Biotic stress: a signiﬁcant contributor to organic aerosol in Europe?

We have investigated the potential impact on organic aerosol formation from biotic stress-induced emissions (SIE) of organic molecules from forests in Europe (North of Lat. 45 ◦ N). Emission estimates for sesquiterpenes (SQT), methyl salicylate (MeSA) and unsaturated C 17 -compounds, due to di ﬀ erent stressors, are based on experiments 5 in the Jülich Plant Atmosphere Chamber (JPAC), combined with estimates of the fraction of stressed trees in Europe based on reported observed tree damage. SIE were introduced in the EMEP MSC-W chemical transport model and secondary organic aerosol (SOA) yields from the SIE were taken from the JPAC experiments. The estimated current-situation SIE in Central and Northern European forests are 10 found to contribute substantially to SOA in large parts of Europe. It is possible that the SIE contributes as much, or more, to organic aerosol than the constitutive biogenic VOC-emissions, at least during some periods. Based on the assumptions in this study, SIE-SOA are estimated to constitute between 50 and 70 % of the total biogenic SOA (BSOA) in a current-situation scenario where the biotic stress in Northern and Central 15 European forests causes large SIE of MeSA and SQT. An alternative current-situation scenario with lower SIE, consisting solely of SQT, leads to lower SIE-SOA, between 20 and 40 % of the total BSOA. Hypothetical future scenarios with increased SIE, due to higher degrees of biotic stress, show that SOA formation


Introduction
The emissions of biogenic volatile organic compounds (BVOC) by forests are the major 10 sources of hydrocarbons to the atmosphere (Guenther et al., 2012;Lamarque et al., 2010;Simpson et al., 1999). Photo-oxidation of BVOC, in the presence of nitrogen oxides (NO x ), contributes to the formation of tropospheric ozone and leads to secondary organic aerosol (SOA) particle formation (Hallquist et al., 2009). Many BVOC, e.g. isoprene, α-pinene, and sesquiterpenes (SQT), are unsaturated and react with all main 15 oxidants in the atmosphere: OH, ozone and NO 3 , while the saturated BVOC preferably react with OH. SOA formation is caused by the gas to particle transformation of some of the oxidation products, depending on e.g. their vapour pressure. Studies using carbon-14 and other tracer compounds have shown that such biogenic SOA (BSOA) is often the major contributor to ambient organic aerosol (OA) at rural, and even some leased in the unperturbed, non-stressed state of the vegetation. In BVOC emission algorithms, the constitutive emissions are assumed to be under the control of meteorological factors, especially temperature and light, as well as phenological cycles (Guenther et al., 2012). The SOA forming potential of these emissions, especially isoprene and monoterpenes (MT), and to a lesser extent SQT, have been extensively investi-10 gated in laboratory studies over many years (e.g. Bonn and Moortgat, 2003;Lee et al., 2006;Hallquist et al., 2009, and references therein;Winterhalter et al., 2009;Donahue et al., 2012;Jaoui et al., 2013).
Algorithms derived from such laboratory data have been applied in atmospheric chemical transport models (CTMs), using a variety of techniques to treat BVOC emis- 15 sions, chemistry and gas-particle partitioning (Bowman et al., 1997;Andersson-Sköld and Simpson, 2001;Schell et al., 2001;Kanakidou et al., 2005;Donahue et al., 2006Donahue et al., , 2009Simpson et al., 2007;Kroll and Seinfeld, 2008;Hallquist et al., 2009;Jimenez et al., 2009;Murphy and Pandis, 2009;Bergström et al., 2012;Li et al., 2013). For many years OA and SOA were largely underestimated by some CTMs compared to field ob-sions as stress-induced emissions (SIE). Berg et al. (2013) investigated the impact of bark beetle infestations on MT emissions and SOA formation in western North America. MT emissions due to bark beetles may both increase (during attack) and decrease (after tree-death) and Berg et al. found that beetle infestations in pine trees can have a significant regional impact on SOA con-10 centrations (up to 40 % increase) during some years; responses may be substantially larger if spruce trees are infested.
Recently, SOA mass yields from the laboratory studies in the Jülich Plant Atmosphere Chamber (JPAC), with real plants as sources, showed that terpenoidic SIE, such as sesquiterpenes, and phenolic BVOC (e.g. methyl salicylate, MeSA), originat- 15 ing downstream of the shikimate pathway (e.g., Wildermuth, 2006), are very efficient in forming SOA (Mentel et al., 2013). Their particle mass yields are 3-4 times larger than those of MT, see Sect. 2.2. In a case where a spruce was infested by Cinara pilicornis (honeydew generating lice) unsaturated C 17 -BVOC were observed with particle mass yields six times higher than those of MT (Mentel et al., 2013). Given that SIE 20 are a ubiquitous source of BVOC and thus SOA in forests, these laboratory findings suggest that SIE may account for a significant fraction of ambient SOA mass. Neglect of the SIE in models might explain some of the discrepancies between observed OA and model predictions.
If the SIE increase in the future (e.g. with increasing frequencies of insect damage, Introduction of the emissions, dependencies on plant history, and adaption to stresses, as well as the scaling of emissions from leaf-level to regional scale (Arneth and Niinemets, 2010;Niinemets et al., 2010). All of these aspects require substantial research, and, as discussed by Arneth and Niinemets (2010), building modules that simulate induced emissions is a difficult if not impossible task at the current level of understanding. 5 SIE are generally neglected in atmospheric models because of (i) a lack of awareness of their possible importance, (ii) lack of suitable data and information about the distribution of stress and its specific effects, and (iii) lack of appropriate SOA formation algorithms. However, given these difficulties, it is still appropriate and important to assess the order of magnitude of such SIE-SOA contributions to ambient aerosol.
Here we want to demonstrate how the neglect of SIE in models affects current SOA predictions and the possible effects of SIE on SOA in the future.
The goal of this model study is to draw attention to the possible importance of SIE emissions, and to make a first estimate of their contribution to SOA formation; both in current conditions and with a projection of what could happen in the future under the 15 assumption that stress to plants becomes more frequent or severe. We combine experimental emission and SOA-formation results from JPAC with estimates of the possible geographical extent of the SIE, and use the EMEP MSC-W regional chemical transport model (Simpson et al., 2012;Bergström et al., 2012) to assess SOA formation over Europe. To construct continental scale emission scenarios for the SIE, we make use 20 of European and national forest damage reports. These are in general based on ocular inspection of defoliation and insect infestation in European forests; the inspections are performed regularly and follow well defined protocols (Lorenz, 2010;Ferretti et al., 2010).
A major strength of this study is that both the emission factors of SIE/constitutive 25 emissions and SIE-SOA mass yields are determined from the same experimental JPAC data for relevant forest species. As shown for new particle formation and SOA yields the results from the JPAC studies can be transferred to atmospheric situations . Uncertainties arise from the estimates of the fraction and spatial Introduction distribution of infested trees, as well as limited knowledge of the seasonal variation of some of the infestations. Despite these uncertainties this work, by use of selected scenarios, clearly shows that SIE and SIE-SOA deserve closer consideration as potentially significant sources of organic aerosol in Europe.

5
This modelling assessment of potential effects of biotic SIE is based on: (1) evaluation of experimentally observed BVOC emissions by insect infested plants and their photochemical conversion to SOA, (2) estimation of the potential fraction of infested trees in European forests, (3) construction of future scenarios with increased fractions of infested trees. The stepwise procedure employed in the present study is described 10 in detail below and a summary of the resulting model scenarios is given in Table 1.

Experimental
SOA mass yields and emission ratios were determined in the same experiments in the JPAC as published in Mentel et al. (2013). In short, JPAC consists of three continuously stirred flow reactors, made of Borosilicate glass, which are placed in temperature con-15 trolled housings. One of these is operated as a reaction chamber and SOA is formed therein by photooxidation-and ozonolysis products  airstream of the plant chamber were averaged for the same time intervals as in which the SIE-SOA mass yields were determined.

BVOC emission factors for infested trees
In order to keep our model results transparent we used a simplified direct approach for preparing the model emission scenarios. We used the standard EMEP emissions for 5 monoterpenes (Simpson et al., 2012) and applied emission ratios for SIE/MT based on experimental data from JPAC (Mentel et al., 2013) to set the SIE in the model. The SIE considered here (SQT and MeSA) are of de novo type (Kleist et al., 2012), i.e. they are emitted in connection with biosynthetic production. The emission ratios were determined in JPAC under steady state conditions as de-10 scribed above. The direct use of JPAC-derived data for application or extrapolation to ambient conditions has been confirmed in earlier studies for particle formation rates and SOA mass yields for Boreal tree species , the chemical composition of the resulting SOA (Kiendler-Scharr et al., 2009b), interaction of isoprene and MT emissions (Kiendler-Scharr et al., 2009a), and the distribution of highly oxi-15 dized aerosol precursors (Ehn et al., 2012). The use of real plant emissions and close to ambient concentrations make us think that the laboratory-derived data from JPAC provide the best-available estimate of SIE-SOA yields for our purposes. From the observations by Mentel et al. (2013), we constructed three biotic stress scenarios. Case 1 treats aphid infestation with enhanced SQT emissions with SQT/MT = 20 2.4 (mass based ratios). Case 2 covers aphid infestations which caused enhanced emissions of SQT (SQT/MT = 4.9) and triggered MeSA emissions via the shikimate pathway (MeSA/MT = 3.75). Case 1 and Case 2 were assumed to be typical for all trees of Boreal and Central European forests. As the SIE emissions are of de novo type, they were only switched on during daytime.

25
A further Case 3 was constructed from an experiment investigating infestation of a spruce by Cinara pilicornis, which led to strong emissions of several unsaturated linear C 17 -BVOC with an emission factor C 17 -BVOC/MT = 18. Mentel et al. (2013)  not determine whether the C 17 -compounds originated from the plant or the infesting insects; the C 17 -BVOC emissions were considered as originating from the coupled plant-insect system. The louse under consideration here, Cinara pilicornis, belongs to the family of bark lice that produce honeydew, which is collected by bees. Such bark lice are of economic 5 interest for beekeepers; observations by beekeepers in Baden-Württemberg (BW) in south west Germany, show that such infestations (or more precisely the honey production from Cinara pilicornis and similar infestations) are strongly varying from year to year and have high seasons during June/July (http://www.stockwaage.de/). Accordingly we constructed a SIE emission pulse of C 17 -BVOC, which was limited in time and spatial extension. As the C 17 -BVOC emissions in JPAC were 2-3 times larger at daytime than during night, we switched them on only during daytime in the model, like the other two SIE. Mentel et al. (2013) focused on tree species from the Boreal region and from Central Europe. Therefore, we have focused our analysis on Northern and Central Europe, and 15 implement SIE for areas north of Lat. 45 • N, although it may be assumed that stress also affects the emissions from plants growing in other regions.

Estimation of fraction of infested trees
Since the observed emission factors only consider infested trees, the fraction of infested trees has to be estimated. It is difficult to estimate the degree of insect in-20 festations on larger scales in real forests. In the present study we chose a relatively simple approach to make what we believe to be a rough but reasonable estimate of the present-day situation. We base the estimate on regular surveys of the European forests. ICP Merilä et al. (2007) stated that about 20 10-12 % of the pines showed a significant degree of defoliation (> 25 %). Similar to the European situation about the same order of trees (about 10 %) showed damage by insects and fungi with in general more importance of fungi. By combining these two types of forest observation, we conclude for our cases that the fraction of trees with significant defoliation (> 25 %) may serve as a first order approach to assess the frac-25 tion of stressed trees in the current situation. Consequently, we adopt these numbers and assume that the fraction of currently stress affected trees is 10 % for Lat. > 60 • N and 20 % between 45 • N and 60 • N. This will constitute our base case for impact of contemporary SIE (Case 1 and Case 2). For the future scenarios our hypothesis is that the degree of infestation may increase if climate changes unfavorably for an established vegetation. Considering that the knowledge about the present-day degree of infestation is limited, it is even more problematic to describe how SIE will develop in the future. However, we use a similar approach as for the current situation but take it a step further and assume that insect infestations may affect trees that today are at the next reported degree of defoliation (> 10 %). This will then include about 2/3 of the trees in Central Europe (Fischer et al., 2012) and about 50 % of the trees in the Boreal forests (Finnish National Report 2007: Merilä et al., 2007). This may be considered as a severe-case scenario of a possible future.

ACPD
These high degrees of infestation were used to illustrate how severe biotic stress can enhance SIE and contribute to SOA and we address these as two extreme future scenarios, Case 1F and 2F. Given the uncertainty of estimating future SIE emissions, the MT emissions were, for simplicity, kept at the current level. 15 The construction of the Case 3 with C 17 -BVOC emission from the Cinara pilicornis infestation is somewhat more indirect. Here we make use of the fact that the honeydew produced by Cinara pilicornis (and other bark lice) is a source of a certain kind of honey, the forest honey. Detailed observational data on forest honey production exist in Baden-Württemberg (see http://www.stockwaage.de/) so we select this region of 20 south west Germany for an episodic test case. The relation between infestation and forest honey production is well known to beekeepers (e.g. see http://stockwaage.de/) and from their statistics we can extract the seasonality and the annual variation of the forest honey production as related to Cinara pilicornis infestation. In a good honey year the infestation is widespread, even if there are some local variations, and lasts through Introduction  Mentel et al. (2013) observed C 17 -BVOC emission from the Cinara pilicornis/Norway spruce system, we assume for simplicity that all spruce in BW are infested and the resulting SIE occur during the months June and July, with the given C 17 -BVOC/MT ratio of 18. The forests in BW consist of 38 % spruce (http://www.mlr. baden-wuerttemberg.de/Die_Baumarten/507.html). Other conifers, mainly fir and pine, 5 make up another 19 % of the forest. The rest are broadleaf species mainly beech (21 %) and oak (7 %). This causes a small flaw in our concept of applying the emission ratios as observed in JPAC, as the emission strength of spruce and the other conifers and specifically of the broadleaf species may be different. For simplicity, we assume that during an active year all spruce are affected and that these also accounts for 38 % of the emissions. As a consequence the C 17 -BVOC/MT emission ratio from the JPAC experiment is weighted by the factor 0.38 in BW. Although the assumption that all spruce trees in BW are heavily infested may be viewed as an extreme case we note that other tree species may also be simultaneously infested by lice and, on the beekeeper web page http://stockwaage.de/index.php/rueckblick, there is indeed a year described (2006) when lice even infested deciduous trees and contributed to honeydew production.

The EMEP MSC-W model
The standard EMEP MSC-W chemical transport model has been described in detail by Simpson et al. (2012) with measurements of many different compounds (e.g. Jonson et al., 2006;Simpson et al., 2006;Fagerli and Aas, 2008;Aas et al., 2012;Bergström et al., 2012;Genberg et al., 2013). For the present study meteorological data from the European Centre for Medium Range Weather Forecasting Integrated Forecasting System (ECMWF-IFS) model (http://www.ecmwf.int/en/research/modelling-and-prediction) were used; 5 all simulations were performed for the (meteorological) year 2007. This study focuses on OA from biogenic emissions. The constitutive emissions of monoterpenes and isoprene are calculated in the model, using near-surface air temperature (T 2 m ) and photosynthetically active radiation. Standard condition emission factors for forests are based on tree species specific monoterpene emission factors for 10 115 different tree species, combined with detailed maps of the distribution of the trees in Europe, as described by Simpson et al. (2012). Such emissions are of course very uncertain given the lack of underlying measurements for European ecosystems, as discussed in more detail in Simpson et al. (1999Simpson et al. ( , 2012 and Keenan et al. (2009), and illustrated for European isoprene emissions (with results from 4 models) in Langner 15 et al. (2012). The EMEP values are believed to be well grounded in recent reviews of emissions rates (Simpson et al., 2012).
Some updates have been done to the model, compared to Bergström et al. (2012). As in the standard EMEP model (Simpson et al., 2012), emissions from open biomass fires were taken from the FINNv1 inventory (Wiedinmyer et al., 2011), and hourly vari-20 ation of anthropogenic emissions were used instead of the simpler day-night system used in earlier EMEP models. For this study, we use updated emissions of primary organic aerosol (POA), with a recently developed inventory for residential combustion of biofuels (Denier van der Gon et al., 2014). In the present study we set the background concentration of organic aerosol to 0.4 µg m −3 , to prevent an overestimation of   Shrivastava et al., 2008); this "ageing" trans- 5 forms the species to lower volatilities that may partition to the particle phase. The base case OA scheme used here is almost identical to the PAP-model in Bergström et al. (2012); the only difference is that a small emission of sesquiterpenes is added (equal to 5 %, by mass, of the daytime MT emissions) based on observed emissions from unstressed plants by Mentel et al. (2009).

10
Very simplified mechanisms for SOA formation from SQT, MeSA and C 17 -BVOC were added to the model. Fixed SOA (mass) yields, based on experimental data (Mentel et al., 2013), were used for these three model compounds. Note that the SOAyield from SQT oxidation (17 mass-%) is based on experimentally determined yields from SQT-emissions from aphid infested Norway Spruce (see Mentel et al., 2013); here 15 we assume the same SOA-yield from all SQT-emissions. For MeSA and C 17 -BVOC the yields are 22 % and 33 %, respectively: where O x is a general oxidant (O 3 or OH; NO 3 may also react with SQT, and possibly with C 17 -BVOC, but, since we only consider day-time SIE in this study, the NO 3 -reaction is of minor importance compared to the fast O 3 and OH reactions). The parentheses around the oxidants indicate that these oxidants are not depleted in the chemical mechanism. As in Simpson et al. (2012)  of the BVOC produces only the compounds specified by the VBS scheme or the fixedyield non-volatile SIE-SOA components. For such compounds, the chemistry is assumed to be "oxidant-neutral", that is, we assume that as much O 3 or OH is reformed in the neglected part of the chemistry, as is consumed in the initial BVOC-reactions. This procedure ensures that the ozone chemistry will be the same as in the standard 5 photochemistry version of the EMEP MSC-W model. SOA-formation from sesquiterpenes is rapid; in the model we use rates based on the β-Caryophyllene chemistry scheme in the Master Chemical Mechanism (MCM v3.2 (Jenkin et al., 2012), via website: http://mcm.leeds.ac.uk/MCM). For the C 17 -BVOC no kinetic information is available. As observed in JPAC, the C 17 -BVOC had a short lifetime 10 with respect to oxidation by O 3 and OH; for simplification, we therefore applied the same OH and O 3 reaction rate coefficients (1.97×10 −10 cm 3 molecule −1 s −1 and 1.16× 10 −14 cm 3 molecule −1 s −1 , respectively) for the C 17 -BVOC as for β-Caryophyllene.
MeSA is much more stable in the atmosphere (Canosa-Mas et al., 2002) and, based on experimental data from JPAC, an OH-reaction rate coefficient of 4 × 15 10 −12 cm 3 molecule −1 s −1 was used. The low reaction rate of MeSA with OH allows for significant MeSA concentrations during the night and since MeSA is a phenolic compound we must also consider the reaction with NO 3 : The rate of the MeSA + NO 3 reaction is not known and neither is the SOA-yield (α) 20 of the reaction; night-time degradation of MeSA by NO 3 reaction could possibly be fast (Canosa-Mas et al., 2002). Canosa-Mas et al. (2002) assumed that MeSA could react as fast with NO 3 as phenol does (k = 3.8 × 10 −12 cm 3 molecule −1 s −1 ), but the MeSA + NO 3 reaction may be slower, because the MeSA molecule may form an internal hydrogen bond between the OH-hydrogen and the ester group leading to an in-25 creased stability compared to phenol. The MeSA + OH reaction, for example, is seven times slower than the phenol + OH reaction (IUPAC: http://www.iupac-kinetic.ch.cam. ac.uk (2008) MeSA+NO 3 reaction is about an order of magnitude slower than the phenol+NO 3 reaction. Details of the measurements regarding the determinations of the rate constants will be published elsewhere.
In the present study we therefore used k = 5.4 × 10 −13 cm 3 molecule −1 s −1 for the MeSA + NO 3 reaction, equal to the phenol + NO 3 reaction rate divided by seven (the 5 scaling factor of the OH reaction). The rate coefficient was combined with two different SOA-yields for the reaction, 0 or 22 mass-%, resulting in two different sensitivity test cases: Canosa-Mas et al. (2002) suggest that photolysis may be the most important daytime loss process for MeSA but other studies have shown that MeSA and related compounds have "striking photostability" (e.g., Acuna et al., 1981) and therefore we neglect this process in the model simulations.
Deposition of gas-phase MeSA is a potentially important loss process since the ox- 15 idation rate is relatively slow. Karl et al. (2008) measured the Henry's law constant for MeSA and obtained a value of ca 34 M atm −1 , that is, MeSA is somewhat more soluble than CH 3 CHO but much less soluble than HCHO. In the standard set-up of the present study we treat MeSA-deposition in the same way as CH 3 CHO (and other higher aldehydes) in the EMEP model (Simpson et al., 2012); this means wet deposi-20 tion is neglected and that the dry deposition is relatively slow. Two sensitivity tests were performed regarding the MeSA deposition: (A) neglecting both dry and wet deposition, (B) assuming dry and wet deposition to be as efficient as for HCHO.

Model emission scenarios
In total five different biotic stress emission scenarios are explored in this study and 25 compared to a reference simulation without stress induced emissions. i.e., biotic stress is assumed to be present during the whole growing season. This is a simplification, since many forms of biotic stress are of more limited duration (e.g. Hakola et al., 2006), but various stressors may be active at different times of the year. In the present study the focus is on getting estimates of the potential relative importance of SIE compared to the constitutive BSOA for long-term OA concentrations.

Case 1 -Sesquiterpene emissions from biotic stress -current situation
The first SIE scenario is based on experimental data for aphid infested Norway spruce with a SQT/MT emission ratio of 2.4. In the EMEP model simulation for Case 1 we apply 10 % of these emissions to all monoterpene emitting plants north of Latitude 60 • N,

Case 1F -Increased degree of infestation -sesquiterpene emissions
The first "future" scenario, Case 1F, use the same biotic stress emission ratios as Case 1 but a larger proportion of the vegetation is assumed to be infested: 50 % in the boreal region (north of 60 • N) and 2/3 of the trees in the 45-60 • N region. This leads to 15 SQT/MT emission ratios of 120 % and 160 %, respectively.

Case 2F -Increased degree of infestation -methyl salicylate + sesquiterpene emissions
The second scenario of increased biotic stress, Case 2F, use the same assumptions as Case 2 regarding the emissions from infested trees and the same proportion of infes- during June and July, leading to large C 17 -BVOC emissions (6.8 × the MT emissions) and substantial SQT emissions (0.38 × MT) during this period. The SOA-yield from the oxidation of C 17 -BVOC is 33 % (Mentel et al., 2013).

Results and discussion
The EMEP MSC-W model for OA was thoroughly evaluated against observations by 10 Bergström et al. (2012). One of the conclusions of that study was that simple VBS based OA models can give reasonably good results for summer conditions. Since the biotic SIE primarily occur during the summer half-year we focus mainly on results for this period here. slightly closer to observations than Case 1 and both SIE-scenarios are in better agreement with observations than reference scenario (Case 0) that neglects biotic stress emissions.

Current situation
Since the model SIE are treated as a simple fraction of the "unstressed" MT emissions an improvement in model results when adding SIE is not a proof that the stress 5 induced emissions are correctly modelled; the model improvement could also be due compensation of underestimated regular BVOC emissions (see the example given for the Swiss site Payerne in Bergström et al., 2012). However, the model results for Hyytiälä show that the additional SIE emissions do not lead to unrealistic model OA concentrations, and that such SIE can have a significant effect on ambient OA levels 10 during the summer period.
Model calculated OM 2.5 (organic matter in PM 2.5 ) and the relative fraction of BSOA, from the model simulation excluding all SIE from vegetation (Case 0), for the summer half-year (April-September) 2007, are shown in Fig. 3. The modelled BSOA is low in most of Europe. The relative contribution of BSOA to modelled regional background  that, even with realistic present-day levels of biotic stress, it is possible that, at least for some periods, the stress induced emissions are more important for organic aerosol production than the constitutive emissions of BVOC.

Future scenarios
We have estimated the potential increase of OM 2.5 due to the much higher degree of 5 infestation assumed in the two future scenarios (Case 1F and 2F). The differences in summer-half-year mean OM 2.5 between Case 1F and Case 1, and between Case 2F and Case 2 are shown in Fig. 5. If the biotic SIE increase to the high levels tested in these scenarios a substantial increase in organic particle mass can be expected. The results from the Case 2F (biotic MeSA+SQT) simulation indicate that SIE-SOA could 10 potentially become an important source of regional background PM 2.5 in large parts of Central/Eastern Europe; the increase in OM 2.5 compared to the present-day Case 2-scenario is larger than 1.5 µg m −3 in parts of Central and most of Eastern Europe.
The Case 1-type scenario, with only SIE of SQT included, have a much lower BSOAforming potential but still the SIE-SOA production may become fairly substantial in the 15 future scenario (Case 1F); the increase compared to the corresponding current situation scenario (Case 1) is above 0.3 µg m −3 in much of Central, Northern and Eastern Europe.

Importance of stress induced MeSA emissions -sensitivity tests
Considering the high emissions of MeSA and high SOA-yield from this component found MeSA mixing ratios of ∼ 100 ppt(v) within and above the canopy of a walnut agroforest.
The amount of SIE-SOA produced in the model in Case 2 is based on the observed SOA formation in the JPAC plant chamber experiments (Mentel et al., 2013); in addition it depends on the assumptions regarding the deposition of MeSA, the MeSA+NO 3 -10 reactivity and the SOA-formation from the NO 3 -reaction. We illustrate the sensitivities in Fig. 7, which shows the mean diurnal variation of SIE-SOA at Hyytiälä for the period March-October.
The modelled SIE-SOA is not very sensitive to the MeSA deposition; the differences between the setups with no deposition or faster deposition (as HCHO) to the base 15 case deposition (as CH 3 CHO) are only about +5 % and −7 %, respectively; similar differences (+3 to +7 % and −3 to −10 %) are seen in the part of the model domain where the SIE are included in Case 2 (the relative differences are larger at longer distances from the SIE regions).
If the MeSA+NO 3 -reaction occurs at the rate tested in this study (k R4 = 5.4 × 20 10 −13 cm 3 molecule −1 s −1 ), and has the same SOA-yield as MeSA+OH, a substantial night-time production of SIE-SOA is seen in the model. most of the region with significant SIE in Case 2, the modelled SIE-SOA is reduced by ca 15-25 %, compared to the base case with only OH-reaction included). The standard Case 2 simulation (as used in Sect. 3.1), which neglects the MeSA + NO 3 reaction, leads to SIE-SOA concentrations between those obtained in the two test case simulations that include the MeSA+NO 3 reaction (with 0 and 5 22 % SOA-yield); the average SIE-SOA (for March-October) are 0.34 µg m −3 (only OH-reaction), 0.27 µg m −3 (incl. NO 3 -reaction with 0 % SOA-yield) and 0.44 µg m −3 (incl. NO 3 -reaction, 22 % SOA-yield). Neglecting the NO 3 -reaction means that more MeSA survive during night-time and can produce SIE-SOA the following day, via OHreaction. As seen in Fig. 6 almost all MeSA is lost during night when the NO 3 -reaction 10 is taken into account.
The SOA formation from the NO 3 reaction is difficult to assess, but if the ambient MeSA levels are indeed often at a level higher than 100 ppt it is worthwhile to study the MeSA -NO 3 yield in more detail. Moreover, recent data indicate very efficient SOA formation from NO 3 reactions of e.g. pool emitted MT (Fry et al., 2009.

Regional bark lice infestation -Case 3
The very large episodic C 17 -BVOC emissions in Baden-Württemberg, simulated in Case 3, lead to a large production of SIE-SOA in the infested region. For the two-month infestation period the average modelled SIE-SOA contribution to PM 2.5 is larger than 3 µg m −3 in BW and above 0.5 µg m −3 in all of southern Germany. For BW the mod-20 elled regional background PM 2.5 concentration is more than twice that in the reference case without SIE, as shown in Fig. 8. Since Case 3 assumes a wide-spread, severe infestation of spruce trees in BW, it could be considered a "worst-case" scenario for lice infestation in BW (or best-case for the honey-production) but it should be pointed out that bark lice may infect other tree species as well and that infestations are likely to 25 occur simultaneously in other regions of Europe. The very high modelled impact of this scenario also indicates that even a much more limited degree of infestation could lead In this study we estimated the degree of stress from forest damage observations in Europe and forest honey production data from beekeepers in south west Germany. Combining these estimates with the plant chamber experiment based stress induced 15 emission factors we constructed different SIE scenarios, and implemented SIE-SOA formation in the EMEP MSC-W model, in order to get a first model estimate of the potential impact of SIE in Europe north of Lat. 45 • N.
The implementation of SIE in the model resulted in less bias and better correlation coefficients, for particulate organic carbon at a forest site in Finland, compared 20 to the standard model simulation with only constitutive biogenic emissions. However, based on these results alone, we cannot draw the conclusion that this is proof of the importance of SIE, since any increase in BVOC emissions, or BSOA yields from "unstressed" BVOC, would have essentially identical effects on total organic aerosol concentrations: we cannot distinguish SIE induced SOA from SOA from unstressed veg-25 etation with higher emission rates. On the other hand we also cannot distinguish between much higher abundance of SIE-SOA already at present and overestimated SOA from constitutive emissions. Further, the fate of the SIE in the atmosphere is uncertain, as shown by the sensitivity study for MeSA, and the modelling of SOA is still so fraught with difficulty that we cannot use model-measurement discrepancy for total OA concentrations to establish the likely level of SIE-SOA.
Having stated this, we want to point out that if our assumptions regarding the mag-5 nitude of the SIE are overall about right, it is possible that, at least for some periods, the SIE including MeSA could be more important for organic aerosol production than the constitutive emissions of BVOC. SIE of SQT have a lower BSOA-forming potential but still the SIE-SOA production from SQT may become fairly substantial in the future in much of Central, Northern and Eastern Europe. The emissions of unsaturated 10 C 17 -BVOC from insect infested vegetation, although episodic and regional, could have a large impact on SOA formation. Measurements using real plant emissions have shown that SIE can have higher potential to form SOA mass than constitutive emissions. On laboratory scale SIE can dominate SOA formation, as is also reflected by the model calculations. But up-scaling 15 of laboratory results is complicated because the contribution of SIE to biogenic emissions in the air over large areas is uncertain. We constructed plausible scenarios, for Central and Northern Europe, by using independent data on European forest systems. This approach is only a first step and may lead to over-or underestimations of the importance of SIE-SOA. However, without consideration of SIE-SOA modeling scenarios suggest that totally non-infested plants are not likely to be common and thus some stress is the normal state of vegetation. Neglecting SIE in modelling therefore is unrealistic.  , 13, 3149-3161, doi:10.5194/acp-13-3149-2013, 2013. 13606, 13607 Berge, E. and Jakobsen, H. A.: A regional scale multi-layer model for the calculation of longterm transport and deposition of air pollution in Europe, Tellus, 50, 205-223, 1998. 13614  Biogeosciences, 5, 761-777, doi:10.5194/bg-5-761-2008, 2008. 13606 Ehn, M., Kleist, E., Junninen, H., Petäjä, T., Lönn, G., Schobesberger, S., Dal Maso, M., Trimborn, A., Kulmala, M., Worsnop, D. R., Wahner, A., Wildt, J., and Gas phase formation of extremely oxidized pinene reaction products in chamber and ambient air, Atmos.