Higher absorption enhancement of black carbon in summer shown by 2-year measurements at the high-altitude mountain site of Pic du Midi Observatory in the French Pyrenees

Black-carbon-containing particles strongly absorb light, causing substantial radiative heating of the atmosphere. The climate-relevant properties of black carbon (BC) are poorly constrained in high-altitude mountain regions, where many complex interactions between BC, radiation, clouds and snow have important climate implications. This study presents 2-year measurements of BC microphysical and optical properties at the Pic du Midi (PDM) research station, a high-altitude observatory located at 2877 m above sea level in the French Pyrenees. Among the long-term monitoring sites in the world, PDM is subject to limited influence from the planetary boundary layer (PBL), making it a suitable site for characterizing the BC in the free troposphere (FT).

The classification of the dominant aerosol type using aerosol spectral optical properties indicates that BC is the predominant aerosol absorption component at PDM and controls the variation in single-scattering albedo (SSA) throughout the 2 years. Single-particle soot photometer (SP2) measurements of refractory BC (rBC) show a mean mass concentration (M_rBC) of 35 ng m⁻³ and a relatively constant rBC core mass-equivalent diameter of about 180 nm, which are typical values for remote mountain sites. Combining the M_rBC with in situ absorption measurements, a rBC mass absorption cross-section (MAC_rBC) of 9.2 ± 3.7 m² g⁻¹ at λ=880 nm has been obtained, which corresponds to an absorption enhancement (E_abs) of ∼2.2 compared to that of bare rBC particles with equal rBC core size distribution. A significant reduction in the $\mathrm{\Delta }{M}_{\text{rBC}}/\mathrm{\Delta }\mathrm{CO}$ ratio when precipitation occurred along the air mass transport suggests wet removal of rBC. However we found that the wet removal process did not affect the rBC size, resulting in unchanged E_abs. We observed a large seasonal contrast in rBC properties with higher M_rBC and E_abs in summer than in winter. In winter a high diurnal variability in M_rBC (E_abs) with higher (lower) values in the middle of the day was linked to the injection of rBC originating from the PBL. On the contrary, in summer, M_rBC showed no diurnal variation despite more frequent PBL conditions, implying that M_rBC fluctuations are rather dominated by regional and long-range transport in the FT. Combining the $\mathrm{\Delta }{M}_{\text{rBC}}/\mathrm{\Delta }\mathrm{CO}$ ratio with air mass transport analysis, we observed additional sources from biomass burning in summer leading to an increase in M_rBC and E_abs. The diurnal pattern of E_abs in summer was opposite to that observed in winter with maximum values of ∼2.9 observed at midday. We suggest that this daily variation may result from a photochemical process driving the rBC mixing state rather than a change in BC emission sources.

Such direct 2-year observations of BC properties provide quantitative constraints for both regional and global climate models and have the potential to close the gap between model-predicted and observed effects of BC on the regional radiation budget and climate. The results demonstrate the complex influence of BC emission sources, transport pathways, atmospheric dynamics and chemical reactivity in driving the light absorption of BC.

1 Introduction

Black carbon (BC) is a light-absorbing carbonaceous aerosol produced by incomplete combustion of fossil fuels, including not only anthropogenic emissions from traffic, residential heating and cooking, power plants, and industries but also natural emissions such as biomass burning (Bond et al., 2013; Bond and Bergstrom, 2006). Recent scientific assessments of the sixth IPCC (Intergovernmental Panel on Climate Change) report (Szopa et al., 2021) estimate that BC is the most absorbing atmospheric aerosol with a best estimate of effective radiative forcing around +0.107 W m⁻², thereby increasing the global mean surface air temperature by 0.063 ^∘C for the period 1750–2019 (Szopa et al., 2021). The contribution of BC to climate change is estimated to be among the highest uncertainties (∼90 %) in climate models, limiting their accuracy (Bellouin et al., 2020). The large uncertainty in BC direct radiative forcing due to BC–radiation interactions can be attributed, in addition to uncertainties in BC emissions and lifetime, to variations in BC optical properties that are neglected by climate models (Matsui et al., 2018).

A crucial factor in estimating the BC radiative effect is the mass absorption cross-section (MAC_BC), which is defined as the light-absorption-equivalent cross-section of BC per unit of mass concentration (M_BC). The MAC_BC can be calculated either by dividing the measured absorption coefficient of BC by its mass concentration or by using Mie's theory and the BC size distribution and coating thickness as input variables. Observations show that the BC radiative forcing is likely underestimated by around 10 % to 40 % in current climate models due to too low a simulated MAC_BC (Bond et al., 2013; Boucher et al., 2016; Matsui et al., 2018; Myhre and Samset, 2015). In situ measurements of MAC_BC have reported a wide range of values, from 3.8 to 58 m² g⁻¹ (Wei et al., 2020). Although such high variability can be attributed, in part, to the determination method of the MAC_BC based on M_BC and absorption measurement techniques, differences in MAC_BC values were found even for the same measurement technique.

Values of MAC_BC depend on BC microphysical and chemical properties, which are related to their emission sources (Schwarz et al., 2008) and the effects of aging processes during transport in the atmosphere (Ko et al., 2020; Laborde et al., 2013; Sedlacek et al., 2022). Freshly emitted BC is made of porous, fractal-like aggregates of nanoparticles (Beeler and Chakrabarty, 2022; China et al., 2013) that can become coated by condensation and/or coagulation with non-BC components (such as sulfate, nitrate and organic components) during atmospheric aging (Fierce et al., 2020). Conversely this coating can be removed through evaporation and/or chemical processing via the production of more volatile substances (Sedlacek et al., 2022). Numerous studies have demonstrated that coating of BC with non-absorbing materials is accompanied by an enhancement of light absorption (E_abs) through the so-called lensing effect (Cappa et al., 2012; Denjean et al., 2020; Yuan et al., 2021; Healy et al., 2015; Liu et al., 2015; McMeeking et al., 2014; Peng et al., 2016; Van de Hulst, 1957; Xie et al., 2019; Schwarz et al., 2006; Yus-Díez et al., 2022). However, most of these measurements were performed in the planetary boundary layer (PBL) and over short periods from a few hours to as long as a season.

Both observations and model simulations pointed out an amplification of the warming rate by greenhouse gases and absorbing aerosols at high mountain sites compared to PBL areas (Gao et al., 2018; Liu et al., 2009; Pepin et al., 2019; Rangwala, 2013). López-Moreno et al. (2014) have shown by running several regional climate models that the occurrence of winter warm events in the Spanish Pyrenees will gradually increase until 2080. This includes an increase in the number of warm days and nights and the number of snow/ice melting days at altitudes above 2000 m above sea level (m a.s.l.). This so-called elevation-dependent warming (EDW) has been reviewed by the Mountain Research Initiative EDW Working Group, 2015, who listed the possible mechanisms behind this phenomenon (Pepin et al., 2015). Among the reasons given, BC is a potential driver of EDW by both absorbing solar radiation in the troposphere and decreasing the surface albedo when deposited on the cryosphere, thereby accelerating snowmelt (Réveillet et al., 2022). In addition, BC has been found to have a higher radiative effect when it is located above clouds rather than near the surface (Samset and Myhre, 2015; Sanroma et al., 2010). All these findings highlight the importance of studying BC at high-altitude mountain sites, where its effects on climate could be even more significant.

This study presents 2-year continuous measurements of BC and aerosol properties conducted during the hygroscopic properties of black carbon (h-BC) campaign at the high-altitude long-term monitoring station Pic du Midi (PDM). Located at 2877 m a.s.l. in the French Pyrenees, PDM has been long identified as a clean remote station (Marenco et al., 1994).

This paper aims to provide a comprehensive picture of the seasonal and diurnal variability in refractory BC (rBC) properties at PDM and to explore the processes driving these properties. Specifically, the following questions are addressed:

1. What are the air mass transport pathways impacting PDM?

2. What is the seasonal variability in aerosol optical properties and dominant aerosol types? What is the specific contribution of rBC to aerosol absorption?

3. How do the microphysical and optical properties of rBC vary on a seasonal and daily basis?

4. What are the roles of wet deposition, the source and the transport pathway in driving rBC absorption?

2 Methods

2.1 Measurement site and observation period

Measurements were performed at the Pic du Midi (PDM; 42.9^∘ N, 0.1^∘ E; 2877 m a.s.l.) mountain research station in the French Pyrenees. This station is part of the Pyrenean Platform for Observation of the Atmosphere (P2OA).¹

As shown in Fig. 1, the site is located 150 km east of the Atlantic coast. The high isolated summit lies around 20 km north of the main ridge of the Pyrenees (on the France–Spain border) and thus closely overlooks the French plain. Using a backward particle dispersion model, Henne et al. (2010) found the influence of local anthropogenic emissions to be very limited at PDM and classified the station in the “mostly remote” category. Collaud Coen et al. (2018) defined an “ABL-Topoindex” as a metric of the atmospheric boundary layer influence for a mountain site. PDM has been found to have a low ABL-Topoindex, similar to other Alpine high-altitude stations. PDM is thus a suitable site to study both the background lower free troposphere (FT) over long timescales and the injection of air masses from the PBL (Hulin et al., 2019; Tsamalis et al., 2014; Fu et al., 2016; Marusczak et al., 2017). Long-term monitoring of numerous meteorological, gas and aerosol parameters has been conducted for nearly 2 decades, notably through the Global Atmosphere Watch (GAW) program of the World Meteorological Organization (WMO), as well as the national research infrastructure ACTRIS-France. Results from the h-BC campaign performed from February 2019 to January 2021 at PDM (in addition to the routine measurements) are presented in this paper.

Figure 1 Geographical location of the Pic du Midi Observatory in the French Pyrenees (© IGN and © Google Earth).

2.2 Instrumentation

2.2.1 Total inlet

All particle-measuring instruments sampled air taken in parallel from a whole air inlet, located 2 m above the building roof. This inlet is used for the long-term observations in mountainous sites and is designed to maintain an isokinetic and laminar flow. The main flow rate was fixed at about 460 L min⁻¹. The splitter was fixed at the end of the stainless tube. The hat of the whole air inlet and the stainless tube were both thermo-regulated in order to avoid frost and gradually regulate the temperature of the sampled air to the measurement room. The air was heated to around 20 ^∘C in order to perform aerosol in situ measurements at a relative humidity lower than 30 %. The instrumental room temperature was regulated at around 20 ^∘C. The annual cycle of the dew point temperature varied between about −10 and +5 ^∘C.

2.2.2 Black carbon measurements

BC can be measured by different methods which are based on different BC properties. Petzold et al. (2013) defined a specific nomenclature for BC according to the BC quantification method. Following the recommendation of the authors, BC quantified by laser-induced incandescence, filter-based absorption and thermal–optical analysis will be referred to as refractory black carbon (rBC), equivalent black carbon (eBC) and elemental carbon (EC), respectively. More general discussion on BC without focusing on its measurement technique will be referred to as BC. The mass concentration of rBC and rBC size distribution were measured by a single-particle soot photometer (SP2; Droplet Measurement Technologies, Longmont, CO, USA). Its operating principles have been described in previous articles (Gao et al., 2007; Moteki and Kondo, 2007; Schwarz et al., 2006). In short, this instrument uses a laser-induced incandescence technique which quantifies rBC mass in single particles. A continuous intracavity laser beam (Nd:YAG; λ=1064 nm) is used to heat rBC-containing particles to their vaporization point. The measured incandescence signal of an individual rBC-containing particle can be converted to a rBC mass, using a calibration curve obtained by recording the incandescence signal peak height of mobility size-selected fullerene soot (Alfa Aesar, lot no. FS12S011) and assuming a BC mass density of 1.8 g cm⁻³ (Moteki and Kondo, 2010). This calibration was performed twice a year and did not change during the 2 years of the measurement campaign.

The SP2 data were processed using a Python code following the method used in the SP2 Toolkit from the Paul Scherrer Institute (Gysel et al., 2009). A comparison of M_rBC resulting from the SP2 Toolkit with our Python processing is presented in Text S1 in the Supplement. The SP2 used in this study measured rBC cores over a size range between 90 and 580 nm. However, the observed size distributions showed that an important fraction (around 12 %) of M_rBC at diameters below 90 nm is not measured by the SP2 (Fig. S2 in the Supplement). Because of these small-mode particles outside the SP2 detection range, the quantification of M_rBC could be underestimated. To compensate for the missing mass, the observed rBC size distributions have been fitted daily using the sum of three individual lognormal distributions to extrapolate rBC size distribution in the range of 1 to 1000 nm. The positions of the three modes were constrained in the following ranges: Mode 1, $\mathrm{50}<D_{\mathrm{g}}<\mathrm{100}$ nm and $\mathrm{1.2}<{\mathit{\sigma }}_{\mathrm{g}}<\mathrm{3}$; Mode 2, $\mathrm{150}<D_{\mathrm{g}}<\mathrm{250}$ nm and $\mathrm{1.3}<{\mathit{\sigma }}_{\mathrm{g}}<\mathrm{2.9}$; and Mode 3, $\mathrm{350}<D_{\mathrm{g}}<\mathrm{500}$ nm and $\mathrm{1}<{\mathit{\sigma }}_{\mathrm{g}}<\mathrm{3}$, with D_g and σ_g the geometric mean diameter and the geometric standard deviation, respectively. Using the fitting procedure, a time-dependent missing mass correction was applied to the observed M_rBC to calculate the overall M_rBC. The average missing mass correction factor over the campaign was 1.2 ± 1.1 (mean value ± SD). More details on the SP2 data procedure can be found in Text S2 in the Supplement. The extent to which the uncertainty in this fitting procedure contributes to the overall M_rBC was quantified by comparing the M_rBC calculated from the observed BC size distribution and the fit curve over the SP2 size range. An excellent match was obtained between the measured and fitted size distribution, resulting in differences of less than 0.2 %. The combined uncertainty in the M_rBC was estimated to be about 24.5 % by calculating the quadratic sum of the measurement uncertainties in sampling flow, anisokinetic sampling errors and the missing mass correction factor.

2.2.3 Aerosol properties

A scanning mobility particle sizer (SMPS), combining a differential mobility analyzer (DMA; model 3071, TSI Inc., Shoreview, MN, USA) and a condensation particle counter (CPC; model 3772, TSI Inc., Shoreview, MN, USA), allowed for the determination of aerosol size distribution between 12.6 and 532.6 nm.

Aerosol scattering coefficients (σ_sca) at three wavelengths (450, 525 and 635 nm) were measured with an integrating nephelometer (model Aurora 3000, Ecotech Pty Ltd, Knoxfield, Australia). A calibration with carbon dioxide and filtered air was performed every 3 months. The instrument measures σ_sca in the angular range 10–170^∘, and the correction of Müller et al. (2011) was used to account for the angular truncation errors.

Aerosol absorption coefficients (σ_abs) were measured by a seven-wavelength Aethalometer (model AE33, Magee Scientific Company, Berkeley, CA, USA; measuring wavelengths: 370, 470, 520, 590, 660, 880 and 950 nm). This instrument measures light attenuation through a filter on which aerosol sample is deposited. The Aethalometer filter loading effect was corrected online by the dual-spot manufacturer correction proposed by Drinovec et al. (2015). The multiple-scattering parameter used to correct the measured attenuation was set to 3.22, according to the value obtained at λ=880 nm by Yus-Díez et al. (2021) at the mountainous site of Montsec d'Ares located less than 200 km from the PDM (see the Table S3 in the Supplement of Yus-Díez et al., 2021). Uncertainty in the corrected σ_abs was estimated to be 35 % (Zanatta et al., 2016). The detection limit of the Aethalometer is 0.039 Mm⁻¹ (corresponding to an equivalent black carbon mass concentration of 0.005 µg m⁻³).

2.2.4 Gas-phase measurements

Two different instruments have been deployed to measure carbon monoxide (CO) with a final time resolution of 1 h: an IR-absorption analyzer (model 48CTL, Thermo Environmental Instruments (TEI), New Delhi, India) placed close to the aerosol instrumentation in order to detect pollution plumes produced locally at PDM and a cavity ring-down spectrometer (CRDS; model G2401, Picarro, Santa Clara, CA, USA), located in an other building, used to measure the background carbon monoxide (CO) concentration (see Sect. 2.3).

A key issue in our study is the distinction between FT- and PBL-influenced air masses. Optical properties of BC depend on its aging and transport pathways in the atmosphere, so it is crucial to determine whether it has been transported over the PBL or in the FT. For this purpose, we routinely monitor the diurnal cycle of radon (²²²Rn) volumetric activity at PDM with a 1500 L high-sensitivity radon monitor (model D1500, Australian Nuclear Science and Technology Organisation (ANSTO), Australia) (Whittlestone and Zahorowski, 1998). Radon is an inert radioactive gas emitted from ice-free soils with a half-life of 3.8 d, making it the most reliable tracer to discriminate between the FT- and PBL-influenced air masses (Chambers et al., 2013).

2.3 Determination of intensive aerosol and BC properties

The spectral dependence of σ_abs was characterized by the absorption Ångström exponent (AAE_{aer,450–635}) calculated between 450 and 635 nm as follows:

$$\begin{array}{}\text{(1)}& {\text{AAE}}_{\text{aer,450–635}}={\displaystyle \frac{-\log \left(\frac{{\mathit{\sigma }}_{\text{abs,450}}}{{\mathit{\sigma }}_{\text{abs,635}}}\right)}{\left(\log \left(\frac{\mathrm{450}}{\mathrm{635}}\right)\right)}}.\end{array}$$

For this calculation, σ_abs,470 and σ_abs,660 from the Aethalometer were adjusted at the wavelengths of 450 and 635 nm measured by the nephelometer using the AAE calculated from the Aethalometer between 370–470 nm and 590–660 nm. AAE_{aer,450–635} provides information about the chemical composition of atmospheric aerosols. Pure BC absorbs radiation across the whole solar spectrum with the same efficiency; thus, it is characterized by an AAE_{aer,450–635} of around 1 (Bond et al., 2013). Conversely light-absorbing organic particles known as brown carbon (BrC), as well as dust particles, generally have an AAE_{aer,450–635} greater than 2 (Sun et al., 2007; Bergstrom et al., 2007; Schuster et al., 2016).

The wavelength dependence of σ_sca can be characterized by the scattering Ångström exponent (SAE_{aer,450–635}) calculated between 450 and 635 nm as

$$\begin{array}{}\text{(2)}& {\text{SAE}}_{\text{aer,450–635}}={\displaystyle \frac{-\log \left(\frac{{\mathit{\sigma }}_{\text{sca,450}}}{{\mathit{\sigma }}_{\text{sca,635}}}\right)}{\left(\log \left(\frac{\mathrm{450}}{\mathrm{635}}\right)\right)}}.\end{array}$$

SAE_{aer,450–635} describes the relative contribution of fine- and coarse-mode particles (Clarke and Kapustin, 2010). Small values of SAE_{aer,450–635} indicate a higher contribution of large aerosol particles (e.g., dust and sea salt), while large values of SAE_{aer,450–635} indicate relatively small aerosol particles (Cappa et al., 2016).

The aerosol single-scattering albedo (SSA_aer,λ) was calculated at the wavelengths of λ=450, 525 and 635 nm using the following equation:

$$\begin{array}{}\text{(3)}& {\text{SSA}}_{\text{aer},\mathit{\lambda }}={\displaystyle \frac{{\mathit{\sigma }}_{\text{sca},\mathit{\lambda }}}{{\mathit{\sigma }}_{\text{sca},\mathit{\lambda }}+{\mathit{\sigma }}_{\text{abs},\mathit{\lambda }}}}.\end{array}$$

SSA_aer,λ describes the relative importance of scattering and absorption to the total light extinction. Thus, it indicates the potential of aerosols to cool or warm the atmosphere. To calculate SSA_aer,λ, σ_abs was first calculated at the proper wavelength λ using the AAE calculated at the closest wavelengths (AAE_{aer,370–470} to retrieve σ_abs,450, AAE_{aer,520–590} for the σ_abs,525 and AAE_{aer,590–660} for the σ_abs,635).

The $\mathrm{\Delta }{M}_{\text{rBC}}/\mathrm{\Delta }\mathrm{CO}$ emission ratio was calculated to provide information on the combustion sources, as well as on BC wet deposition (Baumgardner et al., 2002). First, the background CO concentrations were estimated by taking the rolling 5th percentile of the values in a 14 d time window and then calculating a monthly mean (see Fig. S3 in the Supplement) based on the method of Kanaya et al. (2016). ΔCO was then calculated by subtracting the monthly background CO concentration for any measured hourly CO value. ΔM_rBC was considered to be equal to M_rBC because we assume that the background BC is zero since the atmospheric lifetime of BC is known to be of a few days (Park et al., 2005). By contrast the CO lifetime is estimated to be several days (Bey et al., 2001).

The mass absorption cross-section of rBC (MAC_rBC) was determined as

$$\begin{array}{}\text{(4)}& {\text{MAC}}_{\text{rBC}}={\displaystyle \frac{{\mathit{\sigma }}_{\text{abs,880}}}{M_{\text{rBC}}}}.\end{array}$$

M_rBC values below the 5th and above the 95th percentile were filtered before MAC_rBC calculations to reduce the influence of outliers in statistical analyses. In addition, we filtered out periods when dust was sampled at PDM for the calculation of MAC_rBC since Yus-Díez et al. (2021) observed significant biases in the multiple-scattering correction of the Aethalometer AE33 during such events.

The light-absorption enhancement factor E_abs can be determined as the MAC_rBC value normalized by a reference value for pure, uncoated (bare) rBC:

$$\begin{array}{}\text{(5)}& E_{\text{abs}}={\displaystyle \frac{{\text{MAC}}_{\text{rBC}}}{{\text{MAC}}_{\text{bare,rBC}}}}.\end{array}$$

Three different methods are generally used to estimate MAC_bare,rBC: the first one is to remove the coating of BC with a thermodenuder and measure the corresponding absorption (Cappa et al., 2012; Healy et al., 2015; Yuan et al., 2021); the second one is to extrapolate measurements of MAC_rBC as a function of the measured rBC mixing ratio (Cappa et al., 2019; Yuan et al., 2021); the third one consists in calculating MAC_bare,rBC from the measured rBC size distribution using Mie's theory and the mean geometric rBC diameter (Zanatta et al., 2018; Liu et al., 2017; see Fig. S4 in the Supplement). Here we used the latter method by assuming a rBC refractive index of 1.95–0.79i at λ=880 nm (Bond and Bergstrom, 2006). The calculation of MAC_bare,rBC using Mie's theory assumes a simplified spherical assumption of rBC morphology. However rBC may exhibit complex morphologies whose optical behavior is imperfectly predicted by Mie's theory, introducing a bias into the retrieved MAC_bare,rBC (Saleh et al., 2016). It might be considered that Mie's theory is suitable for estimating the absorption of highly aged rBC, which exhibits an internally mixed core–shell structure. China et al. (2015) used this method to calculate the E_abs of rBC in a high-altitude site of the Azores because the large majority (70 %) of these long-range-transported particles were found to be highly compacted. Several studies found that Mie's scattering model captures basic optical properties of BC in biomass burning plumes (Liu et al., 2017; Denjean et al., 2020). Zanatta et al. (2018) calculated MAC_rBC values of heavily coated rBC particles from the Arctic region using Mie's theory and found results consistent with direct measurements. In addition to the morphology, the MAC_bare,rBC calculation is also very sensitive to the refractive index of rBC core (Sorensen et al., 2018). F. Liu et al. (2020) summarized the changes in MAC values induced by the use of different refractive indexes. They reported deviations of −7 % to −35 % from the MAC_BC value of 7.5m² g⁻¹ recommended by Bond and Bergstrom (2006).

Time periods with high humidity (95 %) or precipitation were filtered before analysis to avoid artifacts in the sampling inlet. Under precipitation some water droplets may indeed enter in the aerosol inlet and change both the inlet cutoff diameter and the measured aerosol size distribution. This would bias all the measured aerosol properties. We also filtered periods where hourly CO concentrations exceeded 200 ppb in order to exclude local pollution events, e.g., due to snow removal of the touristic platform.

All aerosol and gas measurements were converted to standard temperature and pressure (273.15 K and 1013.25 hPa).

2.4 Identification of air mass origins

HYSPLIT (Stein et al., 2015) was used to calculate air masses back trajectories. This model uses 3-hourly atmospheric data from the Global Data Assimilation System (GDAS) of the National Centers for Environmental Prediction (NCEP) at a $\mathrm{1}{}^{\circ }\times \mathrm{1}{}^{\circ }$ spatial resolution. More information can be found at https://www.ready.noaa.gov/index.php (last access: 23 January 2024). A back trajectory was run at 12:00 UTC for each day, going back 72 h, for the 2 years of the campaign. Precipitation rates along the back trajectories were also computed from the HYSPLIT calculations in order to classify days when the air masses arriving at PDM encountered precipitation or not in the past 72 h.

To discriminate between FT- and PBL-influenced air masses (hereafter referred as PBL/FT conditions), we followed a methodology proposed initially by Griffiths et al. (2014), assuming that the diurnal radon increase, which is typically observed at mountain sites during the daytime, is the result of transport of PBL air by thermal anabatic winds up to the summits. The method first consists in ranking the days of the sampling period by decreasing anabatic influence (details in the Supplement and in Griffiths et al., 2014). A threshold rank (here 282; see Fig. S5 and the associated Text S3 in the Supplement) can then be determined to separate days with or without anabatic influence in the daytime.

In our study, it was necessary to select the observation hours strongly influenced by the boundary layer. To do this, we selected the first 200 anabatic days in the ranking and, from these days, all the hours when radon activity was greater than the median value for the current day.

We also needed an ensemble of observation hours with minimum influence of the PBL. In the latter case, we selected hours during the non-anabatic days (i.e., ranked after 282) with radon activity below the median value for the current day.

3 Results

3.1 Meteorology and air mass classification

In the following, seasons are defined as follows: winter (December, January, February), spring (March, April, May), summer (June, July, August) and autumn (September, October, November). The meteorological conditions at PDM during the campaign are described in Fig. S6 and the associated Text S4 in the Supplement.

The backward trajectories performed with the HYSPLIT model for 72 h time periods were assigned to six geographical zones, according to the position of their starting point (shown in Fig. 2): Northwest Europe, Continental Europe, Mediterranean Africa, Atlantic Spain, North Atlantic and Local (within a circular zone of 100 km radius around PDM). The transport to PDM was generally from the west or south, i.e., from the North Atlantic and Atlantic Spain zones. It should also be noted that 99 % of all atmospheric backward trajectories modeled to PDM reveal long-range transport (>100 km).

Figure 2 The 72 h back trajectories of air masses measured at PDM over the measurement period 2019–2020. Geographical boundaries of the sectors used to classify the air mass back trajectories are overlaid.

The analyses of the diurnal cycle of radon concentrations allowed us to determine the FT and PBL conditions prevailing at the site following the methodology presented in Sect. 2.4. Details on the statistical results can be found in Table S1 in the Supplement. Over the campaign, 1149 h was clearly identified as FT-influenced conditions, which represents 56 % of the total classified hours. In winter, FT and PBL conditions occurred roughly 74 % and 26 % of the analyzed time, respectively, against 48 % and 52 % for summer, respectively. These results are broadly in agreement with the previous study by Hulin et al. (2019) at PDM, which quantified around 47 % of the days as PBL-influenced over a 10-year period. The PBL conditions occurred mostly around 15:00 UTC (see Fig. S5), consistent with the dynamics at mountain sites where plain-to-mountain winds and along-valley winds become the strongest in the afternoon (Whiteman, 2000).

3.2 Aerosol optical properties and classification of aerosol types

Figure 3 presents daily time series and statistics of aerosol optical properties over the 2-year measurement period. The average SSA ± GSD (geometric standard deviation) were 0.94 ± 0.06, 0.94 ± 0.07 and 0.95 ± 0.08 at 450, 525 and 635 nm, respectively (Fig. 3a). These values are in the range of those observed at mountain sites in southern Europe (Bukowiecki et al., 2016; Laj et al., 2020; Pandolfi et al., 2014). The mean value ± GSD of σ_abs,880 was 0.27 ± 0.25 Mm⁻¹ (Fig. 3b), which falls in the range of 0.14 to 1.23 Mm⁻¹ obtained at Jungfraujoch and Montsec (Bukowiecki et al., 2016; Pandolfi et al., 2014). The average σ_sca ± GSD values were 15.5 ± 16.1 Mm⁻¹, 13.4 ± 13.9 Mm⁻¹ and 12.2 ± 12.9 Mm⁻¹ at 450, 525 and 635 nm, respectively (Fig. 3c). These weak values of σ_abs and σ_sca can be explained by the remote mountain site type, where almost no aerosols are locally emitted. There was a clear seasonality of aerosol optical properties. SSA at the three wavelengths exhibited the lowest monthly mean values in spring–summer (0.94 ± 0.02 at λ=525 nm) and the highest in autumn–winter (0.99 ± 0.01 at λ=525 nm, as shown in Fig. 3a). Simultaneously, the highest monthly mean SAE values were observed in spring–summer (1.23 ± 0.70) and reached a minimum in the winter (−0.25 ± 0.16) (Fig. 3d). This anticorrelation suggests a higher fraction of absorbing and fine particles relative to purely scattering and coarse particles at PDM during the spring–summer. Interestingly different trends can be observed between the summer and spring seasons. During spring 2019 the decrease in SSA is correlated with a slight enhancement of σ_abs,880 (Fig. 3b) and a decrease in σ_sca at all wavelengths. In summer the increase in σ_abs,880 leads to values multiplied by a factor of 4, while both SSA and SAE remain rather constant. All these parameters combined indicate a similar dominant aerosol type reaching PDM but with a stronger contribution in summer. This noteworthy seasonality of aerosol optical properties has previously been observed at other high mountain sites in Europe (Andrews et al., 2011; Collaud Coen et al., 2011; Laj et al., 2020; Pandolfi et al., 2018). The higher concentration of small and absorbing particles in summer at PDM could be attributed to a higher anthropogenic BC influence favored by strong vertical mixing and a higher PBL height, a higher occurrence of wildfires emitting large amounts of BC and brown carbon (BrC), or a lower precipitation rate.

Figure 3 Time series (left) and statistical distributions (median and 25th and 75th percentiles, right) of aerosol optical properties measured at PDM in 2019–2020 with (a) single-scattering albedo at 450, 525 and 635 nm; (b) absorption coefficient at 880 nm; (c) scattering coefficients at 450, 525 and 635 nm; and (d) scattering Ångström exponent at 450–635 nm. The dots and bars on the time series represent the median and the 25th and 75th percentiles, respectively, with a monthly frequency. Histograms was computed using a 1 d time frequency. Vertical dashed lines represent the seasons' boundaries. The date format is year-month.

In order to investigate these different hypotheses, a classification of the dominant aerosol type sampled at PDM was performed using the spectral dependency of aerosol optical properties. Figure 4 shows AAE as a function of SAE, overlaid with the aerosol classification matrix from Cappa et al. (2016). Aerosols with the highest SSA values (violet points) tend to fall on the left-hand side of the plot with SAE values below 0, indicative of large particles such as marine sea salt, continental dust or highly processed/coated particles. The presence of large marine and dust aerosols is in line with backward trajectories showing a dominant origin of air masses coming from the Atlantic Ocean and Iberian Peninsula as far as north Africa (e.g., Fig. 2). Dust being a strong light absorber, it is expected to lower the mean aerosol SSA. However Fig. 4 shows that SSA values for dust-dominated aerosol (classified as having AAE values above 2) are quite similar to those for remote marine aerosol (classified as having AAE values below 1). Although Europe frequently experiences African dust events (Denjean et al., 2016; Dumont et al., 2023), our results indicate that these dust events were not absorbing enough to substantially lower the aerosol SSA at PDM. This is supported by previous estimates of SSA ranging between 0.90 and 1.00 for dust particles transported in the Mediterranean region (Mallet et al., 2013; Denjean et al., 2016).

Figure 4 Hourly average aerosol absorption Ångström exponent vs. scattering Ångström exponent calculated at 450–635 nm and colored as a function of the single-scattering albedo at 525 nm. The classification of aerosol type by Cappa et al. (2016) is also shown. The points in the dashed zone, representing dust events, were filtered before analyses of BC properties to avoid artifacts in the calculation of MAC_rBC.

There a natural clustering of the most light-absorbing aerosols with SSA<0.9 (pink to yellow points) in the middle of the plot, with sections on the lower side with AAE between 0.5 and 1.5, which Cappa et al. (2016) defined as the sections dominated by BC or a mix of BC and large particles. The success of aerosol classification schemes is largely dependent on uncertainties in AAE attribution for each aerosol species. Although AAE=1 is often considered to be BC such as that in the classification by Cappa et al. (2016), observational and numerical estimates show a wide range of BC AAE from 0.6 to 1.3 (Kirchstetter et al., 2004; Liu et al., 2018) due to the variation in BC core size, coating thickness, composition and morphology (Liu et al., 2018; Zhang et al., 2018). Therefore it is possible that the large range of AAE values observed for the most light-absorbing aerosols was due to different microphysical and chemical properties of the BC sampled at PDM.

Interestingly almost none of the aerosols were classified as strong BrC and BC$/$BrC mixture (AAE and SAE values above 1.5), revealing a very low contribution of BrC to the aerosol absorption at PDM. An explanation for this could be the rapid BrC depletion within the first day of emission, by photobleaching or volatilization that has been observed in several studies (Forrister et al., 2015; Wong et al., 2019; Zeng et al., 2020). Altogether these results suggest that BC was the predominant absorption component of aerosols at PDM and controlled the variation in SSA throughout the 2 observation years.

3.3 rBC sources and properties

3.3.1 rBC mass concentration

Figure 5 shows the time series of the microphysical and optical properties of rBC. The mean M_rBC, shown in Fig. 5a, was 34.8 ± 35.7 (mean ± SD) ng m⁻³, which is a typical level for remote mountain sites. For instance, Sun et al. (2021) observed M_eBC around 20 ng m⁻³ from 9 years of measurements with a multi-angle absorption photometer (MAAP; model 5012, Thermo Scientific) at the Schneefernerhaus station, Zugspitze, Germany (2671 m a.s.l.). Motos et al. (2020) measured M_rBC around 9 ng m⁻³ in summer at Jungfraujoch, Switzerland (3580 m a.s.l.). rBC represented around 7 % ± 5 % of the total aerosol number concentration measured by the SMPS over the campaign. An increase in the rBC number fraction by a factor of 2.5 was found in summer (9 % ± 5 %) compared to winter (4 % ± 3 %). Simultaneously, σ_abs,880 increased by a factor of 4 between winter and summer. Thus, this confirms that rBC contributed to a significant part of the aerosol absorption at PDM.

Figure 5 Time series (left) and statistical distributions (median and 25th and 75th percentiles, right) of rBC properties measured at PDM in 2019–2020: (a) rBC mass concentration, (b) $\mathrm{\Delta }{M}_{\text{rBC}}/\mathrm{\Delta }\mathrm{CO}$ emission ratio, (c) rBC core mass size distribution with geometric diameter in solid green line, and (d) rBC mass absorption cross-section and absorption enhancement at 880 nm. The dots and bars on the time series represent the median and the 25th and 75th percentiles, respectively, with a monthly frequency. Histograms was computed using a 1 d time frequency, as well as the colored background of the rBC core size distribution. Vertical dashed lines represent the seasons' boundaries. The date format is year-month.

3.3.2 rBC emission sources

Figure 6 shows bivariate polar plots obtained by combining wind analysis and M_rBC with a 1 h time resolution in winter and summer. The densities of M_rBC data weighted by M_rBC values and normalized by the maximum density of M_rBC were plotted as a function of the wind direction and speed. The darkest areas of the wind pattern are those where the highest M_rBC was measured with a high occurrence, whereas the lightest zones exhibit the lowest measured M_rBC and/or a low occurrence of measurements. Note that locally emitted pollution at the measurement station was filtered before the analysis, limiting local M_rBC contributions emitted from the PDM station (e.g., Sect. 2.3).

Figure 6 Bivariate polar plots of the density of M_rBC, weighted by M_rBC values as a function of the wind direction and speed in (a) summer and (b) winter. The color scale shows the M_rBC density data weighted by M_rBC. The radial scale shows the wind speed, which increases from the center of the plot radially outwards. Both plots use hourly data. The weighted densities were normalized by their maxima.

In summer, the highest M_rBC values were mainly associated with moderate wind speeds (above 5 m s⁻¹) and from the west and southwest, suggesting a dominant regional transport (Fig. 6a). By contrast in winter (Fig. 6b), the highest M_rBC mainly occurred under more static atmospheric conditions (i.e., for wind speeds below 5 m s⁻¹) and with no evident wind direction dependency. These results suggest that local-scale emissions could be a major contributor to M_rBC in winter, unlike in summer. Further discussion on the role of PBL influence on M_rBC is given in Sect. 3.4.2.

The $\mathrm{\Delta }{M}_{\text{rBC}}/\mathrm{\Delta }\mathrm{CO}$ emission ratio, presented in Fig. 5b, shows a wide range of values from 0 to 10 $\mathrm{ng}{\mathrm{m}}^{-\mathrm{3}}{\mathrm{ppbv}}^{-\mathrm{1}}$, with a mean value of 1.93 ± 2.12 $\mathrm{ng}{\mathrm{m}}^{-\mathrm{3}}{\mathrm{ppbv}}^{-\mathrm{1}}$. Summer ratios were generally higher than winter emission ratios, which could reflect either lower rBC scavenging during transport or different emission sources of rBC between seasons. Indeed, the $\mathrm{\Delta }{M}_{\text{rBC}}/\mathrm{\Delta }\mathrm{CO}$ emission ratio varies as a function of fuel types, combustion efficiencies and wet deposition by precipitation (Baumgardner et al., 2002; Taylor et al., 2014). This explains the high diversity of $\mathrm{\Delta }{M}_{\text{rBC}}/\mathrm{\Delta }\mathrm{CO}$ emission ratios obtained in the literature worldwide, going from 0.5 $\mathrm{ng}{\mathrm{m}}^{-\mathrm{3}}{\mathrm{ppbv}}^{-\mathrm{1}}$ at Jungfraujoch, Switzerland (Liu et al., 2010), to 9 $\mathrm{ng}{\mathrm{m}}^{-\mathrm{3}}{\mathrm{ppbv}}^{-\mathrm{1}}$ in a biomass burning plume above the Texas region during the TexAQS 2006 campaign, USA (Spackman et al., 2010), if only studies using SP2 measurements are considered. Overall $\mathrm{\Delta }{M}_{\text{rBC}}/\mathrm{\Delta }\mathrm{CO}$ from fossil fuel tends to exhibit lower values than that from biomass combustion (Guo et al., 2017; Pan et al., 2011; Zhu et al., 2019). To our knowledge the only available $\mathrm{\Delta }{M}_{\text{rBC}}/\mathrm{\Delta }\mathrm{CO}$ measurements in Europe were performed during airborne measurements in the Cabauw industrial region, the Netherlands, by McMeeking et al. (2010), who found very low values of around 0.8 $\mathrm{ng}{\mathrm{m}}^{-\mathrm{3}}{\mathrm{ppbv}}^{-\mathrm{1}}$. The high time variability in $\mathrm{\Delta }{M}_{\text{rBC}}/\mathrm{\Delta }\mathrm{CO}$ reflects important differences concerning the scavenging processes impacting BC and/or the relative contribution of biomass burning and fossil fuel emissions in the production of BC measured at PDM. These two different factors will be addressed in Sect. 3.4.

3.3.3 rBC mass size distribution

The rBC mass median core size diameter (D_rBC,core) was quite constant during the campaign with a mean geometric diameter of 179 ± 28 nm (Fig. 5c). An exception occurred in early December 2019, when we detected the presence of large rBC particles with D_rBC,core around 400 nm. However observations during this period may be the results of measurement uncertainties due to too low an M_rBC (less than 10 ng m⁻³). The D_rBC,core values obtained at PDM are generally comparable to D_rBC,core, which has been reported ranging from 180 to 225 nm for well-aged background rBC (Liu et al., 2010; McMeeking et al., 2010; Schwarz et al., 2010; Shiraiwa et al., 2008). However, our values are slightly higher than previous observations at Jungfraujoch by Motos et al. (2020), who reported D_rBC,core ranging from 130 to 150 nm in summer and winter. Instead of fitting the SP2 observations with multimodal individual lognormal modes (e.g., Sect. 2.2.2), Motos et al. (2020) only took into account rBC particles with D_rBC,core>50 nm (the lower detection limit of their SP2 version), which may bias the estimated SP2 mode of D_rBC,core (Tinorua et al., 2024).

3.3.4 rBC absorption

The ambient MAC_rBC was around 9.2 ± 3.7 m² g⁻¹ at λ=880 nm (Fig. 5d). Several studies have previously reported MAC_BC values between 8.9 and 13.1 m² g⁻¹ for measurements at λ=637 nm in European mountain stations (Pandolfi et al., 2014; Yus-Díez et al., 2022; Zanatta et al., 2016). By using an AAE of unity, these values can be converted to MAC_BC between 6.4 and 9.5 m² g⁻¹ at λ=880 nm. These studies used different measurement techniques, analysis methods and correction factors from ours for estimating MAC_BC, which makes difficult the comparison of MAC_rBC derived from different instruments. Pandolfi et al. (2014) performed a linear regression between σ_abs,637 measured by a MAAP (multi-angle absorption photometer) and daily M_EC values from off-line filter-based measurements by a Sunset OCEC analyzer. Yus-Díez et al. (2022) and Zanatta et al. (2016) retrieved MAC_BC with these instruments by calculating the ratio between the two parameters instead of a linear regression. Because of the absence of a standard method for quantifying M_BC, the absolute uncertainties in the MAC_BC obtained in the literature are very high, ranging from ±29 % to ±63 % (Zanatta et al., 2016).

In terms of seasonality we found systematically higher values of MAC_rBC in summer (monthly mean ± SD of 10.3 ± 3.3 m² g⁻¹) compared to winter (8.3 ± 3.8 m² g⁻¹). A similar seasonal pattern was observed in Europe at Puy de Dôme (Zanatta et al., 2016, central France) and at Jungfraujoch (Swiss Alps) mountain sites (Motos et al., 2020). An opposite trend was observed at mountain sites affected by strong precipitation during monsoon such as the Tibetan Plateau and Himalayan regions, where both MAC_BC and M_BC exhibit maximum values in winter or autumn (Zhao et al., 2017; Srivastava et al., 2022). The same seasonal pattern with elevated values in winter/autumn compared to summer/spring was observed at several rural and urban sites in the PBL, which was attributed to greater emissions from residential heating combined with a lower PBL height (Zanatta et al., 2016; Kanaya et al., 2016; Yttri et al., 2007). However maximum MAC_BC and M_BC in the PBL during cold periods are not a recurring observation even for the same measurement site. For instance Sun et al. (2022) showed that in Beijing (China), due to the reduction in some predominant BC sources in winter following environmental policies, the annual cycle of M_BC changed over the years between 2012 and 2020.

Variations in MAC_rBC may exist for different reasons. We first address the question of whether the MAC_rBC depends on D_rBC,core in Fig. S7 in the Supplement. There was no clear correlation between MAC_rBC and D_rBC,core, which indicates that the variation in BC size was not the cause of the MAC_rBC variability. This is because D_rBC,core only varied within a relatively narrow range (the 25th and 75th percentiles around 164 and 195 nm) during the campaign. The observed MAC_rBC values were converted to equivalent E_abs by dividing them by a reference MAC for pure, uncoated BC (MAC_bare,BC). While values of MAC_bare,BC are reported in the literature, estimation of campaign-specific MAC_bare,BC allows for more robust determination of E_abs than using values from the literature since MAC_bare,BC is dependent on the size of uncoated BC (Bond and Bergstrom, 2006; Adachi et al., 2007, 2010; Adachi and Buseck, 2013; Cappa et al., 2012). Here the mean MAC_bare,BC was 4.15 m² g⁻¹, with values ranging from 3.90 to 4.37 m² g⁻¹ considering the standard deviation of the mean D_rBC,core, which is in reasonable agreement with literature assessments (F. Liu et al., 2020).

The E_abs values relative to MAC_bare,BC=4.15 m² g⁻¹ at λ=880 nm were significantly greater than unity, with a mean value of 2.2 ± 0.9 (Fig. 5d). Given the remote mountain location and presumable distance from fresh rBC sources, rBC particles reaching PDM may have undergone aging and have gained a consistent coating. Previous studies found an absorption enhancement of BC due to its coating with the aging time (Yus-Díez et al., 2022; Sedlacek et al., 2022; Peng et al., 2016). The most likely cause of the strong E_abs at PDM is a lensing effect due to the internal mixing of rBC with other particles that drives MAC_rBC variability, though we cannot eliminate changes in rBC morphology that can result from coating onto rBC. There was a significant seasonal trend in E_abs with higher values observed in summer, indicating that rBC reaching the PDM station has undergone longer aging processes during this season. These results are consistent with the measurements of Motos et al. (2020) at Jungfraujoch, which also indicated a strong seasonality in the rBC mixing state with larger coating in summer.

Figure 7 further shows the diurnal variation in E_abs for every season. There was a notable opposite diurnal profile between seasons for E_abs, with midday showing a minimum of around 1.7 in winter and a maximum of around 2.9 in summer. Spring and autumn showed intermediate patterns with less regular E_abs throughout the day. These observations suggest that different sources and/or processes drove the seasonal contrast in rBC properties. Nonetheless, it cannot be ruled out that some of the variability in MAC_rBC (and E_abs) can be explained by the use of a fixed multiple-scattering correction applied to calculate σ_abs, since this parameter has been shown to vary as a function of time and atmospheric conditions (Yus-Díez et al., 2021). The following section aims to investigate potential drivers of E_abs variations, including rBC wet scavenging, dominant rBC sources and transport pathways. Particular attention will be paid to winter and summer because these seasons differ greatly, whereas spring and autumn behaviors appear intermediate.

Figure 7 Diurnal cycles of E_abs for each season during the 2019–2020 period. Seasons are defined as follows: winter (December, January, February), spring (March, April, May), summer (June, July, August) and autumn (September, October, November). Boxes, lines, black dots and whiskers indicate the 25th percentile and 75th percentile, median, mean, and 10th percentile and 90th percentile, respectively.

3.4 Investigation of factors influencing rBC properties

3.4.1 The impact of wet scavenging on rBC properties

We first investigated whether E_abs was modulated by a size-dependent rBC wet scavenging process during precipitation along the transport pathway of rBC-containing particles. This hypothesis is based on the fact that the removal of particles is favored for the largest and thickly coated rBC because the activation of aerosols into cloud droplets is predominantly controlled by the particle size (Moteki et al., 2012; Mori et al., 2021; Ohata et al., 2016; Zhang et al., 2021). The wet removal of rBC was investigated by performing a cluster analysis using $\mathrm{\Delta }{M}_{\text{rBC}}/\mathrm{\Delta }\mathrm{CO}$ data for which precipitation occurred or did not along 72 h back trajectories computed by the HYSPLIT model. In order to decrease the influence of the difference sources on $\mathrm{\Delta }{M}_{\text{rBC}}/\mathrm{\Delta }\mathrm{CO}$ compared to the effect of wet scavenging, periods for which the site was under PBL influence were filtered.

Figure 8a shows a median $\mathrm{\Delta }{M}_{\text{rBC}}/\mathrm{\Delta }\mathrm{CO}$ of 0.7 $\mathrm{ng}{\mathrm{m}}^{-\mathrm{3}}{\mathrm{ppbv}}^{-\mathrm{1}}$ for air masses affected by precipitation, against 2.1 $\mathrm{ng}{\mathrm{m}}^{-\mathrm{3}}{\mathrm{ppbv}}^{-\mathrm{1}}$ without precipitation during the transport of the air masses. The reduction in $\mathrm{\Delta }{M}_{\text{rBC}}/\mathrm{\Delta }\mathrm{CO}$ by a factor of 3 suggests that a significant removal process of rBC from the precipitation occurred along the transport pathway, apart from vertical transport from the PBL. As a reminder, time periods when PDM was under precipitation or humidity >95 % have been filtered before the analysis. This result is confirmed by the dependence of $\mathrm{\Delta }{M}_{\text{rBC}}/\mathrm{\Delta }\mathrm{CO}$ on RH in Fig. 8b, where a sudden decline in $\mathrm{\Delta }{M}_{\text{rBC}}/\mathrm{\Delta }\mathrm{CO}$ appeared for the highest RH>80 %, going from a median $\mathrm{\Delta }{M}_{\text{rBC}}/\mathrm{\Delta }\mathrm{CO}$ of between 2.0 and 2.4 $\mathrm{ng}{\mathrm{m}}^{-\mathrm{3}}{\mathrm{ppbv}}^{-\mathrm{1}}$ for RH<80 % to a median $\mathrm{\Delta }{M}_{\text{rBC}}/\mathrm{\Delta }\mathrm{CO}$ of ∼0.4 $\mathrm{ng}{\mathrm{m}}^{-\mathrm{3}}{\mathrm{ppbv}}^{-\mathrm{1}}$ above 80 % RH. This result of high rBC removal by wet deposition is in line with measurements performed in regions at similar altitudes, such as Puy de Dôme, Mt. Tianjing and Mt. Sonnblick, where wet deposition represents 30 % to 70 % of the BC removing processes in the troposphere (Yang et al., 2019). Figure 8c and d show in contrast little influence of precipitation and RH on the rBC absorption enhancement, with a constant median E_abs value of around ∼2.1.

Figure 8 $\mathrm{\Delta }{M}_{\text{rBC}}/\mathrm{\Delta }\mathrm{CO}$ emission ratio and E_abs vs. (a, c) precipitation along the air mass back trajectory calculated with the HYSPLIT model and (b, d) relative humidity measured at PDM. Violin plots represent the probability density function of each parameter. Features of the boxplots are defined in the same way as in Fig. 7. PBL conditions were filtered before analysis.

To better understand the negligible impact of rBC wet scavenging on MAC_rBC, we compared the measured rBC core size distribution of air masses affected or not affected by precipitation during their transport and under high-RH conditions or not (Fig. 9). A 2-fold-lower M_rBC in precipitation conditions compared to that without precipitation provides additional evidence for the dominant role of wet scavenging for rBC. The same result appeared when comparing rBC core size distribution under wet or dry conditions. However no significant change in the mean rBC core diameter was noticed between wet and dry conditions (mean D_rBC,core of 177 and 182 nm, respectively), as well as in the presence of precipitation during the transport of rBC or not (mean D_rBC,core of 178 and 185 nm, respectively).

Figure 9 Mass size distributions of the rBC core measured by the SP2 as a function of the presence or absence of precipitation along the path of air masses arriving at PDM and whether the relative humidity was over 85 % or not. Vertical lines show the geometric mean rBC core diameter colored by the criteria described in the legend.

This result contrasts with previous studies conducted in the FT showing a decrease in rBC size due to wet scavenging (Kondo et al., 2016; Moteki et al., 2012; Taylor et al., 2014; D. Liu et al., 2020). For example, Kondo et al. (2016) found a change in D_rBC,core between 13 and 20 nm depending on the season, while D. Liu et al. (2020) and Moteki et al. (2012) measured rBC cores ∼32 nm lower in air masses affected by wet removal. The insignificant effect of wet scavenging on the modal diameter of rBC core size distribution could be explained by the size of the rBC core sampled at PDM, which was higher than the one described in these studies. At the Jungfraujoch Hoyle et al. (2016) evidenced a threshold diameter of around 90 nm above which a particle is activated into a droplet upon cloud formation. The majority of rBC sampled at PDM exhibited D_rBC,core values above this critical diameter. In addition droplet activation of an aerosol particle occurs when the supersaturation of the surrounding water vapor exceeds a critical value of supersaturation. Thus, it is not likely for freshly emitted BC particles to act as cloud condensation nuclei due to their hydrophobic nature unless the water vapor supersaturation is higher than 2 % (Wittbom et al., 2014), far beyond the actual supersaturation (0.1 %–0.6 %) in ambient air. Furthermore, the rBC wet removal by impaction is also a size-dependent process which could have been responsible for the removal of small rBC particles (Croft et al., 2010). However interstitial scavenging mostly affects particles smaller than 100 nm (Pierce et al., 2015), which have a very low presence at PDM (rBC mean diameter of 180 nm). As the ambient supersaturation varies depending on the environment, it is difficult to conclude whether the insensitivity of rBC size distribution and E_abs to precipitation occurrence was solely due to the presence of large rBC particles at the sampling site or to a high ambient supersaturation in the precipitating clouds. Further measurements of simultaneous rBC wet removal and effective supersaturation are needed to test all these assumptions.

3.4.2 The contrasting seasonal influence of FT and PBL on rBC properties

Figure 10 shows the rBC properties classified by FT and PBL conditions (methodology in Sect. 2.5) and by season. As explained in Sect. 3.3.2, the $\mathrm{\Delta }{M}_{\text{rBC}}/\mathrm{\Delta }\mathrm{CO}$ ratio depends on the condition of combustion (fuel type, efficiency) and wet deposition by precipitation (Baumgardner et al., 2002; Taylor et al., 2014). We observed in Sect. 3.4.1 a large decrease in $\mathrm{\Delta }{M}_{\text{rBC}}/\mathrm{\Delta }\mathrm{CO}$ when precipitation occurred during the transport of the air masses. In order to investigate the influence of rBC sources on rBC properties, precipitation events (air masses for which precipitation occurred along 72 h back trajectories) were removed in this section.

In winter, we measured higher M_rBC values and variability in PBL conditions (39.5, 30.0 and 105 ng m⁻³ for the median and 25th and 75th percentiles, respectively) compared to in FT conditions (33.5, 10.4 and 45.4 ng m⁻³ for the median and 25th and 75th percentiles, respectively) (Fig. 10a). Furthermore the diurnal cycle of M_rBC in winter PBL conditions showed an enhancement in the daytime (Fig. S8 in the Supplement). This trend is consistent with intrusions of pollutants transported from PBL sources through convective mixing. During the night, pollution from the surface is trapped by the low height of the PBL and cannot reach the PDM. At the same time, the cleaner air transported in the FT may contribute to the dilution of M_rBC at the PDM. The higher $\mathrm{\Delta }{M}_{\text{rBC}}/\mathrm{\Delta }\mathrm{CO}$ in PBL conditions (2.5 ± 2.3 $\mathrm{ng}{\mathrm{m}}^{-\mathrm{3}}{\mathrm{ppbv}}^{-\mathrm{1}}$ for the mean ± SD) compared to in FT conditions (0.6 ± 0.1 $\mathrm{ng}{\mathrm{m}}^{-\mathrm{3}}{\mathrm{ppbv}}^{-\mathrm{1}}$ for the mean ± SD) may indicate additional sources from biomass combustion from the valley (Fig. 10b), which could be attributed to either residential wood heating or stubble burning, which is still a common practice in the Pyrenees (González-Olabarria et al., 2015). Figure 10c shows that PBL conditions were associated with lower E_abs values (1.5 ± 0.3 for the mean ± SD) compared to FT conditions (1.9 ± 0.4 for the mean ± SD). Therefore, E_abs was strongly modulated by atmospheric dynamics in winter.

Figure 10 (a) rBC mass concentrations, (b) $\mathrm{\Delta }{M}_{\text{rBC}}/\mathrm{\Delta }\mathrm{CO}$ emission ratio and (c) E_abs as a function of the predominant influence at PDM in winter. The same for summer is given in (d), (e) and (f). Red boxplots represent PBL conditions, and green boxplots represent FT conditions. Precipitation events were filtered before analyses.

During summer, vertical transport from the PBL occurred on about half of the days analyzed in this study. Surprisingly, the M_rBC did not vary between PBL and FT influence, with values of 75.4 ± 33.2 ng m⁻³ (mean ± SD) and 80.2 ± 46.6 ng m⁻³, respectively, meaning that the thermally driven PBL injection did not significantly impact M_rBC measured at PDM (Fig. 10d). This contrasts with our winter observations and most previous surface measurements at mountain sites, where the daytime PBL development has been shown to enhance aerosol mass concentration (Herrmann et al., 2015; Motos et al., 2020). The summer M_rBC values at PDM are twice as high as those observed in winter, which indicates a massive additional regional transport of rBC in the FT and a lower contribution of rBC from PBL injection. $\mathrm{\Delta }{M}_{\text{rBC}}/\mathrm{\Delta }\mathrm{CO}$ values in PBL and FT conditions were close to each other, with values of around 2.8 ± 1.6 $\mathrm{ng}{\mathrm{m}}^{-\mathrm{3}}{\mathrm{ppbv}}^{-\mathrm{1}}$ (mean ± SD) and 3.3 ± 1.7 $\mathrm{ng}{\mathrm{m}}^{-\mathrm{3}}{\mathrm{ppbv}}^{-\mathrm{1}}$, respectively (Fig. 10e). This result indicates that the FT exhibited a significant background load of rBC at the continental scale, thus limiting the relative influence of PBL injection on M_rBC during summer. The resulting E_abs was remarkably similar for PBL vs. FT air mass categories (Fig. 10f). The high rBC loading transported in the FT, coupled with the higher $\mathrm{\Delta }{M}_{\text{rBC}}/\mathrm{\Delta }\mathrm{CO}$ observed in the summertime (Fig. 10d and e), could be due to a strong influence of biomass burning emissions on the background FT in Europe. The majority of trajectories reaching PDM in summer have crossed the Iberian Peninsula and, previously, north Africa and North America (Fig. S9 in the Supplement). In these regions large fire events frequently occur, which may explain the high concentrations of strongly absorbing rBC observed at PDM during summer. This hypothesis is supported by Petetin et al. (2018), who showed that biomass burning aerosol accounts for about 43 %–81 % of the CO concentration in the lower FT in summer using in situ airborne observations of CO from the IAGOS (In-service Aircraft for a Global Observing System) program. The ubiquitous presence of dilute biomass burning in the FT and its significant contribution to aerosol mass loading were also established using airborne measurements of ozone and precursor source tracers from the NASA Atmospheric Tomography Mission (Bourgeois et al., 2021; Schill et al., 2020). Additional measurements of the aerosol chemical composition and in particular of a tracer of biomass burning in the atmosphere such as levoglucosan should be performed at PDM to confirm this.

A question remains about the cause of the diurnal variation in E_abs in summer (Fig. 7c). As shown in Fig. 11, the E_abs increase was not temporally correlated with the wind direction change from west-southwest to south, as evidenced by the 2 h delay between the two events. Furthermore, while the E_abs increase occurs when $\mathrm{\Delta }{M}_{\text{rBC}}/\mathrm{\Delta }\mathrm{CO}$ decreases, the E_abs drop in the afternoon was not accompanied by an increase in $\mathrm{\Delta }{M}_{\text{rBC}}/\mathrm{\Delta }\mathrm{CO}$. Then an increase in E_abs in the morning was most likely due to further aging and the appearance of heavily coated rBC rather than a change in the rBC emission source. Several studies have highlighted the major role of photochemical processes and extensive secondary aerosol generation to promote the light-absorption enhancement of BC (Knox et al., 2009; Krasowsky et al., 2016; H. Liu et al., 2020; Wang et al., 2017; Xu et al., 2018; Yus-Díez et al., 2022). At PDM the enhanced E_abs at noon was accompanied by a shift of the aerosol accumulation mode towards larger sizes, which may be due to the condensation of gaseous species on aerosol particles (Fig. S10 in the Supplement). Simultaneously, a strong elevation of particle number concentration in the diameter range 10–30 nm can be observed, revealing new particle formation most likely produced by photochemical reactions at this time of day. It is thus possible that rBC particles became more coated via condensation of species produced by photochemical reactions at noon. However, it cannot be ruled out that the evolution of aerosol size distribution is a poor indicator of the rBC mixing state.

Figure 11 Hourly variation in (a) E_abs and (b) $\mathrm{\Delta }{M}_{\text{rBC}}/\mathrm{\Delta }\mathrm{CO}$ values in summer. Dots represent mean values, and whiskers are 1 standard deviation. Dots are colored as a function of the wind direction.

4 Summary and implications for climate models

Continuous 2-year measurements of refractory BC (rBC) properties and additional aerosol characteristics have been performed at the high-altitude mountain site Pic du Midi in the French Pyrenees. The classification of the dominant aerosol type using the spectral aerosol optical properties indicates that rBC is the predominant absorption component of aerosols at PDM and controls the variation in SSA throughout the 2 years. The lower SSA in summer (∼0.93) than in winter (∼0.97) is correlated with a higher rBC number fraction, whereas the influence of BrC and dust was found to be negligible.

One key parameter to constrain BC absorption and associated radiative forcing in climate models is the refractive index of BC, and in particular the resulting MAC_BC. It was not clear if BC at high-altitude mountain sites should have a thicker or thinner coating than in urban or plain sites or even should be coated at all. On the one hand, the longer BC lifetime and the low temperature in the free troposphere (FT) favor thicker coating due to enhanced condensation of low-volatility compounds in a colder environment. On the other hand, the low concentrations of particles and gaseous precursors in the FT may limit the coating processes. Our 2-year-long observations show that the overall net effect is a strong absorption enhancement with a mean E_abs value of 2.2 ± 0.9.

The value of 7.5 m² g⁻¹ at λ=550 nm of Bond and Bergstrom (2006) is the most common MAC_BC used in climate models. This recommendation was based on a compilation of experimental results for freshly generated BC at and near sources obtained earlier than the early 2000s. Nevertheless this value is largely under the MAC_rBC found in this study (9.2 m² g⁻¹ at λ=880 nm, which can be converted to 14.7 m² g⁻¹ at λ=550 nm assuming AAE=1). The review by Moteki (2023) has also come to a similar conclusion. The reasons behind this bias should be better understood in the light of observations such as those provided in the present study.

This study has notably shown the high variability in rBC properties measured in a remote site, where rBC has undergone long-range transport and aging. Certain causes of the large variability in MAC_rBC have been eliminated and highlighted:

• Wet deposition is considered to be the main sink of BC, constraining its lifetime and size distribution and thus its atmospheric concentration and optical properties. Our direct $\mathrm{\Delta }{M}_{\text{rBC}}/\mathrm{\Delta }\mathrm{CO}$ measurements show the important role of wet deposition as a sink of rBC with around 67 % removed in the atmosphere by precipitation. However, we found a negligible impact of the rBC wet removal process on both rBC size distribution and E_abs. This result may be due to the combination of large rBC particles reaching PDM (D_rBC,core around 180 nm) and high critical supersaturation in precipitating clouds. The BC wet removal process was found to be one of the most misrepresented processes in the representation of BC in models (Textor et al., 2006; Yu et al., 2019), leading to overestimated BC tropospheric concentrations and lifetimes and, in fine, a higher simulated radiative forcing (Samset et al., 2014; Schwarz et al., 2013). Substantial controversial and ambiguous issues in the wet scavenging processes of BC are apparent in current studies (Yang et al., 2019). Our results suggest that a bulk-wet-deposition parameterization (which does not account for particle-size-dependent scavenging) could realistically represent the actual BC wet scavenging at this site.

• The rBC core was found to have a mean D_rBC,core of 179 ± 28 nm, being reasonably independent of the season and day. There was no clear relationship between MAC_rBC and D_rBC,core, which indicates that the variation in rBC core size was not responsible for the MAC_rBC variability. Similar observations of the rBC core size distribution in the atmosphere provided observational evidence of the stable distribution, with a mode centered at around 200 nm approximately 1 d after emission (Liu et al., 2010; Schwarz et al., 2010; Shiraiwa et al., 2008). This self-similarity could greatly simplify the representation of MAC_BC in model simulations, since a description of the BC mixing state becomes the determinant factor of model performance when estimating BC optical properties and radiative forcing.

• Different timescales of air movements and atmospheric processes affect MAC_rBC throughout the year. MAC_rBC values were found to be higher in summer (geometric mean of 10.3 m² g⁻¹), when the influence of regional-scale motions dominates the rBC load, than in winter (geometric mean of 8.3 m² g⁻¹), when the influence of local-scale motions outweighs the rBC load. There are three possible explanations for this. (i) The plumes traveling in the FT tend to have a longer lifetime, providing sufficient time for rBC aging during transport. In winter this results in a strong diurnal variability in M_rBC (E_abs) with higher (lower) values in the middle of the day linked to the injection of rBC originating from the planetary boundary layer (PBL). However the aging timescale cannot be the only explanatory factor since thermally driven PBL injection did not significantly impact M_rBC and E_abs in summer and higher values have been observed in summer than in winter for similar FT conditions. (ii) The source of rBC emission was different between the winter and summer seasons. Combining $\mathrm{\Delta }{M}_{\text{rBC}}/\mathrm{\Delta }\mathrm{CO}$ with air mass transport analysis, we observed additional sources from biomass burning in summertime, leading to higher M_rBC and E_abs. (iii) Different aging processes occur between seasons, such as photochemical activity, that could explain the observed amplification of light absorption by rBC around noon. However, the latter effect could not be rigorously demonstrated in this study.

The complexity and diversity of BC mixing states in the real atmosphere cannot be represented in climate models, and therefore these models generally use simplified schemes. A fixed e-folding timescale (1–3 d) is commonly used as the turnover time for converting fresh BC particles into aged ones (Myhre et al., 2013). In addition, atmospheric models necessarily approximate the full complexity and diversity of BC composition, which can lead to mismatches with observed E_abs (Fierce et al., 2020). The findings presented here suggest that different dynamic processes governing rBC light absorption occur during the day and night and between summer and winter. A parameterization of BC aging explicitly based on aerosol microphysical processes, in which the conversion rate is considered to vary depending on the environmental conditions (e.g., temperature, photochemical activity) and some key species (e.g., aerosol, coating precursors), may be required to accurately represent the true variability in MAC_BC.