Spatial and Temporal Variability of Carbonaceous Aerosol Absorption in the Po Valley

Received: March 2, 2020
Revised: September 21, 2020
Accepted: September 29, 2020

 Copyright The Author(s). This is an open access article distributed under the terms of the Creative Commons Attribution License (CC BY 4.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are cited.

Cite this article:

Gilardoni, S., Massoli, P., Marinoni, A., Mazzoleni, C., Freedman, A., Lonati, G., De Iuliis, S. and Gianelle, V. (2020). Spatial and Temporal Variability of Carbonaceous Aerosol Absorption in the Po Valley. Aerosol Air Qual. Res. 20: 2624–2639. https://doi.org/10.4209/aaqr.2020.03.0085

HIGHLIGHTS

• We report aerosol light absorption coefficients at two sites in the Po valley, Italy.
• Limited variability of Elemental Carbon Mass Absorption Cross section is observed.
• Wood combustion and secondary organic aerosol contribute to brown carbon (BrC).
• Water-soluble BrC absorption is only 50% of methanol-soluble BrC.

Keywords: Black carbon; Brown carbon; Optical properties; Mass absorption cross section; Climate change.

INTRODUCTION 

Carbonaceous aerosol accounts for more than half of fine atmospheric aerosol mass in urban, rural, and remote locations (Zhang et al., 2007; Bond et al., 2013). An accurate determination of the optical properties (absorption and scattering) and atmospheric concentrations of carbonaceous aerosol is necessary to improve the ability of climate models to describe the interactions of solar radiation with aerosol particles, and thus reduce the uncertainties in climate predictions (IPCC, 2013). Light absorbing carbonaceous aerosol components are identified as black carbon (BC) and brown carbon (BrC) (Andreae and Gelencser, 2006). BC is a known key atmospheric pollutant whose reduction is necessary in order to limit global warming (IPCC, 2018). BC has unique properties due to its refractory nature: it strongly absorbs visible light, is insoluble in water and organic solvents, and is made of aggregates of small carbon spherules with a high carbon to hydrogen ratio (Chakrabarty et al., 2006; Bond et al., 2013; Buseck et al., 2014). The absorption of BC per unit mass, also called the mass absorption cross section (MAC_BC), is the highest among all the light absorbing aerosol species, with values larger than 5 m² g^–1 at 550 nm and centered around 7.5 ± 1.2 m² g^–1 (Bond and Bergstrom 2006; Cross et al., 2010; Bond et al., 2013; Forestieri et al., 2018; Li et al., 2020). BC radiative forcing values used in models range between +0.05 and +0.8 W m^–2, while satellite and ground-based observations point towards higher values (+0.7 to +0.9 W m^–2) (Bond et al., 2013). Part of the discrepancy between the radiative forcing estimated by models and observations is attributed to the variability of MAC_BC, which depends on intra-particle BC mixing state and particle morphology (China et al., 2013; China et al., 2015; Bhandari et al., 2019). The intra-particle mixing of BC particles with other material can lead to the enhancement of the BC absorption cross section and therefore of MAC_BC (Lack et al., 2009). MAC_BC enhancements observed in laboratory experiments vary from 2 to 3.5 ( Zhang et al., 2008; Cross et al., 2010; Shiraiwa et al., 2010; Bond et al., 2013; Saliba et al., 2016), while field observations are characterized by a larger variability that is difficult to incorporate in models (Gustafsson and Ramanathan, 2016). Cappa et al. (2012) observed that aerosol aging in two California regions increased MAC_BC by only about 6%, while Moffet and Prather (2009) determined a MAC_BC enhancement for aged soot in Mexico of 60%. Peng et al. (2016) observed that MAC_BC can be enhanced by 2.4 times, over timescales that differ significantly from clean to polluted environments, potentially explaining the field observation variability. Liu et al. (2015) reported lower enhancement factors (up to 1.5) consistent with Cappa et al. (2012). Discrepancies among laboratory and field observations are also due to the heterogeneity of particle ensemble composition (particle-to-particle differences) and the inaccurate assumption of spherical concentric core-shell structure (Fierce et al., 2020). Limited comparability among measurement techniques contributes to the variability of observed BC optical properties as well (Subramanian et al., 2007; Lack et al., 2008; Moosmuller et al., 2009).

More recently, part of the discrepancy between modeled and observed aerosol radiative forcing has been attributed to the presence of BrC in aerosols (Park et al., 2010; Feng et al., 2013; Jo et al., 2016; Chakrabarty et al., 2018). BrC is a mixture of organic species with different chemical identities, able to absorb both UV and visible light but with a stronger wavelength dependency than BC, although characterized by a lower MAC (Bond et al., 2013). Unlike BC, BrC is mostly soluble in organic solvents (Chen and Bond, 2010). BrC contributes to primary and secondary organic aerosol (POA and SOA, respectively) originating from a variety of sources and chemical transformation pathways (Laskin et al., 2015; Moise et al., 2015). For example, recent studies indicate that tar balls from biomass burning contribute to atmospheric BrC through atmospheric aging (Chakrabarty et al., 2010; Sedlacek et al., 2018; Bhandari et al., 2019b). Despite the radiative forcing of BrC is highly uncertain, the impact of BrC on climate will likely increase in the future due to increase of wildfire frequency (Brown et al., 2018; Saturno et al., 2018; Jia et al., 2019).

Ground-based measurements indicate that BrC can account for up to 20% of carbonaceous aerosol light absorption between 400 and 600 nm (Chung et al., 2012b), although the larger concentrations of BrC at higher altitude can further increase BrC relative absorption (Feng et al., 2013; Zhang et al., 2017). Barnard et al. (2008) indeed estimated that BrC contributes up to 40% of top of the atmosphere aerosol absorption at wavelengths below 500 nm. A limited number of studies using modeling and vertical profile observations indicate that BrC radiative forcing is equal to +0.25 W m^–2, thus partially offsetting the cooling effect of non-absorbing OA (Chung et al., 2012b; Feng et al., 2013). Nevertheless, the scarcity of ambient BrC optical property measurements make its representation in most chemistry and climate models difficult (Forrister et al., 2015; Sumlin et al., 2017; Tsigaridis and Kanakidou, 2018; Browne et al., 2019).

This study investigates the optical properties of BC and BrC in the Po Valley (Italy), one of the European pollution hot-spots. Bordered by the Alps to the north and by the Apennines to the south, the Po Valley is often subject to stagnant atmospheric conditions that favor the accumulation of atmospheric pollutants (Larsen et al., 2012). Across the Po Valley, wood burning for residential heating accounts for up to 50% of primary and secondary carbonaceous aerosol in wintertime (Gilardoni et al., 2011, 2014, 2016). In summer, the intense photochemical activity increases the concentration of SOA over freshly-emitted POA and BC (Saarikoski et al., 2012). Therefore, we can expect different concentrations of BC and BrC across the seasons. These conditions make the area an ideal site to extract the optical properties of BC and BrC and to investigate the effects of atmospheric aging on the optical properties of carbonaceous aerosol in the region.

 
METHODS

 
Sampling Sites

Aerosol optical properties and chemical composition were investigated at an urban site in Milan (MI, 45°28’43”N 9°13’56”E) and at a rural site in the village of Motta Visconti (MV, 45°16’55’’N 8°59’19’’E), about 30 km south-west of Milan. Milan has a population of about 1.3 million people and the sampling site, located inside the university campus, is representative of urban background conditions. Motta Visconti is a small village (population less than 10.000 inhabitants) surrounded by rice and corn fields. Both sites are in the greater Milan metropolitan area and are part of the air quality monitoring network of the Lombardy Environmental Protection Agency (ARPA Lombardia).

The historical emission inventory of the Milan metropolitan area shows that the main sources of BC and OA are road transport (76% of BC and 23% of primary organic carbon, OC) and residential heating (12% of BC and 49% of primary OC) (Inemar, 2014). Diesel cars account for 92% of BC emissions from the road transport sector (Inemar, 2014).

Table 1 reports the sampling periods at each site, covering winter and summer. Meteorological parameters measured during the four experiments are reported in Fig. S1. Wind speed was below 2 m s^–1 most of the time, indicating low horizontal dispersion. Average temperature was 7 °C during winter and 19°C in summer. Relative humidity (RH) was generally high with mean values above 70% both during summer and winter.

 
On-line Optical Measurements

Table 1 lists the instruments used for monitoring the aerosol optical properties during the field observations at both sites. Throughout the four experiments, the aerosol light attenuation coefficient was measured at 880 nm and 370 nm with a 5-minute time resolution using an aethalometer AE22 (Magee Scientific) equipped with a PM₁₀ head. The aethalometer operated at a flow rate of 4 L min^–1 and filter tape advanced when absorbance at 370 nm reached the limit of 0.65.

Aerosol light scattering was measured at 1-minute time resolution, using a 532 nm integrating nephelometer Ecotech M9003, with the exception of the rural winter experiment, when we employed a Ecotech AURORA 3000 operating at three wavelengths (450 nm, 535 nm, 635 nm). Nephelometers were calibrated using filtered air as zero span and CO₂ as high span, in order to derive the calibration curve applied for scattering coefficient calculated by the instruments. The nephelometers sampled without any size cut. However, previous measurements indicate that the majority of aerosols at both sites is in the submicron fraction (Lonati et al., 2011 and Fig. S4). Although the presence of a small number of larger particles could have a significant impact on nephelometer observations, scattering coefficient data were employed mainly for attenuation measurement correction, with negligible effect on particle absorption in this study.

The M9003 measurements were not corrected for truncation error, in accordance with the manufacturer’s indications. Indeed, Müller et al. (2009) demonstrated that, due to a fortuitous interaction of truncation and non-ideal angular illumination, the truncation correction is considered negligible for the typical urban particle size distribution. Concerning the Ecotech AURORA 3000, we applied a truncation correction coefficient for total scattering ranging between 1.006 and 1.106 (mean value 1.050) using the Scattering Ångström Exponent (SAE) and the algorithm developed by Müller et al. (2011). The correction coefficient close to one is consistent with the particle number size distribution being dominated by submicron particles.

During the summer experiment at the Milan urban site, we employed a Cavity Attenuated Phase Shift (CAPS) PM SSA monitor (Aerodyne Inc., Billerica) that provides a fast-response measurement of both optical extinction and scattering coefficients of aerosol particles at 630 nm (Onasch et al., 2015). The aerosol was sampled through a line equipped with PM₁₀ size cut. The ratio of scattering to total extinction, i.e., the single scattering albedo (SSA), was calculated at 1-second time resolution. The scattering channel, which derives its data from a reverse nephelometer incorporated into the extinction measurement cell, was calibrated using polystyrene latex spheres at the beginning of the experiment, and filtered air baseline values were recorded for 45 seconds every 10 minutes for zeroing purposes. The extinction coefficients measured by the monitor are estimated to have an accuracy of ±5%. In the small particle limit (< 250 nm diameter), the SSA values are accurate within ±0.01; however, particle size distributions previously collected at the same location indicate that the accumulation particle mode at the Milan urban site has a geometric mean diameter of about 250 nm and a geometric standard deviation of about 2 (Lonati et al., 2011), corresponding to a CAPS truncation error of 1.04 (Onasch et al., 2015) which was applied to the CAPS data. After applying the truncation error correction, the absorption values were calculated by multiplying the measured extinction by the co-albedo (1-SSA).

In addition, particle light absorption at 670 nm was measured at 15-minute time resolution with a Multi-Angle Absorption Photometer (MAAP 5012 Thermo Fisher) equipped with a PM₁₀ sampling head at the urban site, both in winter and summer.

 
Aethalometer-based Absorption Coefficients

The aethalometer attenuation coefficients (b_ATN) and the nephelometer scattering coefficient (b_scat) were combined to calculate the absorption coefficient (b_abs) according to the correction algorithm developed by Collaud-Coen et al. (2010).

b_abs is the absorption coefficient at a given wavelength (λ), while α, R, and C_ref are correction coefficients considering light scattered by particles on the aethalometer filter tape, filter loading, and multiple scattering, respectively.

It is known that particles deposited on the aethalometer filter tape and able to scatter the instrument beam light decrease the intensity of light reaching the detector, leading to an overestimation of aerosol light absorption (Arnot et al., 2005) The scattering artifact is quantified by the α, which depends on the particle scattering coefficient. During the experiment, α was on the order of 10^–5, and the overestimation of b_abs due to scattering was negligible (below 0.01%).

At the same time, the accumulation of absorbing particles on the filter tape might prevent some particles from absorbing incident light beam, underestimating the actual aerosol absorption. The filter loading correction is introduced by the term R, which depends on the filter attenuation (ATN) and on the particle’s SSA. For the n-th measurement over the s-th filter spot, R_n,s is given by:

where ATN_n is the percentage attenuation, and ω̅_0 n.s is the mean SSA measured over the s-th filter spot. A first estimate of ω̅_0 n.s is calculated by combining the scattering coefficient measured by the nephelometer and the attenuation coefficient. During the experiments, R ranged between 1 and 0.7, indicating that the filter loading effect was significant. If neglected, this could have led to an underestimation of the absorption coefficient by up to 40% over a 15-minute time period (Supplementary material Fig. S2).

The C_ref coefficient corrects multiple light scattering artifacts occurring on the filter tape fibers that can lead to an absorption enhancement. We estimated C_ref by comparing the MAAP b_abs at 670 nm with the attenuation coefficient corrected for the loading effect and extrapolated at 670 nm using the aethalometer Absorption Ångström Exponent (AAE), which describes the wavelength dependence of light absorption. The resulting C_ref was 2.5 in summer and 3.0 in winter, and compared well with the values reported in Collaud-Coen et al. (2010) for aerosol samples with SSA larger than 0.8 (2.5–3.6), as well as with the average of 3.5 observed at several European ACTRIS sites (Muller, personal communication). C_ref is affected by the aerosol type. For example, liquid organic phase, which redistributes on filter fibers, modifies the C_ref value (Lack et al., 2008). In addition, C_ref depends on aerosol single scattering albedo (SSA) and on wavelength (Weingartner et al., 2003; Collaud-Coen et al., 2010). The lack of reference absorption measurements at different wavelengths did not allow a determination of C_ref wavelength dependence, and the same C_ref was used at 880 nm and 370 nm. In addition, since no MAAP measurements were available at Motta Visconti, we used the C_ref calculated at the urban location. This was an acceptable compromise given the similar seasonal SSA values observed at the two sites, as reported in Table 2.

 
Off-line Optical and Chemical Measurements

PM₁ aerosol samples were collected daily during the four field experiments (from 00:00 to 23:59 local time) using an automated Tecora aerosol sampler operating at 1.15 m³ h^–1. Samples were stored at 4°C prior to analysis. For each filter, a 1.5 cm² punch was analysed to quantify organic carbon (OC) and elemental carbon (EC) concentration, by thermo-optical technique, using NIOSH-Like heating protocol (Peterson and Richards, 2002). A second punch (1 cm²) was extracted with water and the solution was analysed by ion chromatography to quantify levoglucosan, a proxy for biomass burning. In addition, daily PM₁₀ aerosol samples were collected routinely at the urban site to measure PM₁₀ mass concentration and its chemical composition. OC and EC concentration were quantified with the same procedure employed for the analysis of PM₁ carbonaceous fraction.

Finally, we quantified daily water-soluble and methanol-soluble BrC concentrations in PM₁ aerosol samples. Two 1 cm² punches were extracted in 5 mL of mQ water and 5 mL of HPLC grade methanol, respectively, by 30-minute ultrasonication. Solutions were then filtered, and the UV-visible light absorption spectra were recorded using a TIDAS E UV/visible light spectrometer (J&M Analitik AG, Germany), equipped with a liquid waveguide capillary cell (LWCC-3050, 0.5 m long, World Precision Instruments, Sarasota, FL) in the range 300 nm–720 nm. The absorption coefficient was then calculated according to the following equation:

where Abs_λ – Abs₇₀₀ is the absorbance at λ corrected for any drift using as reference the absorbance at 700 nm, V_a is the volume of sampled air, V_l is the extraction solvent volume, and l is the optical length of the LWCC cell (Hecobian et al., 2010).

 
RESULTS AND DISCUSSION

 
Aerosol Optical Properties

Fig. 1(a) compares the hourly averaged b_abs values at 630 nm obtained from the MAAP (adjusted to 630 nm using the aethalometer AAE) and from the CAPS PMssa (obtained by difference of extinction and scattering). The data points are color coded by SSA. While the technical principles of the two techniques differ substantially (the CAPS PM_SSA measures optical properties of particles suspended in the air, while the MAAP measurements are filter-based) the correlation coefficient between the datasets is equal to 0.97 and the slope is 0.95 ± 0.01. In addition, when isolating data points corresponding to SSA larger than 0.8 (a condition that can lead to sizeable errors in the b_abs values obtained by difference), we do not observe a statistically significant bias relative to the 1:1 line. Furthermore, the good agreement between the CAPS PM_SSA and MAAP b_abs values suggests that the effects related to the absorption of gas-phase species on the filter and subsequent coating of the filter fibers (Lack et al., 2007; Subramanian et al., 2007), as well as multiple scattering artifacts (Nakayama et al., 2010), can be considered negligible during the present study.

Fig. 1. Comparison of absorption coefficients at 630 nm from CAPS PM_SSA and MAAP measurements (panel a); markers are color-coded as a function of SSA measured by the CAPS PM_SSA. Panel b compares absorption coefficients estimated form aethalometer through the described correction algorithm with the MAAP absorption data, adjusted at 880 nm in summer (triangles) and winter (circles). Measurements here reported were performed at the urban site.

Fig. 1(b) compares the b_abs estimated from aethalometer measurements at the urban site with the b_abs from the MAAP, both in summer and winter. Data are adjusted to 880 nm using the aethalometer AAE. We observed a good agreement between the data obtained from the two techniques, with residuals centered around zero and a standard deviation smaller than 20%. The results confirm the validity of the procedure used in this study to derive the absorption coefficient from aethalometer attenuation measurements, and suggest that the aethalometer correction algorithm is precise.

The box-whisker plots in Fig. 2 show the aethalometer b_abs at 880 nm (panel a) and 370 nm (panel b) measured during the four experiments in Milan (MI) and Motta Visconti (MV), grouped by season (winter in blue and summer in red). The plots report the b_abs, while average absorption coefficients at 880 nm and 370 nm, together with their standard deviations, are reported in Table 2. Higher b_abs values were observed in winter at both sampling sites, likely due to the combined effect of higher emissions of light-absorbing species from residential heating and accumulation of pollutants near the surface due to the frequent atmospheric inversion conditions. In addition, during that particular winter, low wind speed and a small number of rain events limit pollutant dispersion and wet removal. Overall, higher b_abs values were observed at the urban site of Milan independently on the season, indicating a stronger contribution of BC to the urban light-absorbing aerosols.

Fig. 2. Box-whisker plots showing the variability of absorption coefficients at 880 nm (panel a), at 370 nm (panel b) and Absorption Ångström Exponent (AAE panel c) derived from aethalometer measurements in Milan in winter and summer (MIw and MIs, respectively) and in Motta Visconti in winter and summer (MVw and MVs, respectively).

Fig. 2(c) shows the AAE during the four experiments. The AAE in the UV – IR range is calculated from the b_abs coefficients at 370 and 880 nm:

The average AAE was 1.1 and 1.0 at the urban site and 1.2 and 1.1 at the rural site, in winter and summer, respectively. Assuming that 1.1 is an upper bound for AAE_BC (see section “Brown carbon measurements”), our observations of AAE values larger than 1.1 in most of the winter data (58% and 71% of data points in Milan and Motta Visconti, respectively) indicate a consistent presence of BrC in the air samples, especially at the rural site, where AAE values are higher.

Using the average AAE and Eq. (4), we adjusted the b_abs measured during this study to different wavelengths to compare with literature values. Both the absorption coefficient values measured in Milan and their seasonal variation are comparable to those observed at other southern European urban sites. For example, the average b_abs at 637 nm in Rome in winter was 19 Mm^–1 (Costabile et al., 2017), while in Barcelona the monthly b_abs averages varied between 10 Mm^–1 (in summer) and 20 Mm^–1 (in winter) (Ealo et al., 2018). These values compare well with the seasonal averages observed in Milan when adjusted to 637 nm, i.e., 11 and 20 Mm^–1. Conversely, the average b_abs at Motta Visconti at 637 nm (7 and 14 Mm^–1 in summer and winter, respectively) are higher than typical southern European rural coefficients (2–4 Mm^–1) (Zanatta et al., 2016; Ealo et al., 2018). These results confirm the conclusions of a previous study reporting small differences between urban and rural sites in the Po Valley, and indicating that air masses are well-mixed in the basin (Sandrini et al., 2014).

 
Black Carbon Optical Properties

In order to characterize BC optical properties, this study uses EC from thermal-optical measurements as a proxy, even if EC and BC are not equivalent. In fact, BC typically refers to both refractory and non-refractory carbonaceous aerosol strongly absorbing across a wide part of the visible spectrum, while EC indicates only the refractory component of the carbonaceous aerosol (Bond and Bergstrom, 2006). In the PM₁ size fraction, EC concentrations in Milan averaged 1.52 and 0.80 mg m^–3, while in Motta Visconti averaged 0.92 and 0.48 mg m^–3 in winter and summer, respectively. Higher EC concentrations were observed at the urban site, due to a larger contribution of anthropogenic combustion emissions from vehicular traffic both in winter and summer.

MAC_EC (proxy for MAC_BC) is defined as the ratio between the daily average light absorption coefficient and the EC concentration from thermal-optical analysis of PM₁ daily samples:

where the b_abs coefficients and EC measurements are obtained from instrument with two different size cuts (PM₁₀ and PM₁ respectively). However, comparison between PM₁₀ and PM₁ EC measurements at the urban site indicates that, on average, at least 90% of EC is contained in submicron particles (Fig. S3). In conclusion, the error introduced by neglecting the 10% mass fraction of EC larger than 1 m is likely within the uncertainty of EC measurements, typically estimated to be around 30% (Karanasiou et al., 2015). Further, we use the b_abs coefficients measured at 880 nm to remove potential interference from the BrC absorption at shorter wavelengths, and only days characterized by b_abs data availability larger than 70% are considered for this analysis.

The average MAC_EC values for the four measurement periods and their confidence intervals, (± standard deviation) obtained from Eq. (5) are reported in Table 2. MAC_EC values were comparable across sites and seasons. Despite the short duration of the experiments, this result is consistent with previous observations that did not report any seasonal variation of EC optical properties in rural Europe across longer periods (Zanatta et al., 2016).

Furthermore, the MAC_EC measured at Motta Visconti during this study falls in the range of values reported for other European rural background sites adjusted to 880 nm, i.e., 5.7–16.4 m² g^–1 (Genberg et al., 2013, Zanatta et al., 2016). Similarly, the average MAC_EC observed in Milan compares well with measurements in European urban sites, including Barcelona (6.7 m² g^–1) (Reche et al., 2011) and Paris (8–10 m² g^–1 at 950 nm) (Sciare et al., 2011).

Given that the MAC of freshly generated soot ranges between 6.3 and 8.7 m² g^–1 at 550 nm, or to 3.9 and 5.4 m² g^–1 at 880 nm (Bond and Bergstrom 2006), the higher MAC values observed during this and other ambient experiments are likely due to changes in BC optical properties with atmospheric aging, and possibly to intra-particle mixing configuration.

We investigated the effect of intra-particle mixing state on BC optical properties by looking at the variability of MAC_EC as a function of the non-refractory PM₁ to EC ratio (PM_1nr/EC). The PM_1nr/EC ratio can be used as a proxy for BC coating thickness. Zanatta et al. (2016) observed that, in most European sites reporting long term BC and EC measurements, MAC_EC increases with the PM_1nr/EC ratio, with different slopes depending on the location. Fig. 3 reports the measurements performed at the Milan urban site (Fig. 3, panel a) and at the Motta Visconti rural site (Fig. 3, panel b). In Milan, the PM_1nr/EC ratio varies between 5 and 30, while MAC_EC varied by a factor of two, with values ranging from 6 and 12 m² g^–1. The lowest MAC_EC was 6.6 m² g^–1 and was observed on June 21. During that day, the site sampled BC particles emitted from a nearby fire clearly visible from the measurement site as a black smoke plume. It is reasonable to assume that BC particles sampled during such episode were fresh and not internally mixed, explaining the relatively low MAC_EC compared to the PM_1nr/EC ratio of 12. The black dotted lines in Fig. 3 identify orthogonal regression lines of the measured data after removing the June 21 point. After excluding the local fire episode, the linear correlation agrees with observations reported at most European sites and indicates that the PM_1nr/EC ratio explains 75% of the MAC_EC variability observed in Milan. On the contrary, the correlation between PM_1nr/EC and MAC_EC at Motta Visconti was not statistically significant, with a Pearson coefficient of –0.02. The difference in slope and correlation might be related to different intra-particle mixing configurations and the source variability. The variability of fuel type might affect morphology and mixing state of combustion-generated particles, which in turn modify their MAC (Bond et al., 2006), The variability of fuel type can also introduce a bias in the EC thermal-optical measurements (Cavalli et al., 2010), hampering the dependency of MAC on PM_1nr/EC especially in a short lasting experiment. The rural measurement site is in an area characterized by intensive agricultural activities (rice fields) and we cannot exclude the impact of local combustion activities such as agricultural waste burning, mixing with urban outflow. Consistently, there was a larger variability range of the levoglucosan to EC ratio in winter (0.04–0.30) at Motta Visconti than in Milan (0.02–0.06) (levoglucosan concentration in summer was below detection limit during the investigated periods).

Fig. 3. Dependence of elemental carbon mass absorption cross section (MAC_EC) on the non-refractory PM₁ to elemental carbon ratio (PM_1nr/EC), used as a proxy of soot particle coating, during the experiments in Milan (urban site, panel a) and Motta Visconti (rural site, panel b).

 
Brown Carbon (BrC) Measurements

BrC light absorption at 370 nm (b_{abs BrC 370}) is determined from the aethalometer measurements, by subtracting the BC contributions according to the following equation:

This equation implies that the main species contributing to the absorption at 370 nm are BC and BrC.

The use of Eq. (6) to obtain b_{abs BrC 370} requires an assumption about the choice of the AAE_BC values, which are known to depend on BC particle size and their mixing state and to span a wide range. Based on the Mie-theory core-shell model (spherical particles), Lack and Cappa (2010) concluded that BC can show AAE values as high as 1.6, and that AAE smaller than 1.6 cannot exclude a BrC presence. Li et al. (2016) used numerical simulations to calculate the AAE_BC variability range and reported values of 0.9–1.5 for fractal aggregates, and 0.6–1.5 for spheres and spheroids. Conversely, ambient measurements of AAE_BC performed with various methods point towards a narrower range. Chung et al. (2012b) estimated an AAE of 0.7–1 for BC and 1.6–1.8 for BrC between 350 and 970 nm. Zotter et al. (2017) estimated the AAE of BC and wood burning particles (containing both BC and BrC) comparing aethalometer data and radiocarbon measurements and concluded that the best estimate for AAE_BC was 0.9 in the 470–950 nm range. Similarly, Martinsson et al. (2017) validated the assumption of AAE_BC equal to unity using radiocarbon measurements. Laboratory experiment reported AAE_BC of fresh fossil fuel burning emission between 0.8 and 1.1 (Kirchstetter 2004; Sharma et al., 2013; Yuan et al., 2016; Blanco-Alegre et al., 2020). Therefore, we used a range of AAE_BC values collected during field and laboratory experiments (0.7–1.1) and calculated a center value of b_{abs BrC 370} using AAE_BC equal to 0.9. Using literature AAE_BC ambient and laboratory measurement range (Chung et al., 2012a; Martinsson et al., 2017; Zotter et al., 2017), we also calculated the upper and lower bounds of b_{abs BrC 370} using AAE_BC equal to 0.7 and 1.1, respectively.

Although Mie theory calculation might indicate that AAE_BC can be larger than 1.1, it is worth noting that an AAE_BC of 1.1 in this study would lead to many negative Abs_BrC370 values, especially in summer (83% and 98% of hourly data at Milan and Motta Visconti, respectively). This result would not be consistent with the off-line measurements indicating BrC presence during the summertime and wintertime experiments (Fig. 5). Therefore, an AAE_BC equal to or higher than 1.1 is unlikely to represent the optical properties of BC in the Po valley.

Fig. 5. Brown carbon mass absorption cross section (MAC_BrC) versus wavelength and date in Milan in winter (panel a), Milan in summer (panel b), Motta Visconti in winter (panel c), and Motta Visconti in summer (panel d).

Figs. 4(a) and 4(b) report the daily variation of the BrC light absorption coefficient at 370 nm at the urban and rural site, respectively. The shadowed areas indicate the upper and lower bounds of b_{abs BrC 370} based on the different AAE_BC assumptions. The b_{abs BrC 370} coefficient in winter shows maxima during evening and night-time, in agreement with the typical diurnal trend of wood burning for residential heating, previously identified as one of the main sources of OA in the Po Valley (Gilardoni et al., 2014, 2016). Several studies indicate that primary and secondary wood burning OA contribute to aerosol light absorption in the UV and visible range (Habib et al., 2008; Hecobian et al., 2010; Saleh et al., 2014). Even though wood burning emissions are higher during the evening when stoves and fireplaces are used for residential heating, biomass burning OA remains high during the night and decreases during day-time due to the evolution of boundary layer height. Interestingly, the diurnal variation of the b_{abs BrC 370} coefficient in winter was similar at the urban and rural sites due to the similar evolution of the mixing layer, although the rural site showed higher BrC signals in the evening. Consistently, regional inventories in the urban settings (Inemar, 2014).

Fig. 4. Diurnal trends of BrC absorption coefficients at 370 nm (b_{abs BrC 370}) from aethalometer data at the urban site (panel a) and the rural site (panel b), dependence of daily average b_{abs BrC 370} from aethalometer on levoglucosan concentration, a marker of wood burning emissions (panel c), and correlation of b_{abs BrC 370} from aethalometer with methanol soluble BrC absorption coefficient (b_{abs BrC 370}-MeOH) measured off-line with daily resolution (panel d). The shaded areas in panels a and b and the error bars in panels c and d indicate the variability ranges derived from assuming AAE_BC equal to 0.7 and 1.1, while the central values are estimated assuming AAE_BC equal to 0.9.

The rural site in summer was characterized by an increase in the b_{abs BrC 370} during the central part of the day, when solar radiation was higher and the highest concentration of secondary OA is expected (Saarikoski et al., 2012). Ambient and laboratory observations confirm that secondary OA can contribute to aerosol light absorption ( Shapiro et al., 2009; Chang and Thompson, 2010, Hecobian et al., 2010), and therefore can be considered as a form of BrC. Updyke et al. (2012) observed the formation of light-absorbing secondary OA from the reaction of biogenic and anthropogenic precursors with ammonia in the laboratory, suggesting that the significant emission of ammonia from agricultural activities in the Po Valley can contribute to BrC summertime loadings in rural areas, which are significantly impacted by agricultural emissions. At the urban site, b_{abs BrC 370} diurnal trend is less clear. Previous studies reported high night-time BrC in urban areas due to dark-chemistry secondary OA or primary OA accumulating near the surface at night (Hecobian et al., 2010). Still, the low b_{abs BrC 370} coefficients and the lack of specific organic tracers make difficult to reach a definitive conclusion about the specific sources of BrC in summer at both urban and rural sites at different times of the day.

To test the attribution of winter BrC to biomass burning, Fig. 4(c) shows the correlation between the 24-hr wintertime average BrC light absorption at 370 nm (from Eq. (6)) and PM₁ daily concentration of levoglucosan, a wood burning tracer. The data are best fitted by an AAE_BC = 0.9 (R² = 0.70, the solid line), suggesting that BrC in winter in the Po Valley was strongly related to wood burning from residential heating. The linear fitting functions for the lower and upper bounds of AAE_BC are indicated by the dashed gray and green lines in Fig. 4(c). The corresponding correlation coefficient between b_{abs BrC 370} and levoglucosan were 0.68 (assuming AAE_BC = 1.1) and 0.54 (assuming AAE_BC = 0.7).

 
BrC Optical Properties

To investigate the optical properties of BrC, PM₁ OA collected daily on quartz filters was extracted with methanol and water, and the UV-VIS light absorption spectra were measured. Table 2 reports the average absorption coefficients of methanol-soluble and water-soluble BrC, and the corresponding AAE_BrC over the range 330–500 nm. Previous studies show that water extracts less than 70% of the total light-absorbing OA, while methanol extracts more than 90% (Chen and Bond 2010; Laskin et al., 2015; Kumar et al., 2018). The absorption coefficients measured in this study confirm that methanol extracts BrC more efficiently than water, and AAE values indicate that the two solvents extract BrC with different optical properties, and likely different chemical composition.

Fig. 4(d) compares methanol soluble BrC absorption coefficients (b_{abs BrC 370}-MeOH) with aethalometer measurements (b_{abs BrC 370}) averaged over 24-hour sampling period at both rural and urban site, in summer and winter. As in previous graphs, the b_{abs BrC 370} from aethalometer data is calculated using Eq. (6) and assuming AAE_BC = 0.9. The error bars in Fig. 4(d) are calculated using the upper and lower bound of b_{abs BrC 370} assuming AAE_BC equal to 0.7 and 1.1, respectively. Although aethalometer data refer to PM₁₀ aerosol and methanol BrC measurements were performed on PM₁ size fraction, the comparison shows a good correlation between the two parameters for all datasets with an average R² = 0.74 and intercept close to zero (0.7 ± 0.6). However, the orthogonal fit with a slope of 1.8 indicates that a much higher absorption coefficient is measured by the aethalometer. Indeed, particle absorption is expected to be about 1.5–2 times larger than bulk solution absorption for ambient BrC particles in the accumulation mode, due to size effect (Liu et al., 2013). We can exclude the presence of mineral dust as non-carbonaceous light absorbing material, since iron oxides, responsible for dust light absorption, would not be extracted by methanol and would affect exclusively on-line data, leading to a positive line intercept. The corresponding plot showing the water soluble BrC absorption is reported in Supplementary material (Fig. S5).

Assuming that the totality of BrC is extracted by methanol, we calculated the MAC of BrC (MAC_BrC) by dividing the b_{abs BrC}-MeOH at each wavelength by OC concentration (Fig. 5). Averaged MAC_BrC at 370 nm for each site and over each season are reported in Table 2 and vary between 0.3 and 1.2 m² g^–1.

During summer, MAC_BrC was in agreement with values reported for urban and rural sites in summer in the United States (Liu et al., 2013, Zhang et al., 2013) and with MAC of secondary OA from anthropogenic and biogenic emissions (Laskin et al., 2015; Liu et al., 2016). The slightly higher MAC_BrC observed at the urban site in summer might confirm laboratory observations indicating that SOA from anthropogenic aromatic precursors absorbs light more efficiently than SOA from a mixture of anthropogenic and biogenic mixed sources, that might be more characteristic of rural areas (Laskin et al., 2015; Liu et al., 2016).

In winter MAC_BrC was two to four times higher than summer values (1.2 m² g^–1). Chen and Bond (2010) measured a MAC_BrC at 370 nm for laboratory wood burning experiments between 1.5 and 2 m² g^–1. BrC from prescribed fires shows a MAC at 365 nm of 1.3 m² g^–1 (Xie et al., 2017). Finally, MAC_BrC of ambient aerosol in areas strongly impacted by residential wood burning emissions ranges between 1.5 and 1.8 (Cheng et al., 2016; Shen et al., 2017). The MAC_BrC reported for this study using the offline measurements is comparable to the lower bound of the literature range of MAC_BrC from wood burning emissions. This result confirms the impact of biomass combustion as a source of BrC in the Po valley during the cold season.

To verify the consistency of on-line and off-line measurements, Fig. 6(a) compares the campaign average AAE of methanol-soluble BrC with the variability range of the AAE_BrC from aethalometer data. We calculated the aethalometer AAE_BrC by subtracting the BC absorption from the total aerosol absorption at 370 nm and at 550 nm, and assuming AAE_BC equal to 0.9 (lower and upper bounds are calculated with AAE_BC equal to 1.1 and 0.7, respectively). The aethalometer absorption coefficient at 550 nm was extrapolated based on the measured aerosol AAE in the range 370–880 nm. The average AAE_BrC values obtained from off-line and on-line measurements agree well. Over the four experiments AAE of methanol soluble BrC ranged between 2.2 and 4.4, in agreement with the lower bound of values reported in literature (from 2 up to 11) (Zhang et al., 2013; Laskin et al., 2015). The AAE variability is due to the complexity of BrC chemical constituents, which depend on sources and atmospheric processes.

Fig. 6. Panel a: comparison between campaign averages of AAE_BrC from off-line methanol-soluble BrC measurements (blue circles) and the AAE_BrC ranges estimated from on-line aethalometer measurements (gray rectangles); aethalometer-based AAE ranges are calculated assuming AAE_BC equal to 0.7 and 1.1, while central value is calculated assuming AAE_BC equal to 0.9 (open black circles). Panels b and c: dependence of methanol-soluble BrC Absorption Ångström Exponent (AAE_BrC) on the elemental to organic carbon ratio (EC/OC) during the four experiments, in winter and summer, respectively.

Figs. 6(b) and 6(c) investigate the AAE of methanol-soluble BrC as a function of the EC to OC ratio. For ambient measurements, the EC/OC ratio is an indicator of primary versus secondary carbonaceous aerosol, and in winter is also related to the different contribution of wood burning and fossil fuel burning to carbonaceous aerosol loading (Sandrini et al., 2014). In winter (Fig. 6(b)) the AAE increased with the decreasing of the EC/OC ratio, reflecting the change between the urban to the rural sites (with rural sites having lower EC/OC ratio). The overall trend is consistent with laboratory experiments showing an increase of the AAE_BrC associated with aging of wood burning emissions (Saleh et al., 2014, Xie et al., 2017, Kumar et al., 2018), suggesting that during the cold season the variability of AAE is controlled by wood burning aerosol loading and its atmospheric evolutions. Similarly, in summer (Fig. 6(c)) the highest AAE_BrC (~4) is observed for the lowest EC/OC ratios (0.1), in agreement with a larger contribution of secondary organic aerosol. At the urban site, AAE displays smaller variability and lower values compared with the rural data. As already suggested by the higher MAC, urban BrC in summer was likely dominated by anthropogenic SOA from aromatic precursors, whose absorption extends over a larger region of the UV-Vis spectra than aliphatic molecules, leading to a lower AAE compared to biogenic SOA (Zhang et al., 2013). The higher AAE and the lower MAC at the rural location in summer agree with a larger contribution of biogenic SOA.

 
CONCLUSIONS

We investigated the seasonal and spatial variability of BC and BrC absorption across the Po Valley. Ambient measurements of BC optical properties and a better quantification of BrC light absorption are necessary to reduce model uncertainty in describing aerosol-radiation interaction, and to improve model ability to predict the short-term and long-term effects of climate policies.

Light absorption measurements performed during this study indicate that BC concentrations in the urban area are comparable to those observed in other southern European cities; conversely, rural BC concentrations are higher than expected indicating that BC impacts in the Po Valley are not limited to urban agglomerates and urban population. Aggressive measures should be implemented to reduce BC emissions from both traffic and wood combustion, as well as agricultural waste burning.

BC mass absorption cross section was investigated using thermal-optical EC measurements. MAC_EC showed little seasonl and spatial variability, with an average value of 10.0 m² g^–1 at 880 nm. Despite the limited dataset investigated, the analysis of MAC_EC variability suggests that the ratio between non-refractory PM₁ and EC can be used in certain locations as a proxy for BC coating thickness, to describe BC optical properties variability. Nevertheless, such parameterization relies on assumptions (such as intra-particle mixing configuration) that might limit its general application on short time scales, thus caution is recommended when this kind of approach is applied.

Although characterized by lower absorption efficiency, BrC contributes together to wavelength absorption. In the Po Valley, BrC is directly emitted by wood combustion during winter, while it is found in secondary organic aerosol during summer. BrC extraction with methanol and water shows that more than 50% of light absorbing organic aerosol is water insoluble. The water-insoluble fraction is likely dominated by aromatic organic moieties during both seasons, as suggested by the higher AAE of methanol soluble BrC.

MAC_BrC differs significantly depending on the sources, with values for secondary organic aerosol (SOA) that are about two times lower than those for biomass combustion. The accurate determination of MAC for specific organic aerosol sources requires coupling source apportionment analysis with optical property measurements, but such analysis is behind the scope of this study.

Ambient observations collected in this study confirm laboratory measurements indicating that the contribution of SOA to light absorption is not negligible. This has significant climate implications. In fact, the radiative impact of BrC, at wavelength larger than 400 nm, is significant over bright surfaces (Chen et al., 2010) such as clouds or snow. The mechanism of formation of BrC through atmospheric processing of aerosol precursors increases the probability of light absorbing aerosol formation in the upper part of the troposphere, including above clouds, potentially leading to a positive radiative forcing (Zhang et al., 2017). Deposition of BrC on snow and ice can promote their melting and albedo change (Zhang et al., 2019; Beres et al., 2020). In addition, BrC absorption at short wavelength (below 400 nm) reduces atmospheric actinic flux, impacting atmospheric photochemistry and ozone production (Mok et al., 2016). In conclusions, a better knowledge of SOA formation, microphysical, and optical properties is needed to improve the ability of climate models to describe the effect of short-lived climate forcers.

 
ACKNOWLEDGMENTS

The authors would like to acknowledge Fondazione Cariplo (fund number 2015-1195) for supporting the research activities of the Black Carbon Tool (Black CaT) project.