Contents lists available at ScienceDirect

Atmospheric Environment

journal homepage: www.elsevier.com/locate/atmosenv

Short communication

Missing ozone-induced potential aerosol formation in a suburban deciduousforest

T. Nakayamaa,∗, Y. Kurumaa,1, Y. Matsumia, Y. Morinob, K. Satob, H. Tsurumaruc,2,S. Ramasamyc,3, Y. Sakamotob,c,d, S. Katoe, Y. Miyazakif, T. Mochizukif,4, K. Kawamuraf,5,Y. Sadanagag, Y. Nakashimah, K. Matsudah, Y. Kajiib,c,d

a Institute for Space-Earth Environmental Research, Nagoya University, Furo-cho, Chikusa-ku, Nagoya, Aichi, Japanb National Institute for Environmental Studies, 16-2, Onogawa, Tsukuba, Ibaraki, Japanc Graduate School of Global Environmental Studies, Kyoto University, Nihonmatsucho, Sakyo-ku, Kyoto, Kyoto, Japand Graduate School of Human and Environmental Studies, Kyoto University, Nihonmatsucho, Sakyo-ku, Kyoto, Kyoto, Japane Department of Applied Chemistry, Graduate School of Urban Environmental Science, Tokyo Metropolitan University, 1-1 Minami-Osawa, Hachioji, Tokyo, Japanf Institute of Low Temperature Science, Hokkaido University, Kita 8, Nishi 5, Kita-ku, Sapporo, Hokkaido, Japang Department of Applied Chemistry, Osaka Prefecture University, 1-1 Gakuen-cho, Naka-ku, Sakai, Osaka, Japanh Faculty of Agriculture, Tokyo University of Agriculture and Technology, 3-5-8, Saiwai-cho, Fuchu, Tokyo, Japan

G R A P H I C A L A B S T R A C T

A R T I C L E I N F O

Keywords:SOAFlow reactorPotential aerosol formationOzoneSuburban forest

A B S T R A C T

As a new approach to investigating formation processes of secondary organic aerosol (SOA) in the atmosphere,ozone-induced potential aerosol formation was measured in summer at a suburban forest site surrounded bydeciduous trees, near Tokyo, Japan. After passage through a reactor containing high concentrations of ozone,increases in total particle volume (average of 1.4 × 109 nm3/cm3, which corresponds to 17% that of pre-existingparticles) were observed, especially during daytime. The observed aerosol formations were compared with theresults of box model simulations using simultaneously measured concentrations of gaseous and particulatespecies. According to the model, the relative contributions of isoprene, monoterpene, and aromatic hydrocarbonoxidation to SOA formation in the reactor were 24, 21, and 55%, respectively. However, the model could ex-plain, on average, only ∼40% of the observed particle formation, and large discrepancies between the ob-servations and model were found, especially around noon and in the afternoon when the concentrations of

http://dx.doi.org/10.1016/j.atmosenv.2017.10.014Received 13 July 2017; Received in revised form 5 October 2017; Accepted 6 October 2017

∗ Corresponding author.

1 Now at National Institute of Advanced Industrial Science and Technology, 1-1-1, Umezono, Tsukuba, Ibaraki, Japan.2 Now at Institute of Nature and Environmental Technology, Kanazawa University, Kakumamachi, Kanazawa, Ishikawa, Japan.3 Now at National Institute for Environmental Studies, 16-2, Onogawa, Tsukuba, Ibaraki, Japan.4 Now at School of Food and Nutritional Sciences, University of Shizuoka, 52-1 Yada, Shizuoka, Shizuoka, Japan.5 Now at Chubu Institute for Advanced Studies, Chubu University, 1 200 Matsumoto-cho, Kasugai, Aichi, Japan.

E-mail address: nakayama@isee.nagoya-u.ac.jp (T. Nakayama).

Atmospheric Environment 171 (2017) 91–97

Available online 10 October 20171352-2310/ © 2017 Elsevier Ltd. All rights reserved.

MARK

isoprene and oxygenated volatile organic compounds were high. The results suggest a significant contribution ofmissing (unaccounted-for) SOA formation processes from identified and/or unidentified volatile organic com-pounds, especially those emitted during daytime. Further efforts should be made to explore and parameterizethis missing SOA formation to assist in the improvement of atmospheric chemistry and climate models.

1. Introduction

Aerosol particles in the atmosphere are known to have significantimpacts on climate, the atmospheric environment, ecosystems, andhuman health (Boucher, 2015; Lelieveld et al., 2015). Secondary or-ganic aerosol (SOA) particles, which are generated during the oxidationof volatile organic compounds (VOCs) emitted from both biogenic andanthropogenic sources, constitute a large fraction of submicron parti-cles in the atmosphere (Jimenez et al., 2009). However, the formationprocesses of SOA in the atmosphere remain largely uncertain (e.g.Hallquist et al., 2009; Chen et al., 2015; Glasius and Goldstein, 2016).

Formation yields of SOA (defined as the ratio of generated SOA massto the amount of reacted VOC) from various VOCs have been de-termined based on experiments using smog chambers or oxidation flowreactors (Hallquist et al., 2009; Lambe et al., 2015; Bruns et al., 2015;and references therein). Although the dependency of the formationyields on a number of parameters, such as temperature, relative hu-midity (RH), seed particles, co-existing gases, and species and exposuresof oxidants, has been investigated, it is almost impossible to investigateall SOA formation processes in the real atmosphere, including thepossible interactions between different oxidation processes and thevarious parameters.

A number of observational studies have been conducted to estimateSOA formation processes in the atmosphere by measuring tracer com-pounds such as cis-2-methyl-1,3,4-trihydroxy-1-butene, 2-methyl-threitol, and 2-methylerythritol for isoprene-derived SOA; 3-iso-propylglutaric acid, cis-pinonic acid, and pinic acid for monoterpene-derived SOA; and 2,3-dihydroxy-4-oxopentanoic acid for aromatic hy-drocarbon-derived SOA (e.g. Kleindienst et al., 2007a, 2010; Hu et al.,2008; Ding et al., 2012). However, there are potentially large un-certainties in such estimations because the yields of tracer molecules inthe atmosphere may not necessarily be the same as those obtained inlaboratory experiments owing to the complexity of the SOA formationprocess and possible chemical transformation of the tracer molecules inthe real atmosphere.

As an alternative approach to obtaining information on SOA for-mation processes, the monitoring of potential SOA formation after in-ducing ambient air perturbations may be useful. Very recently, Jimenezand coworkers reported potential aerosols formation in a PotentialAerosol Mass (PAM) flow reactor, with perturbations caused by oxidantadditions (Ortega et al., 2016; Palm et al., 2016, 2017). For an urbanarea in the Los Angeles region of the USA in late spring, Ortega et al.(2016) reported that reactor SOA formation under irradiation by ul-traviolet (UV) light, for which VOCs were estimated to be mainly oxi-dized by hydroxyl radicals (OH), peaked at night and correlated withtrimethylbenzene concentrations. At a pine forest in Colorado, USA, insummer, Palm et al. (2016, 2017) also reported greater reactor SOAformation at night for all cases of UV irradiation, ozone (O3) addition,and nitrate radical (NO3) addition; this was mainly due to the oxidationof monoterpenes. Whereas the observed reactor SOA formations for O3

and NO3 additions were consistent with the predicted SOA formationsbased on measured monoterpene and sesquiterpene concentrations, theobserved reactor SOA formation under UV irradiation (OH oxidation)was reported to be 3.4 times larger than that of the model prediction(Palm et al., 2016, 2017).

In the present study, potential aerosol formation with the additionof O3 was investigated by comparing total particle volumes with andwithout passing the particles through an oxidation flow reactor. Thestudy was carried out in summer at a suburban forest site surrounded by

deciduous trees, near Tokyo, Japan. By comparing the observed SOAformation in the reactor with that predicted by the model using thereported SOA formation yields from chamber experiments, the con-tribution of unidentified SOA formation processes is highlighted.Because small forests with deciduous trees are ubiquitous in urban,suburban, and rural areas across the world, the results obtained in thepresent study will be widely applicable to regions having similar en-vironments, where both biogenic and anthropogenic emissions are ex-pected to influence SOA formation. This study aimed to demonstratethat observing potential aerosol formation to reveal unidentified SOAformation processes is useful in improving model predictions for eval-uating impact of aerosol particles.

2. Material and methods

2.1. Observations

Observations were conducted from July 31 to August 8, 2015 as partof the AQUAS-TAMA (Air Quality Study at FM Tama) 2015 field cam-paign at the Field Museum Tama Hill (FM-Tama, 35.6385°N,139.3781°E) measurement station of Tokyo University of Agricultureand Technology (Matsuda et al., 2015). This station is located in asuburban forest (∼0.115 km2) and is surrounded by a residential area∼30 km west of the center of Tokyo (Fig. S1). The dominant treespecies in the forest are Quercus serrata (Japanese konara oak), Quercusacutissima (sawtooths oak), and Cryptomeria japonica (Japanese cedar).

A schematic diagram of the experimental apparatus used to measureaerosol formation potential is shown in Fig. S2. Aerosol particles weresampled through an inlet placed 4.6 m above the ground surface. Thesize distributions of particles between 17.5 and 982.2 nm, which wereeither passed or not passed through the reactor, were simultaneouslymeasured every 5 min using two scanning mobility particle sizers(SMPSs) (TSI, model 3936L72), SMPS-1 and SMPS-2, the sampling andsheath flow rates of which were fixed at 0.2 and 2.0 L per minute (lpm),respectively. These size distributions were measured under dry condi-tions by passing the particles through diffusion dryers with silica gelbefore introduction to the SMPSs. A 33L stainless steel cylindrical cellwas used as the flow reactor. The O3 was generated by irradiating fil-tered oxygen gas (Japan Fine Products,> 99.999%) with ultravioletlight at 184.9 nm from two mercury (Hg) lamps (Hamamatsu, L937-02). The O3 was continuously added to the reactor at a flow rate of0.020 lpm. Note that no particle formation was observed when O3 wasadded after passing ambient air thorough an activated carbon denuder,suggesting that the contribution of impurities in the oxygen cylinder toaerosol formation in the reactor was negligible. The concentration of O3

in the reactor was monitored using a UV absorption analyzer (ThermoScientific, model 49ij). The total flow rate and average residence timein the reactor were 0.6 lpm and 55 min, respectively. The total particlevolume generated in the reactor was obtained by subtracting the totalparticle volume of aerosol that was passed through the reactor fromthat measured without passage through the reactor, after taking ac-count of the delay and dull effects of residence time in the reactor. Theeffects of particle loss and differences in the sensitivity of the twoSMPSs were corrected using the particle size distributions obtainedduring periods when the Hg lamps were turned off (July 31, August 3-4,and August 8). Size dependent particle dry deposition was used for thecorrections. Typical particle losses were between 28% and 48% forparticles with diameters between 100 nm and 500 nm.

During the campaign, ambient concentrations of gaseous nitrogen

T. Nakayama et al. Atmospheric Environment 171 (2017) 91–97

92

oxides (NO and NO2), carbon monoxide (CO), and O3 were monitoredusing commercial instruments. The concentrations of VOCs, such asisoprene, monoterpenes, methylvinylketone (MVK) + methacrolein(MACR), benzene, toluene, C8 aromatics, and C9 aromatics, weremonitored using a proton-transfer-reaction mass spectrometer (PTR-MS, IONICON). Although 2-methyl-3-butene-2-ol (MBO), which isemitted from a few specific species of pines (Goldan et al., 1993; Bakeret al., 1999), has the same molecular weight as isoprene, all signals atm/z = 69 were assumed to be from isoprene in this study (consideringthe tree species around the observation site). The concentrations ofsulfur dioxide (SO2) were estimated by averaging data from the At-mospheric Environmental Regional Observation System of the Ministryof the Environment of Japan (http://soramame.taiki.go.jp/) at fourobservatories in Katakura, Tachi, Atago, and Hashimoto, which arewithin ∼9 km of the FM-Tama site (Fig. S1). In addition, ambient fineparticles with aerodynamic diameters of less than 0.95 μm were sam-pled using a high volume aerosol sampler for chemical analyses. Ty-pically, this was done every 3 h during daytime and for 15 h duringnighttime. Major aerosol components, including elementary carbon(EC), organic carbon (OC), sulfate, nitrate, and ammonium, in thesesamples were quantified. Detailed descriptions of the measurement ofgaseous and aerosol compositions are given in the SupplementaryMaterial (S1). Meteorological data including temperature, RH, andphotosynthetically active radiation were monitored at the top of anobservation tower (30 m above ground surface) located 150 m from theinlet.

2.2. Box model simulation

Box model calculations to simulate chemical reactions and aerosolformation in the reactor were conducted every 1 h using the SAPRC-99chemical mechanism (Carter, 2000) and AERO5 aerosol module(Carlton et al., 2010) after the modifications described below. Althoughno photochemical reactions should occur in the reactor, OH radicals canbe generated through the ozonolysis of unsaturated hydrocarbons, andNO3 radicals can be generated through the reaction of NO2 with O3.Therefore, VOCs can be oxidized by OH, O3, or NO3 in the reactor. Theoriginal AERO5 aerosol module gives the SOA formation yield for OHoxidation (Henze and Seinfeld, 2006) for all oxidation processes ofisoprene. In the present study, the module was modified to give theformation yields for NO3 and O3 oxidations of isoprene. The organic-mass-loading-dependent formation yield reported by Ng et al. (2008)was used for the NO3 oxidation of isoprene, whereas yields reported bySato et al. (2013) and Ren et al. (2017) were used for the O3 oxidationof isoprene. The organic-mass-loading-dependent formation yields andtheir fitting parameters for isoprene-SOA, as used in the present study,are shown in the Supplementary Material (Fig. S3 and Table S1). Threeruns were conducted using different SOA formation yield curves for theO3 oxidation of isoprene. In run-1, the yields reported by Sato et al.(2013) for SOA generated under comparable initial concentrations([O3]ini/[isoprene]ini of 1.0–2.5) in the presence of OH scavenger (CO)were used. In run-2, the yields reported by Ren et al. (2017) for the SOAgenerated under excess O3 conditions ([O3]ini/[isoprene]ini of

Fig. 1. Hourly temporal variations in mixing ra-tios of (a) O3, CO, NO, NO2, and SO2 (b) isoprene(Iso), methylvinylketone (MVK) + methacrolein(MACR), and monoterpenes (Terp), and (c) aro-matic hydrocarbons with carbon atoms of 6 and 7(ARO(C6–C7)) and 8 and 9 (ARO(C8–C9)). (d)Mixing ratio of O3 downstream of the reactor andmass concentration of organic aerosol (OA) andsulfate, and (e) temperature, relative humidity(RH), and photosynthetically active radiation(PAR).

T. Nakayama et al. Atmospheric Environment 171 (2017) 91–97

93

10.7–11.5) in the presence of OH scavenger (cyclohexane) were used.In run-3, the yields reported by Ren et al. (2017) for the SOA generatedunder excess isoprene conditions ([O3]ini/[isoprene]ini of 0.08–0.21) inthe absence of OH scavenger were used.

Observed ambient concentrations of NO, NO2, CO, SO2, water vapor(estimated from temperature and RH), and VOCs, as well as O3 con-centration measured downstream of the reactor, were used as initialconditions in the model. The benzene and toluene were categorized as″ARO(C6–C7)″, whereas C8-aromatics and C9-aromatics were categor-ized as ″ARO(C8–C9)″ in the model. The sum of the concentrations ofindividual compounds was used as the initial concentration for each ofthese categories. Half of the mass concentration, corresponding to m/z = 71 (MVK + MACR), was used as the initial concentrations of MVKand MACR. Hourly initial mass concentrations of organic aerosol (OA),sulfate, nitrate, and ammonium were calculated by multiplying thetotal volume of ambient particles measured with SMPS-1 by the fractionof each composition obtained from filter samples, as described in theSupplementary Material (S2).

3. Results and discussion

As shown in Fig. 1e, the sky was mostly clear during the observationperiod and clear diurnal patterns were observed in temperature, RH,and solar irradiance. No precipitation was recorded during the period.Fig. 1a shows temporal variations in ambient mixing ratios of O3, CO,NO, and NO2. Relatively high concentrations of anthropogenic gasessuch as CO and nitrogen oxides (NOx = NO + NO2) were observedduring nighttime, likely due to their accumulation in the shallow noc-turnal boundary layer. In contrast, concentrations of O3 reached theirmaxima during daytime due to photochemical formation.

Fig. 1b and c shows temporal variations in the ambient mixing ra-tios of isoprene, monoterpenes, ARO(C6–C7), and ARO(C8–C9). Highconcentrations of isoprene were observed during daytime, especiallybetween 8:00 and 14:00, likely due to strong emissions from broad leafplants in the canopy of the forest around the FM-Tama site under highsolar irradiance and leaf temperature conditions (Fig. 1e). Suzuki et al.

(2012) also reported large canopy emissions of isoprene during day-time. In contrast, the highest concentrations of monoterpenes andaromatic hydrocarbons were observed in early morning, especiallybetween 0:00 and 9:00, likely due to (i) continuous emissionsthroughout the day, (ii) a greater oxidation rate during daytime, and(iii) accumulation in the shallow nocturnal boundary layer duringnighttime. Very slow reaction rates of aromatic hydrocarbons with NO3

and O3 would also contribute to high concentrations during nighttime.As shown in Fig. 1d, the concentrations of O3 in the reactor

were>102 times higher than those of SOA precursors throughout theperiod and were expected to correspond to 2–4 equivalent days of ex-posure (assuming ambient O3 concentration of 50 ppbv). OA and sulfatewere dominant ambient aerosol components, with average (± 1σ)fractions of 58% (±12%) and 28% (±11%), respectively, of totalmass concentrations quantified in the present study.

Fig. 2a and b shows temporal variations in the volume-based sizedistributions of ambient particles (measure by SMPS-1) and thosedownstream of the reactor (measured by SMPS-2 and corrected forparticle losses in the reactor), respectively. During the campaign, theaverage (± 1σ) volume-based mean diameter for ambient particlesmeasured by SMPS-1 was 342 (± 45) nm. Temporal variations in thetotal volume of particles are shown in Fig. 2c. Compared with SMPS-1,larger total particle volumes were measured by SMPS-2 during most ofthe observation period, indicating particle formation in the reactor. Onaverage, the total particle volume increased by 1.4 × 109 nm3/cm3,which was 17% that of pre-existing particles. A greater formation ofparticles up to 6–7 × 109 nm3/cm3 (∼80% that of pre-existing parti-cles) was observed during daytime on both August 2 and 7. Note thatduring the daytime (10:00-15:00) on August 2 and 7, the averagemixing ratios of isoprene (6.2 and 15.2 ppbv) were higher than theother days (2.4–6.0 ppbv), whereas the average mixing ratios of NOx

(1.8 and 2.4 ppbv) were lower than the other days (2.5–4.7 ppbv).Assuming a particle density of 1.4 g/cm3 (Kuwata et al., 2012; Nakaoet al., 2013), the average and maximum formations of aerosol in thereactor are estimated to be 2.0 and 10.2 μg/m3, respectively.

Fig. 3a shows simulated temporal variations in the loss rates (units

Fig. 2. Temporal variations in volume-based sizedistributions of particles (a) in ambient air and (b)downstream of the reactor. (c) Total volumes ofparticles, which were calculated from the sizedistributions measured by SMPS-1 and SMPS-2.

T. Nakayama et al. Atmospheric Environment 171 (2017) 91–97

94

of ppbv/hour) of isoprene and monoterpenes in the reactor due tooxidation by OH, O3, and NO3, as well as ARO(C6–C7) and ARO(C8–C9) due to oxidation by OH. The loss rates obtained in run-1 of thebox model are shown in Fig. 3a (those in run-2 and run-3 were almostidentical). Isoprene is expected to be oxidized by OH, O3, and NO3 withcontributions of 12, 45, and 43%, respectively. The results of the si-mulation suggest that loss rates for the NO3 oxidation of isoprene werehigh, especially in the early morning (6:00–11:00) owing to the rela-tively high concentrations of NOx during this period. In contrast, lossrates for OH and O3 oxidation of isoprene were high between 8:00 and14:00, when the concentrations of isoprene peaked. The loss rates ofmonoterpene and aromatic hydrocarbons were high between 0:00 and9:00, when their concentrations were also high. Monoterpenes weremainly oxidized by NO3 in the reactor, whereas aromatic hydrocarbonswere oxidized by OH. Although the concentrations of ARO(C6–C7) andARO(C8–C9) were comparable (Fig. 1c), the loss rate of the latter isconsidered to be higher than that of the former because of greater re-action rates with OH (Calvert et al., 2002).

The simulated temporal variation of OA generated in the reactor(ΔOA) for run-1 of the box model is shown in Fig. 3b, along with theobserved ΔOA, which was calculated assuming a particle density of1.4 g/cm3. The colors in Fig. 3b represent the ΔOA for each oxidationprocess. Note that the formation of inorganic aerosols such as sulfateand nitrate is simulated as being negligible (< 0.2 μg/m3). On average,38% of the observed increase in the mass concentrations of aerosol canbe explained by the sum of model-simulated ΔOA. The relative con-tributions of the oxidation of isoprene, monoterpenes, and aromatichydrocarbons to the ΔOA are estimated to be 24, 21, and 55%, re-spectively. For aromatic hydrocarbons, the relative contribution of ARO(C8–C9) (38%) is approximately twice that of ARO(C6–C7) (17%),mainly due to the difference in loss rate (Fig. 3a).

Fig. 4a shows diurnal patterns of simulated ΔOA for run-1, alongwith observed ΔOA. The simulated ΔOA from the oxidation of mono-terpenes and aromatic hydrocarbons was high from midnight to earlymorning (0:00–9:00). The NO3 oxidation is considered to contributedominantly to the ΔOA from isoprene due to the high SOA formationyields (Fig. S3). The simulated ΔOA from the NO3 oxidation of isoprenewas high from early morning to noon (6:00–11:00). In contrast to thesmall differences in ΔOA between the simulated and observed valuesfrom midnight to early morning, larger differences were observed

around noon and in the afternoon in run-1.The ratio of NOx to HOx possibly contributes to SOA yield. This

effect was included in our model simulations for SOA generated fromaromatic hydrocarbon, but not for SOA generated from monoterpenesand isoprene (Carlton et al., 2010). Because monoterpenes were ex-pected to be dominantly oxidized by NO3 throughout the observation

Fig. 3. Simulated temporal variations in (a) lossrates of VOCs and (b) modeled ΔOA for eachoxidation process in run-1 of the box model. Thetemporal variation of ΔOA estimated based on theobserved size distribution data, assuming a den-sity of 1.4 g/cm3, is also plotted in panel (b) forcomparison. Note that the range of the verticalaxis for simulated ΔOA is half of that for observedΔOA;.

Fig. 4. Diurnal variations in simulated mass concentrations of organic aerosols (ΔOA)generated in the reactor for each oxidation process in (a) run-1, (b) run-2, and (c) run-3 ofthe box model. The diurnal variation in ΔOA, estimated based on the observed size dis-tribution data assuming a density of 1.4 g/cm3, is also plotted in each panel for com-parison.

T. Nakayama et al. Atmospheric Environment 171 (2017) 91–97

95

period (Fig. 3a), contribution of NOx level to the SOA yields for theoxidations of monoterpenes should be minimal. The SOA yields for OHoxidation of isoprene under high NOx condition were reported to be oneorder of magnitude lower than those under low NOx conditions (Carltonet al., 2009). In the present study, greater ΔOA values were observedunder low NOx and high isoprene conditions (Fig. S4). Although thisresult qualitatively agrees with the reported NOx dependence of theSOA yields for OH oxidation of isoprene, the discrepancy in the ΔOAbetween observation and model cannot be explained only by this effectbecause relatively high SOA yields for low NOx conditions (Henze andSeinfeld, 2006) were used in the model for OH oxidation of isoprene(Fig. S3).

Very recently, Ren et al. (2017) reported higher SOA yields for theO3 oxidation of isoprene than those reported in previous studies (Satoet al., 2013; Kleindienst et al., 2007b). If the SOA yield curve based onthe values of Ren et al. (2017) for SOA generated under excess O3

conditions in the presence of OH scavenger is applied in the simulation(run-2), the difference between the simulation and observation be-comes slightly smaller (Fig. 4b); as such, the simulation accounts for40% (on average) of the observation. To test the sensitivity of the SOAyield for the ozonolysis of isoprene to ΔOA, the SOA yield curve basedon the values of Ren et al. (2017) for SOA generated under excessisoprene conditions in the absence of OH scavenger was applied in run-3 (even though their experimental conditions were not comparable toours). In this case, the simulation accounts for 45% (on average) of theobservation. However, even in this case, large differences remainaround noon and in the afternoon (Fig. 4c). These results suggest theexistence of other factors that contribute to the difference in ΔOA be-tween the simulation and observations.

Our results contrast with those of Palm et al. (2017), who reported agreater ΔOA during nighttime and a consistency between observed andsimulated ΔOA during O3 addition to ambient air at a pine forest inColorado, USA. The average ΔOA observed in the present study was 2–3times larger than that reported by Palm et al. (2017) for similarequivalent O3 exposure days. Whereas Palm et al. (2017) suggested thatthe ozonolysis of monoterpene dominantly contributed to the observedΔOA, high concentrations of isoprene and related oxidation products(MVK and MACR) were observed during daytime in the present study.The NO2 concentrations during their campaign at a pine forest in Col-orado (Fry et al., 2013) were ∼5 times lower than those in the presentstudy at the FM-Tama, where a strong influence of anthropogenicemissions from the Tokyo metropolitan area was apparent. These dif-ferences may contribute to the observed differences in the character-istics of ΔOA.

Several potential factors may contribute to the missing ΔOA aroundnoon and in the afternoon. First, VOCs that were not included in ourmodel may generate SOA through oxidation in the reactor. Previousstudies on OH reactivity in forest environments suggest a significantcontribution of oxygenated and unmeasured VOCs to total OH re-activity (e.g. Di Carlo et al., 2004; Kaiser et al., 2016; Ramasamy et al.,2016). For example, MACR, which is expected to be mainly generatedduring the oxidation of isoprene, is considered to generate SOA effec-tively (Brégonzio-Rozier et al., 2015). Although SOA formation fromisoprene via MACR in the reactor was taken into account in the model,SOA formation from pre-existing MACR was not. SOA formation fromsesquiterpenes, which are emitted from leaves, might also have a non-negligible contribution (Hoyle et al., 2011; and references therein).Second, the formation yields of SOA via OH oxidation of isoprene underambient conditions might be larger than the values used in the model.Recently, laboratory and observational studies have reported possibleenhancement of SOA formation from isoprene in the presence of acidicand/or sulfate particles, especially under low NOx conditions (e.g.Surratt et al., 2010; Shilling et al., 2013; Budisulistiorini et al., 2015;Beardsley and Jang, 2016; Rattanavaraha et al., 2016). Althoughaerosol acidity was not directly measured in the present study, sulfateconcentrations (4.3 μg/m3 on average) were significantly high

throughout the observation period and NOx concentrations were rela-tively low when large ΔOA were observed (as mentioned above)(Fig. 1a and d). Third, SOA formation through aqueous phase reactions,which were not included in the model, may also contribute to SOAformation in the reactor. For example, SOA formation through aqueousphase reactions of glyoxal and methylglyoxal has been proposed (Altieriet al., 2006; Herrmann et al., 2015, and references therein). Since ozonewas added under ambient RH conditions in the present study, theseaqueous phase reactions potentially contributed to SOA formation inthe reactor. Forth, SOA yield for each oxidation process of precursorgases may be changed by co-existing gases in the flow reactor thoughthe change in subsequent peroxy radical and nitrate radical reactions.

4. Conclusion

In the present study, potential aerosol formation with ozone addi-tion to ambient air was measured in summer at a suburban forest sitesurrounded by deciduous trees near Tokyo, Japan. Aerosol formation inthe reactor, with an average value of 1.4 × 109 nm3/cm3 (corre-sponding to 2.0 μg/m3), was observed, especially during daytime. Boxmodel simulations, which were based on the SAPRC-99 chemical me-chanism and AERO5 aerosol module with modifications that appliedoxidant-resolved SOA formation yields for isoprene, suggested that 24,21, and 55% of the SOA was formed in the reactor through the oxida-tion of isoprene, monoterpenes, and aromatic hydrocarbons, respec-tively. However, the model could explain only ∼40% of the observedΔOA in the reactor, and a relatively large difference between the ob-served and predicted ΔOA was found around noon and in the afternoon.These results suggest that for a suburban forest environment where theinfluence of both biogenic and anthropogenic emissions is apparent,there exists a significant contribution of missing SOA formation pro-cesses from identified and/or unidentified VOCs, especially during thedaytime. This study demonstrates that potential aerosol formation canprovide useful information on missing SOA formation processes andtheir relation to anthropogenic and biogenic activities. Further ob-servational studies should be made in various environments to exploreand parameterize the missing SOA formation processes. This may helpto improve atmospheric chemistry and climate models.

Acknowledgements

We would like to thank T. Yamasaki (Nagoya Univ.) for his assis-tance with the observation. This work was supported by the Grant-In-Aid for Scientific Research (KAKENHI 25701001, 16H06311,16H02936) from the Ministry of Education, Culture, Sports, Scienceand Technology (MEXT) and the joint research program of the Institutefor Space-Earth Environmental Research, Nagoya University.

Appendix A. Supplementary data

Supplementary data related to this article can be found at http://dx.doi.org/10.1016/j.atmosenv.2017.10.014.

References

Altieri, K.E., Carlton, A.G., Lim, H.J., Turpin, B.J., Seitzinger, S.P., 2006. Evidence foroligomer formation in clouds: reactions of isoprene oxidation products. Environ. Sci.Technol. 40, 4956–4960.

Baker, B., Guenther, A., Greenberg, J., Goldstein, A., Fall, R., 1999. Canopy fluxes of 2-methyl-3-buten-2-ol over a ponderosa pine forest by relaxed eddy accumulation: fielddata and model comparison. J. Geophys. Res. 104, 26107–26114.

Beardsley, R.L., Jang, M., 2016. Simulating the SOA formation of isoprene from parti-tioning and aerosol phase reactions in the presence of inorganics. Atmos. Chem. Phys.16, 5993–6009.

Boucher, O., 2015. Atmospheric Aerosols: Properties and Climate Impacts. Springer, NewYork.

Brégonzio-Rozier, L., et al., 2015. Gaseous products and secondary organic aerosol for-mation during long term oxidation of isoprene and methacrolein. Atmos. Chem. Phys.15, 2953–2968.

T. Nakayama et al. Atmospheric Environment 171 (2017) 91–97

96

Bruns, E.A., et al., 2015. Inter-comparison of laboratory smog chamber and flow reactorsystems on organic aerosol yield and composition. Atmos. Meas. Tech. 8, 2315–2332.

Budisulistiorini, S.H., et al., 2015. Examining the effects of anthropogenic emissions onisoprene-derived secondary organic aerosol formation during the 2013 SouthernOxidant and Aerosol Study (SOAS) at the Look Rock, Tennessee ground site. Atmos.Chem. Phys. 15, 8871–8888.

Calvert, J.K., Atkinson, R., Becker, K.H., Kamens, R.M., Seinfeld, J.H., Wallington, T.J.,Yarwood, G., 2002. The Mechanism of Atmospheric Oxidation of AromaticHydrocarbons. Oxford Univ. Press, New York.

Carlton, A.G., Wiedinmyer, C., Kroll, J.H., 2009. A review of Secondary Organic Aerosol(SOA) formation from isoprene. Atmos. Chem. Phys. 9, 4987–5005.

Carlton, A.G., Bhave, P.V., Napelenok, S.L., Edney, E.O., Sarwar, G., Pinder, R.W., Pouliot,G.A., Houyoux, M., 2010. Treatment of secondary organic aerosol in CMAQv4.7.Environ. Sci. Technol. 44, 8553–8560.

Carter, W., 2000. Final Report to California Air Resources Board. Rep. Univ. of Calif.Riverside, pp. 329.

Chen, Q., et al., 2015. Elemental composition of organic aerosol: the gap between am-bient and laboratory measurements. Geophys. Res. Lett. 42, 4182–4189.

Di Carlo, P.D., et al., 2004. Missing OH reactivity in a forest: evidence for unknownreactive biogenic VOCs. Science 304, 722–725.

Ding, X., Wang, X.M., Gao, B., Fu, X.X., He, Q.F., Zhao, X.Y., Yu, J.Z., Zheng, M., 2012.Tracer-based estimation of secondary organic carbon in the Pearl River Delta, southChina. J. Geophys. Res. 117, D05313.

Fry, J.L., et al., 2013. Observations of gas- and aerosol-phase organic nitrates atBEACHON-RoMBAS 2011. Atmos. Chem. Phys. 13, 8585–8605.

Glasius, M., Goldstein, A.H., 2016. Recent discoveries and future challenges in atmo-spheric organic chemistry. Environ. Sci. Technol. 50, 2754–2764.

Goldan, P.D., Kuster, W.C., Fehsenfeld, F.C., 1993. The observation of a C5 alcoholemission in a North American pine forest. Geophys. Res. Lett. 20, 1039–1042.

Hallquist, M., et al., 2009. The formation, properties and impact of secondary organicaerosol: current and emerging issues. Atmos. Chem. Phys. 9, 5155–5236.

Henze, D.K., Seinfeld, J.H., 2006. Global secondary organic aerosol from isoprene oxi-dation. Geophys. Res. Lett. 33, L09812.

Herrmann, H., Schaefer, T., Tilgner, A., Styler, S.A., Weller, C., Teich, M., Otto, T., 2015.Tropospheric aqueous-phase chemistry: kinetics, mechanisms, and its coupling to achanging gas phase. Chem. Rev. 115, 4259–4334.

Hoyle, C.R., et al., 2011. A review of the anthropogenic influence on biogenic secondaryorganic aerosol. Atmos. Chem. Phys. 11, 321–343.

Hu, D., Bian, Q., Li, T.W.Y., Lau, A.K.H., Yu, J.Z., 2008. Contributions of isoprene,monoterpenes, β-caryophyllene, and toluene to secondary organic aerosols in HongKong during the summer of 2006. J. Geophys. Res. 113, D22206. http://dx.doi.org/10.1029/2008JD010437.

Jimenez, J.L., et al., 2009. Evolution of organic aerosols in the atmosphere. Science 326,1525–1529.

Kaiser, J., et al., 2016. Speciation of OH reactivity above the canopy of an isoprene-dominated forest. Atmos. Chem. Phys. 16, 9349–9359.

Kleindienst, T.E., Jaoui, M., Lewandowski, M., Offenberg, J.H., Lewis, C.W., Bhave, P.V.,Edney, E.O., 2007a. Estimates of the contributions of biogenic and anthropogenichydrocarbons to secondary organic aerosol at a southeastern US location. Atmos.Environ. 41, 8288–8300.

Kleindienst, T.E., Lewandowski, M., Offenberg, J.H., Jaoui, M., Edney, E.O., 2007b.Ozone-isoprene reaction: Re-examination of the formation of secondary organic

aerosol. Geophys. Res. Lett. 34, L01805.Kleindienst, T.E., Lewandowski, M., Offenberg, J.H., Edney, E.O., Jaoui, M., Zheng, M.,

Ding, X., Edgerton, E.S., 2010. Contribution of primary and secondary sources toorganic aerosol and PM2.5 at SEARCH network sites. J. Air &Waste Manage. Assoc.60, 1388–1399.

Kuwata, M., Zorn, S.R., Martin, S.T., 2012. Using elemental ratios to predict the density oforganic material composed of carbon, hydrogen, and oxygen. Environ. Sci. Technol.46, 787–794.

Lambe, A.T., et al., 2015. Effect of oxidant concentration, exposure time, and seed par-ticles on secondary organic aerosol chemical composition and yield. Atmos. Chem.Phys. 15, 3063–3075.

Lelieveld, J., Evans, J.S., Fnais, M., Giannadaki, D., Pozzer, A., 2015. The contribution ofoutdoor air pollution sources to premature mortality on a global scale. Nature 525,367–371.

Matsuda, K., Watanabe, I., Mizukami, K., Ban, S., Takahashi, A., 2015. Dry deposition ofPM2.5 sulfate above a hilly forest using relaxed eddy accumulation. Atmos. Environ.107, 255–261.

Nakao, S., Tang, P., Tang, X., Clark, C.H., Qi, L., Seo, E., Asa-Awuku, A., Cocker III, D.,2013. Density and elemental ratios of secondary organic aerosol: application of adensity prediction method. Atmos. Environ. 68, 273–277.

Ng, N.L., et al., 2008. Secondary organic aerosol (SOA) formation from reaction of iso-prene with nitrate radicals (NO3). Atmos. Chem. Phys. 8, 4117–4140.

Ortega, A.M., et al., 2016. Real-time measurements of secondary organic aerosol for-mation and aging from ambient air in an oxidation flow reactor in the Los Angelesarea. Atmos. Chem. Phys. 16, 7411–7433.

Palm, B.B., et al., 2016. In situ secondary organic aerosol formation from ambient pineforest air using an oxidation flow reactor. Atmos. Chem. Phys. 16, 2943–2970.

Palm, B.B., et al., 2017. Secondary organic aerosol formation from in situ OH, O3, andNO3 oxidation of ambient forest air in an oxidation flow reactor. Atmos. Chem. Phys.17, 5331–5354.

Ramasamy, S., et al., 2016. Total OH reactivity measurement in a BVOC dominatedtemperate forest during a summer campaign, 2014. Atmos. Environ. 131, 41–54.

Rattanavaraha, W., et al., 2016. Assessing the impact of anthropogenic pollution onisoprene-derived secondary organic aerosol formation in PM2.5 collected from theBirmingham, Alabama, ground site during the 2013 Southern Oxidant and AerosolStudy. Atmos. Chem. Phys. 16, 4897–4914.

Ren, Y., Grosselin, B., Daёle, Mellouki, A., 2017. Investigation of the reaction of ozonewith isoprene, methacrolein and methyl vinyl ketone using the HELIOS chamber.Faraday Discuss. 200, 289–311.

Sato, K., Inomata, S., Xing, J.–H., Imamura, T., Uchida, R., Fukuda, S., Nakagawa, K.,Hirokawa, J., Okumura, M., Tohno, S., 2013. Effect of OH radical scavengers onsecondary organic aerosol formation from reactions of isoprene with ozone. Atmos.Environ. 79, 147–154.

Shilling, J.E., et al., 2013. Enhanced SOA formation from mixed anthropogenic andbiogenic emissions during the CARES campaign. Atmos. Chem. Phys. 13, 2091–2113.

Surratt, J.D., Chan, A.W.H., Eddingsaas, N.C., Chan, M.–N., Loza, C.L., Kwan, A.J.,Hersey, S.P., Flagan, R.C., Wennberg, P.O., Seinfeld, J.H., 2010. Reactive inter-mediates revealed in secondary organic aerosol formation from isoprene. Proc. Natl.Acad. Sci. U. S. A. 107, 6640–6645.

Suzuki, K., Okochi, H., Ogata, H., 2012. Biogenic volatile organic compounds in a smallforest near greater Tokyo, Japan. J. Jpn. Soc. Atmos. Environ. 47, 51–57.

T. Nakayama et al. Atmospheric Environment 171 (2017) 91–97

97