Review: Exchanges of volatile organic compounds between terrestrial ecosystems and the atmosphere

Many VOCs are reactive in the atmosphere, may produce secondary organic aerosol (SOA), and keep photochemical ozone concentrations high by VOC-involved reactions. Accumulated studies have shown the importance of terrestrial ecosystems which can be sinks and sources of VOCs. The research progress in the exchange of volatile organic compounds (VOCs) between terrestrial ecosystems and the atmosphere was reviewed in this paper. Representative VOCs emitted from terrestrial ecosystems are low-molecular-weight oxygenated VOCs including methanol, acetone, formic and acetic acids, and terpenoids, including isoprene and monoterpenes. Terpenoid emissions have been intensively investigated from the leaf to the canopy level using advanced analytical systems, including proton-transfer-reaction mass spectrometry. Environmental factors, including temperature, light intensity, carbon dioxide and ozone concentrations, and water stress have been reported to affect terpenoid emissions from plants. The combined effects of these environments influence terpenoid emission additively or interactively, and are important in terms of VOC emission estimates against ongoing climate change. Isoprene is most abundantly released into the atmosphere among VOCs; the potential reasons why some plants release such large amounts of carbon as isoprene were summarized in this study. Among oxygenated VOCs, some compounds, including isoprene oxygenates methacrolein and methyl vinyl ketone, are bidirectionally exchanged, and both atmospheric chemical reactions and reactions under oxidative stress in leaves have been regarded as involved in bidirectional VOC exchanges. Bottom-up process-based models and top-down inverse models have been developed to estimate global and local terpenoid emissions. To validate the accuracy and precision of the models, the collection of additional in-situ ground truth data, such as long-term flux measurement data, at various sites is required. Otherwise, these models may still leave large uncertainties compared with CO2 flux models that can be validated with a large number of ground truth flux data.


Introduction
In addition to greenhouse gases, many kinds of trace gas species are exchanged between terrestrial ecosystems and the atmosphere Fig. 1 . Some of them are emitted from biogenic sources, and others are anthropogenically produced. Several species are exchanged bilaterally.
Some of them are highly reactive in the atmosphere and have a short lifetime. These gases can produce other gas species via chemical reactions, and in some cases, the reactions are important in terms of atmospheric chemistry. More than 100 volatile organic compound VOC species are detectable in the atmosphere, even in forest air. Most of them are at sub-ppb v/v level. VOCs are considered to be involved in producing photochemical ozone and secondary organic aerosols SOAs .
VOC and NO x emissions are believed to form photochemical ozone and degrade air quality in urban and sub-urban areas. NO in the atmosphere is produced from NO 2 during daytime by sunlight wavelength < 398 nm . This NO is easily oxidized by O 3 and peroxides to produce NO 2 . This is the reason why the NO concentration is very low compared with that of NO 2 . Peroxides are produced through a series of chemical reactions between VOCs and OH radicals. If the VOC concentration is high, NO reacts more frequently with the peroxides than with O 3 . This results in maintaining the ozone concentration at a higher level. In urban areas, the reduction of NO x and VOC emissions is essential to decrease the O 3 concentration. SOA is also produced from VOCs and affects human health and the climate. SOAs absorb and scatter shortwave radiation and affect cloud properties, as they serve as cloud condensation nuclei CCN or ice nuclei by providing their surface for water vapor condensation to form liquid droplets and ice particles Shrivastava et al., 2017 . VOCs are degraded in the atmosphere by a number of reactions with OH, ozone, NO 3 , and HO 2 Carlton et al., 2009 . These oxidized products undergo gas-particle conversion processes such as nucleation, condensation, and heterogeneous and multiphase chemical reactions. The formation and growth of SOA depend on the original oxidized products. By mass balance calculation, Goldstein and Galbally 2007 estimated that 510 -910 TgC yr of SOA was formed from globally emitted VOCs with an annual emission of 1300 TgC yr -1 . As annual emission of terpenoids including isoprene and monoterpenes is estimated to be at least more than half of annual VOC emission Arneth et al., 2008 , terrestrial ecosystems may be strongly involved in the formation of photochemical ozone and SOA. Hereinafter, we describe the characteristics of VOC, especially terpenoids, the effects of environmental factors on the exchange of these gases, and the perspective of their behavior affected by ongoing climate change.

Trace gas species and atmospheric impacts 2.1 Isoprene 2.1.1 General
Isoprene is an important secondary metabolite of plants and has a molecular formula of C 5 H 8 . It is volatile b.p. = 34 C and released into the atmosphere immediately after being produced. Isoprene is produced from dimethylallyl diphosphate DMAPP by isoprene synthase in chloroplasts. DMAPP is produced via the mevalonate pathway in cytosol and the methylerythritol 4-phosphate MEP pathway in chloroplasts.
As isoprene is very reactive with OH radicals, lifetime of isoprene in the reaction with OH radicals at a typical concentration 2.0 10 6 molecules cm -3 is 1.4 h Atkinson and Arey, 2003 . Isoprene is less reactive with ozone than with OH radicals, with a lifetime of 1.3 days at a typical ozone concentration 7.0 10 11 molecules cm -3 . Because of its reactivities, isoprene is easily oxidized into methacrolein MAC , methyl vinyl ketone MVK , glycolaldehyde, and other compounds Carlton et al., 2009 . Further reactions of these compounds can lead to the formation of SOA and ozone. Even if the SOA yield from isoprene is ~1 , the large global source strength of isoprene may significantly contribute to the formation of SOA Carlton et al., 2009 . Details of SOA formation from isoprene have been described by Carlton et al. 2009 . Isoprene is emitted from many broadleaved trees and several coniferous trees Table 1 . Isoprene-emitting and non-emitting species can be found even within the same genus. There may be no consistent taxonomic relationships in the presence or absence of isoprene emission Pacifico et al., 2009, but Harley et al. 1999 suggested that the occurrence of isoprene emissions may be involved in evolutionary history. In the genus Quercus, trees in section Cerris are non-emitters, whereas others in sections Rubrae, Quercus, and Protobalanus are isoprene emitters Harley et al., 1999 . In Quercus spp. occurring in East Asia, Tani and Kawawata 2008 revealed that Q. serrata, Q. crispula, Q. dentata, and Q. aliena in the Quercus section were all strong isoprene emitters, but some species in the Cerris section were non-emitters. The presence and absence of isoprene emission is unlikely to be categorized by functional type or climate range Pacifico et al., 2009 . The annual global emission of isoprene is estimated to be 459 -601 TgC by several models Arneth et al., 2008 , which is higher than the annual emission estimates of anthropogenic VOCs 98 -158 TgC y -1 Boucher et al., 2013 .

Cost and benefit of isoprene production
The ratio of carbon emitted in the form of isoprene to photosynthetically fixed carbon is 0.2 -5 depending on temperature and other environmental factors e.g., Tani and Kawawata, 2008 . Isoprene production requires 20 ATP and 14 NADPH per molecule  , and this significant cost is paid by isoprene-emitting species, but not by others. The physiological reason why some plants release such Fig. 1. Trace gas species exchanged between terrestrial ecosystem and atmosphere large amounts of carbon as isoprene is not fully understood and many studies have been conducted. Thermal tolerance seems to be a plausible reason Sharkey et al., 2007 as isoprene emission increases up to ~40 C. Many studies have been undertaken to prove the thermotolerance hypothesis. Singsaas et al. 1997 showed that when air containing several ppm isoprene was passed around leaf discs of an isoprene-emitting plant kudzu Pueraria lobata , the temperature limit of thermal tolerance evaluated by chlorophyll fluorescence increased by as much as 4 C in the least intense light and by as much as 10 C in the more intense light. This suggests that exogenous and endogenous isoprene may help photosynthetic resistance against high temperatures.  used fosmidomycin, which can inhibit isoprene emission without affecting photosynthesis for several hours. The decreased photosynthetic rate of P. lobata caused by a temperature treatment at 46 C for 2 min, recovered less in fosmidomycin-fed leaves than in leaves fed water or fosmidomycin-fed leaves in isoprene-containing air. The non-isoprene-emitting plant Phaseolus vulgaris exhibited increased photosynthetic recovery under a high temperature treatment when air containing 2 -22 ppm isoprene was supplied. These results indirectly suggest that isoprene can mitigate heat stress. Additionally, they show that other short-chain alkenes can also improve thermotolerance, while alkanes reduce thermotolerance.
A genetic approach seems to provide evidence that is more direct. Behnke et al. 2007 produced transgenic grey poplar Populus canescens plants in which the gene expression of isoprene synthase ISPS was silenced by RNA interference RNAi . The authors showed that lack of capacity for isoprene emission decreased the net photosynthetic rate and photosynthetic electron transport of poplars by heat stress. Velikova et al. 2014 suggested that a rearrangement of the chloroplast protein profile occurs in non-isoprene-emitting poplars, probably to minimize the negative stress effects caused by the absence of isoprene. As isoprene synthase was found to bind to the thylakoid membrane Wildermuth and Fall, 1996 , the results of Velikova et al. 2014 may support the idea that isoprene improves or stabilizes the thylakoid membrane structure under heat stress.
Other indirect approaches have been made from an ecosystem perspective. Taylor et al. 2018 estimated the proportional abundance of isoprene-emitting trees at 103 lowland tropical sites and revealed that the proportional abundance increased with the mean annual temperature of the sites. Taylor et al. 2019 compared the photosynthetic temperature responses of 26 co occurring tropical trees and liana species to test whether isoprene emitting species are more tolerant of high temperatures. They revealed that the maximum temperatures for net photosynthesis were ~1.8 C higher in isoprene emitting species than in non-emitters.
The theory of isoprene defense against thermal stress has been expanded to other oxidative stresses Vickers et al., 2009a , as isoprene may act as an antioxidant in leaves owing to its high reactivity. Loreto and Velikova 2001 showed that fosmidomycin-fed leaves of isoprene-emitting plants were less tolerant to ozone. They estimated that isoprene quenched reactive oxygen species ROS , including H 2 O 2 and singlet delta oxygen Zeinali et al., 2016 formed in leaves, and reduced the lipid peroxidation of cellular membranes caused by ozone. Vickers et al. 2009b used transgenic tobacco plants capable of emitting isoprene and found that the isoprene-emitting plants were more resistant to ozone-induced oxidative damage than non-emitting controls. They also found that the content of reduced ascorbate in leaves, which functioned as an antioxidant, was higher in transgenic isoprene-emitting plants than in non-emitting plants, implying that the former contained antioxidant compounds in higher abundance. Isoprene directly reacts with ozone in leaves, producing hydroperoxides that are toxic to plants Hewitt et al., 1990 . However, the isoprene reaction with ozone is not fast Atkinson and Arey, 2003 , meaning that it may not work effectively even though the isoprene concentration inside the leaves is high ~several tens ppm . Isoprene may scavenge ROS generated in plants under other abiotic stresses, including intense light, water deficit, mechanical damage, salinity in soil, and other air pollutants Vickers et al., 2009a . So far, it has become evident that isoprene acts as a ROS scavenging system in leaves.

Monoterpene and other terpenoids
Monoterpene is the general name of a group of secondary metabolites produced by plants. The molecular formula of monoterpenes is C 10 H 16 . Monoterpenes are reactive in the atmosphere and emitted substantially. Monoterpenes are produced from geranyl diphosphate GPP by monoterpene synthase in plastids. Monoterpene alcohols C 10 H 18 O and ketones C 10 H 16 O are also included in the monoterpene family. They are volatile, but their boiling points are > 150 C, suggesting that they are produced in the liquid phase. Most monoterpene-emitting plants have specific storage organs and tissues, including oil glands and trichomes on leaf surfaces and resin ducts inside leaves, stems, or roots. Most coniferous trees have resin ducts and are classified as storage-type monoterpene emitters. Some broad-leaved trees, including those in the Eucalyptus genus and the Lauraceae family Table 1 , are categorized into the storage-type monoterpene emitters. Plants that have no storage organs are exceptions. They emit monoterpenes immediately after producing them. Some Quercus species, including Q. ilex Seufert, 1995 andQ. phillyraeoides Okumura et al., 2008 are of the non-storage type.
Some of the monoterpene compounds are very reactive with OH radicals and ozone Atkinson and Arey, 2003 in the atmosphere, causing the formation of SOA and ozone. The lifetimes of monoterpenes α-pinene, β-pinene, and d-limonene with OH radicals at a typical concentration 2.0 10 6 molecules cm -3 are 2.6 h, 1.8 h, and 49 min, respectively, Atkinson and Arey, 2003 . Their lifetimes in the reaction with ozone at a typical ozone concentration 7.0 10 11 molecules cm -3 are 4.6 h, 1.1 days, and 2.0 h, respectively. The annual global emission of monoterpenes is estimated by several models to be 32 -127 TgC Arneth et al., 2008 .
The molecular formula of sesquiterpenes is C 15 H 24 . Because of their low volatility, their emission rate is generally low. However, they are produced and emitted after being fed to insects Holopainen and Gershenzon, 2010 . Sesquiterpene ketones and aldehydes are also often detected in plant emissions. Diterpenes are rarely detected in the atmosphere and is emitted by a very limited number of tree species. A large amount of kaur-16-ene was reported to be produced and emitted by Cryptomeria japonica Matsunaga et al., 2012 . Many of sesquiterpenes and diterpenes are reactive and low volatile, therefore easily converted to SOA in the atmosphere.

Other biogenic gases
Methanol and acetone are among the most abundant VOCs in the atmosphere Jacob et al., 2002; . Atmospheric methanol is oxidized by OH radicals to produce formaldehyde Riemer et al., 1998 . Methanol and formaldehyde are toxic organic air pollutants, and their emissions are regulated by the United States Environmental Protection Agency EPA and the Ministry of the Environment Japan . Acetone is an important source of HO x radicals Singh et al., 1995 that contribute to the formation of ozone in the atmosphere.
Formic and acetic acids in the gas phase are the dominant organic acids in the atmosphere Kawamura et al., 2000;Mochizuki et al., 2019 . Gaseous formic and acetic acids are adsorbed on existing alkaline particles Alexander et al., 2015 . As formic and acetic acids are highly water soluble, they can alter the hygroscopic properties of atmospheric particles Kanakidou et al., 2005 . It should be noted that organic acids contribute to the acidity of wet deposition e.g., Kawamura et al., 1996 . Atmospheric concentrations of methanol, acetone, and formic and acetic acids range from a few ppt to several ppb see Seco et al., 2007 and the references therein . The lifetimes of gaseous methanol and acetone reacting with OH radicals at a typical concentration 2.0 10 6 molecules cm -3 are 12 and 61 days, respectively Atkinson and Arey, 2003 . The lifetimes of gaseous formic and acetic acids reacting with OH radicals at a typical concentration 2.0 10 6 molecules cm -3 are 13 and 7.2 days, respectively NIST Standard Reference Database Number 69 , suggesting that their lifetimes are much longer than those of reactive hydrocarbons, such as isoprene and monoterpenes. Methanol, acetone, and formic and acetic acids are directly emitted from plant leaves, leaf litter, and dead plant matter Seco et al., 2007 . Briefly, methanol is a product of pectin demethylation in the plant cell wall. Acetone is produced after the decarboxylation of acetoacetate, which is derived from acetyl-CoA in plant leaves. Formic acid is a product of the decarboxylation of glycolic acid during photorespiration. It is also produced through the oxidation of formaldehyde in plant leaves. Acetic acid is a product of the hydrolysis of acetyl-CoA. Several studies have found that emissions of methanol, acetone, and formic and acetic acids from plant leaves varied with temperature, light intensity, and stomatal conductance Nemecek-Marshall et al., 1995;Kesselmeier et al., 1997;Janson and Serves, 2001 . However, the available data on these compounds that are directly emitted by vegetation are limited, and their controlling factors are unclear. Formic and acetic acids are also produced secondarily from the photochemical oxidations of terpenoids, including isoprene and monoterpenes in the atmosphere Paulot et al., 2011 . Methanol, acetone, and formic and acetic acids are emitted from a variety of anthropogenic and biogenic sources. , respectively Paulot et al., 2011 . The emissions of formic and acetic acids from terrestrial vegetation account for 4.5 and 3.0 of the global emissions, respectively; conversely, the secondary production of formic and acetic acids via the photochemical oxidation of biogenic VOCs is estimated to account for 74 and 67 of the global emissions, respectively.

Anthropogenic VOCs
The fossil fuel burning and solvent industry release a wide variety of anthropogenic VOCs, including aliphatic hydrocarbons, aromatic hydrocarbons, ketones, aldehydes, and alcohols. Toluene is the most abundant compound in urban and rural atmosphere. Benzene is also ubiquitous and the second abundant compound. In general, the reactivity of these compounds is low compared with those of isoprene and monoterpenes. However, compounds with double or triple bonds, including acrolein, are more easily oxidized in the atmosphere.
Partitioning of anthropogenic VOCs from air phase to water phase is explained by Henry's law constant. Small amounts of benzene and toluene are partitioned into water phase but oxygenated VOCs including aldehydes, ketones and alcohols are largely partitioned into water phase. As plant leaves include water at a volume ratio of >80 , substantial amounts of these oxygenated compounds are partitioned in leaves. Some of them are metabolized in plant cells Trapp and Karlson, 2001 . Octanol/water partition coefficient is another index that determines the degree of partition of VOCs between plants and outside environments. Toluene is substantially stored in lipids of plant leaves Keymeulen et al., 1997 . Mobility of VOCs within plants from roots to leaves depends on the octanol/water partition coefficient and differs for different VOC species Burken and Schnoor, 1998 .

Isoprene emissions from leaves, branches, and individual
trees Isoprene emissions from plants on the leaf and branch levels have been intensively measured in terms of plant physiology and atmospheric chemistry. By using a real-time isoprene analyzer, such as a chemiluminescence analyzer Hills and Zimmerman, 1990 and a proton-transfer-reaction mass spectrometer PTR-MS Lindinger et al., 1997 , it is evident that isoprene emission is not affected by the degree of stomatal opening. This is different from the CO 2 exchange on the leaf surface that is governed by stomatal conductance. This is because isoprene synthesis is not affected by the intercellular isoprene concentration. When the opening area of the stomata becomes narrow, the intercellular concentration immediately increases to equilibrate the isoprene production rate with its emission rate Fall and Monson, 1992 .
The temperature dependency of isoprene is governed by isoprene synthase. Analogous to most enzymes, its optimum temperature is ~40 C. The isoprene emission rate increases with an increase in temperature up to ~40 C Fig. 2 . The light dependency of isoprene is closely related to the DMAPP production rate, which is affected by CO 2 fixation via photosynthesis. The isoprene emission rate curve against light intensity is very similar to the curve of photosynthetic rate against light Fig , depending on the sun exposure conditions of leaves Harley et al., 1996Harley et al., , 1997Hayward et al., 2004;Tani and Kawawata, 2008 . The equations that express the isoprene emission rate depending on temperature and light intensity were developed by Guenther et al. 1993 and have been called the G93 algorithm. These equations have been widely used to estimate the isoprene emission rate not only on the leaf area and branch bases, but also on the plant and canopy levels.
The effect of elevated CO 2 concentration on plant isoprene emissions has been investigated. Short-term fumigation can decrease the isoprene emission rate compared with ambient CO 2 fumigation e.g., Tingey et al., 1981 . The isoprene precursor DMAPP is produced by the MEP pathway in chloroplasts, and phosphoenolpyruvate PEP is converted to pyruvate, which is a substance of the MEP pathway. The elevated CO 2 may result in a higher concentration of total leaf PEP, but cause a lower PEP concentration in the stroma when compared with ambient CO 2 concentration. This stromal PEP contributes to isoprene production, and this change in this metabolic balance may cause an isoprene emission reduction. The details of this hypothesis are described by Sharkey and Monson 2014 . Many reports suggest that long-term fumigation experiments with elevated CO 2 may also result in the inhibition of isoprene emission Rosenstiel et al., 2003;Scholefield et al., 2004;Monson et al., 2007 , although a few reports have shown that it may increase the isoprene emission capacity Sun et al., 2012; . The inhibition of isoprene emission may be caused by a decreased activity of isoprene synthase Scholefield et al., 2004;Possell and Hewitt, 2011 in addition to the lower PEP concentration in the stroma mentioned above.
Elevated O 3 has been reported to increase isoprene production in several short-term fumigation experiments. Increased isoprene emission under ~300 ppb O 3 quenched H 2 O 2 formed in the ozone-exposed leaves of Phragmites australis and reduced lipid peroxidation in the cellular membranes Loreto and Velikova, 2001;Velikova et al., 2005 . However, this protective effect was not obvious under long-term exposure with a possible concentration of O 3 ambient + 40 ppb in many cases . The isoprene emission rate was reduced by long-term exposure, probably by the suppressed gene expression and protein level of isoprene synthase Calfapietra et al., 2007Calfapietra et al., , 2008 or the decreased production of DMAPP Yuan et al., 2016, Tani et al., 2017 . Under enriched O 3 conditions, biomass was reported to be less decreased in aboveground parts, including leaves, than in the roots, in order to mitigate the decreased photosynthesis on the leaf area basis Agathokleous et al., 2018 . If this is the case, the isoprene emission rate of individual trees may be less decreased than the isoprene emission rate on the leaf area basis.
The effect of water stress on isoprene emission has been intensively investigated. The isoprene emission rate of Q. pubescens Brüggemann and Schnitzler, 2002, Populus alba Brilli et al., 2007, Q. rubra Funk et al., 2005, P. deltoides Funk et al., 2004Pegoraro et al., 2004 , andLiquidambar styraciflua Fang et al., 1996 is reduced by water stress, but the degree of reduction is smaller than that of the photosynthetic reduction. Quercus serrata and Q. crispula, which are major tree species native to East Asia, continue to produce isoprene even when the gross photosynthetic rate becomes lower than the respiration rate during daytime net photosynthetic rate < 0 Tani et al., 2011 . Under normal conditions, DMAPP is supplied mainly from chloroplastic sources, but it may be supplied substantially from an alternative source in the cytosol. The cytosol source may be the primary precursor of isoprene under drought conditions Karl et al., 2002;Schnitzler et al., 2004 . Under severe drought conditions, the alternative source is so limited that the isoprene emission rate is low.
A change in land use may enhance the emission of isoprene from modified vegetation, including spreading bamboo forests Okumura et al., 2018;Chang et al., 2019 and palm oil plantations Wilkinson et al., 2006;Misztal et al., 2011;Dislich et al., 2017 .

Monoterpene emissions from leaf, branch, and individual
tree Monoterpene emissions from non-storage types depend on temperature and light intensity, as does isoprene emission. It increases with temperatures of up to ~40 C and with photosynthetic photon flux density PPFD of up to ~1000 µmol m -2 s -1 Staudt and Seufert, 1995;Staudt and Bertin, 1998 . Monoterpene emissions from storage type plants depend on temperature because monoterpene evaporates from storage organs following the temperature-dependent equation of saturated vapor pressure, which is similar to the Antoine equation. A more simplified equation was proposed by Guenther et al. 1993 , as shown in Fig. 2. In this equation, coefficient β determines the slope of the fitted curve. Coefficient β can be calculated using datasets of the emission rate and temperature, and it is reported to have a wide range of values, that is, 0.02 -0.13 within a strain of C. japonica Miyama et al., 2019 .
The effects of CO 2 , ozone, and water stress on monoterpene emissions from non-storage type trees are similar to those on isoprene emission e.g., Loreto et al., 2001 . However, the effects on monoterpene emissions from storage-type trees are not the same as those on isoprene emission and are not consistent among reports. No effects of elevated CO 2 on the total monoterpene emission rate were observed for the evergreen coniferous tree Pinus sylvestris 700 µmol mol -1 throughout a 5 year period Räisänen et al., 2008

Terpenoid emissions from soil
Leaf litter, roots, forest floor plants, and the microbiological decomposition of organic matter are monoterpene sources Hayward et al., 2001;Leff and Fierer, 2008;Ludley et al., 2009;Tsuruta et al., 2018 . Their source strengths are diverse and differ among vegetation types. As for forest soil, the soil monoterpene emission rates in coniferous P. sylvestris Ketola et al., 2011 andP. sitchensis Hayward et al., 2001 forests were 0.86 nmol m -2 s -1 8 C in soil temperature and 0.08 nmol m -2 s -1 18 C in air chamber temperature , respectively. The soil monoterpene emission rate in a broadleaved Q. ilex forest was 0.004 nmol m -2 s -1 8 C in soil temperature Asensio et al., 2007 . As for the forest floor including ground vegetation e.g., herbaceous plants, fungi, and mosses , the monoterpene emission rate of the floor of P. sylvestris forests was reported to be 1.9 nmol m -2 s -1 28 C in air chamber temperature Aaltonen et al., 2013 and 0.013 nmol m -2 s -1 7.5 C in air chamber temperature Aaltonen et al., 2011. Smolander et al. 2006 reported that the monoterpene concentration in the aerial space of soil in coniferous Picea abies and P. stlvestris forests were significantly higher than that in broadleaved Betula pendula forest. The monoterpene concentration in the aerial space of soil in a coniferous P. mariana black spruce forest was significantly higher than the atmospheric concentration above the forest Morishita et al., 2019. Hayward et al. 2001 reported that monoterpene emission rates differed among soil depths, and needle litter was a strong source of monoterpenes. The monoterpene emission rate of the P. sylvestris forest floor was the highest in autumn Aaltonen et al., 2011 and the observed seasonality was attributed to an increase in litter fall in autumn as the needle litter included monoterpenes. Coniferous forests floor/soil may be an important source of monoterpenes in the atmosphere.
Contrarily, the isoprene emission rate of a P. sylvestris forest floor was 0.0002 nmol m -2 s -1 , which was much lower than the monoterpene emission rate at the same site Aaltonen et al., 2011 . The isoprene emission rate of arctic soil Salix-Eriophorum community measured at air temperature of ~10 C was 0.02 nmol m -2 s -1 Svendsen et al., 2016 .

Leaf uptake of OVOCs
Oxygenated VOCs, including low-molecular-weight alcohols, ketones, and aldehydes, have been reported to be absorbed by plants via stomata Omasa et al., 2000;Tani and Hewitt, 2009;Tani et al., 2013 . At the concentration range of ppb to several tens ppb, measurements based on PTR-MS provide more precise data than does GCMS Tani et al., 2007 . The uptake rate of ketones and aldehydes increases linearly with stomatal conductance g S Tani et al., 2010; , suggesting that the uptake is controlled by stomatal opening. However, its slope may become gentle when g S exceeds a certain level. This phenomenon has been reported for Spathiphyllum fumigated in propionaldehyde and methyl ethyl ketone Tani and Hewitt, 2009 . This is because the ratio of the intercellular concentration to the fumigated concentration of the gas species becomes higher owing to the limitation of metabolic conversion.
Isoprene oxygenates MAC and MVK are also absorbed by trees Tani et al., 2010;Karl et al., 2010 . MAC is metabolized in tissues to produce glutathione conjugates Muramoto et al., 2015 and high-volatile compounds including isobutyl aldehyde, isobutyl alcohol, and methallyl alcohol when tomato plants are fumigated with MAC at the ppm level. Cappellin et al. 2019 recently reported that MVK absorbed by an isoprene-emitting species Quercus rubra was reduced in leaves to emit methyl ethyl ketone MEK or 3-buten-2-ol. Further reactions reduced MEK and 3-buten-2-ol to 2-butanol. A similar result has been reported by Tani et al. 2020 , but they did not observe 3-buten-2-ol emission from non-isoprene-emitting species including three tree and one houseplant species. Cappellin et al. 2019 determined the conversion ratio of MVK to MEK to be 73 , and that of MVK to all volatiles including MEK, 3-buten-2-ol, and 2-butanol to be 97.6 , suggesting that the absorbed MEK was mostly converted to volatiles and scarcely remained in leaves. On the contrary, MVK conversion ratios of the non-isoprene-emitting species determined by Tani et al. 2020 were much lower than those reported by Cappellin et al. 2019 ; i.e. 26 -39 for MEK and 33 -44 for all volatiles. The difference in the conversion ratios from MVK to volatiles between two studies might be attributed to plant materials, i.e. isoprene-emitting or non-emitting species. Isoprene-emitting trees may have more active enzymes for converting MVK to volatiles than non-emitting plants, because cells of the isoprene-emitting plants are always exposed to high concentrations of isoprene oxygenates, including MVK, during daytime. This is because that MVK and MAC are formed in leaves by isoprene reaction under oxidative stress. Jardine et al., 2010 . As reactive carbonyl species including MVK and MACR are deleterious to plants at high concentrations Mano, 2012; Farmer and Mueller, 2013 , it may be safer for the isoprene-emitting plants to release the oxygenates into the atmosphere in the form of other volatiles because the pool size used for accumulating these oxygenates and their metabolites within plant tissues is limited.
It has been reported that some other VOCs are bilaterally exchanged between plants and the atmosphere Niinemets et al., 2014 . Four species of low molecular weight oxygenated VOCs acetaldehyde, formaldehyde, and acetic and formic acids are reported to be both emitted and taken up by plants Kesselmeier, 2001 . The compensation point of the compounds within leaves may vary, depending on their atmospheric concentration and production rate of these compounds within plants Fares et al., 2015 . Using an inverse model, a sink and source analysis of MACR and MVK was conducted by Wada et al. 2020 , but their result suggests that larger amounts of data on spatial distribution of these compounds should be collected to provide a valid estimate.

Flux measurement by micrometeorological techniques
Flux measurement techniques employed for organic gases are theoretically the same as those employed for CO 2 flux. The eddy covariance EC method may be applied for terpenoid fluxes. Fast-response analytical instruments including PTR-MS have been employed Karl et al., 2001 . In terpenoid flux measurements, PTR-MS can distinguish between VOCs of different molecular weights by using quadrupole or time of flight TOF mass filters. The TOF mass filter has the advantage of mass resolution as it allows the estimation of the molecular formula of the detected ions, though we cannot know whether the ions originate from molecular or fragment ions; in most cases they originate from both ion types, however we cannot be certain what percentage comes from molecular ions, which causes quantification and qualification uncertainties. Humidity affects the PTR-MS sensitivity and the fragment ion patterns of the proton-transfer-reaction products of monoterpenes Tani et al., 2004 , and these effects should not be ignored in certain cases. A theoretically determined concentration does not always provide an accurate absolute value of VOC concentrations Kato et al., 2004 and requires correction using the known concentration of the target gas Tani et al., 2003 . To overcome these uncertainties, a recent study tried to improve the data handling method of PTR-MS based flux measurements Millet et al., 2018 .
The relaxed accumulation method REA coupled with the thermal desorption method and gas chromatography mass spectrometry GCMS can provide a more reliable identification of each compound and more accurate concentrations e.g., Mochizuki et al., 2014 . The upward and downward airflows are separately accumulated into reservoirs and the determined concentrations of the two reservoirs are used for the flux calculation. The reservoir type is either canister or adsorbent tube.
The gradient method is also used for many trace gas species including terpenoids Tani et al., 2002 . This method was used in the early years of trace gas flux measurements. However, it is still used if measurement devices with high-time resolution for EC measurement are unavailable and if the gas species cannot be accumulated into REA reservoirs owing to its unstable characteristics. The concentration of the target gas at two different heights within the boundary layer above the tree canopy is measured and the difference between the two concentrations is divided by the height difference and multiplied by the diffusion coefficient of the target gas to determine its flux.
Other methods including the disjunct eddy sampling method Rinne et al., 2000 have been developed. Although several methods to measure trace gas fluxes are available, the EC method may be advantageous from a timesaving point of view if fast-response analytical instruments provide reliable data.

Terpenoid fluxes
Isoprene and monoterpene fluxes have been reported for various forest types including boreal, cool temperate, temperate, Mediterranean, tropical, and savanna forests Table 2 . Broadleaved forests mainly emit isoprene and coniferous forests mainly emit monoterpenes. The canopy flux of isoprene in boreal and cool temperate, temperate and Mediterranean, and tropical , respectively. Relatively high monoterpene fluxes were observed in Mediterranean broadleaved forests and in savanna woodland areas broadleaved trees compared with coniferous forests. Overall, the canopy flux of isoprene is higher than that of monoterpenes.
Isoprene and monoterpenes include 5 and 10 carbon atoms, respectively, in each molecule. The emission of isoprene and monoterpenes is the loss of carbon fixed by photosynthesis. The annual carbon emissions as monoterpenes from a L. kaempferi

Factors affecting ecosystem-level exchange of biogenic VOCs
Generally, canopy-scale isoprene and monoterpene emissions start to increase in the morning and reach a peak around noon or in the afternoon, as they are affected by ambient air temperature and light intensity Laffineur et al., 2011;Mochizuki et al., 2014;Seco et al., 2015;Mochizuki et al., 2020 . However, higher monoterpene fluxes were observed in the morning than in the afternoon Tani et al., 2002 . This is because of the so-called storage effect.
A relatively high isoprene flux was observed in broadleaved mixed forests in the summer Seco et al., 2015 . The monoterpene fluxes of L. kaempferi and P. ponderosa forests are significantly higher in summer than in winter Mochizuki et al., 2014;Holzinger et al., 2006 . The seasonal variation in isoprene and monoterpene fluxes above the forest is not only linked to ambient air temperature variations, but is also involved in emission capacity changes caused by enzymatic activity. The monoterpene flux standardized to 30 C is highest in early summer June in a P. densiflora forest Tani et al., 2002 . High When comparing the terpenoid emissions from the forest floor E floor with canopy scale terpenoid fluxes E canopy , the ratio of E floor /E canopy in a P. sylvestris forest in autumn Aaltonen et al., 2011 and a P. ponderosa forest Greenberg et al., 2012 in the summer was 10 and 1 , respectively. Janson 1993 observed a higher E floor /E canopy ratio 20 -40 under wet conditions. The soil and the forest floor may be important terpenoid emission sources from forest ecosystems, but their importance depends on the vegetation type, season, and environmental conditions. In particular, wet conditions may enhance monoterpene emissions from the soil and litter.
Long-term flux measurements may provide insight into the response of terrestrial ecosystems' terpenoid emissions to complex climate change. However, continuous or intermittent measurements of terpenoid fluxes for years have not been conducted. Compared with CO 2 and H 2 O fluxes, it is difficult to operate the measurement devices automatically in order to measure the terpenoid concentrations. Only PTR-MS may be a candidate, but it requires careful operation and frequent maintenance. A limited number of researchers use PTR-MS for such exclusive monitoring for years.

Combined effect analyses to address climate change effects
Recent studies have focused on the combined effects of two or more environmental parameters on plant terpenoid emissions, in order to estimate the impact of ongoing climate change on terpenoid-related atmospheric chemistry and climate. For this study, leaf, branch, and individual tree measurements were conducted.
A combined experiment of long-term CO 2 and O 3 exposure was conducted by Calfapietra et al. 2008 , the results of which suggest that the decline of isoprene emission under elevated CO 2 and elevated O 3 may be exacerbated by a combination of substrate CO 2 effect and enzyme limitations O 3 effect . The effects of CO 2 and O 3 on monoterpene emissions from B. pendula Vuorinen et al., 2005 andP. sylvestris Sallas et al., 2001 differed by season and monoterpene compounds.
The inhibition of isoprene production under elevated CO 2 has been reported to be mitigated by high temperatures Loreto and Sharkey, 1990;Sun et al., 2013;Potosnak et al., 2014 . However, these experimental results are based on relatively short-term experiments and are not enough to provide an understanding of the mechanism responsible for the plants acclimated to elevated CO 2 or high temperature.
Mochizuki et al. 2018 measured the monoterpene emission rate of C. japonica clone saplings grown at elevated CO 2 conditions in an open-top chamber. The normalized monoterpene emission rates were positively and linearly correlated with soil water content SWC under both the control and elevated CO 2 conditions, but the slope of the regression line between them was significantly higher under the elevated CO 2 conditions, indicating that the monoterpene emissions from plants grown under elevated CO 2 conditions are more sensitive to SWC.
The combined effects of O 3 and nitrogen on plant terpenoid emissions have been investigated Llusià et al., 2014;Carriero et al., 2016;Yuan et al., 2017 . The results of these studies suggest that the effects of O 3 and N on terpenoid emission from plants are additive and not interactive. More complicated effects of elevated CO 2 and O 3 and high temperature on herbivore-induced terpenoid emissions were addressed by Kivimäenpää et al. 2016 andGhimire et al. 2017 . They showed a dramatic increase in total sesquiterpene emissions from the fed shoots of B. pendula and P. sylvestris in combination with high temperature and elevated ozone.

Model estimates and satellite measurements to address climate change effects
Based on the findings of leaf, branch, and individual tree measurements, several process-based models have been developed to estimate global and local terpenoid emissions Arneth et al., 2008 . Most of the models employ the G93 algorithm i.e., the MEGAN model by Guenther et al. 2006 , while others include a model based on energetic requirements for isoprene synthesis Niinemets et al., 1999 . Using the process-based models, coupled with leaf area index LAI , land cover data, vegetation lists classified as terpenoid emission type, and climate data, terpenoid emissions on global Pacifico et al., 2012;Sindelarova et al., 2014 andregional Ren et al., 2014;Bauwens et al., 2018 scales have been estimated for the past, present, and future. The resolution of climate data Ashworth et al., 2010 andplant classification Arneth et al., 2013 has been suggested to affect the accuracy of the terpenoid emission estimates. The tower-flux measurement data from a very limited number of vegetation types were compared with the model estimates for the sites to investigate the validity of the models Sindelarova et al., 2014;Bauwens et al., 2018 . Because of the limited number of flux data at individual sites, the comparisons showed fair agreement. Future estimates of global isoprene emissions based on climate change scenarios indicate that the isoprene emission may increase because of global warming, but this could be inhibited by a CO 2 increase. The degree of change in global isoprene emission estimates depends on the model characteristics, vegetation data, and future scenarios Pacifico et al., 2012;Bauwens et al., 2018 . Ozone and SOA formations have been estimated according to the global warming scenario with a process-based isoprene emission model coupled with an atmospheric chemistry model. Pacifico et al. 2012 suggest that the ozone and SOA burden into the atmosphere may increase by 15 and two-fold, respectively, in the future 2100 -2109 compared with the present levels.
Top-down isoprene emission models have also been developed using space-based formaldehyde HCHO vertical columns, as observed by the ozone monitoring instrument OMI Levelt et al., 2006 and scanning imaging absorption spectrometer for atmospheric chartography SCIAMACHY Bovensmann et al., 1999 . As HCHO is an oxidation product of most VOCs emitted by biomass burning, vegetation, and anthropogenic activities, satellite observations of HCHO may provide information on the underlying VOC sources. Barkley et al. 2013 showed that the annual isoprene emission over tropical South America estimated by the SCIAMACHY column was two times higher than that estimated by the OMI column, the reason of which could not be explained by sampling time of sensors and cloud products by the air-mass factor AMF calculation. Bauwens et al. 2016 compared the inverse analysis using the HCHO vertical column with the MEGAN bottom-up model and discussed the reason for the difference in isoprene emission estimates. Recently, Fu et al. 2019 showed the possibility of the direct measurement of isoprene columns from space using the satellite-borne cross-track infrared sounder CrIS . If the model based on the isoprene column is verified by ground truth data at multiple sites, the approach has the potential to become a more reliable method to estimate isoprene emission sources and their strength from space.
To validate the global and local terpenoid emissions estimated from bottom-up process-based models and top-down satellite estimates, a comparison between the two different methods may not lead to realistic results. To validate the accuracy and precision of the models, much more in-situ ground truth data, such as long-term flux and concentration measurement data at various sites, are required. As accumulated in-situ flux data may indicate a good relationship between the measured flux and environmental factors e.g., Mochizuki et al., 2014 , the data can be used to validate the developed models. Otherwise, these models may still leave large uncertainties compared with global carbon exchange models that can be validated with a large number of ground truth flux data.