Sources of long-lived atmospheric VOCs at the rural boreal forest site , SMEAR II

Introduction Conclusions References

In source areas, VOCs are emitted, for example, from forest fires (de Gouw et al., 2006) or industrial sources or, on the other hand, from densely-populated urban areas: these have a mixture of various source elements such as industry, power plants and vehicles (Baker et al., 2008). The locations of source areas may vary temporally and spatially, due to e.g. seasonal variations in the biogenic activity of plants or variations 15 in anthropogenic activity or meteorological conditions. The SMEAR II (Station for measuring Ecosystem-Atmosphere Relations) site, located in a rural environment in boreal forest in Southern Finland, has been used for two decades to investigate atmospheric processes leading to aerosol particle formation and growth. There is ample evidence that biogenic VOCs contribute to these The observed VOCs may thus have both primary and secondary sources.
The volume mixing ratios (VMRs) of VOCs have earlier been studied at SMEAR II e.g. by Rinne et al. (2005); Ruuskanen et al. (2009); Hakola et al. (2009Hakola et al. ( , 2012 and Hellèn et al. (2004). These studies have only made use of short data sets, with the exception of that by Hakola et al. (2009Hakola et al. ( , 2012, thus not allowing a study of annual to 10 inter-annual variations. Hakola et al. (2012) made continuous measurements using an in situ gas-chromatograph and Hakola et al. (2009) using noon-time samples and laboratory analysis, but in both measured only terpenoids. The analysis of source areas has not been possible without employing larger data sets comprising several compounds.
Thus our aim in this study is to use VOC VMR data covering several years  2011) and to investigate the source profiles and source areas of relatively long-lifetime VOCs (methanol, acetaldehyde, acetone, toluene, benzene) observed at the SMEAR II site. Winters are characterized by stronger anthropogenic influence e.g. from heating whereas biogenic activity is more pronounced in the summer. The specific aims of this study are (1) to investigate long term changes in sources affecting the VOC concentra- 20 tions and to quantify their trends (biogenic and anthropogenic) over a six-year period, (2) to determine the biogenic vs. the anthropogenic influence from defining the source profiles of VOCs in relation to other trace gases for SMEAR II, (3) to identify the source areas of VOCs in air masses arriving from Northern Fennoscandia, Northern Continental Europe and Western Russia within the lifetime of the long-lived VOCs, such as 25 acetone and benzene, observed at SMEAR II, (4) to investigate how these sources coincide with e.g. wildfires and biomass burning, and major urban and industrial areas.
The SMEAR II's focus is atmosphere biosphere interaction and define aerosol formation and growth processes in boreal climate zone. While many studies have focused on Introduction the influence of local to regional sources on the observed trace gases concentrations (Patokoski et al., 2014;Liao et al., 2011;Eerdekens et al., 2009;Hakola et al., 2009Hakola et al., , 2012, this study aims at identifying regional to continental source areas and focuses on characterizing the effect of long range transport.  Hari and Kulmala (2005). The site is located 220 km north-west of 10 Helsinki and 60 km north-east of Tampere which with a population of about 200 000 is the largest city near the site. Continuous long-term measurements of trace gases, aerosol particles and gas exchange between the atmosphere and the biosphere have been carried out at SMEAR II since the mid-1990's (Vesala et al., 1998). The forest surrounding the station is dominated by Scots pine (Pinus sylvestris), sown in 1962 15 (Bäck et al., 2012). There is also some Norway spruce (Picea abies), aspen (Populus tremula) and birch (Betula sp.) at the site (Hari and Kulmala, 2005). Within a square of 40 km × 40 km centred on the station 23 % of the area is covered by pine forests, 26 % by spruce forest, and 21 % by mixed forest (Haapanala et al., 2007). Agriculture and water bodies cover 10 and 13 %, respectively. 20

Instrumentation and sampling
The VOC VMRs were measured with a quadrupole proton transfer reaction mass spectrometer (PTR-MS, Ionicon Analytik GmbH, Austria, Lindinger et al., 1998a). PTR-MS uses the hydronium ion (H 3 O + ) as a primary reactant ion. VOCs with a larger proton affinity than that of water will readily react with H 3 O + ( During the years 2006-2009 the VMRs of the VOCs were measured from a scaffolding tower at a height of 22 m above the ground. From summer 2006 to spring 2007 the measurement sequence consisted of one-hour VMR measurements followed by one-hour disjunct eddy covariance measurements (Rinne et al., 2007). Thus VMR data were obtained every second hour. The sampling protocol changed in March 2007 when 5 a chamber with a Scots pine shoot enclosed was included in the measurement cycle; VOC VMRs were then measured every third hour instead of every second hour. In May 2010 the sampling protocol changed again when the instrument was transported to another measurement hut. At the same time the sampling inlet was moved about 50 m to another tower, 33.6 m above ground. These measurement heights were cho-10 sen for analysis because they are more representative for depicting concentrations due to transport, rather than concentrations inside the canopy. The canopy height was about 16 m.
Nitrogen oxides (NO x ), carbon monoxide (CO), sulphur dioxide (SO 2 ) and ozone (O 3 ) mixing ratio data were used in the analysis as ancillary data. The mixing ratios 15 of NO x were measured with a chemiluminescence technique (TEI 42C TL, Thermo Environmental Instruments, MA, USA) and CO was measured by an infrared light absorption analyzer (HORIBA APMA 360, Horiba, Japan). SO 2 was measured with a fluorescence analyzer (TEI 43 BS, Thermo Environmental Instruments, MA, USA) and O 3 by an ultraviolet light absorption technique (TEI 49, ments, MA, USA). CO, NO x , O 3 and SO 2 were also measured at a height of 33 m except in 2010, when CO was measured at 16.8 m.

Trajectory analysis
HYSPLIT 4 (HYbrid Single Particle Lagrangian Intergrated Trajectory) was used for air mass trajectories (Draxler and Hess, 1998 of the trajectory analysis the VOC VMR data were interpolated using a Piecewise Cubic Hermite Interpolating Polynomial (PCHIP) to cover every hour and thus match the trajectory data. Each time measured VOC VMR data were available at SMEAR II, they were associated with a trajectory arriving at the site at the same time. The path of the back trajectories was considered with a 1 • × 1 • spatial resolution. The VMRs were as-5 sumed to remain constant during the whole transport time. The grid cells over which trajectories traversed prior to the observation of high VMR values at SMEAR II were associated with high values in the source field. Thus the trajectory maps are interpreted so that there is a color code corresponding to VMRs e.g. if there is an area in a map where methanol is observed to be 3 ppb v this means that the measured value 3 ppb v 10 at the site is assumed to come from that area (see . The trajectory analysis was limited to the area between 50 and 75 • N in latitude; 12 and 50 • E in longitude. For reasons of statistical significance, at least 25 trajectories had to cross a grid cell in order for that grid cell to be accepted into the analysis, i.e., grid cells with less than 25 traverses were omitted from the analysis. Finally, all the VOC VMRs at each grid 15 cell were averaged to yield the VOC source field (Stohl et al., 1995;Stohl and Seibert, 1998).

Forest fire locations from satellite observations
The forest fire location data are obtained from FIRMS (Fire Information Resource Managements System), which delivers fire locations and hotspots as globally observed by 20 MODIS (Moderate Resolution Imaging Spectroradiometer). Data have been collected by NASA's Earth observing system (EOS) Terra and Aqua satellites. With these two satellites global data coverage is achieved every 1 to 2 days. sity of Southern California. A basic problem of multivariate receptor models is how to determine the optimal number of sources, the source fingerprints and their contribution from the ambient air VOC measurement data alone. Some additional constraints must be added in order to obtain unique solutions. In Unmix, the composition and contribution of the sources must be non-negative. In addition to this, Unmix searches for 5 periods when the data indicate that the contribution of one of the sources is missing completely or its contribution is minor. The application of Unmix to VOC data obtained by gas chromatographic methods at an urban site in Helsinki is described by Hellén et al. (2003). According to the recommendations for the model, the regression of each of the species explained by the sources (R 2 ) should be over 0.8, while the signal-to-10 noise ratio should be over 2. In this study, Unmix was applied for inorganic trace gas and VOC data, except monoterpenes that were excluded from the source area trajectory analysis (see Sect. 3.5). One-hour medians of trace gases and VOCs were used as input data. Data were filtered by horizontal wind speed, excluding from analysis observations (30 % of data were excluded from analysis) with wind speed below 1 m s −1 .

15
All the results exceeded the recommended R 2 and signal-to-noise values, indicating receptor modelling results that were applicable and valid.

Results and discussion
The VOC VMRs of most compounds which were studied have maxima in summertime and minima in winter (Fig. 1). However, benzene behaves in the opposite way. This is 20 due to a lack of significant biogenic sources of benzene and its shorter atmospheric lifetime in summer. Of the other studied VOCs, methanol, acetone, acetaldehyde and the monoterpenes also have biogenic sources around the measurement site (Rinne et al., 2007), and acetonitrile is emitted by biomass burning. Introduction

Lifetimes of the observed VOCs
To estimate the VOC chemical lifetimes (e-folding times), OH and NO 3 radical concentrations, were estimated, based on data from Hakola et al. (2003). The summertime OH concentration presented by Hakola et al. (2003) agreed well with the observations (Rinne et al., 2012), except in the case of summertime NO 3 , whose radical concentration is at least twice as high as the observations. Thus, for this study, the annual cycle of NO 3 as presented by Hakola et al. (2003) was scaled by dividing it by two. Summerand wintertime median O 3 concentrations were calculated from O 3 measurements at SMEAR II. Photolysis values for summer-and wintertime were calculated following Hellén et al. (2004). Actinic flux values corresponding to the albedo for snow-covered 10 forest in winter (α = 0.8) were used when photolysis values were calculated. The concentrations of oxidants and reaction rate coefficients are presented in Tables 2 and  3.
The calculated atmospheric lifetimes of the studied VOCs (methanol, acetaldehyde, acetone, benzene and toluene) in summer-and wintertime are presented in Table 4.

15
Compared to the lifetimes of monoterpenes (about one day in winter and one hour in summer) these lifetimes were much longer. For most compounds, the atmospheric lifetimes exceed the duration of the back-trajectories used in this analysis. However, in summertime both toluene and acetaldehyde have a lifetime below four days. Thus, for these compounds, the results of the four-day backward trajectory analysis should be 20 interpreted with caution.
The following sections describe the source areas of the studied VOCs: (1) during long-range transport episodes from forest fires, (2) for all VOC VMR data and (3) for summers and winters separately.

25
During the measurement periods two particularly active forest fire episodes with several fire hotspots occurred in Russia, one  They also influenced air quality in Finland (Leino et al., 2014). These fires provide temporally and spatially well-defined sources that can be used to evaluate the ability of the trajectory analysis to identify the source areas. Biomass burning has previously been observed to be a source of VOCs in several field and laboratory studies (e.g. Crutzen and Andreae, 1990;Holzinger et al., 1999;5 de Gouw et al., 2006;Virkkula et al., 2014). Acetonitrile has commonly been used as a marker compound for emissions from biomass burning (de Gouw et al., 2003(de Gouw et al., , 2006Holzinger et al., 1999). The oxygenated VOCs (OVOCs) and aromatic VOCs (benzene and toluene) have also been linked to biomass burning in different studies (Koppmann and Wildt, 2007;de Gouw et al., 2006). Forest fires affect air quality; and the biggest 10 smoke plumes can be seen in satellite images and even reduce visibility in the plume areas.
Forest fires, which were observed during these measurements, occurred In summer 2006 the largest fires were south of Moscow and in Belarus. There were also fires in Karelia, in Vyborg area near St. Petersburg and the Finnish-Russian border. The first forest fire episode occurring within the time-frame of this study was in the 25 period 4-31 August 2006. The trajectory analysis shows the high VOC mixing ratios observed at SMEAR II during this period to have originated from that area (Fig. 3) Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | PM 2.5 ) and higher polycyclic aromatic hydrocarbon (PAH) concentrations at Virolahti, located on the Finnish south coast near the Finnish-Russian border. During summer 2010 a large number of forest fires were located in Northwest Russia. The second forest fire episode within the time frame of this study was in 20 July-31 August 2010. During this period acetonitrile data too was being measured at SMEAR II.

5
Comparing the forest fire locations (Fig. 2) and acetonitrile's mean VMR field from trajectory analysis, one can see that acetonitrile is originating from the general direction of the maximum burning area (Fig. 4). All the other VOCs and trace gases studied also had similar source area distributions to acetonitrile.
For comparison between fire and non-fire VOC VMRs, the mean VMR values of 10 methanol, acetonitrile, benzene, SO 2 and CO were calculated for periods both before and during the forest fire episodes in both years ( processes rather than biomass burning (Seinfeld and Pandis, 1998). In this study, too, the SO 2 concentration in 2006 was actually lower during the forest fire episode than before it; in 2010 it was only slightly elevated during the episode.

Differences in source areas for the whole measurement period
The mean trajectory fields of the VOCs studied (methanol, acetonitrile, acetaldehyde, 5 acetone, benzene and toluene), as well as CO, NO x and SO 2 from the year 2006 to the year 2011 are presented in Fig. 5. From this figure it can be seen that methanol, acetone and acetaldehyde had very similar source areas. There was very good correlation between methanol and acetone (r = 0.86, p = 0). Methanol and acetaldehyde were also correlated to each other (r = 0.66, p = 0). On the other hand, the source areas 10 of benzene, toluene, CO and NO x were also similar to each other and correlated well. Acetonitrile had similarities with both methanol and benzene, correlating, however, better with the methanol group than with the benzene group. The correlation matrix and Pearson's correlation coefficients between all compounds are presented in Fig. A1 (2013) found a similar sharp distinction in SO 2 concentrations between air masses arriving from Eastern and Western Europe. They also found that air masses coming from Central Europe were exposed to more rain and thus wet removal of SO 2 . They speculate that this may be one reason why SO 2 source areas in Central Europe were not separable in the tra-5 jectory fields. However, in this study we found that water-soluble methanol and also NO x were not totally washed away, as there were visible source areas for methanol in the North Sea, Skagerrak and the northern Germany areas, and for NO x in Northern Continental Europe (Fig. 5). We propose that wet deposition does not fully explain the absence of SO 2 in the air masses arriving from Western Europe, as interpreted by Riuttanen et al. (2013), but that the observed difference in SO 2 is probably due to more rigorous emission regulations in the western part of Europe than the eastern part. This interpretation is supported by Vestreng et al. (2007)

Source areas and their seasonal difference
In order to study the possible seasonal changes in the VOC source areas, these were determined separately for the summer (June-August) and the winter (December-February) periods. Data of the short forest fire episodes were removed prior to this analysis so that they would not mask other source areas. Although there was some 25 inter-annual variation in the observed VMRs of VOCs, no clear trends of VMRs were observed during the whole measurement period (Fig. 1) Table 6. These trends are all slightly negative except for that of monoterpenes that seemed to be slightly positive. For monoterpenes the VMR change was 8 % per year for the summer monthly medians. However, one should interpret these trends with care, because they were calculated based on measured summer monthly medians whose trends were not, in fact, statistically significant (i.e. the confidence intervals included zero). Summers were different from each other, e.g. the median temperature between summers in Finland at Hyytiälä were observed to vary from 12.9 to 17 • C, which may have an effect on the emissions of VOCs e.g. methanol, acetone and monoterpenes. Additionally, in the case of monoterpenes' VMRs the change of the sampling location may bias the trend calculated over the whole observation period, as 10 monoterpenes have a short life time (1 h; Table 4). The source areas of the VOCs also varied slightly between years due to variations in the VOC VMRs, deposition and the prevailing paths of arriving air masses. Despite these differences, all five summers (2006)(2007)(2008)(2010)(2011) and two winters (2006-2007 and 2008-2009) were combined in this study, to get as good an areal trajectory 15 data coverage as possible for summers and winters separately. Summer, 2009 was not included in the trajectory analysis because VOC VMR data was then only available for 18 days.
For the evaluation of the VMR source areas, ten rectangular areas were selected for separate analysis of the trajectory fields of both main seasons (summer, winter) (Fig. 6). 20 These areas differ from each other in having e.g. different industry and population densities.
The ten selected areas were as follows: (1)  This interpolated VOC VMR trajectory field is called the background field in the following discussion. By subtracting the latter from the momentary field a differential source field was obtained, with the seasonal trend removed from the source area analysis.
The average values of the differential source field for the ten areas listed above are presented in Fig. 7. Acetonitrile was included in the source area analysis in summer 5 only, when forest fires occurred. In urban areas, VOCs mostly originate from traffic, but are also evaporated from fuels and combustion processes (Reimann and Lewis, 2007;Hellén et al., 2006) and from various industrial processes. The ten areas investigated here were mainly located in similar areas having a lot of industry and/or dense population. The strongest source 10 areas for all the VOCs studied were located in Eastern Europe including Western Russia, Northern Poland, Kaliningrad and the Baltic countries, Karelia and the White Sea. In these areas, calculated mean VMR values differed significantly from values based on monthly medians (Fig. 7). In addition to these source areas, which are common to most of the studied compounds, certain compounds have specific source areas of their 15 own. Methanol is known to be a very abundant VOC in the atmosphere having many different sources, both biogenic and anthropogenic (Jacob et al., 2005). From Fig. 7, it can be seen that nearly all of the selected areas are sources of methanol. Eastern Europe was observed to be a large emitter of OVOCs (methanol, acetaldehyde and acetone) in general. This is in line with earlier observations by Hellén et al. (2004), who 20 reported Eastern Europe to be an important emitter of carbonyls. In addition to these source areas, acetone and acetaldehyde also arrived from the areas of Stockholm, the Skagerrak, the North Sea and the coastal areas and northern Germany: all these areas have traffic emissions and solvent use related to various different industries. The oxidation of hydrocarbons and the primary biogenic emission of acetaldehyde are known 25 to be the major global source of acetaldehyde (Singh et al., 2004); these sources have not, however, been taken to account into this study. Acetaldehyde also has relatively short lifetime (2 days) during summer. These factors can add some uncertainty to this analysis of acetaldehyde. Benzene was also found to have sources in the Kola Penin- sula area, connected with the petrochemical industry and mining. However, the North Sea area with its active petrochemical industry did not appear as a distinguishable source area for aromatic VOCs in this study. As seen earlier, when comparing the VMRs of VOCs before and during forest fires, the mixing ratios at SMEAR II for benzene and acetonitrile were found to be elevated 5 during these episodes (Table 5). In Fig. 7, forest fires seem to be sources of all VOCs, and especially for acetonitrile. Acetonitrile's VMR difference from the background field values was in the year 2010 0.07 ppb v . In other source areas, acetonitrile's VMR difference from values based on the background field was indistinguishable (Fig. 7).
The strongest source areas in the trajectory maps seemed to be of anthropogenic 10 origin, and eight out of ten were located in an easterly or southerly direction from Finland. However, there is a vast boreal forest zone in Northern Europe that is an important emitter of biogenic VOCs (BVOCs). Most BVOCS emitted from boreal forests are short-lived terpenoids, whose high atmospheric reactivity keeps their concentrations relatively low  compared to those of e.g. OVOCs, with longer 15 atmospheric lifetimes. In this study, the forest regions were not identifiable as welldefined source areas, but probably contributed to background levels. During summer there were minor source areas in the Baltic Sea, where there should be no anthropogenic sources. These interesting source areas could be producing VOCs from e.g. algae or cyanobacteria. However, the VMR levels of VOCs originating from algae are 20 low (Kansal et al., 2009) compared to anthropogenic sources. With the current data and analysis it is not possible to identify the source of these marine emissions. In the future, their origin could be clarified by using shorter trajectories and making measurements near the Baltic Sea or by collecting samples from over the Baltic Sea. direction at SMEAR II. The division was made based on the findings of the VOC source areas described above. The sectors were: (1) North (0-5 and 300-360 • ), Urbanized continental (5-210 • ) and Urban and sea (210-300 • ). In all sectors, three distinctive sources were identified: (1) A source containing mainly SO 2 ; this was named the SO 2 source, (2) A source containing toluene, benzene, NO x and CO. These compounds are 5 typical of anthropogenic emissions, and thus the source was named the anthropogenic source, (3) A source containing oxygenated VOCs (OVOCs, methanol, acetone and acetaldehyde), acetonitrile and a portion of CO. This source was related to biomass burning and other biogenic emissions, and was named to biogenic/combustion source. The mean contributions of all these sources in the different sectors were similar 10 in both seasons. The contribution of the anthropogenic source was dominant in winter and the biogenic/combustion source in summer (Fig. 8). The dominance of the biogenic/combustion source in summer can be attributed to two processes. First, the biogenic/combustion source included acetonitrile, and high mixing ratios of acetonitrile were observed during forest fire episodes in summer; secondly, this source in-15 cluded OVOCs which also have biogenic sources and thus higher VMRs during summer (Fig. 1). Biogenic emissions are dominant at SMEAR II in summer. The monoterpenes measured at SMEAR II are mostly emitted from biogenic sources. However, from the time series of monoterpenes it can be seen that there are occasionally notably high VMR peaks, which are known to have an anthropogenic origin (Liao et al., 20 2011) (Fig. 1). Monoterpenes were not included in the trajectory and Unmix analysis because they differ from the other VOCs in this study in having a short lifetime from one hour to several hours and mainly local sources. A comparison of diurnal cycles of the anthropogenic source and monoterpenes is presented in Fig. 9. Data in Fig. 9 are from the urbanized continental sector. Monoterpenes had a considerable diurnal 25 variation during summer, with higher mixing ratios at night, and no variation in winter, as also observed in previous studies in similar ecosystems (Hakola et al., 2000;Rinne et al., 2005). This is due to the diurnal cycle in surface layer mixing and the night-time emissions of monoterpenes from coniferous trees. summertime lifetimes as compared to winter, leading to lower anthropogenic source levels in summer. The contribution of the anthropogenic source in winter was about three times higher in all sectors than in summer. Both summertime and wintertime diurnal cycles of the anthropogenic source show a maximum at night, possibly due to lower night-time mixing in the boundary layer.

Low concentrations from the north, urban influences from continents and
Histograms of the source contributions, together with their mean, median and maximum values both in summer and in winter, are presented in Figs. 10 and 11. Many of these distributions are skewed, having a tail of high contribution values. Thus the mean and median values of these source contributions may have large differences. The skewness of the source distributions also indicates that the simplest statistical pa-10 rameters, such as mean and median, may not adequately describe the distribution of the sources or their contribution to the local atmospheric mixing ratios of these compounds. There were considerable differences between the source distributions from different wind direction sectors. Air masses arriving from the North sector had in general lower source contributions than air masses from the Urbanized continental and 15 the Urban and sea sectors. Particularly in summer there was a tail of high contributions in the Urbanized continental and the Urban and sea sectors for anthropogenic (maximum values were 17.2/4.7) and biogenic/combustion sources (9.8/6.1), as compared to the North sector whose maximum contributions were 1.6 and 4.9 for the anthropogenic and biogenic/combustion sources, respectively. Hence air masses from the 20 north were clearly less polluted with the trace gases studied as compared to the two other sectors. These results combined with the earlier observations in this paper support the conclusion that air masses related to the highest VMRs of long-lived VOCs observed at SMEAR II have their origin in Russia and the Eastern European countries, the Northern part of Continental Europe and Southern and Central Fennoscandia. 15,2015 Sources of long-lived atmospheric VOCs at the rural boreal forest site, SMEAR II

Conclusions
This study has focused on identifying the source areas of the long-lived VOCs (methanol, acetonitrile, acetaldehyde, acetone, benzene) measured at the SMEAR II site in southern Finland, and to investigate the relative influences of biogenic and anthropogenic compounds arriving in Southern Finland from areas outside the country.

5
The analysis is based on a data set which consisted of several years (2006-2011) of VMR measurements of OVOCs, acetonitrile, aromatic VOCs and monoterpenes. Annual trends of VOC VMRs based on summer monthly medians were presented for the measurement period. The trend of VMRs for monoterpenes was slightly positive. All the other VOCs had a small negative trend in their VMRs. Trend calculations showed, 10 however, that none of these trends could be considered as significant. The origin and sources of the VOCs observed were analyzed by trajectory model and a multivariate receptor model. During the measurement period, forest fire episodes occurred in Eastern Europe and Russia. Elevated VMR levels for several VOCs and other trace gases were observed 15 in air masses arriving from areas in which abundant fire counts were observed. This corroborates the applicability of the trajectory analysis as a method for identifying the source areas of these trace gases.
Three sources (labelled SO 2 , biogenic/combustion and anthropogenic) were separated by receptor analysis both in winter and summer. The biogenic/combustion source 20 dominated in summer and the anthropogenic source in winter. Both the trajectory and Unmix analyses showed that air masses coming from a northerly direction were less polluted with the trace gases studied than the air-masses arriving from easterly and westerly directions.