Long-term measurements (2010–2014) of carbonaceous aerosol and carbon monoxide at the Zotino Tall Tower Observatory (ZOTTO) in central Siberia

We present long-term (5-year) measurements of particulate matter with an upper diameter limit of ∼ 10 μm (PM10), elemental carbon (EC), organic carbon (OC), and water-soluble organic carbon (WSOC) in aerosol filter samples collected at the Zotino Tall Tower Observatory in the middle-taiga subzone (Siberia). The data are complemented with carbon monoxide (CO) measurements. Air mass back trajectory analysis and satellite image analysis were used to characterise potential source regions and the transport pathway of haze plumes. Polluted and background periods were selected using a non-parametric statistical approach and analysed separately. In addition, near-pristine air masses were selected based on their EC concentrations being below the detection limit of our thermal–optical instrument. Over the entire sampling campaign, 75 and 48 % of air masses in winter and in summer, respectively, and 42 % in spring and fall are classified as polluted. The observed background concentrations of CO and EC showed a sine-like behaviour with a period of 365± 4 days, mostly due to different degrees of dilution and the removal of polluted air masses arriving at the Zotino Tall Tower Observatory (ZOTTO) from remote sources. Our analysis of the near-pristine conditions shows that the longest periods with clean air masses were observed in summer, with a frequency of 17 %, while in wintertime only 1 % can be classified as a clean. Against a background of low concentrations of CO, EC, and OC in the near-pristine summertime, it was possible to identify pollution plumes that most likely came from crude-oil production sites located in the oil-rich regions of Western Siberia. Overall, our analysis indicates that most of the time the Siberian region is impacted by atmospheric pollution arising from biomass burning and anthropogenic emissions. A relatively clean atmosphere can be observed mainly in summer, when polluted species are removed by precipitation and the aerosol burden returns to near-pristine conditions. Published by Copernicus Publications on behalf of the European Geosciences Union. 14366 E. F. Mikhailov et al.: Long-term measurements (2010–2014) of carbonaceous aerosol

0.7 ng m −3 ppb −1 , respectively, suggesting that the contribution of coal and other fossil fuel burning for heating was dominant. In summertime, pollution plumes arrived at the ZOTTO site from nearby large-scale boreal wildfires, which were observed during the three years from 2011 to 2013. As a result, the seasonal concentrations of CO, PM10, and OC were as high as 670 ± 710 ppb, 59 ± 53 µg m -3 , and 26 ± 27 µg m -3 , respectively, with ∆OC/∆EC of 26.2 ± 0.1 and 5 ∆EC/∆CO of 1.3 ± 0.1 ng m -3 ppb -1 . Agricultural fires from the steppe zone of southern Siberia and northern Kazakhstan also accounted for elevated concentrations of CO and carbonaceous species. For one extreme pollution episode observed on 28 April 2010 the CO, PM10, EC, and OC concentrations were as high as 261 ± 12 ppb, 54.4 ± 3.7, 1.5 ± 0.3, and 18.9 ± 1.2 µg m -3 , respectively, with ∆OC/∆EC = 12.7 ± 2.7 and ∆EC/∆CO = 14.3 ± 4.4 ng m -3 ppb -1 . The observed 10 background concentrations of CO and EC showed a sine-like behavior with a period of 365 ± 4 days, with maximum values in winter of 151 ± 20 ppb and 0.08 ± 0.03 µg m -3 and minimum values in summer of 114 ± 15 ppb and 0.03 ± 0.02 µg m -3 , respectively. The observed background concentrations are mostly due to different degrees of dilution and removal of polluted air masses arriving at ZOTTO from remote sources. Our analysis of the near-pristine conditions shows that 15 the longest periods with clean air masses were observed in summer, with a frequency of 17%, while in wintertime only 1% can be classified as a clean. In summer, variations in the OC/PM ratio during clean periods closely correlated with those in air temperature, which indicates that biogenic sources of OC formation were dominating. Against a background of low concentrations of CO, EC, and OC in the near-pristine summertime it was possible to identify pollution plumes 20 that most likely came from crude oil production sites located in the oil-rich regions of Western Siberia. Overall, our analysis indicates that most of the time the Siberian region is impacted by atmospheric pollution arising from biomass burning and anthropogenic emissions. A relatively clean atmosphere can be observed mainly in summer, when polluted species are removed by precipitation and the aerosol burden returns to near-pristine conditions. 25

Introduction
The Siberian forests cover about 70% of the total area of the Eurasian boreal forest and are an important factor controlling global and regional climate. In turn, climate change causes a response of Siberian ecosystems, which shows up in a redistribution of matter and energy between 30 terrestrial ecosystems and the atmosphere (Goetz et al., 2007;Lappalainen et al., 2016). The atmospheric aerosol over Siberia is of particular interest for several reasons. Firstly, biogenic emissions of volatile organic compounds (VOC) from the vast boreal taiga forest are thought to lead to the formation of secondary organic aerosol (SOA) (Tunved et al., 2006). Secondly, Sibe-Atmos. Chem. Phys. Discuss., doi:10.5194/acp-2017-409, 2017 Manuscript under review for journal Atmos. Chem. Phys. Discussion started: 5 May 2017 c Author(s) 2017. CC-BY 3.0 License. ria has been documented to be an important source region of biomass-burning aerosol particles that are distributed around the globe in the free troposphere (Conard and Ivanova, 1997;Müller et al., 2005;Warneke et al., 2009). Thirdly, Siberia is one of the few possible background regions in the Northern Hemisphere where near-pristine conditions prevail for certain periods of the year (Chi et al., 2013). Such atmospheric observations in remote areas are very important for 5 providing a reference for evaluating anthropogenic impacts in this and other regions (Andreae, 2007;Carslaw et al., 2013;Spracklen and Rap, 2013).
Despite the relatively remote location and low population density, human impacts on Siberian ecosystems are increasingly noticeable. This includes expansion of agriculture at the southern end of the boreal forest zone, forest logging, as well as impacts on steppe and forest fire regimes 10 (Chi et al., 2013;Vasileva et al., 2011;Heintzenberg et al., 2013;Mikhailov et al., 2015a, Panov et al., 2015. In addition, the massive expansion of oil and gas wells, predominantly in northwestern Siberia, is likely to have an impact on aerosol and trace gas emissions through gas flaring and potential leaks in the mining and transport infrastructure (Stohl et al., 2013;Heimann et al., 2014). 15 Typically, two classes of carbonaceous aerosol are commonly present in ambient air -elemental carbon (EC) (often referred as black carbon or soot; Andreae and Gelencsér, 2006) and organic carbon (OC). Carbon-containing components (EC and OC) account for 10% to 70% of atmospheric PM mass (Turpin et al., 2000;Zhang et al., 2007). Both OC and EC are important agents in the climate system, which affect the optical characteristics and thermal balance of the 20 atmosphere both directly, by absorbing and scattering incoming solar radiation, and indirectly, by modifying cloud properties (Haywood and Boucher, 2000;Andreae and Merlet;2001;Pierce et al., 2007, Bond et al., 2013Andreae and Ramanathan, 2013;Alonso-Blanco et al., 2014). EC is the major light-absorbing component of atmospheric aerosol and is the second most important global warming agent after carbon dioxide (Bond et al., 2013). EC aerosols are primary sub- 25 micron particulate matter emitted during incomplete combustion of fossil or bio-fuels. Aged EC aerosols have fractal-like or globular structure and a wide variety of organic and inorganic species adsorbed to them. OC is mainly a light scattering component, however is also known to have light-absorbing fraction called "brown carbon" (BrC) that, similar to BC, has been shown to be an important factor in aerosol radiative forcing (Andreae and Gelencsér, 2006;Andreae and 30 Ramanathan, 2013; Saleh et al., 2014). In addition, due to the lensing effect, organic coatings can substantially modify the optical properties of EC in mixed particles. Thus, laboratory experiments have shown that the light absorption enhancement factor for internal mixtures of EC with Atmos. Chem. Phys. Discuss., doi:10.5194/acp-2017-409, 2017 Manuscript under review for journal Atmos. Chem. Phys. Discussion started: 5 May 2017 c Author(s) 2017. CC-BY 3.0 License. organic and inorganic coatings can range from 1.3 to 3.5 (Schnaiter et al., 2005;Mikhailov et al., 2006;Zhang et al., 2008;Shiraiwa et al., 2010;Pokhrel et al., 2017).
Forest fires and the biogenic activity of coniferous trees and forest litter are the main sources of carbonaceous aerosols emitted into the atmosphere over boreal forests. The molecular composition of OC from biomass burning is still poor understood. It is known only that the 5 thermal decomposition products of plant cellulose and lignin provide the largest contribution to the organic mass. Among them are sugar anhydrides, phenol compounds, alcohols, esters, aldehydes and ketones (Kanakidou et al., 2005;Hoffer et al., 2006;Lappalainen et al., 2009;Mayol-Bracero et al., 2002;Fuzzi et al., 2007). In the spring and summer seasons, biogenic activity increases and organic compounds, mainly isoprene, terpenes, and sesquiterpenes, are emitted to the 10 atmosphere. These compounds form secondary organic aerosols as a result of photochemical oxidation and gas-to-particle conversion (Hallquist et al., 2009). A substantial fraction of boreal forest organic aerosol consists of water-soluble compounds and therefore contributes to hygroscopic growth and cloud condensation nuclei (CCN) activity (Huff Hartz et al., 2005;Ehn et al., 2007;Carrico et al., 2008;Cerully et al., 2011;Jaatinen et al., 2014;Paramonov et al., 2013;Mi-15 khailov et al., 2015b).
Primary biological aerosols (PBA) and their rupture products are another subset of organic particles, which are directly released from the biosphere into the atmosphere in the growing season. They comprise living and dead organisms (e.g., algae, archaea, bacteria), dispersal units (e.g., fungal spores and plant pollen), and various fragments or excretions (e.g., plant debris, 20 brochosomes, and salt particles) (Despres et al., 2012;Fröhlich-Nowoisky et al., 2016). Their particle sizes range from nanometers up to about a tenth of a millimeter. They can serve as nuclei for cloud droplets, ice crystals, and precipitation, thus influencing the hydrological cycle and climate. In pristine air over vegetated regions, bioaerosols are likely to be an essential regulating factor in the formation of precipitation (Pöschl et al., 2010;Pummer et al., 2012Pummer et al., , 2015Steiner et 25 al., 2015). In the atmosphere, bioparticles undergo internal and external mixing with other aerosols, including secondary organic aerosol (SOA) which can influence bioaerosol properties through SOA coatings (Hallquist et al., 2009;Pöschl et al., 2010;Huffman et al., 2012;Pöhlker et al., 2012).
In the northern latitudes of Eurasia, climate change occurs 1.5-2 times faster than the glob- 30 al average (Hansen et al., 2006;Groisman and Soja, 2009;IPCC, 2014). One consequence of these changes is an increase of volatile organic compounds emission, which is a precursor of secondary aerosol formation (Tunved et al., 2006). It is expected that the increase in the concentration of biogenic aerosols will provide negative climate feedbacks, involving aerosol-cloud and Atmos. Chem. Phys. Discuss., doi:10.5194/acp-2017-409, 2017 Manuscript under review for journal Atmos. Chem. Phys. Discussion started: 5 May 2017 c Author(s) 2017. CC-BY 3.0 License.
aerosol-carbon cycle interactions (Kulmala et al., 2004(Kulmala et al., , 2014Andreae et al., 2008). These feedbacks may change the local structure of the atmospheric circulation, cloud properties, and intensify of precipitation (Rosenfeld et al., 2013;Lappalainen et al., 2016). Precipitation, in turn, causes emissions bursts into the atmosphere of primary bioaerosols and submicron degradation products containing hygroscopic water-soluble inorganic ions (potassium, sodium, chlorides, 5 phosphates) and polysaccharides (hexoses, mannitol) (Morris et al., 2014;Bigg et al., 2015;Huffman et al., 2013). As temperatures increase, population outbreaks of tree-damaging insects can occur more frequently. The emissions of organic substances from damaged trees are significantly higher than from healthy trees (Bergström et al., 2014). For example, in the boreal environment, trees damaged by the pine borer (Neodiprion sertifer) emit 11 times more monoter-10 penes and 20 times more sesquiterpenes compared to healthy trees. As a result, the total mass of aerosols increased by 480% (local maximum), and the concentration of cloud condensation nuclei increased by 45% (Joutsensaari, 2015). It is also projected that increased growth of the forest area resulting from the surface temperature increase will be accompanied by more frequent forest fires (Shvidenko et al., 2011), which are powerful sources of aerosol particles and greenhouse 15 gases (Paris et al., 2009;Janhäll et al., 2010;Smolyakov et al., 2014). In addition, industrial production and steady increase in oil and gas production in Siberia will further increase the concentrations of aerosols in the Siberian air basin. It is therefore expected that the role of biogenic, pyrogenic, and anthropogenic aerosol emissions will grow with increasing temperatures and their influence on climate change will be important both on the regional and global level (Kulmala et 20 al., 2014;Lappalainen et al., 2016). In order to assess the magnitude and sign of these climatic effects, as well as to predict the possible consequences for Siberian ecosystems, it is necessary to provide a comprehensive long-term monitoring of the burden and composition of atmospheric aerosols in the region.
In 2006, the 300-m tower of the Zotino Tall Tower Observatory (ZOTTO) was established 25 in Central Siberia (Heimann et al., 2014). The background character and the geographical location of this station are appropriate conditions for studying atmospheric transport and coincident chemical transformation of polluted air at a wide range of spatial and temporal scales, particularly for assessing the potential influence of emissions from various natural and anthropogenic sources on surface air composition over the large territory of Siberia. The remote location of 30 ZOTTO in the middle of the Siberian taiga forest also makes it highly suitable for investigating the exchange of trace gases with this ecosystem and the production of aerosol by the boreal forest. Continuous measurements of comprehensive sets of atmospheric constituents in the gas and particle phase together with meteorological parameters have been carried out at ZOTTO since Atmos. Chem. Phys. Discuss., doi:10.5194/acp-2017-409, 2017 Manuscript under review for journal Atmos. Chem. Phys. Discussion started: 5 May 2017 c Author(s) 2017. CC-BY 3.0 License. (Kozlova et al., 2008Heintzenberg et al., 2008). With regards to aerosol, a study of the representativeness of the ZOTTO facility and first analyses of the particle size distribution data can be found in Heintzenberg et al. (2008) and Heintzenberg and Birmili (2010); a statistical analysis of particle size distribution, particle absorption, and carbon monoxide (CO) data taken from the first four years of operation of the ZOTTO facility (September 2006 to January 2010) is 5 given in Heintzenberg et al. (2011), together with seasonally dependent major air mass pathways and the related particle size distributions. An extended statistical analysis of aerosol properties including scattering coefficients, Ångström exponents, single scattering albedo, and backscattering ratios, as well as a CO data set with seasonal, weekly, and diurnal variations between September 2006 and December 2011can be found in Chi et al. (2013). 10 In 2010, a filter-based sampler was mounted at the ZOTTO station for aerosol chemical analysis. In this study, we present the time series of carbonaceous aerosol measurements coupled with CO data for 5 years (2010 -2014). We investigate the seasonal variations of EC, OC, WSOC (water soluble organic carbon), and CO. We analyze polluted, background, and nearpristine periods as well as the most pronounced pollution events and their sources observed over 15 the entire sampling campaign. The methodology is described in Sect. 2, and the seasonal features of temporal variations and air mass origins during pollution periods are discussed in Sect. 3.4 and 3.5. The background and near-pristine air masses and their characteristics are discussed in Sect. 3.6. 20 2. Methods

Aerosols sampling
The aerosols samples were collected from April 2010 to June 2014 at the Zotino Tall Tower Observatory facility, which is located near the Yenisei river at the eastern edge of the West Siberia Lowland in the boreal zone (60.8 °N and 89.4 °E, 114 m asl), about 600 km north of the closest 25 large city, Krasnoyarsk (950,000 inhabitants); the nearest village (Zotino) is about 20 km east of the site. The site lies in a vast region of boreal coniferous forest and bogs, and the ecosystem in the light taiga around the station is dominated by Pinus sylvestris forest stands (about 20 m height) on lichen covered sandy soils. The heart of the station is a 300-m tower, which was designed for long-term atmospheric observations and where the sampled air masses are representa- 30 tive of a very large fetch area. A more detailed description of the ZOTTO facility is given elsewhere (Heimann et al., 2014). The climate is dominated by a large seasonal temperature cycle reaching from minima below −55 °C in winter to maxima above 30 °C in summer.
Ambient air was sampled through a stainless steel inlet pipe with an internal diameter of about 2.9 cm, reaching to the top of the tower at 300 m above ground. The inlet was designed for a laminar nominal sampling flow of 40 L min -1 (Birmili et al., 2007). Pre-installation calibration showed that particles with diameter D p > 50 nm are nearly perfectly transmitted through this pipe (Heintzenberg et al., 2008). Additional test measurements with supermicron aerosol particles 5 have shown that the upper transmission size limit for the inlet system is ~10 μm. Thus, carbonaceous species concentrations obtained in this study refer to aerosol particles with an upper limit of ~10µm.
Aerosols were collected directly from the inlet line on 47-mm quartz fiber filters (2500QATUP, Pallflex) at a flow rate of 20 L min -1 using a home-made sampler. One of the dif-10 ficulties in long-term filter sampling is the decision about sampling hours. On one hand, the aerosol concentration at ZOTTO is very low during the near-pristine periods, therefore days or even weeks of sampling time are needed in order to obtain enough material on the filter for analysis; on the other hand, extremely high aerosol concentrations were observed during pollution episodes (e.g., biomass burning) and higher time resolution was needed for better characterization 15 of such episodes and to avoid overloading of the filters. As a result, the sampling time for each filter varied from 10 hours during pollution events to 480 hours during clean periods, the median sampling time was 104 hours. A total of 292 samples were collected between April 2010 and June 2014. The exposed filters were sealed in aluminum foil and then placed in Ziploc bags. The samples were stored at -18 ºC before being analyzed.

Carbon monoxide measurements
Carbon monoxide (CO) was measured by UV resonance fluorescence, using a Fast-CO-Monitor (model AL 5002, Aerolaser GmbH, Germany). Details of the experimental setup and calibration 25 are described elsewhere (Chi et al., 2013). The original CO data, measured with a frequency of 3 s, were converted to 1-h averages to minimize uncertainties inherent in the data analysis methodology. These CO concentrations were further averaged over the aerosol sampling intervals.
Technical problems occurred with the CO monitor during the following periods, resulting in gaps in the CO time series (Fig. 2
The aerosol mass concentrations were determined gravimetrically using a Mettler-Toledo micro balance model XP6 with 0.6 μg sensitivity. Before being weighed, the filters were equilibrated for 24 h at a constant temperature of 23 °C and a relative humidity between 35 and 45 %. Each filter was weighed at least three times before and after sampling. An anti-static U-Electrode (Mettler-Toledo) was used to remove electrostatic charge before weighing. The uncer-5 tainty (1 standard deviation) for the PM determination is estimated to be 3.5 μg for 47-mm quartz filters.

Organic carbon and elemental carbon analysis
Organic carbon (OC), elemental carbon (EC), and total carbon (TC = OC + EC) were measured 10 by a thermal-optical transmission (TOT) technique (Birch and Cary, 1996), using a thermaloptical carbon analyzer from Sunset Laboratory (OR, U.S.A.). The temperature protocol used was NIOSH5040 (National Institute for Occupational Safety and Health) with a preset maximum of 870 ºC (Birch, 1998). The uncertainty in the OC, EC, and TC measurement is provided for each individual filter sample by the calculation program. The uncertainty is made up of a con- 15 stant part (which is 0.2 μg C cm -2 for OC and EC and 0.3 μg C cm -2 for TC) and of a variable part, which amounts to 5% of the OC, EC, or TC mass loading. To correct for the positive artifact in the OC determination, two quartz filters in series were used (Maenhaut and Claeys, 2007).
Both filters were pre-baked at 850 ºC. The carbon loading on the second filter was subtracted from that on the first filter. Water soluble organic carbon (WSOC) was determined by soaking 20 part of the filter in water (18.2 MΩ cm, Direct-Q3 UV, Millipore) for 12 hours; after drying the remaining carbon in the filter was measured using the Sunset instrument. Organic matter (OM) was estimated as 1.8·OC. The same OC-to-OM conversion factor of 1.8 had been used in the SMEARII (Finland) (Maenhaut et al., 2011a) and K-puszta (Hungary) (Maenhaut et al., 2008) remote coniferous forest sites, providing the best agreement in the aerosol chemical mass closure 25 calculations. As a result, the TCM -total carbonaceous matter was calculated as TCM = 1.8·OC+EC.
It needs to be noted that the OC/EC analysis is very sensitive to the temperature protocol used and to the optical correction method (OC/EC split). NIOSH and IMPROVE (Interagency Monitoring of Protected Visual Environment) are most widely applied thermal protocols, which 30 differ in their temperature ramping regime and charring correction. The discrepancy between NOISH-and IMPROVE-derived EC concentrations may vary in the range of a factor of 1.2 -2, depending on the source and aging of the samples (Chow et al., 2001;Cheng et al., 2014;Wu et Atmos. Chem. Phys. Discuss., doi:10.5194/acp-2017-409, 2017 Manuscript under review for journal Atmos. Chem. Phys. Discussion started: 5 May 2017 c Author(s) 2017. CC-BY 3.0 License.
al., 2016). Therefore, wherever possible, we compare OC/EC results from studies using the same or similar analytical methods.
In addition to the thermal-optical method, a single-wavelength (574 nm) Particle Soot Absorption Photometer (PSAP, Radiance Research, Seattle, USA) was used in this study for measuring the particulate light absorption coefficient to estimate equivalent black carbon (BC e ) con-5 centrations (Andreae and Gelencsér, 2006). Details about this method and the methodology used for data analysis and interpretation correction can be found in Chi et al. (2013). In this study, PSAP and TOT measurements cover the date range from 19 April 2010 to 15 May 2012, therefore the PSAP data will be mainly used to quantify the correction factor (BC e /EC ratio) between by the two methods used. The BC e mass concentration (µg m -3 ) can be calculated based on the where σ (Mm -1 ) is the PSAP-measured absorption coefficient, and α abs (m 2 g -1 ) is the mass absorption efficiency. The commonly used value of α abs is 10 m 2 g -1 , as recommended by the PSAP manufacturer in their manual.

Ancillary products
Air mass trajectories were calculated with the Hybrid Single-Particle Lagrangian Integrated Trajectories (HYSPLIT) model (Stein et al., 2015) using the NCEP/ NCAR meteorological archive data produced by the National Center for Environmental Prediction (NCEP) and the National 20 Center for Atmospheric Research (NCAR) (2.5° horizontal resolution, 17 pressure levels) (Kalnay et al., 1996). The trajectories were calculated an arrival height of 300 m above ground level, which corresponds to the aerosol sampling height at the ZOTTO station. The NCEP/NCAR meteorological dataset was also used to estimate the mixing layer depth (MLD) and accumulated precipitation along the trajectory (APT). We used the APT data as a measure of wet removal of 25 aged carbonaceous species (Kondo et al., 2011a;Matsui et al., 2011;Kanaya et al., 2016). In some cases, the APT values were averaged by dividing the total precipitation along the all trajectories for a selected time period by the number of these trajectories ( ing of the flare sites associated with production of crude oil (Elvidge et al., 2011;Anejionu et al., 2015).

Results and discussion
3.1 PM, TCM, and regional meteorology 5 The time series of PM10 and TCM concentrations together with daily averaged meteorological parameters during the sampling period are shown in Fig. 1. Visual analysis indicates that the highest concentrations of PM10 and TCM, observed in the spring and summer periods (Fig. 1a), correspond to high temperatures and lower levels of precipitation (Fig. 1b). It is also seen that TCM was the dominant species in particulate mass during summertime. As will be shown below, Our measurements of the carbonaceous species can be compared with those from the Hyytiälä (Finland) and K-puszta (Hungary) remote European forested sites, which used the same EC/OC thermal protocol (Chi, 2009;Maenhaut et al., 2011b). Table 1 shows that the summertime EC concentration at ZOTTO (0.13 µg m -3 ) is comparable to measurements at Hyytiälä (0.12 20 µg m -3 ) and K-puszta (0.16 -0.19 µg m -3 ), whereas the median OC and WSOC concentrations at ZOTTO are higher than those at European sites. Elevated concentrations of the organic species at the ZOTTO site can be attributed to the strong influence of biomass burning events during the summer season. This is particularly reflected in the wide concentration range of OC and WSOC, varying from 0.3 to 106 µg m -3 and 0.2 to 45 µg m -3 , respectively (Table 1). In summertime, the 25 WSOC/OC ratio at the ZOTTO site was 65 ± 13%. This value is close to that for the Hyytiälä boreal forest site (62±9%) and slightly higher than for the K-puszta station (57±11%) ( Table 1).
The obtained WSOC/OC ratio is typical for nonurban sites surrounded by pine forest. It has been documented that in the summer season aged secondary organic aerosols produced by monoterpene oxidation as well as biomass burning organic particulate matter contain a high fraction of 30 WSOC, ranging from 50% to 80% (Saxena et al., 1995;Kiss et al., 2002;Pöschl, 2005;Pio et al., 2007;Timonen et al., 2008). within the concentration range observed at the European high-altitude sites and lower than those from Chinese stations, whereas the yearly-average OC and WSOC concentrations at ZOTTO noticeably exceed those obtained at both the European and Chinese sites. As a result, our EC/TC ratio is lower than those at the high-altitude stations, probably because of higher contributions from VOC oxidation at ZOTTO. Note, as mentioned above, that the EC discrepancy can be par-5 tially caused by the different thermal protocols used for EC/OC analysis.
Overall, the comparative analysis of the carbonaceous species concentrations suggests that the aerosols sampled at ZOTTO during our study period were generally representative of a fairly clean region.

Estimation of polluted, background, and clean periods
Siberia is a highly diverse region, where relatively clean periods (PM10 < 1 µg m -3 ) alternate with heavily polluted intervals (PM10 > 50 µg m -3 ) (Fig.1a). Air masses that are influenced by regional emission sources are not representative for well-mixed background air. Data filtering is therefore an important step in analyzing the data. In our analysis, we therefore differentiate pol- 15 luted, background, and clean periods.
We refer to "background" conditions as an atmospheric state without the detectable influence of local or regional pollution sources, but affected by emissions from natural origin as well as by pollution transported from very distant sources (Andreae, 2007;Chi et al., 2013;Mikhailov et al., 2015b). To define the concentrations of the measured species representing background air 20 (or the other way around to define the polluted periods) we made use of a non-parametric statistical approach named REBS (Robust Extraction of Background Signal). This technique has been previously applied for the identification of background CO concentrations for long-term measurements at the high alpine background site at Jungfraujoch (Ruckstuhl et al., 2012), at a global background station in China , and at the ZOTTO site (Chi et al., 2013). Here, 25 we use a bandwidth of three months following the suggestion of Ruckstuhl et al. (2012). Figure 2 shows the time series of the CO mixing ratio, EC, OC, and PM10 concentrations at ZOTTO, with the red baseline indicating REBS background concentrations g i (t i ). All concentrations C i (t i ) ≤ g i (t i )+3σ are classified as "background" measurements; all other observations are classified as "polluted". Here, σ is the goodness of fit (Ruckstuhl et al., 2012). 30 In contrast to background conditions, the term "pristine" implies that all aerosol sources arise from natural emissions. It has been argued that regions in which aerosols are totally unperturbed by air pollution no longer exist in today's atmosphere (Andreae, 2007). Therefore, the terms "near-pristine" or "clean" are commonly used to denote relatively clean air. Both EC (or BC) and CO are indicators of combustion and pollution, and their concentrations are frequently used to separate near-pristine from polluted periods (Andreae, 2007;Chi et al., 2013;Hamilton et al., 2014). It should be noted, however, that it is not always possible to define a pristine environment based on CO concentrations at a particular site, since CO can be produced by biogenic sources directly or by means of CH 4 and NMVOCs (nonmethane volatile organic compounds) study. Accordingly, the frequencies associated with clean periods are significantly lower (Table   20 2). Besides more strong pollution events observed from 2010 to 2014, our low values also arise from a more rigid criterion used for clean period detection. In Chi et al.'s (2013) work, clean periods were selected based on the aerosol absorption coefficients, i.e., the time intervals when their values were below 1 Mm −1 ( ~ 0.1 µgC m -3 ) were classified as clean. In contrast, in this study the clean periods were defined by EC concentrations below the detection limit of   the different concentrations are mostly due to different degrees of dilution and removal. We concur with their suggestion that the differences between polluted and "clean" concentrations in the cold seasons (Table 2) are mainly the result of changes in meteorological conditions (wind direction, precipitations). Note also that across all seasons the WSOC/OC ratio ranged from 0.5 to 0.7 (Table 2). This interval is typical for aged primary and secondary organic aerosols, which have 5 undergone chemical transformations (oxidation, nitration, hydrolysis, and photolysis) during long-range transport (Pöschl, 2005;Timonen et al., 2008;Jimenez et al., 2009;Saarino et al., 2010). 10 As described in Sect. 3.3, the entire ZOTTO data set was separated into polluted and background periods using the REBS method. In order to evaluate the degree of pollution, we used the Enhancement Factor (EnF), defined here as the ratio of the median concentration of a species during a polluted period to the background concentration. The monthly variations of EnF for PM, EC, and OC are shown in Fig. 3. It is seen that the EnF of all species correlate quite well, indi- 15 cating high values of the EnF during the summer months and lower values in the cold season.

Characterization of polluted air
Particularly, the EnF for EC indicates that minimal levels of pollution were observed in March and April with EnF of ~1.7, followed by the winter months (EnF~ 3), while maximal pollution is present in summer with EnF = 20, 51, and 5.4 for June, July, and August, respectively. In general, Figure 3 clearly illustrates that during the whole year the air masses arriving at ZOTTO 20 contain pollution components from regional and local sources.
Biomass burning and fossil fuel combustion processes emit CO, OC, and EC (BC), but the emission ratio between these species differs by combustion type and burning condition (Kondo et al., 2006;Wang et al., 2011;Zhang et al., 2013). We used the ∆OC/∆EC and ∆EC/∆CO relationships observed at ZOTTO as an additional indicator of the origin of pollution emissions (An-  fire (e.g., flaming vs. smoldering) the OC/EC ratio for boreal forest fires varies over a wide range (Table 3). From prescribed burns conducted in boreal forests (northern Ontario, Canada), Mazurek et al. (1991) reported a range of 10-18 and 21-95 for full-flaming and smoldering fire conditions, respectively. A chamber experiment conducted with typical Siberian biomass (pine) showed that the OC/EC ratios for PM10 in fresh (aged) smoke for flaming and smoldering are  (126), respectively (Popovicheva et al., 2015). In addition, the OC/EC ratios for fresh debris smoke were found to be 0.6 and 35 in flaming and smoldering fires, respectively. In mixed boreal wildfires, the mean OC/EC ratios for PM10 in fresh smoke plumes were reported as 6.7 (Saarnio et al., 2010), 6.5 (Popovicheva et al., 2015), and 15 (Lee at al., 2005). For aged smoke particles, Yan et al. (2008) reported a value of 25.6, which is close to our summer aver- Unlike the ∆EC-∆OC relationship, the correlation between ∆EC and ∆CO is not as strong 25 and varies greatly from season to season. The strongest correlation is found in the summer season (R 2 =0.85; Fig. 5c), followed by the wintertime (R 2 =0.32; Fig. 5a), while no significant relationship between ∆EC and ∆CO was observed in spring and fall, with R 2 =0.02 ( therefore reasonable that the strength of the correlation between EC and CO depends on the season and the history of the air mass before arriving at ZOTTO. The strong correlation obtained in the summer season and the ∆EC/∆CO slope of 1.3±0.1 ng m −3 ppb −1 suggest that both components were emitted into to the atmosphere from the same nearby biomass burning sources. EC/CO emission ratios of 1.7±0.8 and 3.4±1.6 ng m −3 ppb −1 for smoldering and flaming combus-5 tion phases, respectively, were obtained during summer aircraft measurements in fresh boreal forest fire plumes (Kondo et al., 2011a) (Table 3). Therefore, our low overall summer ∆EC/∆OC ratio of 1.3 ng m −3 ppb −1 suggests that the air masses arriving at ZOTTO were mainly influenced by smoldering fires.
The ∆EC/∆CO slope averaged over all winter polluted periods is 3.3 ± 0.8 ng m −3 ppb −1 10 ( Fig.5a). This is within the range of 1.4 -5.4 ng m −3 ppb −1 obtained in winter at the Hedo site (Okinawa Island, Japan) due to long-range transport of pollution plumes from East Asia (Verma et al., 2011), but on average it is lower than the values reported for industrial and urban regions of East Asia and North America (Table 3). Most likely in wintertime scavenging by ice crystal precipitation (Cozic et al., 2007) and dry deposition of EC has resulted in decreasing ∆EC/∆CO 15 ratios during the air mass transport from the pollution sources to the ZOTTO site.
The lack of correlation (R 2~ 0.02-0.03) between ∆EC and ∆CO concentrations in spring (Fig. 5b) and in fall (Fig. 5d) is an indicator of the long-range transport of the polluted air masses. Since in the transition seasons the secondary CO sources and sinks are small, it is reasonable to assume that wet deposition of EC as well as variable source types are responsible for the poor relation-20 ship between EC and CO. In addition, CO from long-range transport from Europe may contribute a substantial fraction of CO at the ZOTTO site (Chi et al., 2013).

Pollution episodes
Next we consider in detail some specific pollution events that occurred during the sampling peri-  Table 4. For all ∆X i /∆X j ratios presented in Table 4, each species concentration enhancement, ∆X i , was obtained as the difference between measured and REBS-derived concentra- 30 tions. In addition, we also included the mixed layer depth (MLD), height above ground level (AGL), and accumulated precipitation (APT) averaged along the air mass trajectories.

Polluted winter air
Atmos. Chem. Phys. Discuss., doi: 10.5194/acp-2017-409, 2017 Manuscript under review for journal Atmos. Chem. Phys. Discussion started: 5 May 2017 c Author(s) 2017. CC-BY 3.0 License. Figure 6 shows winter polluted 5-day air mass trajectories arriving at ZOTTO in the periods of 25-30 December 2010 (Fig. 5a, orange), 10-15 January 2011 (Fig. 5a, grey) and from 27 December 2011 to 9 January 2012 (Fig. 5b). It is seen that all trajectories passed over the southwestern and southern regions with densely populated industrialized areas, and therefore air masses moving across this area are likely to have accumulated anthropogenic emissions. During these peri-5 ods the trajectory heights ( Fig. 5 c,d) were less than or comparable to the mixed layer depth (Table 4), indicating that the polluted air masses were trapped under a very low inversion layer. In addition, the precipitation rates along the trajectories were moderate (Table 4). These meteorological conditions contributed to the relatively high CO, PM10, and carbonaceous species concentrations observed during the winter pollution periods. As a result, the overall ∆OC/∆EC slope 10 for the winter pollution periods is 3.9 ± 0.6 (R 2 =0.75) ( Table 3). This value is consistent with emission ratios found in urban areas, and represents a mixture between aged regional haze aerosols with OC/EC ranging from 3 to 7 and local emissions with a ratio of about 1.4 (Andreae et al., 2008a;Meng et al., 2007;Li et al., 2012;Cheng et al., 2011;Pan et al., 2012) (Table 3).

Polluted summer air
Previous studies based on CO measurements during 2007 and 2008 (Vasileva et al., 2011) and with these earlier studies. Table 4 shows that within the selected summer periods, CO, PM10, by a factor of 20 or even more (Fig. 3), and suggesting therefore that the biomass burning events were close to the ZOTTO site. As an example, Fig. 7 shows fires detected by MODIS in June-July 2012 and trajectories that passed over them. It is evident that the fires were very close to our site, resulting in a strong correlations between ∆EC, ∆OC, and ∆CO (R 2 >0.9) (Fig. 8). The

5
∆OC/∆EC and ∆EC/∆CO ratios from the fire pollution episodes are 26.2 ± 0.1 and 1.4 ± 0.1 ng m -3 ppb -1 , respectively (Table 3). As discussed in Section 4.5 and shown in Table 3, these values are in the range of emission ratios found in boreal forest fires dominated by smoldering. In addition, the ∆TCM/∆PM ratio was close 1, indicating that carbonaceous material was the dominant component in PM10. The ∆WSOC/∆OC ratio ranged from 0.47 to 0.68 and did not correlate 10 with meteorological conditions (Table 3), probably due to the short-distance transport of the pollution plumes.

Polluted air in the transition seasons
In the spring season, strongly elevated concentrations of CO, PM10, and carbonaceous species 15 were observed on 28 April 2010 (Table 4). Figure 9a shows large-scale fires as observed by the MODIS satellite. In this period, the fires were mainly located in the middle belt of European Russia and in the area covering the south and southwest of Siberia and northern Kazakhstan (Fig.   7a). The HYSPLIT 4-day trajectories show that the air masses that arrived at ZOTTO on 28 April 2010 had passed over these fire zones. In particular, the low-altitude air masses that arrived 20 at ZOTTO from 18h to 24h on 28 April had passed over fires located in southern Siberia (Novosibirsk region), whereas the high-altitude airflows that arrived between 00h and 16h of 28 April 2010 had traveled over the fires in the European part of Russia and southwestern Siberia (Tyumen region) (Fig. 9a, c). As a result the CO, PM10, OC, and EC concentrations reached 261 ppb, 54 µg m -3 , 18.9 µg m -3 and 1.5 µg m -3 (  Andreae et al., 2001Andreae et al., , 2004Trentmann et al., 2006;Rosenfeld et al., 2007).
The large-scale biomass burning that occurred in April -May in southern Russia, including Siberia and northern Kazakhstan, is caused by agricultural fires started by farmers clearing the fields.
Agricultural prescribed burning in Russia is estimated to total 30 million ha annually, of which  (Shvidenko et al., 1995). These burns often escape and cause forest wildfires. The air masses impacted by these fires can be transported over long distances. For example, Siberian and Kazakhstan agricultural emission plumes were sampled by flights over northern Alaska (Warneke et al., 2009;Kondo et al., 2011a) and the Arctic (Matsui et al., 2011).

5
Our ∆OC/∆EC ratio obtained for agricultural plumes during pollution episodes is 12.7 ± 2.7, which is slightly higher than values reported for prescribed fires during field and chamber experiments ( Table 3). The reason could be chemical and physical ageing processes during longrange transport of polluted air masses from the sources to ZOTTO. Adsorption, condensation and cloud processing of semi-volatile organic compounds tend to increase the OC abundance in the  (Table 3). It should be noted that in combustion experiments the emission factor for different carbon species strongly depends on the type of residue combusted, initial moisture content, burning phase, and analytical instrument used for 20 EC analysis. As a result, different studies yield a wide range of OC/EC and EC/CO ratios, as shown in Table 3 ("Agricultural fires" section). In addition, the long-range transport of agricultural plumes can change these ratios significantly due to aerosol aging and wet scavenging.   Figure 10 shows the scatterplot between PSAP-derived ВС e and thermal-optical EC obtained in this study during 19 April 2010 -15 May 2012. The BC e concentration was calculated from Eq.
(1) with the commonly used value of α abs of 10 m 2 g -1 . A good correlation between EC and BC e for the entire period with R 2 = 0.90 is observed, with a BC e /EC slope of 1.67 ± 0.05. Thus, our ∆EC/∆CO ratio of 14.3 ± 4.4 ng m -3 ppb -1 obtained for agricultural fires (Table 4) can be con-20 verted to a ∆BC e /∆CO ratio of 23.9 ± 7.3 ng m -3 ppb -1 . This value is within the range of 21.8 ng m -3 ppb -1 and 29.8 ng m -3 ppb -1 reported by Chi et al. (2013) and Cristofanelli et al. (2013), respectively, using filter-based light absorption techniques (Table 3).
In the fall season, elevated concentrations of CO, PM10, and carbonaceous species were observed during 14 -15 October 2010. Figure 9b shows that air masses arriving at ZOTTO were 25 impacted by large-scale fires and anthropogenic emissions from industrialized urban sites. As in the case of the spring pollution episodes, the observed fires are caused by agricultural burning of crop, pasture, and hayfield residues. Figure 9d shows that about half of the 4-day HYSPLIT trajectories stay well below 300 m, that is, the air masses moved mainly within the boundary layer  lationship considering that both species come to ZOTTO from remote combustion sources and that atmospheric processes exert different influences on the two species, as discussed above.

Characterization of background air
In contrast, the OC (Fig.11c) and PM10 (Fig.11d) concentrations do not exhibit clearly pro- inputs of pollution aerosols in winter. As noted above, aerosol aging and precipitation could further modify the proportion between different species. It is known that aged organic species and water soluble inorganic compounds are removed from the atmosphere by wet deposition more efficiently than EC (Cerqueira et al., 2010;Witkowska et al., 2016). In addition, during the transition from winter to summer differential scavenging of the aerosol species can occur in mixed-5 phase clouds due to the Wegener-Bergeron-Findeisen effect (Wegener, 1911;Bergeron, 1935;Findeisen, 1938). In this type of clouds, the scavenging efficiency of EC decreased more than a factor of five (Cozic et al., 2007;Qi et al., 2016). 10 Over the entire measurement period, the cleanest air masses were observed in summertime, as long as biomass-burning pollution did not affect the site. We selected three clean periods based on the selection criterion that EC concentrations had to be below the detection limit of 0.02 ± 0.01 µgC m -3 . Table 5 summarizes the dates and chemical species concentrations obtained during these time periods. The lowest concentrations were observed from 7 July to 1 August 2010. 15 For this period, the EC concentration was below the detection limit and the CO mixing ratio was as low as 93 ± 11ppb (Table 5). Backward trajectories computed for this time period together with fire maps (Fig. 12a, b) show that the trajectories either did not pass through the fires (panels a.2, a.6, a.8, a9, a.10) or just barely touched them ( Fig. 12a; panels a.3 -a.5, a.7, a.11-a.14).

Near-pristine air
Given the low CO concentration, a measureable pollution input from the fires is not to be ex-20 pected. In addition, any biomass burning aerosol that may have been transported in the boundary layer ( Table 5, column 1 and 2) could have been effectively removed via precipitation (APT = 7.3 mm) (Kanaya et al., 2016;Ohata et al., 2016). Figure 12c shows that precipitation with different intensity occurred along the air mass trajectories before they reached the measuring site.
As a result the PM10 and OC concentrations were as low as 0.77 ± 0.09 and 0.29 ± 0.02 µg m -3 , 25 respectively. The TCM/PM ratio was the highest among the clean periods ( For the other relatively clean periods, i.e., 6 -10 July 2011 and 15 -27 June 2013, the PM10 and carbonaceous species concentrations were much higher than those presented above (Table 5). Most likely, the air masses arriving at ZOTTO during these time intervals were partly 30 influenced by remote pollution sources. Figures 13a,b show that the location of some hot spots (bold red circles) coincides with gas and oil field production, suggesting that these hot spots represent gas flaring (Elvidge et al., 2011;Anejionu et al., 2015), which is a significant source of EC and other carbonaceous species (Stohl et al., 2013). Therefore, the air masses coming over the land surface from the northwest (Fig. 13c, d) could have accumulated gas flaring products emitted from oil production sites in the West Siberia Basin. In addition, the southeastern air masses arriving at ZOTTO on 19 June 2013 (Fig.13b) passed over forest fires and most likely were also enriched with biomass combustion products. On the other hand, for some time intervals precipitation along the air mass trajectories was light and scattered (Fig. 13e, f) indicating 1.14 and 1.46 µg m -3 ) concentrations (Table 5).
In the growing season, forested boreal regions emit biogenic volatile organic compounds orological conditions (wind direction, precipitation) obscured any effect of temperature variations on the particle mass concentration. However, Fig. 15 shows that the ratio OC/PM, which eliminates the effects of transport and removal processes that affect the absolute mass concentration, shows a strong dependence on temperature. There is a clear exponential temperature dependence of the OC/PM ratio, with R 2 =0.66, which follows the exponential increase of the mon- BC e determination, respectively. Accordingly, the BC e /EC ratio of 1.67 ± 0.05 obtained in this study was applied to convert the BC e -derived concentrations at ATTO into EC values. Table 6 summarizes the carbonaceous species concentrations obtained at the Amazonian sites and at  Our analysis of the near pristine conditions, based on the EC detection limit of 0.02 ± 0.01 µgC m -3 shows that the longest periods with clean air masses were observed in summer (fre-5 quency of 17%). For this period, the variations in the OC/PM ratio closely correlated with those in air temperature, which implies that biogenic sources of OC formation are dominating. Against a low concentration of CO, EC, and OC observed in the near-pristine summer episodes it was possible to identify pollution plumes that most likely arrived at ZOTTO from crude oil production sites located in the oil-rich regions of Western Siberia. 10 Over the entire measurement campaign, the WSOC/OC ratio remain relatively stable and ranged from 0.5 to 0.7, indicating that partitioning between water soluble and water insoluble organic matter is not significantly dependent on the source of the polluted air mass. Most likely, aging processes and precipitation scavenging control the ratio between them.
Further studies at ZOTTO may investigate the link between aerosol and climate: e.g., the 15 influence of rising temperatures on an enhancement of biogenic SOA formation, as well as a quantification of anthropogenic emissions, which produce feedbacks on cloud formation, tropospheric radiation balance, and precipitation patterns (Kulmala et al., 2004;Goetz et al., 2007).
The average CO, PM10, and carbonaceous species concentrations (± st. dev.) together with WSOC/OC and TCM/PM ratios obtained during the cleanest periods in the growing season. MLD and AGL are median and (25th -75th) quartiles of the mixed layer depth and air mass trajectory height above ground level, respectively. 5.3 ± 0.2 0.018 ± 0.010 1.4 6 ± 0.08 1.22 ± 0.14 0.84 ± 0.11 0.50 ± 0.03 Table 6 Average background concentrations of carbonaceous species (± st. dev.) or (min -max) as well as their ratios obtained at Fazenda Nossa Senhora (FNS, Rondônia, Brazil) and at the ATTO site (Amazonas, Brazil) during the wet period and those measured in summertime at ZOTTO.    Table 2). The ∆EC/∆CO slope is in µg m -3 ppb -1 .  Air temperature ( 0 C) ME rate (µg g -1 h -1 ) ME (Isidorov,1985) Exponential fit Figure 15. OC/PM ratios in PM10 for pristine conditions vs. averaged ground-level temperature (± st. dev.) estimated from the meteorological profiles along each trajectory. The coefficient of determination of the fit (R 2 ) is 0.66. The insert shows monoterpene emission (ME) rate from Scots pine vs. temperature (Isidorov et al., 1985) and the exponential fit: ME rate = 0.121·exp[0.179·T(°C)] with R 2 = 0.99.