Influence of Biomass Burning and Anthropogenic Emissions on Ozone, Carbon Monoxide and Black Carbon at the Mt. Cimone Gaw-wmo Global Station (italy, 2165 M A.s.l.)

This work investigates the variability of ozone (O 3), carbon monoxide (CO) and equivalent black carbon (BC) at the Italian Climate Observatory " O. Vittori " (ICO-OV), part of the Mt. Cimone global GAW-WMO station (Italy). For this purpose, ICO-OV observations carried out in the period January 2007–June 2009, have been analyzed and correlated with the outputs of the FLEXPART La-grangian dispersion model to specifically evaluate the influence of biomass burning (BB) and anthropogenic emissions younger than 20 days. During the investigation period, the average O 3 , CO and BC at ICO-OV were 54 ± 3 ppb, 122 ± 7 ppb and 213 ± 34 ng m −3 (mean ± expanded uncertainty with p < 95 %), with clear seasonal cycles characterized by summer maxima and winter minima for O 3 and BC and spring maximum and summer minimum for CO. According to FLEXPART outputs, BB impact is maximized during the warm months from July to September but appeared to have a significant contribution to the observed tracers only during specific transport events. We char-acterised in detail five " representative " events with respect to transport scales (i.e. global, regional and local), source regions and O 3 , CO and BC variations. For these events, very large variability of enhancement ratios O 3 /CO (from −0.22 to 0.71) and BC/CO (from 2.69 to 29.83 ng m −3 ppb −1) were observed. CO contributions related with anthropogenic emissions (CO ant) contributed to 17.4 % of the mean CO value observed at ICO-OV, with the warm months appearing particularly affected by transport events of air-masses rich in anthro-pogenic pollution. The proportion of tracer variability that is described by FLEXPART CO ant peaked to 37 % (in May– September) for CO, 19 % (in May–September) for O 3 and 32 % (in January–April) for BC. During May–September, the analysis of the correlation among CO, O 3 and BC as a function of the CO ant indicated that ICO-OV was influenced by air-masses rich in anthropogenic pollution transported from the regional to the global scale. On the other side, CO and O 3 were negatively correlated during October–December, when FLEXPART does not show significant presence of recent anthropogenic emissions and only a few observations are characterized by enhanced BC. Such behaviour may be attributed to an ensemble of processes concurrent in enhancing O 3 with low CO (upper troposphere/lower stratosphere intrusions) and to …


Introduction
Tropospheric carbon monoxide (CO), ozone (O 3 ) and black carbon (BC) have a substantial impact on climate and air quality.CO, primarily emitted from combustion processes (also formed in substantial amounts from methane P. Cristofanelli et al.: Influence of biomass burning and anthropogenic emissions oxidation), strongly influences the abundance of the OH radical and thereby alters the lifetime of methane and other greenhouse gases (e.g., Seinfeld and Pandis, 1998;Forster et al., 2007).O 3 is involved in photochemical reactions (e.g.Crutzen et al., 1999;Volz-Thomas et al., 2002) and in determining the overall oxidation capacity of the troposphere (Gauss et al., 2003).Moreover, it is a harmful pollutant (Conti et al., 2005) and an efficient greenhouse gas (Forster et al., 2007).Due to its direct impact on solar and thermal radiation, BC was recognised as an important factor contributing to global warming (Andreae and Gelencsér, 2006 and references therein).Additionally, BC causes an indirect radiative forcing by modifying the concentration and size spectrum of cloud droplets (e.g.Lohmann et al., 2000;Forster et al., 2007) and it decreases the surface albedo of snow (e.g.Flanner et al., 2009).
Southern Europe and the Mediterranean basin (hereafter SE/MB) are recognised as a hot-spot region both in terms of climate change (e.g.Forster et al., 2007) and air-quality (Monks et al., 2009), also representing a major crossroad of different air mass transport processes (Lelieveld et al., 2002;Millàn et al., 2006;Duncan et al., 2008).The transport of polluted air masses from Europe and other continents can influence the variability of trace gases and aerosol over SE/MB.In particular, large amounts of anthropogenic pollutants emitted in continental Europe are transported towards the SE/MB especially during summer (e.g.Duncan et al., 2008;Henne et al., 2005), when intense photochemical O 3 production also occurs (see for instance Vautard et al., 2005).Moreover, in the past decades many studies have revealed significant intercontinental transport of air-masses rich in anthropogenic pollutants or biomass burning emissions from North America to Europe (e.g.Stohl and Trickl, 1999;Stohl et al., 2003, Auvray andBay, 2005;Huntrieser et al., 2005).In particular, biomass burning (BB) represents a major source of atmospheric pollutants and climate altering species (e.g.Crutzen and Andreae, 1990;Simmonds et al., 2005) and forest wildfires have a considerable impact on the variability of CO, O 3 and BC in the Northern Hemisphere (e.g.Novelli et al., 2003;van der Werf et al., 2006;Bond et al., 2011).BC produced by boreal wildfires accounts for 10 % of the annual anthropogenic BC emissions in the Northern Hemisphere (Bond et al., 2004).Previous studies have shown that the atmospheric compounds directly emitted by BB or produced by photochemical processes occurring within BB plumes can be transported over long distances, thus affecting both air quality and climate from local to global scales (e.g.Val Martin et al., 2006 and references therein).SE/MB can also be affected by large wildfire events in Europe, especially during summer (e.g.Pace et al., 2006;Cristofanelli et al., 2007;Turquety et al., 2009).
The pollutant budget in the Mediterranean basin may also vary due to climate change.A dryer and warmer climate in the Mediterranean region, as projected by climate models, can lead both to an increase in pollution sources such as wildfire events and to enhanced photochemical O 3 production (Solomon et al., 2007) with a dramatic change in pollutant transport patterns over Europe (Forster et al., 2007).
In this work, we quantify the variability of O 3 , CO and BC over SE/MB from daily to seasonal timescales: observations are acquired on a routine basis at the Italian Climate Observatory "O.Vittori" (ICO-OV) of the Mt.Cimone global GAW-WMO station (CMN, 44 • 11 N, 10 • 42 E, 2165 m a.s.l.) during the period 2007-2009.The ICO-OV location is considered representative for the baseline conditions of the Mediterranean free troposphere (Bonasoni et al., 2000;Fischer et al., 2003) and only during the warm months an influence of airmass transport from the regional boundary layer is detected due to convective processes and breeze regimes (Fischer et al., 2003;Van Dingenen et al., 2005).Three years of observations are analysed here to identify the role of BB (as done in Cristofanelli et al., 2009 for a specific event) and anthropogenic pollution transport to the observed compound variability.The Lagrangian atmospheric transport model FLEX-PART provides an estimate of the emission age of BB and anthropogenic emissions at ICO-OV and, coupled to the observations, can allow to quantify the roles played by these emission types to the atmospheric composition variability.

Observations
At the ICO-OV, tropospheric O 3 measurements have been carried out continuously since 1996 using a UV-photometric analyser (Dasibi 1108).The accuracy and quality of measurements (sampling time: 1 min, combined standard uncertainty less than ±2 ppb in the range 1-100 ppb) and sampling procedures are guaranteed within the GAW requirements (WMO, 2002).In particular, the O 3 analyser working at ICO-OV was traced back to SRP#15 Standard Reference Photometer at EMPA (Swiss Federal Laboratories for Materials Testing and Research).
Since February 2007, CO mixing ratio is measured with a custom gas chromatograph equipped with a reduction gas detector (GC-RGD), Trace Analytical RGD2.The instrument has been running continuously since January 2007.Every 15 min, an air sample is injected into the gas chromatograph for separation, and then analysed for CO via mercury oxide reduction and detection of mercury vapour by UV absorption.Each analysis sample is alternated with a calibration sample by means of real air working standards representative for ambient air mixing ratios for the Northern Hemispheric troposphere.The working standards were prepared at Max-Planck-Institute for Biogeochemistry in Jena and indirectly referenced against NOAA 2004 scale.This guarantees a continuous check of the detector calibration (Novelli, 1999) with an accuracy, in term of daily relative standard deviation over repeated analysis of the working standards (48 runs) of 0.72 %, total expanded uncertainty of 1.51 % (k = 2).RGD detectors are well-known to be affected by non-linear response.At CMN, a set of calibration standard with different CO mixing ratios were unavailable.Thus, we tested the nonlinearity of the GC-RGD system by inter-comparison with a GC-FID (Agilent GC6890) which was working in parallel at the measurement site since 2008, January.Even if only for the limited range of the typical mixing ratios observed at CMN, correction functions were obtained (by mean of linear correlation between the two data-sets) and applied to the data series presented in this work.
Continuous measurements of equivalent BC (hereafter BC) are obtained by a multi-angle absorption photometer (MAAP 5012, Thermo Electron Corporation) since March 2006.The reduction of light transmission at 670 nm, multiple reflection intensities, and air sample volume are continuously integrated over the sample run period to provide a real time data output (1 min resolution, variable integration time) of BC concentration (Petzold et al., 2002).
To neglect the variability related with annual to seasonal scales, not captured by FLEXPART, we calculated tracer anomalies ( O 3 , CO, BC) by subtracting a 21-day running means from the 3-h averaged values.These anomalies were used for the combined analyses with FLEXPART outputs.During the period January 2007-June 2009, for which FLEXPART outputs are available (with only a data gap in July 2008), data coverage of 76 %, 97 % and 72 % have been achieved for CO, O 3 and BC, respectively.

FLEXPART simulations
The Lagrangian particle dispersion model FLEXPART (Stohl et al., 2005) was used to quantify the impact of anthropogenic and BB emissions on the observed CO, BC and O 3 variability at ICO-OV.To this aim, 20 days backward (socalled "retroplume") simulations were done from ICO-OV location with a time resolution of 3 h (http://transport.nilu.no/projects/eucaari).
Every simulation consists of 40 000 particles released in the volume of air sampled.The backward simulations are done with full turbulence and convection parameterizations.The FLEXPART retroplume output is proportional to the residence time of the particles in a given volume of air and corresponds to a potential emission sensitivity.When convolved with the gridded CO emission fluxes from an emission inventory, maps of potential source contributions are obtained.Integrating these maps over the globe or over certain regions of specific interest (e.g., Europe, North America), yields a model-calculated CO tracer mixing ratio at the ICO-OV location.Since FLEXPART does not account for CO emissions older than 20 days, this model-calculated CO can only explain a fraction of the CO mixing ratio observed at the ICO-OV, thus providing useful hints about specific contributions still not mixed within the tropospheric background.FLEX-PART considers both BB (hereinafter CO fire ) and anthro-pogenic CO emissions (hereinafter CO ant ).CO fire emissions are estimated from fire observations obtained from MODIS on board the Aqua and Terra NASA satellites (Justice et al., 2002;Giglio et al., 2003).Hot spots detected with a confidence of 75 % or greater are used and combined with a land inventory with 1-km resolution to provide an emissions inventory.It is assumed that area burned is 180 ha per fire detection.A parameterization was applied to determine the biomass available to burning, the fraction actually burned, and the emission factors, which are all dependent on land use (see Stohl et al., 2007, for a description).For CO ant , EDGAR version 3.2 emission inventory for the year 2000 (fast track) on a 1 × 1 degree grid was used outside North America and Europe.Over Europe, the EMEP emission inventory for the year 2005 with a resolution of 0.5 degree was used.Over most of North America, the inventory of Frost et al. (2006) was used.
For the purpose of systematically evaluating the influence at ICO-OV of BB and anthropogenic emissions with age less than 20 days and thus still not fully mixed within the atmospheric background, we analysed the time series of CO fire and CO ant emitted over the global domain.
FLEXPART also provides the information on the time spectrum of emission.In order to have a synthetic diagnostic, we defined a "dominant" emission age for BB and anthropogenic emissions as the time (prior to observation and expressed in days) when the total CO fire and CO ant contributions reached 75 % of their 20-day integrated value.Alternatively, emission age was also estimated by identifying the time when CO showed the maximum increase along the backward Lagrangian simulation.However, since no substantial differences were observed, only the first method is used in this paper.O 3 , CO and BC ICO-OV observations are averaged over 3-h to compare directly to FLEXPART data.

Carbon monoxide (CO)
Figure 1 reports the annual cycle of CO.During 2007-2009, the average CO at ICO-OV was 122 ± 7 ppb (in terms of mean ± expanded uncertainty) at a confidence level of 95 %.The annual CO cycles were characterised by the highest monthly values from February to April and minima from June to September (only months with data coverage larger than 15 days have been reported).Increasing CO is observed during autumn.Not-negligible inter-annual variability characterised CO levels at ICO-OV.In particular, high CO was observed during spring and autumn 2007 with mixing ratios exceeding by 20 ppb the average values during 2008-2009.Both in term of annual shape and temporal variability, the CO seasonal cycles at ICO-OV are well comparable with those at other high-altitude mountain sites in Europe (Zelwegger et al., 2009;Gilge et al., 2010).The observed annual cycle results from varying contributions by natural and anthropogenic emissions, production vs. removal processes and atmospheric transport (e.g.Yurganov et al., 1999;Chevalier et al., 2008;Gilge et al., 2010).In particular, the slower chemical destruction leads to a longer CO lifetime in winter and subsequent accumulation, thus explaining the observed late winter/early spring peak.Long-range transport can play a significant role in determining the seasonal cycle: in fact, as pointed out by Pfister et al. (2006), Asian and North American anthropogenic CO sources can significantly impact tropospheric Europe between January and March in addition to continental emissions.Another possible contribution that will be explored later comes from BB emissions that strongly impact CO variability during specific episodes (e.g.Yurganov et al., 2005;Cristofanelli et al., 2009;Zellweger et al., 2009).

Surface ozone (O 3 )
The simultaneously observed average O 3 for the years 2007-2009 at ICO-OV was 54 ± 3 ppb, in good agreement with earlier measurements (Cristofanelli et al., 2006).As reported by Fig. 1, the annual O 3 cycle is characterised by a winter minimum and two peaks: (1) in spring (in April, for the three years here considered) and (2) in summer (from July to August).This is a common feature for high-altitude measurement sites in Europe (e.g.Chevalier et al., 2007;Gilge et al., 2010).The summer peak is usually explained by increased photochemical production of O 3 in the lower troposphere: summer fair and warm weather conditions contribute to direct O 3 production from anthropogenic precursors and favour the export of polluted air masses from the boundary layer to the free troposphere of continental Europe (Henne et al, 2005;Cristofanelli et al., 2007).At middle latitudes in the Northern Hemisphere, the spring O 3 maximum can be related to the accumulation of precursors (as seen at ICO-OV for CO) during winter and subsequent O 3 production (Monks, 2000;Virganzan, 2004).Stratosphere-to-troposphere transport (STT) might also increase O 3 at ICO-OV during spring (Stohl et al., 2000), even if more recent analyses (Cristofanelli et al., 2006) also indicate high STT influence during autumn and winter.

Equivalent black carbon (BC)
The average monthly BC concentration during the year 2007-2009 was 213 ± 34 ng m −3 .This value is comparable with typical average levels recorded at European continental background sites (Van Dingenen et al., 2005;Putaud et al., 2004).Similarly to O 3 (Fig. 1), the annual cycle is characterised by the presence of a winter minimum (December-January) and maxima in spring (April-May) and late summer (August-September).This behaviour probably reflects the higher efficiency of uplift during warmer months and the subsequent transport of polluted air masses from the boundary layer to the measurement site (Marinoni et al., 2008;Cristofanelli et al., 2007).Interestingly, a secondary minimum in BC was observed during June when wet deposition is expected to peak at CMN, due to maximum rainfall.Large variability characterises the BC annual cycle with monthly values spanning almost an order of magnitude from 50 ng m −3 to 400 ng m −3 .

Role of biomass burning
The identification of BB events at ICO-OV is done here based on a combination of observation and model analysis.
The first step is to quantify the monthly mean contribution of CO fire at ICO-OV to provide an estimate of the mean impact and its seasonal variability.Figure 2 shows that FLEX-PART CO fire was low for the most part of the investigation period with an annual average value of 2.5 ppb.It contributed only up to 2.6 % to the mean CO observed at ICO-OV during 2007-2009.Only few events (about 3 % of the entire dataset) were characterised by CO fire exceeding 10 ppb and the highest CO fire values (up to 200 ppb) were observed at the end of August 2007 when a strong BB plume was detected at ICO-OV (see Cristofanelli et al., 2009).The CO fire contribution shows a clear annual cycle with the highest values during the warm months: maximum monthly average values were observed in July/September (4.2 and 3.6 ppb, respectively) and August (8.6 ppb) and a secondary maximum in late spring (4.0 ppb in May).As deduced by the increase in the lowest percentiles of the monthly distributions (Fig. 2), these higher values were due both to an increase of the CO fire baseline value and to the occurrence of specific events, as testified by the large upper percentiles.
To identify at ICO-OV the presence of BB plumes, we considered the time periods for which FLEXPART CO fire continuously exceeded 10 ppb for more than 12 h.Figure 3 reports CO observed at ICO-OV (black points) together with calculated FLEXPART CO fire (red line): time periods for which the selection criterion is fulfilled are highlighted with coloured dots, scaled on the fire emission age deduced by FLEXPART outputs.Following FLEXPART, CO fire enhancements occur sporadically and for a limited period of time.By considering only the period with CO fire exceeding 10 ppb, 16 events with a duration ranging from 12 to 72 h were identified, for a total of 21 days over the three years.Based on the FLEXPART analyses, for these events the transport of BB emissions occurs at very different scales: six events originated from the SE/MB region, two events from the Balkan peninsula, one event from Russia.Long range transport may have a likely role with 5 events from North America, one from Central America and one from equatorial Africa.The majority of BB events (11) occurred from May to September, thus confirming that transport of BB emissions mostly occurred during warm months.Over the 16 events detected by FLEXPART outputs, those having "local" BB source were characterized for the greatest part (6/8 events) by clear CO enhancements ( CO > 0 ppb) in respect to the average behaviour obtained by applying the 21-day running mean, those having "global" sources were characterized by negative or non-significant CO (7/7) and the only event that was attributed to emission occurring at "regional" scale showed a positive CO.
In this paper, we decided to describe in greater detail five "representative" events (24-25 April 2007, 28 August-1 September 2007;23-24 March 2009;28 March-1 April 2009, 1-4 May 2009), that might resume the main features of the 16 events we detected by the FLEXPART analysis.In particular, according to FLEXPART, these events are due to: (i) local transport of BB emissions from central Alps (event I) and north Africa, south Italy and Greece (event II), (ii) regional transport from Russia (event V), (iii) global scale transport from North America (event III) and west equatorial Africa (event IV).Supplement presents the CO fire source contributions maps for each of these events (Figs.S1-S5), while Table 1 reports the CO, O 3 and BC: CO fire contributions ranged from 10.1 to 47.5 % of the observed CO mixing ratio.Events I, II and V are characterized by clear CO enhancements ( CO > 0 ppb) with respect to the average behaviour obtained by applying the 21-day running mean.Events III and IV, related with long range transport of BB emissions, were instead characterised by CO decreases ( CO < 0 ppb).The actual CO due to a more aged and distant BB may be highly variable due to a combination of processes that is difficult to evaluate and that may lead to the negative observed CO anomaly.FLEXPART showed to be accurate in capturing long-range transport, including BB plume (e.g.Stohl et al., 1998Stohl et al., , 2006Stohl et al., , 2007;;Forster et al., 2001;Damoah et al., 2006;Lapina et al., 2008).However, the simulated transport can be inaccurate for individual cases and it is therefore important to use additional data to confirm a long-range transport episode.For these reasons, we performed some additional analysis to have a qualitative assessment on the reliability of the FLEXPART analysis for these BB long range transport events.As shown by Fig. 3b, event III is characterized by an increase of CO up to 170 ppb.Nevertheless, due to the relatively high CO values occurring before and after the event, CO was negative.Event IV is also characterized by CO up to 160 ppb.However a further enhancement in CO occurred 24 h after the possible arrival of the BB plume.Even if possible error in the timing of the plume arrival at CMN cannot be ruled out, nevertheless no clear BB signature can be seen in O 3 and BC data for this latter CO increase (not shown).It is not possible to disentangle the role of uncertainty in air parcel position and timing related with transport simulation, that may induce an erroneous attribution, and the limits in detectability of the event from observations, due to a highly variable background that may hide the CO increase related with BB emissions.Nevertheless, even if it is difficult to provide a definitive attribution of those events, the arguments supporting a realistic impact of BB emissions for these two events are (1) the   Real et al. (2007) showed that the mixing ratio of CO in forest fire plumes decreases by 50 % during a 5-day transport due to atmospheric mixing.Additionally, for the event IV, injection of BB emissions into tropical air masses with low pre-existing CO mixing ratios may lead to CO values still lower than those typically observed in the middle latitudes (WMO, 2011).Thus, the advection of this plume can lead to a decrease of observed CO at ICO-OV, even in the presence of a substantial BB contribution.This could partially explain the significant negative correlation existing between the FLEXPART CO fire and the CO mixing ratio observed at ICO-OV during the episode IV (see Table 1).
O 3 has a large variability among the identified fire plumes, with mean mixing ratios ranging from 45.2 to 88.5 ppb: overall, 4/5 events show O 3 enhancement from the 21-day filtered time series ( O 3 > 0 ppb).Figure 4a shows the relationship between CO and O 3 mixing ratios observed at ICO-OV during the five events.In order to remove the correlation due to the diurnal cycle of CO and O 3 at ICO-OV (see e.g.Cristofanelli et al., 2009), here we considered the residuals of 3-h Table 1.List of the five "representative" BB events at ICO-OV as deduced by FLEXPART analyses.For each event we reported: start and end dates ("Date"), transport-scale ("Scale"), the emission region ("Origin"), the range (minimum, mean and maximum values) of representative emission ages during the event ("Emission age"), the averaged fraction of FLEXPART CO fire contribution ("CO fire fraction"), the linear correlation coefficient between CO and CO fire ("R"), the mean CO and CO at ICO-OV ("Mean CO"; "Mean CO"), the mean O 3 and O 3 at ICO-OV ("Mean O 3 "; "Mean O 3 "), the mean BC and BC at ICO-OV ("Mean BC"; "Mean BC"), the calculated O 3 -CO and BC-CO enhancement ratios ("O 3 -CO ER"; "BC-CO ER").The different scatter of data among the detected events (with R values ranging from −0.14 for the event III to 0.93 for the event I) is indicative of a great variability of chemical reactivity in air-masses potentially impacted by BB.
Previous studies (see Val Martin et al., 2006 and references therein) showed that the slope of the linear fit of O 3 versus CO can be used to evaluate the O 3 /CO enhancement ratio (ER), i.e about the amount of O 3 molecule produced per CO molecule emitted.We calculated the linear fit by applying the reduced major axis (RMA) technique to four events characterised by O 3 -CO linear correlation coefficients larger than 0.30 (see Table 1).The range of the positive O 3 /CO ER (0.27-0.71) was in relatively good agreement with previous characterization of BB plumes. Val Martin et al. (2006) reported ER ranging from 0.14 to 0.89 in boreal fire plumes observed at the Azores Islands, while Bertschi and Jaffe (2005) found ER from 0.15 to 0.84 in plumes from Asian boreal wildfires over Northwest US, suggesting that significant O 3 production can occur in BB plumes probably due to the slow recycling of PAN, HNO 3 and organic nitrates (Honrath et al., 2004;Bertschi and Jaffe, 2005).A negative value of -0.22 is observed during event IV.Negative or not-significant O 3 /CO ER have been already observed in aged BB plumes by Val Martin et al. (2006) and Real et al. (2007).This was attributed to the high aerosol concentrations within fire emission plumes or other processes affecting night-time chemistry (i.e.hydrolysis of N 2 O 5 ) as well as O 3 loss by reaction with organic compounds during long-range transport.Real et al. (2006) showed that the strong aerosol light extinction in dense BB smoke plumes can decrease O 3 formation efficiency.In our case, a strong Saharan dust transport event was also detected at ICO-OV by in-situ and satellite (MODIS) observations.This could possibly lead to heterogeneous removal of O 3 and its precursors (via HNO 3 and NO 3 depletion) and by decreasing photolysis rate (e.g.He and Carmichel, 1999;Balis et al., 2000;Harrison et al., 2001).Depletion of O 3 in Saharan dust layers was systematically observed at ICO-OV by Bonasoni et al. (2004).
BC concentrations are highly variable during the events, ranging from 44.1 to 1072.3 ng m −3 (Table 1).BC concentrations showed positive anomalies with respect to the 21 days average for events I, II, III and V and negative for event IV which was characterised by the lowest BC concentrations indicating that removal processes may have been important.BC res as a function of CO res is reported in Fig. 4b and all the analysed events show significant positive correlation; the linear correlation coefficient R ranges from 0.31 (event IV) to 0.94 (event II).The BC res /CO res slopes were very similar, except for event IV.RMA analysis gives BC/CO ERs value ranging from 2.69 to 29.83 (Table 1).Previous observations of BB (from North America and Siberia) at baseline stations in the Azores (Val Martin et al., 2006), Ireland (Forster at al., 2001) an Japan (Kaneyasu et al., 2007) reported BC/CO ERs ranging from 0.5 to 8.4.Warneke et al. (2009) and Kondo et al. (2011) reported ERs of 8.5 ± 5.4 and 10 ± 5 ng m −3 ppbv −1 for BB in Asia (Sibera and Kazakhstan).Pan et al. (2011), reported BC/CO ratio of 10.3 ± 0.3 and 11.6 ± 0.5 ng m −3 ppb −1 for BB plumes observed at the background station Mt.Huang (1840 m a.s.l.) in east China.Thus, while events I, II and IV showed ERs in good agreement with already published values, events III and V were characterised by significant higher values.It is not entirely clear why the BC-CO slopes were so large, but different explanations could be considered, especially for event V. Firstly, these high BC/CO ERs may be explained by high combustion efficiency of the source fires (e.g.Kondo et al.,Fig. 4. Relationship between observed CO res , O res 3 and BC res during the five representative BB events listed in Table 1. 2011).Secondly, Pan et al. (2011) showed that BC/CO ERs are strongly influenced by rain washout and/or large environmental RH along the transport pathways.In particular, dry conditions dominated at ICO-OV during event V, with RH values well below 60 %.Moreover, as deduced by the analysis of meteorological analyses based on the Global Data Assimilation System (GDAS) by the National Centre for Environmental Prediction (NCEP), also along the transport path, dry conditions prevailed for this event (average RH values: 59 %), suggesting that rain washout or removal processes in high RH environment were unlikely for this specific event.
Finally, it cannot be completely excluded anthropogenic contribution within BB plumes, which can increase the slope of the BC-CO correlation.As reported by Cristofanelli et al. (2009), at ICO-OV, anthropogenic pollution plumes from the polluted Po basin can be characterised by ERs higher than 14 ng m −3 ppb −1 .FLEXPART retroplume analyses (here not shown) indicated significant footprint emission sensitivity over the central/northern France and the eastern Po basin for the air-masses reaching ICO-OV during event III and V, respectively.Thus, it is possible that mixing with polluted airmasses occurred 1-2 days before the arriving of air-masses at the measurement site.
To provide a more robust characterization of the O 3 /CO and BC/CO ER as a function of the BB plume ages, we considered the 16 BB events identified by FLEXPART (Fig. 5).By excluding event IV (as discussed before) and a further very old BB event (representative emission age: 273 h), we observed for O 3 a good qualitative agreement with previous investigations showing a general increase of ER values with the ageing of BB plumes (e.g.Pfister et al., 2006;Val Martin et al., 2006;Real et al., 2007;Cristofanelli et al, 2009), probably indicating the role of PAN, HNO 3 and organic nitrates, which favor the photochemical formation of O 3 .On the other side a general decrease of ERs with increasing emission age has been observed for BC, possibly related to a more effi- cient scavenging of aerosol particles during long atmospheric transport.On the other side we did not obtain a clear relationship between the type of vegetation burnt or the different source regions with the ERs.This could indicate that the plume aging is the main parameters for determining ERs in relatively aged (i.e.older than 2 days) BB plumes.However, a number of uncertainties related with the actual type of combustion (e.g.smoldering vs. flaming), actual composition of the fuel, its moisture and structure (see Andrea and Merlet, 2001) can affect our results.

Role of anthropogenic emissions
We turn now to the role of anthropogenic emissions in affecting atmospheric composition at ICO-OV based on the observations of O 3 , CO, and BC and on the CO ant calculated by FLEXPART.The 20-day old FLEXPART CO ant has an average value of 22.4 ppb, which corresponds to 17.4 % of the mean CO value observed at ICO-OV during the same period.It is difficult to recognise a clear yearly cycle of CO ant (Fig. 6): the highest monthly CO ant is found for April (27.7 ppb) and the lowest ones in autumn and early winter (January 16.8 ppb, October 19.3 ppb and November 18.5 ppb).In particular, April 2007 was characterised by higher average CO ant (+25 %) than April 2008 and 2009, in good agreement with the highest observed CO at ICO-OV (see Fig. 1).This suggests that the transport variability of anthropogenic CO simulated by FLEXPART can explain part of the inter-annual CO variability observed at ICO-OV.
Larger upper percentiles of FLEXPART CO ant indicate that spring and summer (from April to September) are more affected by episodes with air-masses rich in anthropogenic CO.These events are typically due to relatively fresh emissions.The emission age median value for April-September is only 72 h, whereas it is 114 h from October to February.This can be explained by the more efficient export of CO from the planetary boundary layer to the altitude of ICO-OV during summer.FLEXPART also gives high CO ant in December with upper quantiles similar to those seen for the warm months.However, the median emission age (129 h) is much higher than in summer.Even if a significant fraction of these high CO ant were still related with moderately aged (less than 96 h) European emissions, transport from North America significantly increased the emission age at ICO-OV during specific periods (e.g.3-11 December 2008).This is in agreement with previous work showing that inter-continental transport can be a relevant source of polluted air-masses over Europe (e.g.Stohl and Trickl, 1999;Trickl et al., 2003).
With the aim of systematically investigate the source regions of CO ant as a function of the representative emission ages, we calculated the average contribution to the total CO ant at Mt. Cimone for different source regions tagged back by FLEXPART (i.e.Europe, Asia, North America and other continents).This analysis (see Fig. 7) suggests that the contributions related to the long-range transport from outside Europe became predominant for emission ages larger than 120 h.In particular, anthropogenic emissions from North America appeared to play a predominant role when the CO ant emission age exceeded 192 h.For emission ages older than 120 days not negligible contributions were also tagged to emissions from Asia (ranging from 4.7 to 11.7 %, as a function of emission age) as well as from other continents (ranging from 3.9 to 11.5 %), with predominant contribution related with transport of anthropogenic emission from Africa.
Figure 8 reports the observed anomalies of CO, O 3 for three seasons corresponding to the yearly observed CO maximum (January-April), minimum (May-September) and transition period (October-December) for the 30-month investigation period.CO and O 3 have been reported as a function of observed BC concentrations and CO ant emission age (coloured scales).In order to reduce the effect of diurnal variations caused by thermal circulations and local photochemistry, data have been averaged over a 12 h period centred around 00:00 and 12:00 UTC+1.Correlation between CO and O 3 is strong and positive in May-September (R : 0.79), indicating photochemical O 3 formation.In particular, enhanced CO and O 3 are observed with the presence of high BC concentration (> 400 ng m −3 ).This indicates that polluted air-masses influenced by combustion emissions Figure 8 also shows that, for May-September, the greatest fraction of observations tagged to CO ant < 7 days were mostly associated with enhanced CO, O 3 and BC.How-ever, positive CO and O 3 anomalies with enhanced BC values were also present with older CO ant emission age (up to 15 days), indicating that anthropogenic emissions occurring at very different temporal scales affected tracer values at ICO-OV.In January-April, it is less straightforward to identify clear signatures of CO ant emission age with observed atmospheric composition, even if positive CO still occur with younger CO ant emission ages.However, the highest BC concentrations (400 ng m −3 ) were mostly associated with emission ages ranging from 7 to 11 days, thus suggesting a contribution by slower or farther polluted air-mass transport.From October to December, FLEXPART calculations do not show the presence of significant recent emissions (i.e. less than 5 days), indicating again that during these months ICO-OV is mostly influenced by emissions with longer times of atmospheric transport.In order to quantify the amount of observed variability that can be ascribed to transport diagnosed by FLEXPART, Fig. 9 reports the seasonal correlation between CO, O 3 and BC with CO ant for 24-h time average.The days influenced by the BB transport events identified in the previous Section were not treated in this analysis.Correlation between FLEXPART CO ant with observed CO increases with the length of the time averaging window between 3 and 24 h because the effect of short timing errors in modelled contributions is reduced by increasing the average time window.Correlations decrease for even longer averaging times, so we present the results obtained for the 24 h analysis.FLEXPART CO ant correctly reproduce the absolute range of CO variability observed at ICO-OV (−30/30 ppb): in fact, as reported by Fig. 7, CO ant ranged from 0 to 60 ppb.The slope of the linear correlation between CO and CO ant varies from 0.24 (January-April) to 0.70 (May-September).The calculated CO-CO ant linear correlation coefficients (R) are statistically significant at the 95 % confidence level for all the seasonal aggregations.This suggests that an important fraction of the observed CO variability can be explained by anthropogenic emissions, especially in May-September when the strongest linear coefficient has been found (R = 0.61).
Maximum CO ant -O 3 correlation is also obtained in May-September, indicating the important role of anthropogenic emissions for O 3 formation during the warm months.As for CO, the highest positive O 3 were tagged to emissions younger than 5-7 days even if contribution from older emissions cannot be neglected, indicating that anthropogenic emissions transported on very different scales affected atmospheric composition at ICO-OV.In agreement with the results shown in Fig. 6, for O 3 the linear correlation is negative (and significant at the 95 % confidence level) during October-December and the few observations tagged to young anthropogenic emissions (blue dots) were generally characterised by negative O 3 further stressing the possible role played by NO titration in the local-regional European PBL during these months.
For BC, the linear correlation with CO ant is highest in January-April (R : 0.57): as suggested by FLEXPART it is likely that enhancement in aerosol concentrations is due to relatively aged anthropogenic emissions.Significant correlation can be seen also in May-September with BC enhancements above 400 ng m −3 mainly related to emission ages lower than 7 days.In October-December BC and CO ant appear to be almost uncorrelated with BC concentrations never increased up to 300 ng m −3 and minimum values of CO ant (see Fig. 6).

Conclusions
We quantified the variability of O 3 , CO and BC at the Mt.Cimone global GAW-WMO station (ICO-OV) and we investigated the influence of biomass burning (BB) and anthropogenic emissions younger than 20 days.
In 2007-2009, the average O 3 , CO and BC at ICO-OV were 54 ± 3 ppb, 122 ± 7 ppb and 213 ± 34 ng m −3 (mean ± expanded uncertainty with p < 95 %), with marked seasonal cycles characterized by summer maxima -winter minima for O 3 and BC, and a spring maximum -summer minimum for CO.
According to the FLEXPART simulations, impact of 20day old BB emissions is maximized during the warm months from May to September, however BB emissions traced back by FLEXPART had a significant impact on the observed tracers only during 16 specific events (21 days over the 30 months here considered).Very large variability of enhancement ratios for O 3 /CO and for BC/CO were observed at ICO-OV during these events even if we observed a general increasing (decreasing) trend of O 3 /CO (BC/CO) with increasing emission ages.
As indicated by FLEXPART, 20-day old anthropogenic CO contributions can explain 17.4 % of the mean CO value observed at ICO-OV during the investigating period, with the warm months (from April to September) appearing to be particularly affected by transport events of air-masses rich in anthropogenic CO.Overall, the correlation analysis of ICO-OV tracers with CO ant indicated that the fraction of observed variability ascribed to 20-day old anthropogenic emissions ranged from 22 % (in January-April) to 37 % (in May-September) for CO, from 1 % (in October-December) to 19 % (in May-September) for O 3 and from less than 1 % (in October-December) to 32 % (in January-April) for BC.
The analysis of the seasonal correlation among CO, O 3 and BC observed at the ICO-OV as a function of the CO ant emission age, allowed to identify three main regimes.During May-September, significant positive correlation (R : 0.79) has been found for CO and O 3 , with the highest CO and O 3 values observed in presence of high BC concentration (> 400 ng m −3 ).On the other side, CO and O 3 were negatively correlated (R : −0.51) during October-December: such behaviour may be attributed to an ensemble of processes concurrent in enhancing O 3 with low CO (i.e upper troposphere/lower stratosphere intrusions) and O 3 titration with NO in polluted air-masses along with low photochemical activity.An intermediate situation occurs in January-April with CO and O 3 almost uncorrelated and BC enhancements mainly associated to relatively old (> 7 days) anthropogenic emissions.
This work lead to the conclusion that trace gas and BC concentrations at the ICO-OV, which can be considered representative for SE/MB, are partially driven by local to global scale transport of anthropogenic emissions while, as diagnosed by FLEXPART, BB appeared to have a minor impact on the annual mean.However, it should be clear in mind that these values represent minimal estimates about BB influence, since FLEXPART cannot quantify the contributions coming from the well-mixed background.FLEXPART showed to be accurate in capturing long-range transport, including BB plume (e.g.Stohl et al., 1998Stohl et al., , 2006Stohl et al., , 2007;;Forster et al., 2001;Damoah et al., 2006;Lapina et al., 2008), even if, especially for individual events of long-range BB transport, some uncertainties can affect the identification of the BB transport event at CMN, due to air-mass transport simulation, identification of fires and estimate of emissions.However, as also shown by MOPITT satellite measurements for two case studies of long-range BB transport, FLEXPART seems to be able in describing the impact of BB emissions.Nevertheless, further work is needed to explore the validity of the results here obtained for other sites in the SE/MB region, that can be affected by different transport patterns and/or emissions.

Fig. 1 .
Fig. 1.Monthly means of CO, O 3 and BC at ICO-OV for the years 2007, 2008 and 2009.The error bars denote the average expanded uncertainties with p < 95 %.

Fig. 3a .
Fig. 3a.Time series of observed CO at ICO-OV (black dots) with FLEXPART CO fire (red line).Observations at ICO-OV are tagged with the FLEXPART fire emission age (colour scale) when FLEXPART fire CO exceeded 10 ppb.FLEXPART CO fire is plotted with a positive constant bias of 50 ppb to reduce the extent of y-axis.

Fig. 3b .
Fig. 3b.Time series of observed CO concentration at ICO-OV (black dots) with FLEXPART CO fire (red line) for 11 March-6 April 2009.Observations at ICO-OV are tagged with the FLEXPART fire emission age (colour scale).FLEXPART CO is plotted with a positive bias of 50 ppb to reduce the extent of y-axis.

Fig. 5 .
Fig. 5. ERs of O 3 (red) and BC (black) relative to CO observed for the 16 BB events detected at CMN by FLEXPART.

Fig. 6 .
Fig. 6.Monthly distributions of FLEXPART daily CO ant (upper plot) with related representative emission ages (bottom plot) at ICO-OV during the period January 2007-July 2009.The boxes and whiskers denote the 10th, 25th, 50th, 75th and 90th percentiles.Bold lines represent average value.Data at 3-h resolution have been shown.

Fig. 7 .
Fig. 7. Fraction of CO ant at CMN (y-axis) as a function of source regions (stacked coloured bars) and representative emission age (x-axis).

Fig. 8 .
Fig. 8. Observed CO and O 3 anomalies ( CO and O 3 ) for January-April (top), May-September (middle) and October-December (bottom).Left column shows data tagged with BC concentrations observed at ICO-OV, while right column shows data tagged with emission age of CO ant from FLEXPART.Linear correlation coefficients between CO and O 3 (R) are reported for each plot in the left column.Data are averaged over a 24-h time window.

Fig. 9 .
Fig.9.Observed CO (first row), O 3 (second row) and BC (third row) versus FLEXPART CO ant for January-April (left column), May-September (middle column) and October-December (right column).Data are averaged over 24 h and are tagged with emission age of CO ant (colored scale).The linear correlation coefficients (corr) is reported for all the scatter-plots while the slope values (slope) are only reported for the CO-CO ant correlation.
BC res ) with respect to the mean diurnal variation estimated in the 21 days around the BB event.