Influence of local air pollution on the deposition of 1 peroxyacetyl nitrate to a nutrient-poor natural grassland 2 ecosystem

19 Dry deposition of peroxyacetyl nitrate (PAN) is known to have a phytotoxic impact on plants 20 under photochemical smog conditions, but it may also lead to higher productivity and threaten 21 species richness of vulnerable ecosystems in remote regions. However, underlying 22 mechanisms or controlling factors for PAN deposition are not well understood and studies on 23 dry deposition of PAN are limited. In this study, we investigate the impact of PAN deposition 24 on a nutrient-poor natural grassland ecosystem situated at the edge of an urban and 25 industrialized region in Germany. PAN mixing ratios were measured within a 3.5 months 26 summer to early autumn period. In addition, PAN fluxes were determined with the modified 27 Bowen ratio technique for a selected period. The evaluation of both stomatal and non- stomatal deposition pathways was used to model PAN deposition over the entire summer- autumn period. We found that air masses at the site were influenced by two contrasting 3 pollution regimes, which lead to median diurnal PAN mixing ratios ranging between 50 and 4 300 ppt during unpolluted and between 200 and 600 ppt during polluted episodes. The 5 measured PAN fluxes showed a clear diurnal cycle with maximal deposition fluxes of 6 ~ -0.1 nmol m -2 s -1 (corresponding to a deposition velocity of 0.3 cm s -1 ) during daytime and a 7 significant non-stomatal contribution was found. The ratio of PAN to ozone deposition 8 velocities was found to be ~0.1, which is much larger than assumed by current deposition 9 models. The modelled PAN flux over the entire period revealed that PAN deposition over an 10 entire day was 333 µg m -2 d -1 under unpolluted and 518 µg m -2 d -1 under polluted episodes. Besides, thermochemical decomposition PAN deposition accounted for 32% under unpolluted 12 episodes and 22% under polluted episodes of the total atmospheric PAN loss. However, the impact of PAN deposition as a nitrogen source to the nutrient-poor grassland was estimated to be only minor, under both unpolluted and polluted episodes.

and biospheric nitrogen cycle through dry deposition (Singh, 1987). Besides, locally produced 25 PAN may also impact on ecosystems downwind of pollution sources. While high PAN 26 mixing ratios (> 15 ppb), prevailing under strong photochemical smog conditions, PAN is 27 known to be phytotoxic and may harm plant tissues significantly (Temple and Taylor, 1983), 28 the impact of PAN deposition under less extreme conditions and for lower PAN mixing ratios 1 is not yet clear. As a nitrogen source, PAN deposition may also lead to higher productivity 2 and may threaten species richness especially in vulnerable ecosystems (Stevens et al., 2010). 3 Previous studies on the surface-atmosphere exchange of PAN showed that PAN is deposited 4 to vegetation. On the one hand, chamber experiments on PAN uptake on both leaf and plant 5 level (Okano et al., 1990;Sparks et al., 2003;Teklemariam and Sparks, 2004) found a direct 6 relationship between PAN uptake and stomatal conductance. They suggest that stomatal 7 uptake is the major pathway of PAN into leaves. On the other hand, previous studies have 8 also shown the existence of non-stomatal deposition of PAN, mainly associated with the 9 uptake by the leaf cuticles ( habitats, where additional nitrogen input via deposition may play a significant role, are often 21 dominated by grass species rather than trees. In this case study, we investigate the influence 22 of polluted and non-polluted air masses on the dry deposition of PAN at a nutrient-poor 23 natural grassland ecosystem in Central Europe. PAN mixing ratios were measured and 24 analysed over a three months period under two contrasting pollution regimes. For a selected 25 period, we also derived PAN fluxes with the flux-gradient approach, employing a newly 26 developed flux measurements system for PAN . In addition, fluxes of 27 O 3 , which has similarities to PAN in terms of its formation and deposition and thus is 28 important for model applications, were determined by eddy covariance. Based on our 29 approaches, we estimate the contribution of stomatal and non-stomatal deposition pathways 30 for PAN and compare these results to those obtained for O 3 . 31 rate is about one order of magnitude lower than of the reaction of O 3 with NO. The quality 1 scheme of Foken and Wichura (1996) was used to exclude periods with significant non-2 stationarity or poor developed turbulence. Data for which the footprint area of the flux 3 measurement (calculated with a Lagrangian forward stochastic model from Rannik et al., 4 2000) included less than 80% of the natural grassland area were omitted. 5 The mixing ratio difference of O 3 between 4.0 and 0.8 m a.g.l. was determined using a 6 differential UV absorption O 3 analyser (49i, Thermo Environmental, USA, modified 7 according to Cazorla and Brune (2010); see Moravek et al. (2014) for details on operation). 8 Absolute O 3 mixing ratios at both heights were derived from a vertical profiles system, which 9 also measured NO and NO 2  ( ) and the non-stomatal ( ) conductance) was obtained for both PAN and O 3 from the 23 measured deposition velocity ( , i.e. the flux normalized by the concentration at ) and 24 the estimated aerodynamic ( ) and quasi-laminar boundary layer ( ) resistances (see 25 Garland (1977) and Hicks et al. (1987), respectively): 26 In case processes in the leaf mesophyll (or surrounding components) do not limit the trace gas 1 exchange (i.e. 1⁄~0), as in the case of water vapour or O 3 , simply equals the sum of 2 for the ratio of their molecular diffusivities to the molecular diffusivity of water vapour. Due 7 to its longer molecular structure, the diffusivity of PAN is lower ( ~ 0.87•10 -5 m 2 s -1 ) 8 than for O 3 ( 3~ 1.40•10 -5 m 2 s -1 ), which results in , representing all non-stomatal deposition pathways, e.g., to leaf cuticles, soil or 18 water surfaces, was derived by the difference between and (Eq. (3)). 19 The findings on from the partitioning of were used to model PAN deposition 20 fluxes for the entire period from 29 June to 21 October. Applying the resistive scheme given 21 in Eq. (3), the modelled PAN flux ( ) was derived as 22 where ,, and were determined as described above over the entire period. Here, 23 and represent the molar air density and the PAN mixing ratio, respectively, at the 24 height of the eddy covariance measurements ( ) 25

Determination of PAN loss by thermochemical decomposition 26
Next to dry deposition process, other sink terms impact the measured surface PAN mixing 27 ratios. While PAN photolysis and reaction with the hydroxyl radical (OH) are expected to be very low at altitudes below 7 km (Talukdar et al., 1995), thermochemical decomposition of 1 PAN (back reaction of R1) has to be considered. Thermochemical decomposition of PAN 2 increases exponentially with temperature and is more efficient at high NO/NO 2 ratios as PA 3 reacts faster with NO than with NO 2 to reform PAN. Hence, the time scale of PAN towards 4 thermochemical decomposition ( ℎ ) is given by (Orlando et al., 1992;Shepson et al., 5 1992) as 6 In addition, loss of PA due to uptake by fog droplets can have an influence on the 7 thermochemical decomposition of PAN (see Roberts et al., 1996;Villalta et al., 1996). 8 However, since at night, when fog conditions may have occurred, the thermochemical 9 decomposition of PAN was limited by the low temperatures, this effect was neglected for this 10 study.
The height dependent functions of ( ) and ℎ ( ) were approximated by logarithmic 16 interpolation between the available measurement heights of the required parameters and 17 ( ) was assumed to be constant with height (see Sect. 2.2). 18 The thermochemical PAN loss over the entire atmospheric boundary layer, represented as a 19 flux ( ℎ ), was obtained by integrating Eq. (6) from zero level to the height of the 20 boundary layer (ℎ ) 21 Assuming a well-mixed boundary layer, the measured PAN concentration and NO/NO 2 ratio 22 were taken as an average value for the whole boundary layer. As ℎ is very sensitive to 23 temperature, we assumed a dry adiabatic lapse rate of temperature with height. The height of 24 the nocturnal boundary layer was estimated from the nocturnal decline of O 3 and the 1 corresponding measured 3 from the relation given by (Shepson et al., 1992) (see also 2 where 3 ( 0 ) and 3 ( 1 ) are the O 3 mixing ratios at the start and end of the considered time 4 interval, respectively. Since ℎ _ ℎ was determined from a boundary layer budget 5 approach, it might not agree well with the real boundary layer height, as the nocturnal 6 boundary layer might be significantly stratified. Instead, ℎ _ ℎ represents the theoretical 7 depth of a mixed boundary layer, which was required in Eq. (7) to assume constant trace gas 8 mixing ratios with height. The development of the diurnal boundary layer (ℎ _ ) after 9 dawn was modelled using the measured sensible heat surface flux and a simple encroachment 10 approach implemented in the mixed layer model MXLCH (Vilà-Guerau de Arellano et al., 11 2011). 12

Meteorological conditions: Classification of low and high NO x episodes 14
The field experiment was dominated by wind directions from south west. These air masses 15 were associated with relatively low levels of NO x (= NO + NO 2 ) (ranging mainly between 1 16 and 10 ppb). Air masses from north easterly directions were much less frequent, but were 17 often enriched with NO x with values ranging mainly between 10 and 30 ppb (Fig. 1). This 18 enrichment was mainly caused by advection from NO x sources originated from the City of 19 Mainz, nearby motor ways and other sources in the densely populated and industrialised 20 Rhine-Main region. In contrast, the south west sector is dominated by farming without major 21 industrial activity, thus representing an area with much less air pollution. Consequently, the 22 occurrence of low and high NO x situations during the field experiment was directly coupled to 23 the wind direction and could be attributed to two contrasting synoptic conditions: 24 (1) Episodes under deep pressure influence and south westerly wind directions yielded low 25 NO x conditions. They were characterized by higher wind speeds, frequent cloud coverage, a 26 mainly neutrally stratified boundary layer and typically lasted from 2 to 5 days. 27 (2) Sunny, convectively driven episodes with low wind speeds and, therefore, also varying 1 wind directions resulted in high NO x conditions, in cases when the wind direction was not 2 from the south west sector. In contrast to the low NO x conditions, these periods occurred 3 sometimes as very isolated events and were associated with an unstable boundary layer during 4 daytime and a stable stratification during nighttime. 5 For the further evaluation, entire days were selected and classified according to wind speed 6 and wind direction. In total 20 days were classified as low NO x and 27 days as high NO x 7 conditions. The diurnal averages of the meteorological conditions and micrometeorological 8 characteristics during these days are displayed in Fig

Characterisation of PAN under low and high NO x conditions 24
The diurnal cycle of PAN mixing ratios was closely linked to the diurnal cycle of O 3 . As for 25 O 3 , PAN mixing ratios increase after dawn to the maximum in the afternoon, with median 26 values of 300 ppt under low and of 600 ppt under high NO x conditions, respectively (Fig. 2j). 27 The maximum is followed by a steady decrease over night to median values just before dawn 28 of about 50 ppt under low NO x and 200 ppt under high NO x conditions. 29 The major reason for the much higher PAN levels during high NO x conditions, are the 30 elevated NO 2 mixing ratios, which occurred especially during nighttime and declined with the 31 onset of photolysis after dawn and the clearing of the nocturnal boundary layer. Comparing 1 the diurnal evolution of PAN and O 3 mixing ratios, we find a higher PAN/O 3 ratio under high 2 NO x conditions at all times throughout the diurnal cycle. During peak PAN and O 3 mixing 3 ratios in the afternoon, the PAN/O 3 ratio was 0.003 and 0.006 during low and high NO x , 4 conditions, respectively. Since photolytic production of O 3 from NO 2 was similar for both 5 conditions, a large PAN/O 3 ratio implies a higher abundance of PA as a precursor of PAN 6 (Zhang et al., 2009). Although no direct measurements of PA were available, the very low 7 abundance of volatile organic compounds measured at the site (e.g., isoprene < 0.7 ppb, 8 monoterpene < 0.3 ppb, J. Kesselmeier, personal communication, 2013) suggests that these 9 higher levels of PA during high NO x conditions primarily originated from anthropogenic non-10 methane hydrocarbons (NMHCs). Hence, PAN mixing ratios at the site were mainly 11 influenced by advection from nearby pollution sources from north easterly directions. 12 The timescale for thermochemical decomposition of PAN, ℎ , ranged for both low and 13 high NO x conditions mainly between 4 and 20 days at night (Fig. 2k). During daytime, 14 ranged between 2 h and nearly one day (median ~5 h) for low NO x conditions, but were 15 significantly lower during high NO x conditions (ranging between 30 min and 5 h; median 16 ~2 h) caused by both on average higher NO/NO 2 ratios in the morning and higher 17 temperatures in the afternoon. 18 ≈ 1 cm s -1 ) measured at two different pine forest sites in the USA during summer 7 (Table 1) ratios vary considerably, which might be attributed to a large extent to the error 12 of the applied measurement methods and the assumptions made. It has to be noted that is 13 height dependent, which can make its comparison between different studies difficult. 14 However, the ratio The chemical flux divergence between and 0 due to thermochemical decomposition of 18 PAN (Eq. (6)) was found to be very small with the highest median value of 0.007 nmol m -2 s -1 19 at noon (Fig. 4a). In contrast, for the O 3 flux, the loss term due to reaction with NO and the 20 production by NO 2 photolysis were significantly higher between 6:00 and 11:00 CET and led 21 to a small net production of O 3 during daytime, which was corrected for in the presented 22

fluxes. 23
The overall canopy conductance for PAN ( ), representing the flux normalized by the 24 concentration at 0 , shows a mean diurnal cycle with its maximum during daytime (Fig. 4c-d). 25 The midday median values were around 0.4 cm s -1 and were similar to values observed for 26 O 3 . 27

Stomatal uptake 28
During nighttime values were zero due to stomata closure (Fig. 4c). With the onset of 29 radiation in the morning increases and reaches its maximum of 0.26 cm s -1 at 30 11:00 CET. As both and 3 differ only by the PAN and O 3 diffusivities (see 1 Sect. 2.4), they show the same pattern, while 3 is larger by a factor of 1.6 due to the faster 2 diffusivity of O 3 . Due to an increased vapour pressure deficit in the afternoon the maximum 3 values of and 3 are slightly skewed towards the morning. 4 The existence of a mesophyllic resistance limiting the stomatal uptake of PAN, as it was 5 found by Teklemariam and Sparks (2004) or by (Sparks et al., 2003) at high stomatal 6 conductance, cannot be validated from our data. Only if the modelled values exceeded 7 the experimentally determined values, a limitation could be suspected. It is suggested 8 that the mesophyllic uptake of PAN is lower than for O 3 , as there are less reaction sites for 9 PAN within the plant cell and its reaction with proteins is slower, although the mesophyll 10 biochemistry for PAN assimilation is not clearly understood (Doskey et al., 2004). 11

Non-stomatal deposition 12
According to the MBR flux measurements at our site, the non-stomatal sink played a major a compound that reacts fast with substances in the leaf cuticles such as protein thiols (Mudd, 29 1982). Due to the poor solubility of PAN in water ( * = 4.1 M atm -1 , see Kames and 1 Schurath (1995)) the first term of the right side of Eq. (9) can be neglected and only the 2 reactivity index, 0 , is of significant importance. According to Wesely (1989)  deposition by Hill (1971) and Garland (1977). This contradicts our findings by both the MBR 6 and the NBLB method, which observed at least equal or even higher non-stomatal deposition 7 for PAN than for O 3 , and supports the statement by Turnipseed et al. (2006) that current 8 deposition models may significantly underestimate PAN non-stomatal deposition. 9

PAN deposition fluxes for low and high NO x conditions 10
To evaluate the PAN deposition under both low and high NO x conditions as well as its 11 potential influence on the natural grassland ecosystem and its role for the atmospheric N r 12 budget, the PAN deposition flux was modelled for the entire period from 29 June to 13 21 October (see Sect. 2.4). For this, we used the bulk value for of 0.28 cm s -1 14 (Sect. 3.3.3) for both low and high NO x , as we found this to be the best estimate from our 15 data. The obtained median diurnal cycles of for low and high NO x conditions 16 ( Fig. 5) reveal that the total deposition (i.e. stomatal + non-stomatal) was more than twice as 17 high during high NO x (~-0.1 nmol m -2 s -1 ) than during low NO x (~-0.05 nmol m -2 s -1 ) 18 conditions, which is mainly attributed to the higher PAN mixing ratios during high NO x 19 conditions. Median midday deposition velocities were very similar during both episodes 20 ( ≈ 0.5 cm s -1 ). As already discussed in Sect. 3.3.3, the non-stomatal pathway was 21 significant, which is reflected by a daytime fraction of ⁄ of 0.7 during low NO x 22 and 0.6 during high NO x conditions. As about half of the grassland vegetation was senescing 23 or was already dead, reaction on plant surfaces may be a reason for the large non-stomatal 24 fraction. 25 The importance of PAN deposition as a loss process of PAN from the atmosphere is 26 determined by comparison to the magnitude of the thermochemical decomposition of PAN in 27 the boundary layer (Eq. (7)). Due to the lower temperatures and the lack of NO at night, the 28 nocturnal thermochemical loss was insignificant during both low and high NO x conditions. 29 Using the boundary layer budget approach (Eq. (8)), we found ℎ _ ℎ to be on average 30 200 m (Fig. 5). In contrast, during daytime the thermochemical loss constituted the largest 31 PAN sink, during both low and high NO x conditions. After dawn, ℎ _ grew during high 1 NO x conditions on average up to 1200 m, whereas its development was slightly suppressed 2 during low NO x conditions. The modelled boundary layer height was compared for selected 3 days to the boundary layer height obtained from a WRF model. The WRF model yielded 4 slightly higher daytime maximum values ranging from 1100 up to 1700 m. When the 5 boundary was well mixed (11)(12)(13)(14)(15)(16)(17), the thermochemical loss during high NO x conditions 6 was about 3.5 times higher than during low NO x conditions. This was caused by a 7 combination of (a) the higher PAN mixing ratios (effect: 59 %), (b) the reduced reaction time 8 scale due to higher temperatures and larger NO to NO 2 ratios (effect: 34 %) and to some 9 extend also by (c) the higher boundary layer (effect: 7 %). A summary of the relevant 10 parameters for nighttime and daytime conditions is given in Table 2 (2010) evaluated the effect of total inorganic nitrogen deposition on grasslands across Europe 1 and found that species richness decreased with sites that were subject to higher nitrogen 2 deposition. The observed PAN removal via dry deposition (i.e., ) over one entire day was 3 in this study 333 µg m -2 d -1 during low and 518 µg m -2 d -1 during high NO x conditions 4 ( Table 2). This is much lower than the total nitrogen deposition observed at the sites reported 5 by Stevens et al. (2010) ranging between 4.7 and 104.2 mg m -2 d -1 (equivalent to 2 and 6 44 kg N ha -1 a -1 ), which suggests that PAN deposition under both low and high NO x does not 7 play a critical role on plant species richness at our site. Moreover, PAN mixing ratios 8 observed at our site were significantly below the threshold given for phytotoxic effect on 9 plants (between 15 and 25 ppb, see Temple and Taylor, 1983 = 0.28 cm s -1 ). This resulted in an equal or even higher non-stomatal conductance for 16 PAN than for O 3 , most likely suggesting an underestimation of PAN deposition by current 17 models. We did not find a relation of the non-stomatal conductance for PAN with other 18 quantities, such as relative humidity. However, it cannot be fully excluded that this may also 19 be attributed to the limited PAN flux data above the flux detection limit. The modelled 20 stomatal uptake did not exceed the overall deposition, suggesting that stomatal uptake is not 21 limited by further, not-considered resistances. 22 PAN deposition at our measurement site was governed by two contrasting pollution regimes, 23 (1) low NO x episodes with clean air from south westerly directions and (2) high NO x episodes 24 with more polluted air masses from the north eastern sector. Under high NO x conditions, 25 locally produced PAN from the industrialized region was advected to the site, leading to PAN 26 mixing ratios which were a factor of two to four higher than under low NO x conditions. 27 Hence, PAN deposition during these episodes was larger with daytime maxima