Increased PM Concentrations during a Combined Wildfire and Saharan Dust Event Observed at High-Altitude Sonnblick Observatory , Austria

A period of increased particulate matter concentrations was observed at the high-altitude Sonnblick Observatory in August 2013. Trajectory analysis, wildfire maps and the evaluation of aerosol measurements revealed a combined and sometimes alternating influence of long-range transport of Saharan dust and emissions of wildfires. The occurrence of Saharan dust was confirmed by an increase of coarse particle number concentration and a negative exponent of the single scattering albedo wavelength dependence, determined by Nephelometer and Aethalometer measurements. During time periods less influenced by Saharan dust, number concentration of accumulation mode particles increased and a marked correlation of aerosol mass concentrations and CO mixing ratios was observed. By analyzing the wavelength dependence of the absorption coefficients determined with a seven wavelength Aethalometer, the influence of the two aerosol sources was decoupled. Therefore, absorption exponents of 3 and 1.3 were assumed for Saharan dust and wildfires, respectively. Mass concentrations of particulate matter caused by Saharan dust and wildfire emissions were estimated, with the contribution of Saharan dust to overall particulate matter mass ranging from 5% to 80%.


INTRODUCTION
The great potential of mountain observatories to be used as platforms for monitoring background concentrations of aerosolized particulate matter and atmospheric trace gases is well documented (e.g., Galasyn et al., 1987;Kasper and Puxbaum, 1998;Nyeki et al., 1998).Depending on the season and time of the day, air masses can be representative of the free troposphere or the sites are immersed in the planetary boundary layer (e.g., Van Dingenen et al., 2005;Collaud Coen et al., 2011;Gallagher et al., 2011).Furthermore regional or long range transport can lead to elevated concentration levels.Long range transport of Saharan dust and its climatology are well documented (Collaud Coen et al., 2004).Furthermore, wildfires have been shown to increase concentrations of primary and secondary particulate matter at mountain sites (Wigder et al., 2013).
The Sahara is the most powerful source for the emission of soil dust in the world (Swap et al., 1996;Goudie and Middleton, 2001) and transport to Europe is observed regularly (Stuut et al., 2009;Su and Toon, 2011).The transport of Saharan dust to Europe and especially the high Alpine regions has been addressed long ago (Wagenbach et al., 1989;DeAngelis and Gaudichet, 1991).A long time series based on ice core data has been reported by Thevenon et al. (2012).A review about the occurrence and implications of Saharan dust storms and subsequent long range transport of soil dust is given by Goudie and Middleton (2001).Saharan dust in the atmosphere influences the radiative budget and consequently radiative forcing (Diaz et al., 2001;Meloni et al., 2004).Acting as cloud or ice nuclei, dust particles interact with cloud formation, again influencing the radiative budget, but also changing precipitation patterns (Rosenfeld et al., 2001;Mahowlad and Kiehl, 2003;Karydis et al., 2011).Deposited Saharan dust influences nutrient cylces (Avila et al., 1998;Swap et al., 1992) and bacterial population in glacial ice (Zhang et al., 2008).Saharan dust events lead to PM concentrations above the existing regulatory limit values in the Mediterranean region (Matassoni et al., 2009;Salvador et al., 2013).
A number of methods exist to allow the identification of Saharan dust events (SDEs), based on ambient air sampling (e.g., Viana et al., 2010).An on-line SDE detection method was described by Collaud Coen et al. (2004), based on the investigation of the single scattering albedo wavelength dependence.A two year data set measured at Jungfraujoch showed the predominant occurrence of SDEs in spring (March-June) and autumn (October-November).Occasional but intense events are also reported during summer months.
Biomass burning during wildfires emits primary aerosols and is also responsible for secondary particulate matter having a strong impact at the regional scale (Alves et al., 2010), but as emissions get transported to regions far from the source (Duck et al., 2007;Petzold et al., 2007;Ulevicius et al., 2010) impacts are observed on the global scale as well.
The emission factors largely depend on the type of biomass burnt as well as on burning conditions and conditions during plume transport (Reid et al., 2005a, b;Diapouli et al., 2014).Most of the particulate matter emitted by biomass burning can be attributed to the accumulation mode and generally organic matter accounts for most of the particulate matter mass (Reid et al., 2005a;Alves et al., 2010).Due to their size, physical properties and chemical composition, particles emitted due to burning activities interact with the radiative budget (Lewis et al., 2008;Spracklen et al., 2011), and are also known to act as cloud condensation nuclei, thus influencing cloud formation and precipitation patterns (Lee et al., 2010, Engelhart et al., 2012;Hennigan et al., 2012).Furthermore, global nutrient cycling (Furutani et al., 2010) can be influenced.Considering that, due to the expected climatic trend, more wildfires can be expected in the future (Spracklen et al., 2009;Moreira et al., 2010), this impact of particulate matter emitted by wildfires is of increasing importance.
Wildfire plumes impacting background sites have been identified based on CO mixing ratios and elevated concentration levels of particulate matter (Wigder et al., 2013).
In August 2013 elevated concentrations of particulate matter were observed at the Sonnblick Observatory, a high Alpine background site in the Austrian Alps, for a time period of seven days.Trajectories generally reveal a southwestern flow with a large fraction of trajectories originating over North Africa.In the afternoon of August 4, 2013 the permanent staff of the observatory noticed the odour of wood smoke outdoors and recorded this impression in the QM-log.Wildfire maps as well as information from local newspapers confirmed this observation and showed the occurrence of fires in Southern Austria and Northern Italy.
Combining these pieces of information, the Sonnblick Observatory site was obviously influenced by long range transport of Saharan dust, as well as by emissions from wildfires occurring in the surrounding regions (50-200 km) during the time period of August 3 to 9, 2013.This gives us the opportunity to investigate to which extent these events can be distinguished by the ongoing highly time resolved measurements.Our scientific interests can be formulated as follows: Is it possible to distinguish the influence of the two aerosol sources using on-line instrumental data and do various methods already described in literature give similar results?Is it possible to decouple the influence of the respective sources on aerosol absorption and on particulate matter mass?

Site Description
The continuously manned Sonnblick Observatory (SBO) is situated at the summit of Mt.Sonnblick at 3.106 m asl in the Austrian Alps (12°57'E, 47°03'N), within the core zone of Hohe Tauern National Park.The peak of Mt.Sonnblick is surrounded by large glacier fields and located approx.1000 m above the tree line.The only access to SBO is via a ropeway restricted to the transport of observatory supplies and personnel or via a five hour hike from the valley floor.To minimize emissions, the observatory is supplied with electricity via a permanent line from the valley.Internal air of the observatory is disposed of through a vent inside a 20 m high meteorological tower.In the larger context the observatory is located in the center of Europe.

Sampling and Instrumentation
Aerosol sampling is performed via a whole air inlet designed according to GAW guidelines (WMO/GAW Report No. 153, 2003) and recommended for high alpine mountain sites, which are frequently impacted by clouds and experience freezing and rime formation.The inlet temperature is controlled and kept at + 20°C to prevent it from freezing over by impacting super cooled droplets.Its upper cut-off size is 20 µm at a wind speed of 20 m s -1 .Activated aerosol particles will evaporate inside the inlet and the residues are included in subsequent sampling.Furthermore losses due to evaporation of semi-volatile fraction might occur.All measurements are performed under dry conditions.Sampling is performed after distributing the air flow via an isokinetic flow splitter to minimize particle loss.No extra size cuts were installed for the single instruments, thus we only refer to particulate matter collected behind a whole air inlet.According to Weingartner et al. (1999), characterizing an inlet system being the prototype of the recommended GAW-inlet, quantitative sampling of accumulation mode particles can be expected.Regarding particles > 1 µm losses will occur.Thus, concentrations of mineral dust have to be regarded as a lower limit to actual conditions at the site.Aerosol characterization included the determination of aerosol mass based on a combination of nephelometry and β-attenuation (Thermo Scientific Sharp 5030 instruction manual, 2013).Just in one case of the discussion, the mass concentration determined by nephelometry only will be used for comparison purposes.Furthermore, the total number concentration of particles was determined using a condensation particle counter (TSI CPC 3022A) and particle size distributions were measured in the size range between 0.3 and > 2.5 µm with an optical particle counter (Klotz, TCC-3).A three-wavelength polar Nephelometer (Ecotech Aurora 4000) was used for the determination of the light scattering coefficients at 450, 525 and 635 nm.Absorption coefficients were determined at seven wavelengths (370, 470, 525, 590, 660, 880 and 940 nm) with an Aethalometer (Magee Scientific AE 33; Drinovec et al., 2015).All measurements are reported as 30 min averages, denoted timestamps are in UTC and aerosol concentrations given are calculated for 273K and 1013 mbar.
Meteorological data are measured by ZAMG (Central Institute for Meteorology and Geodynamics) while trace gas concentrations are determined by the Umweltbundesamt (Austrian Environmental Protection Agency) via an independent manifold used for gas sampling.Here we refer to CO data measured by a Horiba APMA-360 CO monitor (NDIR).

Data Treatment
All data collected by above-mentioned instruments are recorded routinely in the Observatory's database.The basic calculations on aerosol data, e.g., the determination of the scattering-, absorption-and single scattering albedo exponents are performed within this database.Advanced statistical analyses were performed using The R Project for Statistical Computing (R Core Team, 2014).All Standard Major Axis (SMA) regressions were calculated with the lmodel2 package (Pierre Legendre, 2014).The overdetermined linear system of equations (Eq.( 7)) in the form of Ax ≈ b was transformed into the Gaussian Normal Equation A T A x = A T b and solved in R. Back trajectories were calculated with the Lagrangian transport and dispersion model FLEXTRA V5.0 (Stohl et al., 1998(Stohl et al., , 1999;;Seibert et al., 2004;Stohl et al., 2005) using meteorological data from ECMWF.

Determination of Saharan Dust Events Based on SSA
Dust episodes have been identified following the method described by Collaud Coen et al. (2004) and developed for the Jungfraujoch site, which uses the same type of inlet as the Sonnblick observatory.Based on the scattering and absorption coefficients determined at the site, the single scattering albedo (SSA) was calculated.The wavelength dependence of the single scattering albedo is then fitted with a power-law dependence: to obtain α SSA , the single scattering albedo exponent.Based on theoretical considerations and confirmed by experimental data Collaud Coen et al. (2004) interpret the inversion of the SSA wavelength dependence, seen as the change in α SSA sign, as an indication for mineral dust affecting the site.They define a Saharan dust event (SDE) as a time period exhibiting a negative α SSA for more than 3 hours.We modified this method slightly, as we did not base our calculations on the wavelength used for the absorption measurements, but started with interpolating the absorption coefficients with a power-law dependence to calculate the respective absorption coefficients at the three wavelengths for which scattering coefficients are available.The SSA was then calculated based on those data-pairs.This allowed us to investigate the wavelength dependence of SSA in the range from 450 to 635 nm.Furthermore, we increased the time resolution for the identification of SDEs from 3 hours to 30 minutes, as we do not use the procedure for monitoring purposes, but investigate a rather short time period of one week which was obviously influenced by other aerosol sources as well.The increased time resolution of the Aethalometer (AE33) allowed this modification.

Particulate Matter Concentrations Determined in August 2013
Aerosol mass and number concentrations at the Sonnblick Observatory are generally very low, reflecting the background status of the site with prevalent clean air conditions.During summer, concentrations tend to increase due to the frequent immersion of the site in a modified mixing layer.Based on 30 min averages, aerosol mass concentrations determined in August 2013, ranged from 0.4 to 30.7 µg m -3 (5th-95th percentile), with an average value of 9.4 µg m -3 .Elevated concentrations were determined in the time period from August 3 to August 9, 2013, when the average mass concentration increased to 22.4 µg m -3 .Table 1 summarizes average values of mass concentrations, condensation particle count (CP-count) and number concentrations > 0.3 and > 2.5 µm as well as percentiles for August 2013, the time period August 3 to August 9, as well as the remaining days in August, highlighting the special conditions from August 3 to August 9.
A more detailed picture of the time period from August 3 to August 9 is given in Fig. 2(a), showing the particulate mass concentration and identified SDEs.Concentrations of particulate matter start to increase during August 3 and, after a temporary decline on August 4, remain above 17 µg m -3 until August 9, when precipitation sets in.During that time period, elevated mass concentrations were often associated with the influence of Saharan dust, determined with the negative α SSA value.However, according to α SSA , the influence of Saharan dust was not permanent.Furthermore Collaud Coen et al. (2004) report that values of α SSA usually fall between -0.1 and -0.5 when Saharan dust is present.In our case α SSA below -0.1 is observed during the first three days of the given time period, only.On the other hand, with the exemption of a few hours on August 4, α SSA remains at or below 0.1 during the whole period of elevated particulate mass concentrations.This is rather low compared to the rest of August, with an average α SSA of 0.21 and the 90% range between 0.02 and 0.43.

Trajectory Analysis and Wildfire Maps
240 h back trajectories have been calculated for arrival at station height (SBO) every 3 hours over the period of investigation.The trajectory paths generally reveal a southwestern flow with a large fraction of trajectories originating over North Africa, where they show residence times of a week and more before traveling via Spain to the SBO.Thus the general influence of long range transport of Saharan dust to Sonnblick could be confirmed by trajectories, but no conclusive results could be deduced about short term variations and the intensity of this influence.During    the last hours of travel to SBO, air masses pass over Northern Italy and Southern Austria (Carinthia).Thus, emissions originating in these regions will be incorporated as well.
Analysis of wildfire maps (Fig. 1, NASA Worldview) showed the occurrence of fires in Southern Austria and Northern Italy starting with August 4.The fires continued during the following days with varying intensity.
Particle Size Distribution Fig. 2(b) shows CP-count and number concentrations of particles > 2.5 µm as well as > 0.3 µm.As expected, the number concentration of coarse particles (> 2.5 µm) shows a strong increase when Saharan dust is indicated by α SSA .During these periods mass concentrations and number concentrations of coarse particles are strongly correlated (r 2 = 0.809), while a lower correlation is found for the remaining time (r 2 = 0.597).The influence of Saharan dust on number size distribution becomes more evident when the relative contribution of coarse particles to number concentrations of particles > 0.3 µm are considered.During time periods characterized as SDEs, coarse particles account for 1.6 to 4.3% (10 th and 90 th percentile) of number concentration > 0.3 µm, while the contribution is just 0.4 to 1.7% (10 th and 90 th percentile) for the remaining time.
Until the morning of August 5, the number concentration of particles > 0.3 µm remains low and rather constant, while coarse particles increase at times when Saharan dust is indicated using α SSA .Later, this interrelation between coarse particles and a negative α SSA remains, but additionally concentrations of particles > 0.3 µm increase.They show some fluctuation, being inversely correlated with number concentrations of coarse particles most of the time.This indicates the alternating influence of the two aerosol sources, Saharan dust and wildfires.Wildfires will lead to an enhanced formation of secondary aerosols, i.e., accumulation mode particles.The most pronounced influence of accumulation mode particles was observed in the afternoon and evening of August 5, simultaneously with elevated mixing ratios of CO, as will be discussed further down on.
Comparing CP-count and number concentration of particles > 0.3 µm, taken as a proxy for secondary aerosols, three well-defined time periods become visible.They can already be deduced from the temporal trends given in Fig. 2(b), but the direct comparison is shown in the scatter plot in Fig. 3. From August 3 until August 5 (10:30) CP-count ranges from 1000 to 3000 cm -3 .Number concentrations of particles > 0.3 µm remain rather low and are not correlated with the CP-count.The following period, August 5 (14:30) until August 6 (8:00), showing elevated CO concentrations indicating a marked influence of wildfires, is characterized by a strong increase of number concentrations of particles > 0.3 µm and a marked correlation with CP-count (r 2 = 0.761).The conditions observed during the remaining time period (August 6 (9:00) until the onset of precipitation on August 9 (18:00)) lie between the two groups described before.Obviously, the combined and alternating influence of both emission sources of particulate matter can be observed.
If we assume that in addition to the negative SSA exponent being indicative for a SDE, that the value of α SSA allows to deduce the intensity of such an event, then the influence of Saharan dust tends to become smaller over the course of time.As mass concentrations and number concentrations of coarse particles increase perpetually, a variation of the absolute contribution of dust particles is not straightforward.However, an increasing influence of the second aerosol source could be possible, which would influence α SSA as well.

Correlation with CO Data
Wigder et al. ( 2013) evaluate a long term data set of the impact of emissions from regional wildfires to particulate matter concentrations observed at Mt. Bachelor.Emissions from wildfires are accompanied with elevated concentrations of CO and normalized enhancement ratios of particulate matter expressed as ∆PM 1 /∆CO are given.Some influence of the distance of the fires on the enhancement ratio was determined, pointing to an increase of mass due to formation of secondary aerosols with increasing transport time, as long as distances < 540 km are considered.The correlation of CO mixing ratios and PM concentrations is given in Fig. 4. Based on the whole observation period, the correlation shows a lot of scatter (r 2 = 0.311, Fig. 4(a)).Eliminating time periods with a pronounced influence of Saharan dust (identified by a negative SSA exponent), a more conclusive trend can be deduced, and the data set seperates into three groups (Fig. 4(b)), which are consistent with the differentiation performed before based on particle number concentrations > 0.3 µm and CP-count.Before August 5, CO concentrations are rather low and correlation with PM concentrations is poor.Later, when we expect the influence of fires to be most pronounced (based on the contribution of particles > 0.3 µm to CP-count), mass concentrations of particulate matter and CO correlate pretty well (r 2 = 0.844) with a slope of 0.28.This enhancement of ∆PM/∆CO is within the range reported by Wigder et al. (2013) and references therein.After August 6, when we expect a combined or alternating influence of wildfires and Saharan dust, a similar enhancement of ∆PM/∆CO is observed, although the scatter increases strongly.These results lead to assuming that conditions without Saharan dust can be approximated by conditions typical for emissions from wildfires during the investigated period.In Fig. 4(c) particulate mass is also expressed by the Nephelometer signal of the Sharp monitor.This signal is not adjusted by the ratio of the measurements of scattering coefficients and the attenuation of β-radiation (Thermo Scientific Sharp 5030 instruction manual, 2013).As we do not know the exact algorithm for data processing performed within the Sharp monitor, we do not attempt to interpret this result any further.However, the adjustment of the Nephelometer signal with the internal calibration factor determined for the preceding time interval (averaging time of several hours) seems to influence the variations in this combined event.Fialho et al. (2005) present an approach to decouple the influence of Saharan dust and black carbon particles to aerosol absorption coefficients at Pico mountain in the Azores Island.Analyzing the wavelength dependence of the absorption coefficients, the best-fit aerosol absorption exponents for BC (-1.0 ± 0.1) and dust particles (-2.9 ± 0.2) were deduced, which allowed an estimate of the aerosol absorption due to BC and dust, respectively.(Note that Fialho et al. (2005) do not report the exponents as negative of the slope of a plot of the absorption coefficients versus wavelength on a log-log plot.In the present paper we use the negative of the slope to be consistent with the method used for the fit of the wavelength dependence of the SSA.)Based on the assumption that, during the time period under investigation, aerosol absorption coefficients at Sonnblick are only driven by Saharan dust and wildfires, we adopt this method to separate and quantify the contribution of dust and wildfires to aerosol absorption and in a second step, we also separate and quantify the contribution of the two sources to particulate matter mass determined during the mixed event in August 2013.

Evaluation of the Absorption Exponent
The typical dependence of the aerosol absorption coefficient σ on the wavelength λ, characteristic for Saharan dust, was deduced from the entire dataset of absorption measurements with an Aethalometer conducted at Sonnblick.These measurements comprise a two year time period (March 2013 until April 2015) and were evaluated to determine the absorption exponent (α ABS ): The frequency distribution of α ABS , based on 30 min averages, yielded 3.0 as an extreme value, still occurring repeatedly (Fig. 5(a)).Thus, an α ABS of 3.0 is regarded as typical for pure dust.This value is in close agreement with the value reported by Fialho et al. (2005) and references therein.However, in a subsequent paper Fialho et al. (2006) change α ABS , based on long term data determined for Pico mountain, to -4.0.At Sonnblick values close to an α ABS of 3.3 were determined as well, when the long term data set is considered, but, as they occur in less than one percent of cases only, we regard an α ABS of 3.0 as the more robust value.Considering only the time interval from August 3 to 9, 2013, the maximum value of α ABS is 2.5 (Fig. 5(b)), indicating that no period of 'pure' Saharan dust was observed during this event.
The absorption exponent characteristic for wildfires cannot be deduced from the long term data set of absorption measurements at Sonnblick.Containing considerable amounts of brown carbon, biomass burning aerosols will definitely differ from α ABS used by Fialho et al. (2005) to characterize black carbon originating from fossil fuel combustion.The variation of α ABS reported for various types of biomass burning is wide and is strongly influenced by the type of wood burnt as well as burning conditions and, considering the methodological approach, the wavelength range considered.Bergstrom et al. (2007) report α ABS for emissions from biomass burning ranging from 1 to 3. A value of 2.0 ± 0.4 is observed by Ulevicuis et al. (2010), Kirchstetter et al. (2004) report 1.8 for a savannah fire.Sandradewi et al. (2008) suggest an absorption exponent of 1.8-1.9 to be indicative for aerosols originating from domestic wood burning.The absorption exponent may depend on the aerosol ageing, as shown in chamber experiments (Saleh et al., 2013).
As we expect a marked influence of wildfires, we investigate the frequency distribution of the absorption exponent during the period in August 2013 (Fig. 5(b)), which differs from the one characteristic for the aerosol observed at Sonnblick (Fig. 5(a)).Absorption exponent values range from 1.1 to 2.5.Excluding data with pronounced influence of Saharan dust, determined according to a negative α SSA , the frequency distribution of the absorption exponent shows peaks at 1.7 and a second maximum at 1.35 (Fig. 5(c)).The value of 1.7 would be in good agreement with literature data reported for wildfire emissions.However, as we want to set a model based on two influencing factors only, we decided to use 1.3 as characteristic for the emission source 'wildfires' for the time period we investigate here.Obviously, air masses being indicative for wildfires are also influenced by other emissions sources shifting the absorption exponent to smaller values.
Using an α ABS,SD of 3.0 for Saharan dust and α ABS,WF of ) ) a linear model can be used to approximate K SD × C SD and K WF × C WF at a given point in time t as slope and intercept when applying the least square fitting technique to the considered wavelength range (470-660 nm).Thus the relative contributions of the two aerosol sources can be expressed as the products of K SD × C SD (t) and K WF × C WF (t), i.e., the product of an empirical constant accounting for the instrument and the optical properties of the particulate matter and the concentration at a given time step.
To estimate the uncertainty of the method explained above the absorption coefficients for any given time period were recalculated by using the calculated parameters K SD × C SD (t) and K WF × C WF (t) to determine the absorption coefficients for dust and wildfire aerosols ; , ( ) ( ) ; , ( ) ( ) and adding those up to the total absorption coefficient.Again, calculations were made for the wavelength range of 470 to 660 nm.The differences between the calculated and measured absorption coefficients at 470, 525, 590 and 660 nm (Delta ABS) were calculated for every time step and are given in Fig. 6.They remain well below 0.25 Mm -1 during most of the time.Compared to median values of the absorption coefficient above 4.5 Mm -1 , this reflects the suitability of the assumptions made.Variations are most pronounced during the last 24 hours of the event.
In Fig. 2(c) the temporal trends of K SD × C SD (t) and K WF × C WF (t) are superimposed to aerosol mass concentrations and SDEs.Opposing trends of K SD × C SD (t) and K WF × C WF (t) can be seen most of the time.Furthermore, a good agreement between the methods used for source apportionment of Saharan dust is visible.Time periods affected by Saharan dust, determined by a negative α SSA , are always accompanied by a temporary increase of the product K SD × C SD (t).When α SSA denotes a period unaffected by Saharan dust this is reflected by an increase of K WF × C WF (t).
The products K SD × C SD (t) and K WF × C WF (t) allow to separate the relative contribution of Saharan dust or particulate matter originating from wildfires to the overall aerosol absorption coefficient.In case of Saharan dust, absorption will be caused by the contribution of iron oxide, while emissions from wildfires contain elevated amounts of black and brown carbon.As the empirical constants K SD and K WF are unknown, it is not straightforward to calculate the concentrations of particulate matter related to either Saharan dust or wildfire emissions.However, assuming that the chemical composition and physical characteristics of Saharan dust transported to the site during the six day period remains rather invariant and the same is valid for particulate matter related to wildfires (what might be the more critical assumption), we can determine an approximation of K SD and K WF and derive a proxy for particulate matter mass related to either of the aerosol sources, i.e., C SD (t) and C WF (t).
Assuming that with C SD and C WF as the respective mass fractions calculated according to using the products K WF × C WF and K SD × C SD obtained from Eq. ( 4).One can write Eq. ( 7) for every single point in time as Note that the closure to particulate matter mass concentration (PM) does not only account for the absorbing species, but the total particulate matter mass related to transport from either the Sahara or wildfires.Solving Eq. ( 10) for the time period between August 3 12:00 and August 9 23:00, the overdetermined linear system of 296 equations yields 0.0317 µm 3 m 2 g -1 for K SD and 0.147 µm 1.3 m 2 g -1 for K WF .Introducing these values in Eqs. ( 8) and ( 9) yields the proxy for C WF (t) and C SD (t).Summarizing these concentrations, the calculated particulate matter mass concentration is obtained and can be compared to the measured mass concentration as given in Fig. 7 as a temporal trend and Fig. 8 as a scatter plot.The calculated mass concentrations reflect the general temporal trend very well.Comparing calculated and measured mass concentrations results deviate up to ± 12 µg m -3 .Considering the 90% interval deviations become much smaller (between -7.4 and +7.7 µg m -3 ).Relating these absolute values to mass concentrations a relative error between -28% and +35% (90% interval) is obtained.It goes up to -44% and + 68% when the maximum differences are considered.The slope between the measured PM and the one reconstructed from the wildfire and Saharan dust contributions is 1.08, indicating a good reconstruction.The deviations get largest during the last two days of the event, when meteorological conditions at the site changed.Obviously, the assumption that the chemical composition relating absorbing species to overall aerosol mass remains constant, is not always applicable.
The contribution of the calculated mass concentrations of Saharan dust to the overall particulate matter mass across the whole time period range from a few percent to more than 80%.In Fig. 9, this share is plotted versus the single scattering albedo exponent α SSA .Despite some scatter a relationship becomes visible, indicating that the value of α SSA can also be used to get an approximation of the relative contribution of Saharan dust to mass concentrations.This is in agreement with the findings mentioned earlier in the text.α SSA gets closer to zero after August 5, even when Saharan dust is indicated by elevated concentrations of coarse particles.When the SSA exponent gets negative the average contribution of Saharan dust to overall particulate matter mass is 36%.

CONCLUSIONS
The analysis of a combined event of long range transport of Saharan dust and particulate matter originating from wildfires allowed us to distinguish the contributions of the two aerosol sources.The determination of time periods affected by Saharan dust based on a negative exponent of the single scattering albedo (Collaud Coen et al., 2004), gave the same results as the decoupling of the influence of Saharan dust and wildfires based on the wavelength dependence of the absorption coefficients (Fialho et al., 2005).Based on the measurements on Sonnblick absorption exponents of 3 and 1.3 were assumed for Saharan dust and wildfires, respectively.
To quantify the mass contributions of either Saharan dust or wildfires a mass closure was performed with the results of the decoupling process, the products of K SD × C SD (t) and K WF × C WF (t) and the particulate mass concentrations determined at Sonnblick.During the time period under investigation the contribution of Saharan dust to overall particulate matter mass was 5% to 80%, the remaining share formed by particulate matter originating from wildfires.Relating the contribution of Saharan dust to the exponent of the single scattering albedo showed a clear relationship, indicating that the value of α SSA can also be used to get an approximation of the relative contribution of Saharan dust to mass concentrations.
Furthermore, the enhancement of particulate matter in relation to the CO mixing ratio could be determined.Regarding time periods when the influence of particulate matter originating from wildfires was dominant this ratio equals 0.28-0.36µg PM/ppb CO.

Fig. 2 .
Fig. 2. Temporal trends of particulate matter mass and number concentrations determined from Aug. 3 until 10 2013.Time periods with a negative exponent of the SSA are denoted as SDEs and are shaded in grey.K × C(WF) and K × C(SD) are the products of the empirical constants K characterizing the instruments and optical properties of particulate matter aerosols related to wildfires or Saharan dust and the concentrations of particulate matter, respectively.

Fig. 4 .
Fig. 4. Correlation of PM concentrations and CO mixing ratios for selected time periods.

Fig. 5 .
Fig. 5. Frequency distributions of the absorption exponents.The selection of non-SDE episodes in (c) is based on the exponent of the single scattering albedo.

Fig. 6 .
Fig. 6.Temporal trends of the deviation between measured and calculated absorption coefficients for the wavelengths 470 nm, 525 nm, 590 nm and 660 nm.

Fig. 7 .
Fig. 7. Temporal trends of measured (PM mass), calculated SD and WF particulate matter mass concentrations as well as calculate total mass (SD + WF mass).

Fig. 9 .
Fig. 9. Correlation of the contribution of calculated mass concentrations of Saharan dust to measured mass concentrations versus the single scattering albedo exponent.

Table 1 .
Mass and number concentrations of particulate matter determined at SBO in August 2013.