Interactive comment on “ Detection of Saharan dust and biomass burning events using near real-time intensive aerosol optical properties in the northwestern Mediterranean ”

1. This paper by Ealo et al. presents a very interesting idea for the real-time detection of dust and biomass burning events. However, one major concern I see with this technique is the difficulty to differentiate between the dust and biomass burning events, both dust and biomass being strong absorber in UV. This issue might be bigger in summer when the co-occurrence of SDE and Wildfire events may be highly probable. Due to re-circulation, these events may not be differentiated over prolonged time scales. 2. This technique make use of intrinsic properties of the aerosol species like Absorption, Scattering and Single Scattering Albedo Angstrom Exponents. However, these properties are influenced by environmental factors like temperature, RH, aerosol aging time,

Aerosol Mass Spectrometer (ACSM) measurements. A wildfire episode was identified at MSY, showing AAE values up to 2 when daily BB contributions to BC and OM were 73% and 78% respectively.

Introduction
Atmospheric aerosols play an important role in our environment affecting air quality and health (Pope and Dockery, 2006), contributing to the largest uncertainties to the total radiative forcing (IPCC 2007(IPCC , 2013. Aerosol affects climate by 5 perturbation on the Earth's radiative budget, directly through absorption and scattering of solar and terrestrial radiation, and indirectly by acting as cloud condensation nuclei (Twomey et al., 1984;Albrecht, 1989). Most particles scatter the sunlight, causing a net cooling at the top of the atmosphere (TOA), whereas black carbon (BC) absorbs solar radiation in the whole visible spectrum, thus causing a net warming at the TOA (Jacobson, 2001;Ramanathan and Carmichael, 2008;Bond et al., 2013). Absorbing particles can modify the radiation fluxes directly, by absorption of shortwave solar radiation and semi-the high occurrence of Saharan dust events (SDE), especially during the summer period, also contribute strongly to the increment of PM10 levels in the WMB (Rodríguez et al., 2001(Rodríguez et al., , 2015Querol et al., 2009;Pey et al., 2013a). In fact, more than 70 % of the exceedances of the PM10 daily limit value (2008/50/CE European Directive) at most regional background sites of Spain have been attributed to dust outbreaks (Escudero et al., 2007a). Thus, all these processes lead to a radiative forcing in the WMB among the highest in the word (Jacobson, 2001). Nevertheless there is a large uncertainty in the total 15 radiative forcing by atmospheric aerosols in the Mediterranean area (Mallet et al., 2013). The high occurrence and intensity of SDE in the WMB give us the opportunity to look deeply into the characterization of the optical properties of mineral dust when mixed with local aerosols. Despite several studies published on physical and chemical properties of mineral dust in the WMB region (Rodríguez et al., 2001;Escudero et al., 2007b;Querol et al., 2009;Pey et al., 2013a), very few have studied how SDE affect the aerosol intensive optical properties 2014a;Valenzuela et al., 2015) 20 Possibly related to the scarce use of biomass burning for domestic heating in the Mediterranean region compared to Central and Northern Europe, very few studies have been published describing BrC effects on intensive aerosol optical properties in the WMB. However, recent studies have estimated that biomass burning sources in the WMB may contribute more than expected to the measured ambient elemental carbon (EC) and organic carbon (OC) concentrations 2015;Reche et al., 2012;Viana et al., 2013;Pandolfi et al., 2014b). In these studies the biomass 25 burning source was characterized by means of techniques such as positive matrix factorization (PMF) on AMS (Aerosol Mass Spectrometer) or ACSM (Aerosol Chemical Speciation Monitor) data, filter-based analysis of 14C and/or specific chemical tracers such as levoglucosan or K+. Nevertheless, only few studies have used multi-wavelength Aethalometer data (Sandradewi et al., 2008b) in the WMB (Segura et al., 2014).
The main aim of this work is to provide a deep characterization of the intensive optical properties of atmospheric aerosols in 30 the WMB under specific pollution episodes (SDE and BB). Thus, here we evaluate the feasibility of using the intensive aerosol optical properties for the near real-time detection of specific atmospheric events in the WMB. A sensitivity study aimed at calibrating the measured intensive aerosol optical properties is presented and discussed. We show that this calibration is needed to take into account the effects of local pollution on the intensive optical properties during SDE and BB Atmos. Chem. Phys. Discuss., doi:10.5194/acp-2015-902, 2016 Manuscript under review for journal Atmos. Chem. Phys. Published: 18 January 2016 c Author(s) 2016. CC-BY 3.0 License. events. Moreover, we provide the range of variability of the calculated intensive optical properties as a function of the intensity of these events. This information is a valuable input for models studying the radiative effects of atmospheric aerosols in this very peculiar area. With this aim we used high-quality data collected at two stations located in the WMB: Montseny (MSY, regional background station; 720 m a.s.l.) and Montsec (MSA, remote station; 1500 m a.s.l.).  (Ripoll et al., 2014). In situ optical aerosol properties measured at these two sites were performed following the standards required 15 by GAW and ACTRIS networks.

models.
A detailed description of the main meteorological processes affecting the area under study can be found in Pérez et al., (2008);Pey et al., (2010); Pandolfi et al., (2014a); Ripoll et al., (2014). This study is focused in the atmospheric scenarios 25 affecting significantly the concentrations of pollutants in the WMB: North African (NAF), summer regional (REG) and Atlantic advections (AA). SDE, driven by NAF air masses, are more frequent from March to October strongly contributing to increase PM 10 . The summer REG scenarios favour the dispersion of the pollutants around the emission sources and the transport and accumulation of pollutants through the regional recirculation of air masses (Millán et al., 1997). Often REG occur after SDE causing important effects on air quality as shown later. Atlantic advections (AA) affect the WMB 30 throughout the year but mainly in winter. Fresh and clean air masses from the Atlantic clear out the previously accumulated Atmos. Chem. Phys. Discuss., doi:10.5194/acp-2015-902, 2016 Manuscript under review for journal Atmos. Chem. Phys. Published: 18 January 2016 c Author(s) 2016. CC-BY 3.0 License. stagnated air masses, leading to lower pollutant concentrations at regional scale. The seasonal distribution of the main atmospheric episodes throughout the year is very similar at MSY and MSA. However, during colder periods MSA high altitude station is frequently within the free troposphere conditions whereas MSY station is frequently affected by regional/local emission sources being often within the planetary boundary layer PBL (Pandolfi et al., 2014 a, b).
The African dust contribution to PM 10 (%dust) at MSY was calculated by the statistical methodology described in Escudero  et al. (2007b) and Pey et al. (2013). This method is based on the application of 30 days moving 40th percentile to the daily PM 10 data series, after excluding those days impacted by African dust. For those days affected by African dust the percentile value is assumed to be the theoretical background concentration of PM if African dust did not occur. After that, the African dust daily contribution is obtained as the difference between the experimental PM 10 concentration and the calculated 40th percentile value.

Aerosol absorption and Equivalent black carbon (BC) concentration measurements
Aerosol light absorption coefficient (σ ap ) at 637 nm (Müller et al., 2011a) was measured at 1 min resolution with a Multi Angle Absorption Photometer (MAAP, model 5012, Thermo). BC mass concentrations (Petzold et al., 2013) were calculated assuming a constant mass absorption cross section (MAC) of 6.6 m 2 g −1 (Petzold and Schönlinner, 2004). The detection limit 15 of the MAAP instrument is lower than 100 ng m −3 over 2 min integration.
Aerosol light absorption coefficients (σ ap ) at seven different wavelengths (370,470,520,590,660,880 and 950 nm) were obtained every 1 min at both stations by means of Aethalometer instruments (models AE-31 and AE-33). At MSA site the AE-33 (Drinovec et al., 2015) was equipped with a PM 2.5 cut-off inlet until March 2014 and with a PM 10 cut-off inlet afterwards. Absorption measurements at MSY station were carried out with a PM 10 cut-off inlet using an AE-31 20 Aethalometer model from June 2012 to June 2013, then replaced with an AE-33 model. Absorption measurements from the AE-31 were corrected for loading and scattering effects according to Weingartner et al. (2003). The site-specific AE-31 multiple scattering correction factor (C) at MSY was obtained by comparing with measurements from MAAP and it was estimated in around 3.6. Data was normalized to standard conditions (273K, 1013 hPa). Multi-wavelength aerosol absorption measurements used in this work cover a period of 2.

Aerosol scattering measurements
Aerosols light scattering (σ sp ) and hemispheric backscattering (σ bsp ) coefficients were measured at each site every 5 min at three different wavelengths (450, 525 and 635 nm) with a LED-based integrating nephelometer (model Aurora 3000, ECOTECH Pty,Ltd, Knoxfield, Australia). Calibration of the nephelometer was performed three times per year by using CO 2 30 as span gas while zero adjusts were performed once per day by using internally filtered particle free air. A relative humidity (RH) threshold was set following the ACTRIS recommendations (RH<40%). Scattering measurements were corrected for Atmos. Chem. Phys. Discuss., doi:10.5194/acp-2015-902, 2016 Manuscript under review for journal Atmos. Chem. Phys. Published: 18 January 2016 c Author(s) 2016. CC-BY 3.0 License. truncation due to non-ideal detection of scattered radiation following the procedure described in (Müller et al., 2011b).
Multi-wavelength aerosol scattering measurements used in this work cover a period of 5 years at MSY (from January 2010 to December 2014) and 3.5 years at MSA (from July 2011 to December 2014).

PM measurements
Real-time PM concentrations were continuously measured at 30 and 5 min resolution by optical particle counters (OPC) using GRIMM spectrometers (GRIMM 180 at MSY, and GRIMM 1107 and GRIMM 1129 at MSA). Concentrations were corrected by comparison with 24 h gravimetric mass measurements of PMx . For gravimetric measurements 24h PMx samples were collected every 4 days on 150 mm quartz micro-fiber filters (Pallflex QAT) with highvolume (Hi-Vol) samplers (DIGITEL DH80 and/or MCV CAV-A/MSb at 30 m 3 h −1 ).

10
The extensive and intensive aerosol optical properties and the equations used to derive the intensive properties are reported in Table 1 and briefly commented below.
In order to study some of the aforementioned intensive optical properties over a wider spectral range, the 3λ scattering measurements from nephelometer were derived at the 7 Aethalometer wavelengths using the SAE calculated from 3λ measured scattering. Once scattering was obtained at the 7λ, we estimated SSA and SSAAE at these 7λ.

15
The extensive and intensive aerosol optical properties and the equations used to derive the intensive properties are reported in Table 1 and briefly commented below.
In order to study some of the aforementioned intensive optical properties over a wider spectral range, the 3λ scattering measurements from nephelometer were derived at the 7 Aethalometer wavelengths using the SAE calculated from 3λ measured scattering. Once scattering was obtained at the 7λ, we estimated SSA and SSAAE at these 7λ.

20
a) The SAE depends on the physical properties of aerosols and mainly on the size of the particles. Generally, SAE lower than 1 or higher than 2 indicate that the scattering is dominated by larger or finer particles, respectively (Seinfeld and Pandis, 1998;Schuster et al., 2006). In this study SAE was estimated from a linear fit of 3λ scattering measured in the 450-635 nm range.
b) The g parameter (Delene and Ogren, 2002;Andrews et al., 2006) is defined as the cosine-weighted average of the 25 phase function which is the probability of radiation being scattered in a given direction. Values of g can range from -1 for 180° backwards scattering to +1 for complete forward scattering (0°). A value of 0.7 is commonly used in radiative transfer models (Ogren et al., 2006). c) The AAE provides information about the chemical composition of atmospheric aerosols. BC absorbs radiation in the whole solar spectrum with the same efficiency, thus it is characterized by AAE values around 1 (Kirchstetter et 30 al., 2004;Kim et al., 2012). Conversely, BrC and mineral dust show strong light absorption in the blue to ultraviolet Atmos. Chem. Phys. Discuss., doi:10.5194/acp-2015-902, 2016 Manuscript under review for journal Atmos. Chem. Phys. Published: 18 January 2016 c Author(s) 2016. CC-BY 3.0 License. spectrum leading to AAE values up to 3 and 6.5 respectively (Kirchstetter, 2004;Chen and Bond, 2010;Kim et al., 2012;and Petzold et al., 2009). AAE was estimated from a linear fit of 7λ absorption measured in the 370-950 nm range.
d) The SSA parameter is defined as the ratio between the scattering and the extinction coefficients at a given wavelength and describes the relative importance of scattering and absorption on radiation. Thus the SSA parameter 5 indicates the potential of aerosols for cooling or warming the atmosphere. A detailed description of SSA at both MSY and MSA was presented by Pandolfi et al., (2011) and(2014a), respectively. Nevertheless in this work the SSA is used with the main objective of calculating SSAAE.
e) The wavelength dependence of the SSA is known as the SSAAE and it is defined as SSAAE=(1-SSA)*(SAE-AAE) (Moosmüller and Chakrabarty, 2011). This parameter provides general information about the type of sampled aerosols integrating both physical and chemical properties, and it has been proposed as a good indicator for the presence of Saharan dust in the atmosphere (Collaud Coen et al., 2004). The Saharan dust outbreaks change the intensive optical properties of sampled aerosols causing a reduction of SAE and an increase of AAE, resulting in a negative SSAAE during these events. Therefore this parameter can be used to asses which type of aerosol is dominating the scattering and the absorption. For example Collaud Coen et al. (2004) reported measurements 15 performed at the high altitude alpine station Jungfraujoch (Switzerland) and showed that the SSAAE was able to detect 100% of Saharan dust outbreaks compared with 80% and around 40% of events detected using SAE and AAE, respectively. Other works have used SSAAE to distinguish between the two important sources of UV absorbing aerosols, biomass burning and Saharan dust, as is detailed in (Russell et al., 2010). The SSAAE was estimated from a linear fit of 7λ-SSA calculated in the 370-950 nm range (Table 1).

The Aethalometer model
The Aethalometer (AE) model allows the detection of fossil fuel combustion (FF) and biomass burning (BB) contributions to the total BC concentrations taking advantage of the different spectral absorption efficiency of the main markers of these two sources: BC for FF combustion and BrC for BB (Sandradewi et al., 2008b). The AE model has also been applied for FF and BB source apportionment to total carbonaceous material (CM total =OM+BC) and to organic matter (OM) (Favez et al., 2010).

25
Light absorption measurements at 370-450 nm and 880-950 nm are used due to the fact that BC from FF combustion has a weak dependence on wavelength whereas BrC from BB shows enhanced absorption at shorter wavelengths. Here we applied the AE model to absorption measurements performed at 370 nm and 950 nm.
The AE model is usually applied selecting AAE values around 0.8-1.1 for BC from FF combustion (AAE ff ) and around 1.6-2.2 for BB (AAE bb ). It is known that the AE method may lead to high uncertainties in the estimation of biomass burning 30 contribution due to the high variability of AAE bb depending on the wood burned combustion regime and on the internal mixing with non-absorbing materials (Lewis et al., 2008;Harrison et al., 2013). Thus, AAE ff and AAE bb are usually chosen Atmos. Chem. Phys. Discuss., doi:10.5194/acp-2015-902, 2016 Manuscript under review for journal Atmos. Chem. Phys. by comparing the AE model outputs with FF and BB contributions to BC and/or OM from other techniques such as chemical mass balance (CMB) model on off-line filter measurements, positive matrix factorization (PMF) model on AMS and/or ACSM data or 14 C technique (Favez et al., 2010;Herich et al., 2011;Crippa et al., 2013). Here we followed a similar procedure to calibrate the AE model: the optimal AAE ff and AAE bb were selected comparing results from the AE model with those obtained from PMF on simultaneous ACSM hourly data at MSY station for 1 year (Minguillón et al., 2015). Then, the 5 optimal AAE ff and AAE bb for MSY were applied to MSA Aethalometer model.
In this work, CM total was calculated as the sum of BC concentration measured by MAAP (637 nm) and OM measured by ACSM. Following equations (1-3), CM total was expressed as the sum of carbonaceous material from FF combustion (CM ff ), carbonaceous material from BB emissions (CM bb ) and non-combustion organic aerosols (OA). At MSY station, OA may account for a large contribution mainly in summer and includes principally organic aerosols from biogenic origin as reported 10 in Minguillón et al. (2011) and Pandolfi et al. (2014b). Thus, we included the constant C 3 in contrast to previous studies where it was negligible assuming a low contribution of OA sources. CM ff and CM bb were then expressed as the product of the constants (C 1 and C 2 ) multiplied by the aerosol absorption due to FF at 950 nm (b abs,ff,950 ) and the aerosol absorption due to BB at 370 nm (b abs,bb,370 ), respectively. The b abs,ff,950 and b abs,bb,370 were calculated for different values of AAE ff and AAE bb following the equations reported in Sandradewi et al. (2008b) and then used in eqs. 1-3 for OM source apportionment.

15
Finally, the constants C 1 , C 2 and C 3 , which related the light absorption to the particulate mass, were calculated by multilinear

General features
Mean, standard deviation, median, minimum, maximum, skewness and percentiles (5,25,50,75,95) of hourly extensive and 25 intensive aerosol optical properties used in this work are reported in Tables S1a and S1b. Although the periods considered at the two stations were different, time coverage was sufficiently large to allow for a characterization of the mean aerosol optical properties at the two sites. Mean values of scattering, backscattering and PM 10 concentrations at both sites were consistent with previous studies performed at these stations (Pandolfi et al., , 2014aRipoll et al., 2014Ripoll et al., , 2015b. Higher σ sp and σ bsp were on average measured at MSY consistent with higher PM 10 concentrations due to the larger impact of

Detection of Saharan dust outbreaks using aerosol intensive optical properties
As already observed, SDE can be detected using optical properties measurements taking advantage of the changes that mineral dust causes in the spectral dependence of aerosol scattering and absorption (Collaud Coen et al., 2004). In fact SDE scenarios are characterized by a decrease of SAE, as a consequence of the predominance of coarse particles, and an increase of AAE due to the enhanced absorption in the UV spectrum by mineral dust. Therefore, the scatter plot between AAE and 15 SAE (called Ångström matrix) is useful to detect periods dominated by SDE (Russell et al., 2010). Here we calibrate, based on the available tools, the Ångström matrices for MSY and MSA in order to use them for SDE detection.
The Ångström matrix for MSY and MSA (Fig. 2b, e) showed dominance of coarse material (high % of PM 1-10 in PM 10 ) related to low values of SAE (roughly lower than 1) and larger values of AAE (approximately higher than 1.3) during SDE.
In order to demonstrate that these SAE and AAE limits were mainly related with the presence of mineral dust from Africa in 20 the area under study, the Ångström matrices were also weighted by the occurrence of the three main atmospheric scenarios The blue spot area displayed in the Ångström matrix for MSA station (Fig. 2e) showed AAE-SAE pairs characterized by low 5 contribution of PM 1-10 to PM 10 ∼%1-10, which are mainly represented by AA scenarios. These AA scenarios, some of them related with free troposphere conditions in MSA during winter, lead to a cleaner environment free of pollutants characterized by finer and relatively darker particles in the Ångström matrix. Conversely, a predominance of REG scenarios is seen at MSY (yellow dots in Fig. 1a), related to larger contribution of PM 1-10 to PM 10 (40-80%) (Fig. 2b). REG episodes, mainly related to pollution scenarios, are characterized by local (affecting lower altitude regions driven by the breeze patterns) to regional (reaching higher altitude locations driven by larger circulations and upslope winds) atmospheric circulations transporting fine particles from the urbanized/industrialized coastline. Mean SAE ranged between 1.5-3 and 1.  by Ogren et al. (2006) for other in situ measurements. Therefore, given that SAE parameter presents larger variability than g in relation to changes in %PM 1-10 , we conclude that SAE is a better proxy for estimating aerosol size. Despite this, providing experimental variability ranges for g is important given that the asymmetry parameter is commonly used in radiative transfer models (Ogren et al., 2006).
As already mentioned, the SSAAE has been identified as a good indicator for Saharan dust outbreaks at mountain top sites 25 being negative during these types of events (Collaud Coen et al., 2004). The SSAAE is a useful parameter, which can be used together with the Ångström matrix in order to characterize mineral dust at different emplacements with the aim to identify SDE in real-time. Similarly to what already observed for the Ångström matrices, our results showed that the feasibility of detecting SDE by means of SSAAE depended on both the location and altitude of the measurement station, which determines the aerosol background concentration, and the intensity of the SDE.
30 Figure 3a showed a relationship between SSAAE and %dust at MSY for those days affected by SDE. At MSA, where %dust was not calculated due to limitations of the methodology, SSSAE did correlate with percentage of coarse particles in PM 10 ( Fig. 3b). SSAAE became negative for most of the SDE identified at MSA accounting for 85% detection of these events.
However SSAAE showed more frequently positive values near to zero at MSY, detecting 50% of SDE due to a larger

15
(highlighted by the blue rectangle). Thus, the local and regional recirculation of air masses under the REG episode, often lasting for a few days, recirculated an aerosol mixture dominated by coarse Saharan particles in the atmosphere at a level able to cause the SSAAE be negative even in absence of African air mass advection (Fig. S2). The evidence that mineral dust can recirculate under dry conditions in summer for a few days after the SDE is of high relevance for air quality. Thus, near-real-time aerosol optical parameters such as SSAAE are very useful to detect mineral dust in the atmosphere even after 20 the end of the event.

Calculation of the constants from the Aethalometer model
In order to test the stability of the AE model for our emplacement (MSY), C 1 , C 2 and C 3 were calculated varying (Table 2): a) AAE bb between 1.8 and 2.2 (for a fixed AAE ff =1), and b) AAE ff between 0.9 and 1.1 (for a fixed AAE bb =2). In the first 25 case (a) C 1 showed a very low variability keeping values around 1.05±0.01 g m −2 ,whereas C 2 showed a higher variability ranging between 0.28 g m −2 (AAE bb =1.8) to 0.24 g m −2 (AAE bb =2.2). In our work C 3 , which represents the contribution from non-combustion OM, was estimated in around 0.31±0.04 µg m −3 . These results were consistent with previous studies dealing with AE source apportionment to OM and reporting less variability for C 1 compared to C 2 (Sandradewi et al., 2008b;Favez et al., 2010). In another study (Herich et al., 2011) the AE model was not applied to OM mainly due to the high variability

30
(i.e. model instability) observed for C 1 from different model outputs. In the second case (b), C 1 changed only little (less than 10%) ranging between 1.01 g m −2 (AAE ff =0.9) to 1.09 g m −2 (AAE ff =1.1) for a fixed AAE bb of 2. As reported bellow, AAE bb for our environment was set to 2 by comparison with ancillary experimental measurements, whereas AAE ff was set to 1 as in Roveredo, Switzerland) set C 1 to a fixed value of 0.26 g m −2 , being this parameter less variable, and C 2 was estimated around 0.7-0.8 g m −2 . Differences between the constants were due to the larger use of biofuel for domestic heating in these later 5 locations, leading to higher contribution of BB to BC compared to FF combustion sources (and probably less effect of FF sources). Contrary to our emplacement where results indicated (as shown later) higher contribution from FF sources compared to BB for both BC and OM.
Given the large differences our constants C 1 and C 2 showed compared to previous studies for different environments, we applied here a similar procedure as described in Herich et al. (2011). Thus, we simulated CM total using C 1 and C 2 from Sandradewi et al. (2008b) and Favez et al. (2010), and b abs,ff,(λ1) and b abs,bb,(λ2) as derived from our measurements. As expected the results showed very low correlation between calculated and measured CM (R 2 =0.009; slope=0.65) compared to R 2 =1 and slope=1 using our calculated constants C 1 , C 2 and C 3 . Therefore we conclude that calculation of the specific constants of the model for the area under study is required in order to successfully perform the Aethalometer model.
Moreover, we calculated C 1 , C 2 and C 3 for two more different cases: (a) including only the winter season in order to account 15 for a larger contribution of BB emissions and to reduce the influence of non-combustion OM and SOA formation which maximize in summer at MSY station , and (b) excluding SDE from the database which could overlap with BrC being both, BB and mineral dust, important absorbers in the UV. The differences for C 1 , C 2 and C 3 calculated between these two cases and the whole period (June 2012-July 2013, Table 2) in case (a) were lower than 10%, 20% and 15%, respectively. These differences were around 3%, 6% and 34%, respectively, for the case (b). Given that the 20 AE model outputs have been estimated having errors as high as 50% (Favez et al., 2010) and given that we are continuously measuring absorption with the AE instrument at MSY and MSA without ACSM data, the model was calibrated using 1 year data set in order to apply the AE model at any other period without ancillary measurements.  Cubison et al. (2011); and b) an OM ff /HOA ratio of 4.4 (R 2 =0.6) which is consistent with 5 90% portion of SOA found at MSY in previous studies . The correlations were only moderate mainly due to the variable SOA formation, which is partially driven by the environmental conditions, as opposed to the primary OA emissions. Moreover, it should be noted that the slopes and R 2 in Table 3 were obtained using hourly averages.

Validation of the Aethalometer model with simultaneous experimental data
Scatterplots by bins (Fig 4) showed that the relationships had slopes in agreement with those reported in Table 3 but much higher R 2 (0.97).

10
The relationship between OM bb and BBOA calculated only for the winter period using hourly data showed R 2 =0.4 and F=0.96. The slope was close to the unity due to the lower SOA formation in winter, consequence of less photochemistry activity, and the prevalence of primary emissions. Experimental measurements of Nitrogen dioxide (NO 2 ), which is mainly related to fossil fuel emissions, agrees well (R 2 =0.64) with BC ff obtained from AAE bb =2 and AAE ff =1 for the winter period at MSY (Fig. 5).
Besides uncertainties in determining FF and BB contributions from the Aethalometer model, results from sensitivity test analysis showed good agreement with experimental measurements and good stability of the model. We have shown that the constants C 1 , C 2 and C 3 depend on the relative contributions of FF and BB, thus these constants are site-dependent and should be calculated for each measurement emplacement. Moreover, a calibration of the model is necessary to determine the most suitable AAE ff and AAE bb pair for a reliable estimation of fossil fuel and biomass burning contributions. Interestingly 20 AAE bb and AAE ff chosen in this work were the same as in other studies, suggesting a stable value of AAE=2 for characterizing BB emissions within the model. Our results showed that the higher AAE bb the lower the estimated BC bb contribution, which ranged between 35-45% depending on the AAE bb used (1.8-2.2). average AAE was higher at MSY than at MSA during winter months (December-January) suggesting higher relative BB contribution at MSY compared to MSA in winter ( Fig. 6e and S3a). This was likely due to the fact that MSA station is often above the polluted PBL in winter whereas MSY, located at lower altitude, is usually within the PBL and frequently affected by local pollutants accumulated under winter anticyclonic conditions (Pandolfi et al., 2014b;Ripoll et al., 2015b). Low values of AAE during the day and higher at night at both sites resulted mainly from the development of sea and mountain 5 breezes, favouring the transport of anthropogenic pollutants from the urbanized/industrialized coastline and valleys to inland areas and leading to an increase of AAE during the warmest hours of the day (Fig. 6a).

Seasonal and daily variation of fossil fuel and biomass burning contribution to BC and OM at Montseny and
The measured BC was well reproduced by the sum of BC ff and BC bb contributions from the AE model showing slightly overestimation, by 11% and 15% at MSY (Fig. 6c, g) and MSA (Fig. 6d, h), respectively, on annual average. However, measured OM is underestimated by the sum of OM ff and OM bb at MSY, due to the large contribution of carbonaceous 10 material from non-combustion sources (C 3 ) during the warmer months (27%) (Fig. 6f). This difference was mainly driven by biogenic sources which are expected to have important contribution in our measurement emplacement, particularly in summer due to the SOA formation. Then C 3 time variation was well reproduced by the model showing larger contribution during the summer period. Nevertheless, based on the available previous studies performed at MSY 2015;Pandolfi et al., 2014b), C 3 contribution might be slightly underestimated due to possible apportionment within 15 OM ff and/or OM bb . It should also be note that some SOA UV absorbing compounds originated from anthropogenic sources, such as nitroaromatic compounds which are the major contributors to the light absorption of the toluene SOA (Laskin et al., 2015), may be partially apportioned within OM bb , possibly resulting in an overestimation of this later.
Interestingly, a relationship was observed between AAE and the relative contribution of BC bb to BC concentrations at MSY and MSA (Fig. 7). AAE increased up to 1.5 when %BC bb was higher than around 50% of the total measured BC. The 20 intercept of the linear fit was 1.01 and 1.15 at MSY and MSA, respectively, pointing to BC from FF sources as main absorber in absence of biomass burning events. Therefore, we can clearly appreciate the effect of BrC from biomass burning on AAE even if the mean BC bb contributions (0.13 µg m −3 and 0.06 µg m −3 ) at MSY and MSA, respectively, to the total BC were quite low (36% and 40%). Mean OM bb concentration at MSY was 0.9 µg m −3 , accounting for a 30% contribution to total OM.

25
The prominent increase of FF contribution at MSY and MSA in summer, when both stations are within the PBL and dominated by similar atmospheric circulations, is in agreement with lower AAE values. Stronger summer recirculation processes which are strengthened by sea and mountain breezes favour the transport of pollutants toward regional areas inland. Daily variation of both BC and OM is mainly driven by FF combustion from Barcelona anthropogenic sources. The daily cycle is more pronounced at MSY as a consequence of the proximity to Barcelona Metropolitan Area and the lower 30 altitude compared to MSA. Despite OM is mainly driven by biogenic sources during the summer period at MSY, significant FF contribution is registered during the warmest hours of the day (Fig. S3b). However BB sources time variation, from both BC and OM, are leaded by local atmospheric processes as domestic heating turning into a dominant source during the colder months at both stations. Thus, during winter, BC bb and BC ff showed almost the same contribution reaching the maximum values in the afternoon (Fig. S3c). Conversely OM ff daily cycle is decoupled from OM bb , showing this later larger concentrations during the night given that it is mainly leaded by BB emissions from domestic heating emitted during the colder hours, and also possibly as a result of SOA formation after the OM was emitted (Fig. S3b). Note that during the night OM bb concentration does not present large variations, possibly because it remains as a residual layer above the thermal inversion.

5
FF contribution to OM and BC was found to be significant at MSY, according to the large values obtained for C 1 constant in the Aethalometer model. In order to compare the results with different source apportionment methods, the fossil fuel and non-fossil fuel contribution to EC (EC ff , EC non_ff ) and OC (OC ff , OC non_ff ) reported by Minguillón et al. (2011)  winter and summer, respectively, whereas the AE model resulted in a OM ff contribution of 39% and 58%, respectively. We also saw a OM bb contribution around twice more than OC non-fossil fuel. The apparently overestimation of OM bb and OM ff , particularly in summer, compared to the available results from 14 C might be possibly leaded by the partially apportionment of non-combustion carbonaceous material and SOA anthropogenic within OM bb and/or OM ff , as we commented above.
A second assessment of the AE model results was carried out by comparison with OA source apportionment results reported 20 by Minguillón et al. (2015) for winter (28 October-7 April 2013) and summer (14 Juny-9 October 2012) at MSY based on ACSM measurements. The agreement needs to be evaluated considering the different outputs from each method; thus whereas the ACSM OA source apportionment identifies the contribution of primary fossil fuel (HOA) and biomass burning (BBOA) contributions, the AE model calculates the total (including the SOA) fossil fuel (OM ff ) and biomass burning (OM bb ) contributions. HOA contribution was 12% and 13% for winter and summer, whereas OM ff accounted for 47% and 59%.

25
BBOA was identified only in winter with a contribution of 28%, and OM bb contribution was 37% for the same period. These results are in agreement assuming the ratios OM ff -to-HOA and OM bb -to-BBOA based on SOA-to-POA proportion, used in the previous section 5.4.3 in order to calibrate the Aethalometer model and fit the most suitable AAE ff and AAE bb representative of our environment.
An interesting wildfire episode detected at MSY took place the 23th of July 2012 with AAE increasing significantly up to 2 30 and the lowest value at 1.3 (Fig. 8). BB sources dominated BC and OM contributions accounting for 73% and 78% respectively, until the breezes were developed and transported pollutants from urban areas toward the station during the warmest hours of the day, resulting in a decrease of the AAE. As we shown previously for the whole dataset, good Atmos. Chem. Phys. Discuss., doi:10.5194/acp-2015-902, 2016 Manuscript under review for journal Atmos. Chem. Phys. Published: 18 January 2016 c Author(s) 2016. CC-BY 3.0 License. agreement was found between measured and simulated BC. Conversely OM was slightly underestimated during the sunlight hours likely due to biogenic emissions and SOA formation by photochemical reactions.

Conclusions
The present work shows the variations of the intensive aerosol optical properties measured at regional (Montseny) and continental (Montsec) background stations in the WMB. We have studied the feasibility of using the near real-time optical 5 measurements performed at these stations for the detection of specific atmospheric pollution episodes affecting the WMB: Saharan dust and biomass burning.
The Ångström matrix revealed that Saharan dust events (SDE) in the WMB were characterized by SAE on average lower than 1 due to the larger size of mineral dust particles and AAE values higher than 1.3 (up to 2.5 depending on the intensity of SDE) indicating absorption in the UV by iron oxide contained within the mineral dust. Linear relationships were found 10 between AAE and increasing %dust at MSY (0.7) and %PM 1-10 at MSA (0.4) confirming the enhanced absorption in the UV due to mineral dust from SDE. Interestingly, SAE showed higher sensitivity than g to characterize the size of aerosols, ranging this latter between 0.55-0.75 and 0.50-0.70 at MSY and MSA respectively during SDE.
Feasibility of detecting SDE by means of SSAAE depended on both the location and altitude of the measurement station, which determines the aerosol background concentration, and the intensity of the SDE. Better results were shown at higher 15 altitude locations, at MSA were detected most of the SDE (85%), whereas at MSY, with a larger exposure to anthropogenic pollutants, the detection of SDE depended mainly on the intensity of the Saharan dust outbreak. At MSY site 50% of SDE were detected, which were unequivocally identified when the relative contribution of mineral dust to PM 10 was higher than 60%.
The proximity to anthropogenic sources of mainly fine particles can prevent both the Ångström matrix and the SSAAE 20 parameter from detecting SDE. We have shown that transport of anthropogenic pollutants (mainly finer particles and precursors) from the urbanized/industrialized coastline towards regional areas inland can hinder the effect of mineral dust on the intensive aerosol optical properties during less intense SDE. We have also shown that regional atmospheric circulations occurring after SDE may favour the resuspension of mineral dust at regional level in the WMB. Thus mineral dust can remain in the atmosphere for a few days after the SDE. This fact is highly relevant for air quality since SDE frequently 25 promote exceedances in the PM 10   We have demonstrated the potential of in situ aerosol optical measurements, from both Nephelometer and Aethalometer 25 instruments, for detecting specific air pollution scenarios in near real-time. This is possible given the high sensitivity of particular intensive aerosol optical parameters to characterize different types of atmospheric aerosols. However, it is necessary to perform a previous sensitivity test in order to evaluate and calibrate the intensive optical properties for detecting specific pollution episodes at different emplacements.