The impact of stay-home policies during Coronavirus-19 pandemic on the chemical and toxicological characteristics of ambient PM2.5 in the metropolitan area of Milan, Italy.

The goal of this study was to characterize changes in components and toxicological properties of PM2.5 during the nationwide 2019-Coronavirus (COVID-19) lockdown restrictions in Milan, Italy. Time-integrated PM2.5 filters were collected at a residential site in Milan metropolitan area from April 11th to June 3rd at 2020, encompassing full-lockdown (FL), the followed partial-lockdown (PL2), and full-relaxation (FR) periods of COVID-19 restrictions. The collected filters were analyzed for elemental and organic carbon (EC/OC), water-soluble organic carbon (WSOC), individual organic species (e.g., polycyclic aromatic hydrocarbons (PAHs), and levoglucosan), and metals. According to online data, nitrogen dioxide (NO2) and benzene (C6H6) levels significantly decreased during the entire COVID-19 period compared to the same time span in 2019, mainly due to the government-backed shutdowns and curtailed road traffic. Similarly, with a few exceptions, surrogates of tailpipe emissions (e.g., traffic-associated PAHs) as well as re-suspended road dust (e.g., Fe, Mn, Cu, Cr, and Ti) were relatively lower during FL and PL2 periods in comparison with year 2019, whereas an increasing trend in mass concentration of mentioned species was observed from FL to PL2 and FR phases due to the gradual lifting of lockdown restrictions. In contrast, comparable concentrations of ambient PM2.5 and black carbon (BC) between lockdown period and the same time span in 2019 were attributed to the interplay between decreased road traffic and elevated domestic biomass burning as a result of adopted stay-home strategies. Finally, the curtailed road traffic during FL and PL2 periods led to ~25% drop in the PM2.5 oxidative potential (measured via 2',7'-dichlorodihydrofluorescein (DCFH) and dithiothreitol (DTT) assays) with respect to the FR period as well as the same time span in 2019. The results of this study provide insights into the changes in components and oxidative potential of PM2.5 in the absence of road traffic during COVID-19 restrictions.

Introduction

The recent outbreak of novel Coronavirus (COVID-19), caused by severe acute respiratory syndrome Coronavirus 2 (SARS-CoV-2), has been declared as a worldwide health emergency, rapidly spreading throughout different parts of the globe (Jain and Sharma, 2020; Le et al., 2020; Wu et al., 2020; Zangari et al., 2020). In addition to transmission by macro-droplets (>5 μm) and fomites (Allen and Marr, 2020; Chia et al., 2020; Ong et al., 2020), the pathogenic and contagious characteristics of SARS-CoV-2 along with its relatively high residence time in the atmosphere (with a lifetime of 1–3 h in aerosols (van Doremalen et al., 2020; Setti et al., 2020a)) facilitate efficient transmission of the virus among humans, resulting in respiratory disorders, and in severe cases, mortality (Grasselli et al., 2020; Wu et al., 2020; Yang et al., 2020). The unprecedented pace of COVID-19 transmission from China to Europe, United States of America, and other parts of the world has led to globally more than 1 million deaths as of 9th October 2020 (WHO, 2020). In an effort to restrain the rapid spread of this infectious pathogen, governments have adopted various prevention and control strategies such as social distancing, businesses shutdown, and city-wide lockdowns (Anjum, 2020; Collivignarelli et al., 2020). Due to the implementation of abovementioned strategies, several recent studies have reported improvements in urban air quality, mainly attributed to the significant decreases in road traffic and (to some extent) industrial emissions (Bao and Zhang, 2020; Raffaelli et al., 2020; Tobías et al., 2020). Among different criteria air pollutants, ambient fine particulate matter (PM2.5, particle with aerodynamic diameter <2.5 μm) is of great importance due to its distinct physio-chemical characteristics and well-established adverse health consequences (e.g., respiratory and cardiovascular disease) (Brook et al., 2010; Gauderman et al., 2002; Pope et al., 2015). Therefore, the current COVID-19 shutdowns around the world have provided a unique opportunity for researchers to investigate the changes in the chemical/toxicological characteristics of ambient PM2.5, in light of a reduction in major urban activity source emissions.

Recent studies have reported a significant reduction in PM2.5 and gaseous pollutant levels (e.g., carbon monoxide (CO), nitrogen oxides (NOx), and benzene (C6H6)) during COVID-19 lockdown period across polluted cities (e.g., New York, Milan, and Beijing), due to the drastic decline in anthropogenic emissions (Chauhan and Singh, 2020; Collivignarelli et al., 2020; Dutheil et al., 2020; Jain and Sharma, 2020; Setti et al., 2020b; Xu et al., 2020). For instance, Chauhan and Singh (2020) observed significant reductions in ambient PM2.5 levels in New York, Beijing, and Delhi (by almost 32%, 50%, and 35%, respectively), during the COVID-19 period in comparison with the same time span in 2019. In another study, Collivignarelli et al. (2020) reported respectively ~47% and ~71% reductions in PM2.5 and black carbon (BC) mass concentrations at Milan area during the lockdown period (i.e., March 23rd, 2020 to April 5th, 2020) compared to the normal conditions (i.e., February 7th, 2020 to February 20th, 2020). In contrast to the global reductions in PM2.5 levels, a few studies have reported unchanged/increased ambient PM2.5 concentrations during COVID-19 lockdown (Huang et al., 2020; Le et al., 2020; Lv et al., 2020). For example, Le et al. (2020) attributed the enhanced concentration of PM2.5 in northern China during COVID-19 period to the impact of meteorological factors (particularly high relative humidity) on the formation of secondary aerosols. Additionally, it was documented that decreased traffic emissions in eastern China cannot offset the severe haze pollution due to the elevated levels of atmospheric oxidants, and in turn, higher rate of secondary aerosols formation (Huang et al., 2020; Lv et al., 2020). Some researchers have recently hypothesized probable association between enhanced rates of Coronavirus-related mortality and high background concentration of air pollutants in different urban environments (e.g., China, Italy) (Dutheil et al., 2020; Frontera et al., 2020; Zoran et al., 2020a, Zoran et al., 2020b). Conticini et al. (2020) and Wu et al. (2020) postulated that long-term exposure to PM2.5 provokes the vulnerability of individuals to COVID-19 infectious pathogens, although this hypothesis requires further investigation to reveal the direct impact of ambient PM on the biological and physio-chemical mechanisms of virus transmission. To the best of our knowledge, neither of the recently published studies have provided any insights on the changes in either the chemical composition or the toxicity of PM, resulting from the adopted lockdown restrictions during the COVID-19 period.

The Po Valley, located in the Lombardy region of northern Italy, has suffered from persistent air pollution challenges during the last decades (Conticini et al., 2020; Lonati et al., 2008; Zoran et al., 2020a, Zoran et al., 2020b). Previous studies have identified vehicular emissions, domestic biomass burning, industrial emissions, and formation of secondary organic aerosols (SOA) as well as secondary inorganic aerosols (SIA) as the major sources of ambient PM2.5 in Po Valley region (Daher et al., 2012; Decesari et al., 2017; Gilardoni et al., 2011; Pecorari et al., 2014; Perrone et al., 2012; Squizzato et al., 2013), which, combined with the particular topographical and meteorological conditions (facilitating stagnation and fog formation especially during fall and winter), have led to the deterioration of the air quality in the area. The high mountains of the western and central-eastern Alps along with the wintertime stagnant atmospheric conditions limit the horizontal and vertical dispersion of air pollutants emitted in the valley (Caserini et al., 2017; Cermak et al., 2009; Decesari et al., 2017, Decesari et al., 2001; Finardi et al., 2014; Frontera et al., 2020; Squizzato et al., 2016; Tositti et al., 2014). In addition to the severe air pollution episodes in the Lombardy region, this area has been the epicenter of COVID-19 pandemic in Europe, with unmatched rates of confirmed infectious individuals and lethality (Contini and Costabile, 2020; Frontera et al., 2020; Ogen, 2020). Although the strict lockdown policies and in turn curtailed traffic decreased the ambient PM2.5 levels in this area with respect to the pre-pandemic period (Collivignarelli et al., 2020), the contribution of domestic biomass burning emissions to total PM2.5 may have surged due to the stay-home strategies (Sicard et al., 2020). Previous studies have documented the extensive domestic biomass burning (for residential heating purposes) as a major source of pollution, elevating the PM2.5 mass concentration in the area during the colder periods of the year (Paglione et al., 2020; Ricciardelli et al., 2017; Tositti et al., 2014). In particular, the recently published study by Hakimzadeh et al. (2020) attributed a dominant fraction of PM2.5 oxidative potential in the Lombardy area to the domestic biomass burning emissions. It is therefore important to evaluate the ambient PM2.5 components and toxicological characteristics during the COVID-19 lockdown period to investigate the impact of the decreased traffic as well as the potentially increased domestic biomass burning emissions in this time period on the overall air quality in the area.

Accordingly, this study sought to characterize ambient PM2.5 components and oxidative potential in an urban background site in the Milan metropolitan area due to the major COVID-19 related stay-home policies. Ambient PM2.5 samples were collected within three periods, including major COVID-19 national lockdown (i.e., full-lockdown), the followed partial-lockdown, and full-relaxation (i.e., post-lockdown with limited restrictions) phases. The collected filters were analyzed, and the dithiothreitol (DTT) as well as 2′,7′-dichlorodihydrofluorescein (DCFH) in vitro assays were deployed to determine the oxidative potential of these samples. While several recent efforts have investigated the ambient concentrations of pollutants during COVID-19 restrictions in comparison with the pre-pandemic period in 2020 (similar to the study by Collivignarelli et al. (2020) in Milan area), our analysis considered the trends in components and toxicological characteristics of PM2.5 with respect to the same time span in 2019. This approach enabled us to investigate the impact of COVID-19 stay-home policies on atmospheric PM2.5 regardless of variations in meteorological factors from winter (i.e., pre-pandemic phase) to spring and summer (i.e., full-lockdown and full-relaxation periods).

Methodology

Sampling site, instrumentation and collection period

Sampling was carried out at two locations across Milan, Italy: (I) an institutional network's air quality station (i.e., Milano – via Pascal) for continuous air pollution monitoring, and (II) a residential area located at the suburban site of Bareggio for time-integrated PM2.5 filter sampling. The town of Bareggio, a suburban area located 14 km to the north-west of the Milan's metropolitan center (see Fig. S1), has been previously elected as an urban background district to determine public exposure to the baseline levels of ambient PM2.5 in the Lombardy region (Hakimzadeh et al., 2020; Mousavi et al., 2019).

Daily average mass concentrations of PM2.5, nitrogen dioxide (NO2), C6H6, and BC were obtained within the period of January to early-June at 2019 and 2020 from the Environmental Protection Agency of Lombardy (Agenzia Regionale per la Protezione Ambientale (ARPA)) website for Milano – via Pascal air quality station (45°28′42″N 9°13′54″E) as the closest sampling unit to the Bareggio area (Fig. S1). It should be noted that the Milano – via Pascal station is located in a small urban playground/little garden inside the court of University of Milan area called “Citta Studi”, therefore not directly impacted by vehicular emissions and it is also officially classified “urban background” station, as part of the national air quality network and used for PM2.5 exposure assessment within the National Evaluation Programme. The investigation period in this study has encompassed the pre-pandemic conditions prior to the COVID-19 restrictions (PP), the first partial-lockdown period (PL1), full-lockdown (FL), the second partial-lockdown (PL2), and full-relaxation phase with limited restrictions (FR). Further details regarding the start and end date of each phase as well as the corresponding adopted restrictions are provided in Table S1.

Given the unprecedented pandemic circumstances along with the extensive wintertime domestic biomass burning in the Milan metropolitan area which undoubtedly influences PM2.5 levels/toxicity (as discussed earlier in Hakimzadeh et al. (2020) and Daher et al. (2012)), filter sampling campaign was conducted from April 11th at 2020 to investigate the impacts of COVID-19 stay-home policies (with a special focus on traffic restrictions) on the PM2.5 components and toxicological characteristics. Thus, ambient PM2.5 samples were collected within three distinct periods: (I) FL spanning from April 11th to May 4th, (II) PL2 encompassing May 5th to May 18th, and (III) FR with limited restrictions from May 19th to June 3rd. By excluding PM2.5 filter sampling during the colder months of COVID-19 pandemic (i.e., February–March), we sought to minimize the substantial impact of extensive wintertime domestic biomass burning emissions on our sampling and analysis, and evaluate solely the impact of COVID-19 traffic restrictions on the PM2.5 oxidative potential.

Bi-weekly time-integrated PM2.5 samples were collected on prebaked quartz filters (37 mm, Pall Life Sciences, 2-μm pore size, Ann Arbor, MI) by means of the Sioutas Personal Cascade Impactor Samplers™ (PCISs, SKC Inc., Eighty-Four, PA, USA) (Misra et al., 2002; Singh et al., 2003). Although the PCIS has 4 different impaction stages (A through D), only the first stage of the PCIS (i.e., stage A) was used to remove particles >2.5 μm from the air stream, followed by an “after-filter” stage collecting PM2.5 on prebaked 37-mm quartz filters at an operational flowrate of 9 l per minute (LPM). The surface of stage “A” was covered with a thin layer of grease (Super Lude, NY, USA) to prevent the particles bouncing off the impaction surface. A microbalance (MT5, Mettler Toledo Inc., Columbus, OH) with a precision of 0.001 mg was used to determine the PM2.5 collected mass on the quartz filters as the difference between the pre-sampling and post-sampling weight of filters, after equilibration under temperature of 22–24 °C, and relative humidity of 40–50%, following US EPA's protocols for PM gravimetric analysis (US EPA, 2016, US EPA, 2008). Finally, the meteorological parameters (i.e., temperature, relative humidity (RH), precipitation, wind speed, and wind direction) were obtained from the Milano – Piazzale Zavattari weather station, located ~12 km to the west of our time-integrated sampling site. The monthly average values of abovementioned meteorological parameters and the corresponding wind rose patterns during the study period are reported in Tables S2 and Fig. S2, respectively. Further discussion regarding the meteorological characteristics of our sampling area can be found in Section 3.1.

Analysis

The collected PM2.5 samples were analyzed for elemental carbon (EC), organic carbon (OC), water-soluble organic carbon (WSOC), individual organic species, metals, and trace elements by the Wisconsin State Lab of Hygiene (WSLH). In summary, 1 cm2 punch from each filter was used in thermo-optical transmittance (TOT) analysis by a model-4-semi-continuous OC/EC field analyzer (Sunset Laboratory Inc., USA) to measure the OC/EC content of the PM2.5 samples (Birch and Cary, 1996). After extracting a quarter section of the collected filters in ultrapure water and filtering (0.22 μm pore size) the aqueous suspension, the WSOC fraction of samples was quantified by the means of a Sievers 900 Total Organic Carbon Analyzer (Stone et al., 2008; Sullivan et al., 2004). In addition, inductively coupled plasma mass spectroscopy (ICP-MS) analysis was employed on another ¼ section of each filter to measure the metal and trace element components of PM2.5 samples (Herner et al., 2006). Finally, speciated organic compounds including levoglucosan and polycyclic aromatic hydrocarbons (PAHs) were measured by means of gas chromatography/mass spectrometry (GC/MS) method (Schauer et al., 1999).

PM2.5 oxidative potential measurement

The fluorogenic 2′,7′-dichlorodihydrofluorescein (DCFH) and dithiothreitol (DTT) assays were employed as two of the well-established methods in the literature to measure the oxidative potential of PM2.5 samples (Fuller et al., 2014; King and Weber, 2013; Landreman et al., 2008; Sauvain et al., 2013; Verma et al., 2015; Vreeland et al., 2017; Wang et al., 2019). For the case of DCFH alveolar macrophage assay, the collected PM2.5 samples were initially extracted in 1.00 ml of sterilized Milli-Q water at room temperature and in dark environment, followed by agitation (for 16 h) and 30 min of sonication. Cultures of rat alveolar macrophage cells (NR8383, American Type Culture Collection) were exposed to the aqueous PM2.5 suspensions (derived from the filter extraction procedure) along with 2′,7′-dichlorodihydrofluorescein diacetate (DCFH-DA) probe as a fluorescent detector to quantify the rate of reactive oxygen species (ROS) generation, and thus, PM oxidative activity. Upon reacting with cellular esterases, the fluorescent DCFH-DA is converted to non-fluorescent DCFH, followed by oxidization to the highly fluorescent 2′,7′-dichlorofluorescein (DCF) due to the generation of ROS within the cellular structure. The DCF production rate (i.e., a reliable proxy of PM oxidative stress) is determined via a microplate reader in fluorescence units per PM mass (FU/μg PM). Zymosan, a fungal–glucan, was used for macrophage activation to trigger a robust immuno-chemical cellular response. Recorded fluorescence data were control-corrected to the Zymosan response, and reported as intrinsic oxidative potential in normalized units of μg Zymosan/mg PM (Landreman et al., 2008). To account for the effect of atmospheric dilution, the extrinsic oxidative potential of PM2.5 was derived by normalizing the measured Zymosan response to the volume of sampled air (Fang et al., 2015; Tuet et al., 2016). In contrast, the DTT assay measures the linear consumption rate of dithiothreitol as a metric of the oxidative potential of PM. In this method, the PM2.5 liquid suspensions (derived from the aqueous extraction of quartz filters) were reacted with a mixture of potassium phosphate (KPO4) buffers and DTT. Based on designated time intervals (i.e., 0, 15, 30, 45, and 60 min), trichloroacetic acid was added to vials of the incubation mixture for quenching the reaction, followed by recording the absorbance via a plate reader. The DTT rate of depletion was determined by converting the recorded absorbance to the remained DTT in units of nmol.min−1. The reported oxidative potential values in this study (for both DCFH and DTT assays) have been blank corrected to account for the potential uncertainties associated with the mentioned procedures (e.g., filter extraction, sonication). More details regarding the methodology of DCFH and DTT assays can be found in Shafer et al. (2016) and Cho et al. (2005), respectively.

Results and discussion

Impact of COVID-19 stay-home strategies on ambient levels of atmospheric pollutants

Fig. 1(a–d) indicates the weekly box plots of PM2.5, BC, NO2, and C6H6 mass concentrations from January to early-June of 2020 at the Milano – via Pascal urban background station. According to the figure, a significant (Pvalue < 0.0001) decreasing trend was observed in the ambient levels of abovementioned air pollutants during the investigation period. For example, the average concentrations of PM2.5, and BC decreased as much as ~78% and ~90% from January to May, respectively. All of the investigated pollutants decreased in concentration significantly (Pvalue < 0.0001), moving from the PP to the lockdown period (including PL1, FL, and PL2 phases). For example, the average NO2 concentrations decreased by ~58% from the PP period (51.94 ± 8.00 μg/m3) to the lockdown phase (21.44 ± 12.05 μg/m3). Although the significant decline in the ambient levels of pollutants from PP to the lockdown period could be attributed to the national-wide lockdown and COVID-19 related traffic restrictions, it is necessary to decouple the impacts of changes in PM emission rates (e.g., road traffic, domestic biomass burning) from the temporal variation of meteorological factors (e.g., boundary layer dilution properties) on the ambient levels of pollutants (Daher et al., 2012; Hakimzadeh et al., 2020; Marcazzan et al., 2001). Atmospheric mixing height is associated with meteorological parameters including vertical profiles of potential temperature, relative humidity, and wind speed (Ferrero et al., 2011). In summer, higher wind speed develops a broader mixing layer which facilitates the dispersion of pollutants in the atmosphere. Conversely, winter is characterized by stable weather conditions and weak atmospheric mixing due to the persistent thermal inversions along with fog situations when considerable amounts of air pollutants trap in the lower layers of the atmosphere (Marcazzan et al., 2001; Perrino et al., 2014).

Fig. 1

Temporal trends in the concentrations of (a) PM2.5; (b) BC; (c) NO2; and (d) C6H6 from January 2020 to early-June 2020. Each box plot corresponds to the period of one week during pre-pandemic (PP), full-lockdown (FL), partial-lockdowns (PL1 and PL2), and full-relaxation (FR).

To further evaluate the exclusive impact of the adopted lockdown strategies on the local air quality, the ambient concentrations of PM2.5, BC, NO2, and C6H6 are depicted during the lockdown period (i.e., early-March to mid-May 2020) and contrasted to the same time span in year 2019 (Fig. 2 ). It should be noted that the meteorological parameters (i.e., temperature, relative humidity, precipitation, and wind speed) were comparable (Pvalue in the range of 0.15–0.72) between the years of 2019 and 2020, thus precluding any significant influence of meteorological factors on atmospheric dilution and air pollutant levels between these years (see Table S2 and Fig. S2). Ambient temperature and relative humidity are not expected to induce significant variations in the rate of secondary aerosols formation from 2019 to 2020 based on the comparable temperature and relative humidity during the investigation periods of these consecutive years (Fang et al., 2019; Pun et al., 2006). As shown in Fig. 2, the levels of PM2.5 and BC were comparable (Pvalue in the range of 0.10–0.75) between the lockdown period (i.e., PL1, FL, and PL2) and the same time span in year 2019. Our observations are not in accordance with the results of recent studies across the Po Valley area, attributing the decreased levels of atmospheric air pollutants during the COVID-19 solely to lockdown restrictions (Collivignarelli et al., 2020; Sicard et al., 2020; Zoran et al., 2020a, Zoran et al., 2020b). The observed trends in our study might be attributed to the impact of enhanced domestic heating and biomass burning counterbalancing the effect of curtailed road traffic during the COVID-19 restrictions (as suggested by Sicard et al. (2020)). Both PM2.5, and BC originate from vehicular and biomass burning emissions (Al Madhoun et al., 2011; Briggs and Long, 2016; Masiol et al., 2019; Mousavi et al., 2019; Zhang et al., 2015), and the relative contributions of these distinct pollution sources determine the overall concentration of each species during the lockdown period.

Fig. 2

Temporal trends in the concentrations of (a) PM2.5; (b) BC; (c) NO2; and (d) C6H6 during lockdown phase (i.e., PL1, FL, and PL2) of 2020 and the corresponding period in 2019.

In contrast to the case of PM2.5 and BC, the NO2 and C6H6 concentrations decreased significantly (Pvalue < 0.05) during the lockdown period with respect to the same time span in 2019. For instance, the average NO2 levels decreased by ~35% during the lockdown period (21.44 ± 12.05 μg/m3) in comparison with the same time span in 2019 (33.17 ± 14.49 μg/m3). Previous studies have documented road traffic (particularly diesel engines) and fossil fuel combustion as major sources of NO2 (Afzal et al., 2012; Pepe et al., 2019; Zhang et al., 2018). The same sources but focusing on gasoline engines are associated with C6H6 (Keenan et al., 2010; Whaley et al., 2020). Consequently, in line with the finding of recent studies (Anjum, 2020; Bauwens et al., 2020; Collivignarelli et al., 2020; Dutheil et al., 2020; Le et al., 2020; Otmani et al., 2020; Sicard et al., 2020), the significant decrease in NO2 and C6H6 levels during the lockdown phase with respect to the same period in 2019 can be attributed to the marked reduction of emissions from road traffic.

Investigation of domestic biomass burning emissions during the lockdown period

In order to evaluate the rate of increase in biomass burning emissions during the COVID-19 shutdown phase, we have estimated the mass concentration of BC originating from domestic biomass burning during the study period. Fig. 3 shows the temporal trends in the weekly box plots of BC to NO2 ratio (both in units of μg/m3) during the investigation period for the years of 2019 and 2020. According to the figure, the average BC/NO2 values follow a decreasing trend, starting from ~0.12 in early-January 2020 and approaching asymptotically a value of approximately 0.040 in the warmer weeks of the campaign, when residential heating emissions (i.e., domestic biomass burning) are assumed to be negligible. Similar observations were also made in the year 2019, indicating an average BC to NO2 ratio of ~0.040 in the warmer season. Therefore, we selected 0.040 as an approximate upper ratio of non-biomass burning BC (BCnb) to NO2 in the metropolitan area of Milan. In concert with our observation, Kim et al. (2004) reported 0.034 as the BC/NO2 ratio during their experimental measurements in the vicinity of busy roadways in the San Francisco metropolitan area. In addition, as demonstrated in Fig. S3, the BC/NO2 ratio was ~0.032 in central Los Angeles (i.e., an urban environment heavily affected by vehicular emissions (Kozawa et al., 2009; Soleimanian et al., 2019a)) in 2019. Finally, according to the European Monitoring and Evaluation Program (EMEP) emission inventory database for Italy, the average ratio of BC/NO2 was found to be ~0.031 within the 2015–2017 period (EMEP, 2019). The slightly higher ratio of BCnb/NO2 (i.e., 0.040) in our sampling site compared with those of urban environments impacted by vehicular emissions can be attributed to the higher fraction of diesel cars in Milan's fleet along with the impact of non-traffic related (excluding biomass burning) combustion sources (e.g., cooking, local industries, and off-road equipment), emitting BC without major contributions to the NO2 emissions in the Po Valley region.

Fig. 3

Temporal variations in the BC/NO2 ratio during the investigation period for (a) 2019; and (b) 2020.

Considering the BCnb/NO2 ratio of 0.040 in our sampling location, the non-biomass burning fraction of BC (BCnb) was estimated as 4.0% of NO2 concentrations. Pepe et al. (2019) reported that the ambient NO2 in the Milan area is predominantly originated from vehicular emissions and fossil fuel combustion along with negligible contribution (~1%) of biomass burning. In addition, since the NO2 fluctuations are mostly explained by non-biomass burning combustion sources, the excess increase in the BC/NO2 ratio during the colder periods of the year can be attributed to the impact of increased domestic biomass burning. Subsequently, the biomass burning originated BC (BCbb) was derived by subtracting BCnb from total BC concentration. Fig. 4 shows the weekly box plots of BCbb and BCnb mass concentrations during January to early-June of 2020. As shown in the figure, BCbb mass concentrations were significantly higher (Pvalue < 0.0001) during the colder months of the year (i.e., January and February) compared to the warmer period. In addition, while domestic biomass burning emissions contributed to the majority (i.e., 61 ± 14%) of total BC mass concentrations in January, the contribution of non-biomass burning emissions to the ambient BC levels increased from ~40% in January to ~82% in May. These observations are in very good agreement with the findings of a previous study in the same area, attributing nearly 65% and 20% of total BC concentration to the domestic biomass burning during the winter and summer seasons, respectively (Mousavi et al., 2019). Although Fig. 4(b) represents BCnb mass concentrations derived from ambient NO2 levels shown in Fig. 1(c), there are a few differences between the two graphs (e.g., the 21st and 22nd weeks). For these few instances (caused by the NO2 and BC measurement uncertainties), estimated BCnb was marginally higher than total measured BC, and thus, we assumed equal concentrations of BCnb and total BC with a negligible contribution of BCbb.

Fig. 4

Weekly box plots of estimated (a) BCbb; and (b) BCnb mass concentrations during January to early-June of 2020.

Fig. 5 shows the concentrations of BCbb and BCnb during the lockdown period (i.e., PL1, FL, and PL2) with respect to the same time span in year 2019. According to Fig. 5(a), while the BCbb concentrations were negligible within the early-March to mid-May period in 2019, relatively higher concentrations of this pollutant were observed during the same period (i.e., lockdown period) in 2020, further demonstrating the impact of domestic biomass burning during the lockdown period, when stay-home strategies have presumably increased the residential heating emissions. Moreover, as shown in Fig. 5(b), BCnb showed a statistically significant (Pvalue < 0.05) decrease from 1.26 ± 0.79 μg/m3 in 2019 to 0.84 ± 0.48 μg/m3 during the lockdown period (i.e., PL1, FL, and PL2) in 2020, underscoring the effect of COVID-19 restrictions on decreased road traffic. Therefore, as discussed in Section 3.1, the comparable concentrations of PM2.5 and BC during the lockdown phase with respect to the similar period in 2019 were probably related to the increased domestic biomass burning emissions despite the lower road traffic in the Po Valley area during the COVID-19 restrictions (Sicard et al., 2020).

Fig. 5

Temporal variations in mass concentrations of (a) BCbb; and (b) BCnb during the lockdown phase (i.e., PL1, FL, and PL2) of 2020 and the corresponding period in 2019.

Impact of COVID-19 restrictions on PM2.5 components

Carbonaceous aerosols

Fig. 6 illustrates the mass concentration and content of carbonaceous fractions, including EC, OC, and WSOC in our sampling site during FL, PL2, and FR periods. As shown in the figure, PM2.5-bound EC (as a marker of traffic and biomass burning emissions (Saffari et al., 2013; Schauer, 2003)) remained almost constant during the investigation period (in the range of 0.24–0.29 μg/m3). The comparable levels of EC during FL, PL2, and FR periods are in agreement with the observed trends in BC mass concentration (as discussed in Section 3.2) and demonstrate the interplay between relative increase in domestic biomass burning and curtailed road traffic emissions in total EC concentration within lockdown phase. In addition, the measured EC levels during FL and PL2 periods were within the range of previously reported values at the same sampling site and time span in 2019 (i.e., ~0.35 μg/m3) as well as several locations across Po Valley (in the range of 0.20–0.55 μg/m3) (Daher et al., 2012; Hakimzadeh et al., 2020; Ricciardelli et al., 2017), further underscoring that lockdown restrictions have not led to significant reductions in ambient EC concentrations.

Fig. 6

The elemental carbon (EC), organic carbon (OC), and water-soluble organic carbon (WSOC) fractions of PM2.5 during full-lockdown (FL), second partial-lockdown (PL2), and full-relaxation (FR) periods: (a) normalized by the air volume; and (b) normalized by PM2.5 mass.

According to Fig. 6(a), PM2.5-bound OC mass concentration increased from 2.35 ± 0.19 μg/m3 during FL and PL2 to 2.79 ± 0.14 μg/m3 in FR phase. Likewise, ambient WSOC concentration increased by almost 40% from FL and PL2 (1.37 ± 0.40 μg/m3) to FR period (1.92 ± 0.16 μg/m3). To further investigate the observed trends in OC and WSOC levels, the biomass burning fraction of WSOC (WSOCbb) was quantified following the suggested approach by Fine et al. (2004). In this method, biomass burning originated OC (OCbb) was estimated based on the levoglucosan/OCbb ratio of 0.135 derived from wood smoke source profile (Fine et al., 2004). OCbb was multiplied by a factor of 0.71 to estimate WSOCbb, whereas the difference in mass concentrations of total WSOC and WSOCbb is ascribed to the non-biomass-associated WSOC (WSOCnb) (Sannigrahi et al., 2006; Stone et al., 2008). As shown in Fig. 6(a), WSOCbb decreased by a factor of 2.6 from ~0.29 μg/m3 in FL to ~0.11 μg/m3 during FR period. The comparison between measured WSOCbb levels and the previously reported value of ~0.08 μg/m3 at the same sampling site in 2019 (Hakimzadeh et al., 2020) validates our hypothesis regarding relatively higher residential heating emissions during stay-home strategies with respect to 2019. Further discussion about enhanced domestic biomass burning within lockdown period is provided in Section 3.3.1 of the manuscript. In contrast, the PM2.5-bound WSOCnb, which is mainly associated with SOA (Sun et al., 2011; Weber et al., 2007), increased from FL and PL2 (~1.11 μg/m3) to FR period (~1.79 μg/m3). It has been shown that modest formation of secondary aerosols during COVID-19 lockdown may be attributed to the reduced emissions of their precursors from primary sources (Lv et al., 2020). Thus, the observed trend in WSOCnb can be justified by the enhanced levels of anthropogenic SOA precursors emitted during FR period along with higher photochemistry within the warmer months of the year (Daher et al., 2012; Prévôt et al., 2004; Ricciardelli et al., 2017; Unger, 2012). On the other hand, water-insoluble organic carbon (WIOC), which is dominantly originated from combustion sources (Saffari et al., 2015; Snyder et al., 2009), exhibited consistent concentrations during the investigation period (in the range of 0.87–1.01 μg/m3). Although we expected increased levels of WIOC within the FR period (compared to FL and PL2), higher ambient temperature during FR can lead to evaporation of insoluble semi-volatile organic species emitted from road traffic (Shirmohammadi et al., 2016), reducing the impact of increased traffic on WIOC concentrations. Therefore, based on the abovementioned discussion, the enhanced concentration of PM2.5-bound OC during FR period (as opposed to FL and PL2) resulted from the higher formation of SOA due to the elevated photo-oxidation of SOA precursors. Finally, it is noteworthy that OC concentration during FL and PL2 (~2.35 μg/m3) was slightly lower than previously reported values across Po Valley within spring/summer season (in the range of 2.5–3.5 μg/m3), underscoring the decreased OC levels during lockdown restrictions (Hakimzadeh et al., 2020; Perrino et al., 2014; Ricciardelli et al., 2017).

Individual organic species

Fig. 7 shows the total PM2.5-bound PAHs and levoglucosan mass concentrations/fractions during COVID-19 period in our sampling site. According to the literature, levoglucosan has been identified as a tracer of biomass burning emissions (Simoneit, 2002), whereas PAHs are dominantly semi-volatile species originating from incomplete combustion of fossil fuel as well as wood burning emissions (Alves et al., 2015; Delgado-Saborit et al., 2011; Galarneau, 2008). As shown in Fig. 7(a), levoglucosan mass concentration decreased by a factor of almost 2.5 from ~56 ng/m3 in FL to ~22 ng/m3 during FR phase. The ambient levoglucosan concentration measured during FL and PL2 periods was higher than the previously reported values within identical time span at the same sampling location in 2019 (i.e., ~16 ng/m3) as well as Milan city center in 2010 (i.e., ~18 ng/m3) (Daher et al., 2012; Hakimzadeh et al., 2020). Moreover, none of the mentioned studies identified an explicit decreasing/increasing trend in levoglucosan mass concentration from April to June. Therefore, the elevated levels of levoglucosan during COVID-19 lockdown period further corroborate the results of our analyses in Section 3.2, suggesting relative increase in domestic biomass burning due to the adopted stay-home strategies.

Fig. 7

Temporal trends in levoglucosan and total PAHs concentrations during COVID-19 period normalized by (a) air volume; and (b) PM2.5 mass content. FL, PL2, and FR refer to full-lockdown, second partial-lockdown, and full-relaxation periods, respectively.

According to Fig. 7(a), total PM2.5-bound PAHs concentrations were comparable (in the range of 1.41–1.61 ng/m3) during FL, PL2, and FR periods at our sampling site. The relatively consistent levels of total PAHs during the investigation period can be attributed to the cumulative impact of domestic biomass burning and traffic sources to these organic species. For instance, the ambient mass concentration of indeno[123-cd]pyrene, a known surrogate of vehicle exhaust emission (Marr et al., 2006), exhibited an approximate 3-fold increase from 0.22 ng/m3 in FL to 0.73 ng/m3 during FR phase. Conversely, benzo(b)fluoranthene as a chemical marker of biomass burning combustion, decreased by 37% from ~0.46 ng/m3 in FL to ~0.29 ng/m3 within FR phase. Therefore, the impact of curtailed traffic on total PAHs mass concentrations within lockdown phase has to some extent been counterbalanced by the elevated residential biomass burning during stay-home strategies, leading to consistent levels of total PAHs from FL to PL2 and FR periods. In addition to the mentioned interplay between the contribution of vehicular and domestic biomass burning emissions to total PAHs mass concentrations, higher ambient temperature during FR phase might have precluded the conversion of semi-volatile PAHs into the particle phase (Pehnec et al., 2016; Soleimanian et al., 2020), thus attenuating the impact of increased emissions from mobile sources on total PAHs levels.

Redox-active metals

Fig. 8 shows the concentration of selected metals and trace elements, including titanium (Ti), chromium (Cr), manganese (Mn), iron (Fe), nickel (Ni), and cupper (Cu) during the investigation period. These metals, characterized as redox-active ones responsible for PM oxidative potential (Decesari et al., 2017; Lai et al., 2016; Ntziachristos et al., 2007; Yuan et al., 2001), are well-known tracers of re-suspended road dust particles originating from soil, tire and brake wear, and road abrasion (Adamiec et al., 2016; Al-Shidi et al., 2020; Dall'Osto et al., 2008; Godri et al., 2011; Harrison et al., 2012; Soleimanian et al., 2019b). In addition, Ni can be emitted as a result of fuel combustion in industrial sectors, power plants, vehicles engines, and to some extent residential heating (Masiol et al., 2020; Moreno et al., 2010; Zhang et al., 2014). According to Fig. 8(a), the ambient concentrations of Ti, Cr, Mn, Fe, and Cu increased considerably from FL to PL2 and FR period. Because of the association between these redox-active metals and road traffic in Milan area (Hakimzadeh et al., 2020; Vecchi et al., 2004), the observed trends are attributed to the increase in road traffic from FL to FR due to the gradual lifting of COVID-19 restrictions. Additionally, higher ambient temperature and lower relative humidity during FR phase (as opposed to FL) facilitate the resuspension rate of open surface mineral/road dust particles through vehicle movement (Branis and Safranek, 2011; Charron and Harrison, 2005), further elevating concentrations of these species during FL period. In contrast to the observed trends in ambient concentrations of metal elements, Ni levels were almost consistent between FL, PL2, and FR periods (in the range of 0.70–0.84 ng/m3), which can be attributed to the fact that, unlike road traffic and domestic biomass burning, industrial sectors/power plants operations have not been significantly affected by COVID-19 restrictions, so the variations in Ni concentrations from FL to PL2 and FR are rather minimal.

Fig. 8

PM2.5-bound redox-active metals concentrations measured during full-lockdown (FL), second partial-lockdown (PL2), and full-relaxation (FR) periods: (a) normalized by the air volume; and (b) normalized by PM2.5 mass.

Finally, the concentration of selected metal elements during FL and PL2 periods was lower than previously reported values in Milan metropolitan area during spring/summer season (Hakimzadeh et al., 2020; Vecchi et al., 2004). For instance, comparing our findings with those of Hakimzadeh et al. (2020) at the same sampling site in 2019, we noted consistent levels of Mn, Fe, Ti and Cu between FR phase and identical time span in 2019, while significant reductions (by on average 77 ± 33%) in concentrations of mentioned metals were observed during the FL period (with respect to FR phase). Similarly, Fe and Ti decreased by almost 2.5 times during the FL period in comparison with the reported values for the Milan area during summer season (Vecchi et al., 2004).

PM2.5 oxidative potential

Earlier studies across the Po Valley have shown that PM oxidative potential is influenced by the concentrations of metals (e.g., Cr, Mn, Fe, Ni, Cu) and carbonaceous species (e.g., OC, EC) (Hakimzadeh et al., 2020; Longhin et al., 2020; Pietrogrande et al., 2019; Visentin et al., 2016). These species are mainly associated with road traffic, domestic biomass burning, oil combustion, and SOA across Italy (Larsen et al., 2012; Pietrogrande et al., 2019); however, it has been shown that the oxidative potential of PM in the Po Valley region, and in particular in the Milan metropolitan area, is elevated due to the high density of anthropogenic emission sources in conjunction with specific stable atmospheric conditions, favoring the accumulation and aging of particles in the atmosphere (Perrone et al., 2016, Perrone et al., 2010; Pietrogrande et al., 2019). Fig. 9 shows PM2.5 oxidative potential during COVID-19 restrictions measured by means of DCFH in vitro and DTT assays. For the case of DCFH (Fig. 9(a)), the extrinsic PM2.5 oxidative potential increased from 55.2 ± 6.0 μg Zymosan/m3 in FL period to 62.3 ± 6.5 μg Zymosan/m3 during PL2 and to 74.1 ± 6.3 μg Zymosan/m3 within FR phase. Similarly, the mass-based PM2.5 oxidative potential increased from FL and PL2 (5056 ± 762 μg Zymosan/mg PM) to FR phase (6241 ± 536 μg Zymosan/mg PM). The intrinsic PM2.5 toxicity during FR was in very good agreement with previously reported values (i.e., 6037 ± 549 μg Zymosan/mg PM) within the same time period in 2019 for Milan metropolitan area (Hakimzadeh et al., 2020). Comparing the mass-based PM2.5 oxidative potential with reported values in the literature, the measured intrinsic toxicity of collected samples during FL and PL2 periods was almost half of the detected PM2.5-induced ROS across the traffic impacted areas of Milan city center (i.e., ~11,000 μg Zymosan/mg PM) as well as Athens, Greece (i.e., ~13,000 μg Zymosan/mg PM) (Daher et al., 2012; Taghvaee et al., 2019). However, the recorded PM2.5 oxidative potential at our sampling site was drastically higher than several cities across the world, such as Bologna (~1800 μg Zymosan/mg PM), Denver (~2000 μg Zymosan/mg PM), Thessaloniki, Greece (~750 μg Zymosan/mg PM), and even Los Angeles freeways (~3500 μg Zymosan/mg PM) (Decesari et al., 2017; Saffari et al., 2014; Shirmohammadi et al., 2017). This observation revealed that PM2.5 toxicity in the urban background site of Milan metropolitan area was still higher than several crowded cities across the world, despite the significant drop in road traffic during COVID-19 restrictions. Similar to the results of DCFH assay, curtailed PM2.5 oxidative potential during city-wide shutdown strategies was also verified by the DTT assay, in which the recorded values increased from 0.70 ± 0.15 nmol/min.m3 in FL and PL2 to 0.99 ± 0.20 nmol/min.m3 during FR period. It is noteworthy that measured PM2.5 oxidative potential via DTT assay during the FR phase is consistent with the previously reported DTT consumption rate during summer 2019 at the same sampling site (i.e., 0.85 ± 0.10 nmol/min.m3) (Hakimzadeh et al., 2020), but significantly higher than annual average DTT reported for Rome (i.e., ~0.23 nmol/min.m3) (Jedynska et al., 2017).

Fig. 9

Air volume-based (extrinsic) and mass-based (intrinsic) oxidative potential of ambient PM2.5 during the investigation period measured by the means of (a) DCFH macrophage; and (b) DTT assay (FL: full-lockdown; PL2: second partial-lockdown; FR: full-relaxation).

The observed trends in PM2.5 oxidative potential measured by both DCFH and DTT assays clearly demonstrated a reduction in the toxicity of PM2.5 due to the COVID-19 restrictions. Although total PM2.5 mass concentration was not significantly impacted following lockdown strategies (as discussed in Section 3.1), curtailed road traffic during FL and PL2 periods have led to noticeable reductions in PM2.5-bound OC levels as well as redox-active metals concentrations, leading to lower PM2.5 toxicity during lockdown restrictions. Finally, as shown in Fig. 9, while DCFH macrophage indicated a slight increase (by ~14%) in PM2.5 oxidative potential from FL to PL2 phase, the DTT assay showed comparable PM2.5 toxicity between the mentioned periods. This observation can be ascribed to the distinct characteristics of DCFH-DA and DTT probes, which demonstrate different levels of sensitivity to toxic PM species (Jovanovic et al., 2019; Sauvain et al., 2013). Therefore, the mentioned discrepancies are within the experimental uncertainties in quantifying PM oxidative potential via various assays.

Summary and conclusions

In this study, we investigated the chemical properties and oxidative potential of ambient PM2.5 during the COVID-19 lockdown period in the metropolitan area of Milan. Significant reductions (Pvalue < 0.05) were observed in the ambient concentrations of NO2 and C6H6 (as surrogates of vehicular emissions) during the lockdown phase in comparison with the same time period in 2019, due mainly to the restricted road traffic. Similarly, COVID-19 traffic restrictions have led to reduced levels of OC, and traffic-originated PAHs (produced from engine combustion) as well as redox-active metal elements (stemmed from re-suspended road dust) during the lockdown period. The concentration of selected PM2.5-bound metals increased (by 40% to 180%) from FL to PL2 and FR periods, coinciding with the gradual lifting of COVID-19 restrictions. On the other hand, PM2.5 and BC levels exhibited comparable (Pvalue = 0.10–0.75) concentrations between the lockdown phase and the same period in the year 2019. This observation can be attributed to the increased domestic residential heating during stay-home strategies which to some extent counterbalanced the reduction in traffic-induced emissions at our sampling area. Of particular note, the PM2.5 oxidative potential (measured by means of DCFH and DTT assays) decreased by ~25% during FL and PL2 periods as opposed to the year 2019 (as well as FR phase), due mainly to the curtailed road traffic within lockdown restrictions. The results of this study provide insights into the changes in composition and oxidative potential of ambient PM2.5 in the absence of primary emissions from road traffic during COVID-19 restrictions in the metropolitan area of Milan.

CRediT authorship contribution statement

Abdulmalik Altuwayjiri: Conceptualization, Methodology, Data curation, Writing - original draft. Ehsan Soleimanian: Conceptualization, Methodology, Validation, Visualization, Writing - review & editing. Silvia Moroni: Data curation, Resources, Writing - review & editing. Paolo Palomba: Data curation, Resources. Alessandro Borgini: Methodology, Writing - review & editing. Cinzia De Marco: Methodology, Writing - review & editing. Ario A. Ruprecht: Conceptualization, Methodology, Data curation, Resources, Supervision, Writing - review & editing. Constantinos Sioutas: Conceptualization, Funding acquisition, Project administration, Resources, Supervision, Writing - review & editing.

Declaration of competing interest

The authors of this paper declare that there is no conflict of interest.