge of background and local contribution to mobile monitoring is important for highly polluted regions. This meta-data can be further implied for the sensitive analysis of the representativeness of mobile measurements and to obtain robust modeling prediction in space and time.

We can draw several conclusions by comparing the decomposed concentration and ambient data, NO2 is more affected by traffic emissions. Our data highlight the importance of non-traffic sources for temporal variation in the early evening. While traffic is a dominant source of NO2, non-traffic sources appear to be the dominant source for PM2.5. The importance of non-traffic sources suggests that merely regulating vehicle emissions may be insufficient to reduce PM2.5 concentrations to meet the new WHO guidelines. The findings highlight hyperlocal air monitoring is a complementary tool to provide meta-data for the research community. It enhances understanding of the implications of air quality policies and the effectiveness of air pollution mitigation plans from multiple emission sources. Furthermore, it enables providing air quality assessment to local environmental departments and policymakers for developing targeted strategies for mitigating traffic-related air pollution at a localized level.