Leaf level emissions of volatile organic compounds ( VOC ) from some Amazonian and Mediterranean plants

Emission inventories defining regional and global biogenic volatile organic compounds (VOC) emission strengths are needed to determine the impact of VOC on atmospheric chemistry (oxidative capacity) and physics (secondary organic aerosol formation and effects). The aim of this work was to contribute with measurements of tree species from the poorly described tropical vegetation in direct comparison with the quite well-investigated, highly heterogeneous emissions from Mediterranean vegetation. VOC emission from sixteen plant species from the Mediterranean area were compared with twelve plant species from different environments of the Amazon basin by an emission screening at leaf level using branch enclosures. Analysis of the volatile organics was performed online by a proton-transferreaction mass spectrometer (PTR-MS) and offline by collection on adsorbent tubes and subsequent gas chromatographic analysis. Isoprene was the most dominant compound emitted followed by monoterpenes, methanol and acetone. The average loss rates of VOC carbon in relation to the net CO2 assimilation were found below 4 % and indicating normal unstressed plant behavior. Most of the Mediterranean species emitted a large variety of monoterpenes, whereas only five tropical species were identified as monoterpene emitters exhibiting a quite conservative emission pattern (α-pinene< limonene< sabinene< ß-pinene). Mediterranean plants showed additional emissions of sesquiterpenes. In the case of Amazonian plants no sesquiterpenes were detected. However, missing of sesquiterpenes may also be due to a lack of sensitivity of the measuring systems. Furthermore, our screening activities cover only 1 % of tree species of such tropical areas as estimated based on recent biodiversity reports. Methanol emissions, an indicator of growth, were found to be common in most of the tropical and Mediterranean species. A f w species from both ecosystems showed acetone emissions. The observed heterogeneous emissions, including reactive VOC species which are not easily detected by flux measurements, give reason to perform more screening at leaf level and, whenever possible, within the forests under ambient conditions.


Introduction
Vegetation is the main producer of volatile organic compounds (VOC) affecting chemical and physical properties of the atmosphere (Atkinson and Arey, 2003).Biogenic VOC regulates the oxidative capacity of the atmosphere and has an indirect impact on the lifetime of greenhouse gases.Furthermore, they are involved in the formation and growth of secondary organic aerosols (SOA), which in turn affect cloud development and precipitation (Andreae and Crutzen, 1997;Wang et al., 1998;Collins et al., 2002;Lelieveld et al., 2002;Claeys et al., 2004;Pöschl et al., 2010;Sjostedt et al., 2011).
Emission inventories and model calculations have been developed to define regional and global biogenic VOC emission strength, however they rely mostly on studies of VOC emissions from vegetation from temperate areas of North America and Europe (Guenther et al., 1995;Kinnee et al., 1997; Published by Copernicus Publications on behalf of the European Geosciences Union. A. Bracho-Nunez et al.: Leaf level emissions of volatile organic compounds Simpson, 1999;Guenther et al., 2006), or on studies of only one VOC species, i.e. isoprene (Lerdau and Keller, 1997;Lerdau andThroop, 1999, 2000;Harley et al., 2004;Pegoraro et al., 2006;Oku et al., 2008;Ferreira et al., 2010;Misztal et al., 2010).Because of the heterogeneity of ecoregions, species diversity, inaccessibility, and logistical and methodological difficulties, the number of investigations made in tropical regions is limited (Geron et al., 2002;Kesselmeier et al., 2002Kesselmeier et al., , 2009;;Kuhn et al., 2002aKuhn et al., , b, 2004Kuhn et al., , 2007;;Harley et al., 2004).Yet, VOC emissions from tropical forests are of special interest because their contribution to the global VOC budget (estimated to be 30 %) is disproportional as compared to their terrestrial area fraction of only 7 % (Guenther et al., 1995).In contrast to the tropics, Mediterranean plant species have already been intensively studied and are known to emit a great variety of VOC (Owen et al., 1997;Kesselmeier and Staudt, 1999;Owen et al., 2001Owen et al., , 2002;;Llusia et al., 2002;Simon et al., 2006).
In general, most of these studies concentrated on the emission of isoprene and monoterpenes.Less information is available on the emission of short-chain oxygenated compounds (oxVOCs) such as formaldehyde, acetaldehyde, acetone, methanol, ethanol and formic and acetic acids (Seco et al., 2007).Emissions of these oxygenated compounds were identified only recently as being a large source of carbon (150-500 Tg C yr −1 ) to the atmosphere (Singh et al., 2001).Other compounds of high interest are sesquiterpenes (Jardine et al., 2011), but they are not easily measured due to their high reactivity (Fuentes et al., 2000).Sesquiterpenes and other highly reactive compounds may constitute a considerable amount of the unmeasured VOC (Goldstein et al., 2004) as evidenced by indirect approaches (Di Carlo et al., 2004;Kuhn et al., 2007).
The present study aims to contribute towards a more complete assessment of VOC released at leaf level, contrasting flux studies which may miss the more reactive VOC species.We chose the Amazonian vegetation because it is a poorly investigated region and compared it with the much better described Mediterranean ecosystem which exhibits a very heterogeneous emission composition.This study was performed as a doctoral thesis (Bracho-Nunez, 2010) which can be consulted at the given URL for more details.To enable a comparison of the VOC emissions between tropical and Mediterranean vegetation over a broad range of VOC species, several different methods for characterization and quantification of VOC were used: a proton-transfer-reaction mass spectrometer (PTR-MS) for the detection of all VOC with proton affinity higher than water; an online gas-chromatograph with a flame ionization detector (GC-FID) for the determination of isoprene emissions; and an offline GC-FID along with an offline gas-chromatograph coupled to a mass spectrometer (GC-MS) for the determination of higher VOC, including sesquiterpenes.

Tropical vegetation
VOC emission screening was carried out with a total of twelve different tree species common for terra firme and floodplain areas in the Amazonas region (Table 1).Three representative tree species were derived from the upland forest (terra firme) ecosystem, eight from the white water floodplain region (várzea) which is regularly inundated with nutrient-rich and sediment-loaded "white" water and three from the (igapó) floodplain region regularly inundated with nutrient-poor "black" water rich in humic matter, or clear-water with an intermediate amount of nutrients (Prance, 1979;Sioli, 1954Sioli, , 1956)).Two of the plant species, Vatairea guianensis and Hevea spruceana, are found in the igapó and várzea as well.
The tropical screening experiment was a cooperative effort between the Max Planck Institute for Chemistry and the Instituto Nacional de Pesquisas da Amazônia (INPA, Manaus), and was performed at the INPA Campus forest in Manaus, Brazil.One-or two-year-old saplings of Garcinia macrophylla, Hevea spruceana and Vatairea guianensis from igapó, Hura crepitans, Pouteria glomerata, Pseudobombax munguba, Ocotea cymbarum, Pachira insignis, Zygia jurana, Hevea spruceana and Vatairea guianensis from várzea, and Scleronema micranthum, Hevea brasiliensis and Hevea guianensis from terra firme were measured in 2006 and 2007.The terra firme species Hevea brasiliensis, Hevea guianensis and Scleronema micranthum were collected at the Reserva Ducke, which is situated 40 km from Manaus.90 % of the Reserve's area is covered by primary vegetation characteristic of the Central Amazon terra firme (Gomes and Mello-Silva, 2006).The várzea species Hevea spruceana, Hura crepitans, Ocotea cymbarum, Pachira insignis, Pouteria glomerata, Pseudobombax munguba, Vatairea guianensis and Zygia juruana were collected at the bank of the Ilha da Marchantaria (03 • 15 S, 59 • 58 W), an island located in the Solimões River.The igapó species Garcinia macrophylla, Hevea spruceana and Vatairea guianensis were collected at the bank of the Tarumã Mirim (03 • 08 S, 60 • 01 W) an affluent of the Rio Negro.Effects of different floodplain environments on trace gas emission quality and quantity were investigated comparing plant individuals growing in várzea and igapó as well, such as Hevea spruceana and Vatairea guianensis.Collected plants were potted in soil obtained from the sites of the plant's origin and allowed to adapt to the new conditions for at least one month before making measurements.Plants were kept under sunlit conditions, protected by mosquito nets, and were irrigated daily.Daily mean temperature conditions of 29 • C ± 1.6 • C and 27.8 • C ± 1.3 • C were recorded during the dry and wet season, respectively.A total of three replicates for each species were measured with some exceptions (Table 7).Plant incubation in dynamic flowthrough-enclosures started at 8.00 p.m. (local time) on the evening before measurements were performed the next day to allow tree individuals to adapt to the enclosure system.

Mediterranean vegetation
All plants were collected in March 2008 from the surroundings of Montpellier, France, and were potted and maintained in a greenhouse of the institute at an approximate day/night temperature of 25/15 • C. All together a total of sixteen plant species (all 2-3 yr old) including deciduous and nondeciduous trees, shrubs, grasses and palm trees typical for the Mediterranean area were studied at the CEFE-CNRS in Montpellier (France) during the months of April to July 2008 (Table 2).Again, for each species there were three replicates.

Tropical vegetation
For the measurement of tropical plants an enclosure system developed at the Max Planck Institute for Chemistry in Mainz, Germany was used.This enclosure system has been described in detail elsewhere (Schäfer et al., 1992;Kesselmeier et al., 1996;Kuhn et al., 2002a, b).Two identical branch cuvettes made of fully light-permeable FEP Teflon foil (Norton, 50 µm thickness, Saint-Gobain Performance Plastics, Germany) were flushed with ozone-free ambient air.One cuvette was used as the reference "empty" cuvette and a second cuvette enclosed a branch or the complete plant above ground.Ambient air was scrubbed of small particles using Teflon filters (Zefluor Teflon filters, 2 µm pore size, Gelman Science, USA) and of ozone with an ozone scrubber composed by ten copper nets coated with MnO 2 (Ansyco, Germany) placed in a Teflon tube to prevent oxidant inter-  ferences inside the enclosure.A combination of three Teflon membrane pumps (Vacuubrand, Germany) was used to pump the filtered ambient air to both cuvettes.The air flow to each cuvette was monitored by an in-line flow meter (EL-Flow, 50 L min −1 , Bronkhorst Hi-Tec, Germany).Two different cuvette sizes with volumes of 9 L and 100 L were used depending on the size of the plant to be enclosed.The cuvette flow was controlled by a needle valve and adjusted to 10 L min −1 for the small cuvette and to 20-40 L min −1 for the larger cuvette.
Mixing of air within the cuvette was achieved with a Teflon-coated fan.Cuvette temperature and relative humidity was measured in bypassed air with a commercial sensor (Model Rotronics YA-100F, Walz, Germany).Exchange of VOC, CO 2, and H 2 O was monitored by direct sampling from the sample and reference cuvette.All sample tubing was made of Teflon and maintained at a constant temperature of 45 • C, which was always higher than the cuvette or ambient temperature in order to avoid condensation in the lines.
CO 2 and H 2 O exchange were measured with a CO 2 / H 2 O infrared gas analyzer (LI-COR Inc. 7000, Lincoln, Nebraska, USA).The analyzer was operated in differential mode receiving an analog signal of the absolute concentration measured by a second CO 2 / H 2 O infrared gas analyzer (LI-COR Inc. 7000, Lincoln, Nebraska, USA) as reference.Pressurized nitrogen gas (N 2 5.0, Messer Griesheim, Germany) was used as the reference gas for the CO 2 and H 2 O zero-point calibration of the infrared gas analyzers.The gas flow to the instrument was supplied by a custom-made pump unit (membrane pump, 12 Volt; KNF-NEUBERGER, Freiburg, Germany) and adjusted to 0.5 L min −1 by a rotameter (Omega, USA).Calibration of the analyzer was accomplished prior to the experiments by use of a calibration gas standard for the calibration of CO 2 (512 ± 2 ppm CO 2 in synthetic air, LI-COR, Lincoln, Nebraska, USA) and a dew point generator for the calibration of water channel (Li 610; LI-COR, Table 3. Leaf temperature (T leaf ) and photosynthetic active radiation (PAR) of tropical plants measured under semi-controlled conditions.Enclosure temperatures were not significantly different as compared to leaf temperatures with 1 • C lower mean temperatures at most.Averages ± standard deviation of measurements with n tree individuals.Each replicate represents an average of data points measured every minute during one day measurement by a PAR < 300 µmol m −2 s −1 (this PAR value was determined with light curves (data not shown) and represents the photosynthetic saturation point of all tropical plants).Lincoln, Nebraska, USA).At the end of each experiment the calibration of the analyzer was checked and the signal response was corrected for sensitivity and zero drifts as a function of time and temperature.Assimilation, transpiration and stomatal conductance were calculated according to Pearcy et al. (1989).All measurements of tropical vegetation were performed under semi-controlled conditions following ambient relative humidity, temperature and CO 2 .Relative humidity ranged between 77 ± 9 % and 91 ± 7 % during the dry and wet season and ambient CO 2 concentrations were found in the range of 340-404 ppm.Leaf temperatures varied between 31 to 36 • C (Table 3).Garcinia macrophylla, Hevea spruceana (igapó), Hura crepitans, and Pseudobombax munguba were measured under ambient light conditions (Table 3).For remaining plant species additional photosynthetic active radiation (PAR) was constantly provided by an LED system consisting of four double chains of a mixture of red, blue and white light constructed and designed by the electronics department of the MPI for Chemistry in Mainz, Germany.This LED system was placed perpendicular to the cuvette and supported a constant light regime around 500 µmol m −2 s −1 (Table 3).Gaps between the LED groups were closed with reflecting film in order to obtain a homogenous distribution of the light in the cuvette.The irradiation was monitored with a quantum sensor (Model SB 190, LiCor, USA) inside the cu-vette at different heights before and after the measurements and found to be homogeneous for the area of the whole enclosed plant.The irradiation value determined in the center of the cuvette was taken as the incident PAR.In the course of all other gas exchange measurements the same sensor was installed next to the chamber system.Light intensities and the cuvette and leaf temperatures measured with thermocouples of type E (Chrom-Constantan, OMEGA) were recorded with a data logger CR23X (Campbell Scientific Ltd., Shepsherd, UK).All other micrometeorological and physiological parameters were monitored and recorded with a in-house control unit (V25, Max-Planck Institute for Chemistry, Mainz, Germany).

Mediterranean vegetation
Gas exchange of Mediterranean vegetation was measured using a dynamic temperature-and light-controlled chamber system (Staudt et al., 2004;Bracho-Nunez et al., 2011).This custom-made gas exchange chamber with a volume of approximately 105 mL was constantly flushed with air at a flow of 650 mL min −1 .The air was purified and dried by a clean air generator (AIRMOPURE, Chromatotec, France) and then re-humidified by passing a variable portion of the air stream through a water bubbler.All tubing was made of Teflon and sample lines were maintained at a constant temperature of 45 • C. Chamber and plants were illuminated with white light (OSRAM 1000 W) filtered by a 5 cm water bath.PAR was measured with a quantum sensor (LiCor, PAR-SB 190, Lincoln, NE, USA) located next to the chamber system.Leaf and chamber air temperatures were monitored with two thermocouples of type E (Chrom-Constantan, OMEGA).For all measurements of Mediterranean vegetation, temperature and light was kept constant at standard conditions (leaf temperature = 30 ± 1 • C and PAR = 1055 ± 36 µmol m −2 s −1 ).Relative humidity was also maintained constant at 48 ± 13 %.All data were stored on a 21X data logger (Campbell Scientific Ltd., Shepsherd, UK).Photosynthesis and transpiration were measured by directing a constant portion of the inlet and outlet air through a CO 2 /H 2 O infrared gas analyzer (LI-COR Inc. 7000, Lincoln, Nebraska, USA).Ambient CO 2 concentrations varied between 318 and 380 ppm during the experiments.
For the measurements, one or several terminal leaves of an individual plant were placed perpendicular to the light to ensure homogenous light distribution on the adaxial surface of the leaves.For most of the plants, it was necessary to remove some leaves to enable proper placement in the chamber; this was done at least one week before any measurements to minimize disturbance effects.In order to ensure adaptation of the plants to the chamber environment, all species were placed in the chamber at least 1 h before the measurements, or until the VOC signals measured with PTR-MS were found constant.Leaves of the conifer Pinus halepensis and the aromatic shrub Rosmarinus officinalis possess glands and ducts  (Niinemets et al., 2011).Therefore, as based on our experiences during previous work, the leaves of these species were enclosed at least 12 h before making measurements.

GC techniques
VOC emissions from tropical vegetation were measured using two different methods, (i) by collection on cartridges using an automatic sampler, and subsequent analysis with an offline gas chromatographic (GC) method equipped with a flame ionization detector (FID; Kesselmeier et al., 2002;Kuhn et al., 2002b) in the MPIC laboratory in Mainz, Germany; and (ii) by online PTR-MS as described below.In the case of Mediterranean vegetation three GC-instruments were used: an AirmoVOC C2-C6 online GC-FID (Chromatotec, France) for the detection of isoprene and apolar light VOCs, a Chrompack CP9003 offline GC-FID equipped with a Chrompack TCT4002 thermo-desorber (all Varian Inc.) for the detection of higher VOCs such as monoterpenes and sesquiterpenes, and a Varian CP3800/Saturn2000 GC-MS equipped with a Perkin-Elmer Turbomatrix thermo-desorber to confirm PTR-MS signals by related GC peaks.A detailed description of the GC systems is reported by Bracho-Nunez et al. (2011).
In the case of tropical vegetation screening, the PTR-MS probing switched between the reference and the sam-ple chamber, taking into account ten cycles of the reference and sample chamber, each.The instrument background signals were determined with air samples after passing over a heated platinum catalyst maintained at 350 • C (Parker Co., USA).The background signal was subtracted from the reference and sample chamber signals.This procedure was repeated from 8.00 p.m. to 6.00 p.m. of the following day.For VOC emission rate calculations the reference chamber signals (with background subtracted) were subtracted from the sample chamber signals (with background subtracted).Mediterranean vegetation screening by PTR-MS measurements were performed similarly as described above with a slightly changed sampling protocol.Ten cycles of enclosure measurements switching between sample and empty reference cuvette were followed by three cycles of instrument background measurements.This process was repeated 3 to 10 times.
The PTR-MS instrument was calibrated using a VOC standard mixture in nitrogen (Deuste Steininger GmbH, Germany) containing some target VOCs (isoprene, αpinene, methanol and acetone) at concentrations of 300 nmol mol −1 ± 10 %.For calibration, the gas was further diluted with synthetic air to final concentrations of 0.5-10 nmol mol −1 .For the quantification of sesquiterpenes a calculation of simple ion-molecule reaction kinetics was used as described elsewhere (Hansel et al., 1995;Wisthaler et al., 2001).Since the quantification of monoterpenes and sesquiterpenes with the PTR-MS is usually very difficult due to their tendency to fragment into different compounds depending on the VOC species (Tani et al., 2003;Demarcke et al., 2009), these data were always complemented by GC-FID or GC-MS (Bracho-Nunez et al., 2011).The detection limit of the PTR-MS was estimated as being 1.96 times the standard deviation of the empty chamber concentrations (at the 95 % confidence level) and was typically 1.6 nmol mol −1 for methanol, 579 pmol mol −1 for acetone, 493 pmol mol −1 for isoprene, 201 pmol mol −1 for monoterpenes, and 94 pmol mol −1 for sesquiterpenes, and the detection limits of the GCs ranged between 0.1 and 0.4 ng L −1 .

Leaf area and dry weight determination
Projected leaf areas were determined either with an optical area meter (Delta-T Devices Ltd., Cambridge, UK) or by copying the shape of the measured leaves on a sheet that was subsequently scanned using the software Size (Version 1.10R, Müller-Software) (Kesselmeier et al., 1996).To obtain leaf dry weights, leaves were dried at 60 • C for at least 48 h before weighing.Errors were estimated at 0.2 % and 2 % for dry weight and leaf area determination, respectively.

Data treatments
The emission rate E s [µg g −1 h −1 ] of each compound was calculated according to Eq. ( 1) based on the measured concentration difference ( c = c sample chamber − −c empty or reference chamber ) in [nmol L −1 ], the chamber flush rate Q in [L h −1 ], the leaf dry weight (dw) in [g] and the molecular mass M in [g mol −1 ].
For VOC emissions measured under nonstandard light and temperature conditions (the conditions for most of the measurements on tropical species, Table 3), the phenomenological algorithm described by Guenther et al. (1993) was used to standardize the actual VOC emission rates to standard emission rates, often referred to as the basal emission rate.This algorithm describes the light and temperature dependencies for VOC originating from DMAPP (Dimethylallyl pyrophosphate), such as isoprene and monoterpenes and distinguishes between emissions from VOC storage pools and newly synthesized VOC.Since acetone and methanol also showed light and temperature dependency during this study, emission rates of these compounds could also be described with this algorithm.Methanol emission depends on stomatal conductance and accumulation occurs in the stomatal cavities during night.Thus, in the morning a methanol peak was detected on some occasions, due to the burst of the nocturnally accumulated methanol after stomatal opening.Though this burst can reach higher emission rate values than observed during the rest of the day, it does not dominate the daily emissions as it lasts only a short time.Nevertheless, it could affect the determination of E s derived from linear regression analysis according to the corresponding algorithm (Guenther et al., 1993).Therefore, this morning peak was excluded for the calculation of E s .Our strategy implies that the VOC emitted from the tropical plants were de-novo-synthesized without substantial storage after production.This assumption could be confirmed by the absence of VOC emissions in the dark even at elevated temperatures.

Statistics
One-way Anova tests were carried out in order to detect possible differences among two or more independent groups.Mean VOC emissions (methanol, acetone, isoprene or monoterpenes separately) from tropical plant species were tested against those emitted from Mediterranean plant species and the same species from different ecosystems (igapó vs. várzea).The statistically significant difference effect in ANOVA was proven by the Tukey's test.

Screening Tropical vegetation
In ten out of twelve plant species screened, we were able to observe VOC emissions (Table 4).Considering the total amount of VOC emissions of the ten emitters, eight species were identified as high emitters (< 10 µg g −1 h −1 ) and two species, Scleronema micranthum and Hevea guianensis, as low to moderate emitters (1-10 µg g −1 h −1 ).It is important to note that saplings were investigated.Mature, naturally growing plants may exhibit other emission qualities and quantities.However, their investigation in the field is difficult.Therefore, our screening results may be regarded as a first step adding to our current knowledge.
All isoprene-emitting plant species were high emitters (< 10 µg g −1 h −1 ) (Table 4).Four plant species were found to emit monoterpenes (Table 4).Only slight differences were found comparing one and the same species derived from várzea and igapó, i.e.Hevea spruceana and Vatairea guianensis.The relatively conservative composition of the monoterpenes emitted by all monoterpene-emitting tropical plant species studied was similar with α-pinene as the most abundant monoterpene followed by limonene, sabinene or βpinene and other monoterpenes, except for Hevea guianensis that emitted only α-pinene and camphene (Table 5).It is interesting to see all Hevea species investigated so far being identified as monoterpene emitters in close agreement with earlier reports about Hevea brasiliensis (Klinger et al., 2002;Baker et al., 2005).
In addition to isoprenoid emissions a release of oxygenated VOCs, methanol and acetone, was found with several of the investigated plant species.Methanol emissions were detected in the case of seven species with emission rates varying between 1.5 to 20.2 µg g −1 h −1 (Table 4).Low emissions of acetone could be detected in the case of Hevea spruceana originating from both ecosystems (< 1.5 µg g −1 h −1 ).
A noticeable release of mass 73 was detected from Garcina macrophylla during daytime conditions (light).Mass 73 may be regarded as the protonated methyl ethyl ketone (MEK), an oxidation product of isoprene.This interpretation would be in accordance with the high emissions of isoprene found for this plant.Furthermore, production within plants can no longer be excluded (Jardine et al., 2013).But other compounds, such as methylglyoxal (Holzinger et al., 2007) or 2-methyl propanal (Jardine et al., 2010) could also be made responsible for this mass.

Screening Mediterranean vegetation
As in the case of tropical plants, the Mediterranean plants investigated were also saplings showing emission rates that may differ as compared to mature, naturally growing individuals.All sixteen Mediterranean plant species investigated were found to emit VOC, nine of which being high emitters (11.0-69.1 µg g −1 h −1 ), whereas the rest of the plants were moderate emitters, showing VOC emissions in the range of 2.6 and 9.3 µg g −1 h −1 (Table 4).
< 10 µg g −1 h −1 ) and the other eight as low emitters (E s < 1 µg g −1 h −1 ) (Table 4).Fourteen out of the sixteen investigated Mediterranean plants emitted monoterpenes in significant amounts (Table 4).Nine of them were classified as moderate to high monoterpene emitters with emissions ranging between 1 and 36 µg g −1 h −1 .Furthermore, PTR-MS measurements also indicated low monoterpene emissions for Chamaerops humilis, Cistus albidus, Coronilla valentina and Prunus persica (< 1 µg g −1 h −1 ).However this could not be confirmed by GC analysis.This discrepancy might have been caused by decomposition of more labile monoterpene species, an artifact of cartridge sampling for GC analysis, or by the emission of many different monoterpene species at rates below the detection limit for GC analysis.Alterna-tively, the masses classified as monoterpene (m81 and m137) by PTR-MS might be related to other VOC species.For example, interferences of sesquiterpene fragments like (−)Ecaryophyllene (m81) and α-humulene (m137) have been reported to potentially contribute 7.5 and 6 % for each mass (Demarcke et al., 2009).The monoterpene composition of the emissions from Mediterranean vegetation was highly species specific and very variable among the plant species, as summarized in Table 5.It should be noted that Mediterranean plants showed up to 14 different monoterpene species.Furthermore, the variability of monoterpene species among the different plant species was very high as compared to tropical vegetation.
Not only isoprenoids were emitted from Mediterranean plants.Oxygenated VOC such as methanol and acetone were also observed.All investigated plant species except Chamaerops humilis emitted methanol (Table 4).The range of emission rates was very large (between 1 and 14 µg g −1 h −1 ).For some species low acetone emissions were detected (Table 4).Most of these emission rates were < 1 µg g −1 h −1 , except for Brachypodium retusum and Ficus carica.

Comparison of VOC emissions from tropical and Mediterranean plant species
Most of the investigated tropical and Mediterranean plants exhibited substantial VOC emissions.Of striking difference was the higher number of monoterpene emitters in the case of the Mediterranean vegetation.Only four out of eight tropical plant species were emitting monoterpenes.In contrast, monoterpenes dominated emissions in 14 out  of 16 Mediterranean species.Furthermore, the monoterpene species composition emitted from the investigated tropical plants was quite conservative, in contrast to the great variation of the pattern observed among Mediterranean plants.Sesquiterpenes were detected only in emissions from Mediterranean vegetation.However, although our experiments did not detect any sesquiterpene emissions from tropical vegetation, their existence in Amazonian forest air has been recently demonstrated (Jardine et al., 2011).Emissions of methanol were found to be widely distributed in plants of both ecosystems.This can be understood as a consequence of growth (Fall, 2003).In addition, acetone emissions were found only in one tropical plant species, whereas five of the investigated Mediterranean species emitted this VOC species.

Plant physiology
The net exchange of CO 2 was monitored along all measurements to note the physiological activities of the enclosed plants.The data confirmed that all plants investigated were photosynthetically active (Table 7).Net photosynthetic rates (NPR) were found in the range of 1.8-8.2µmol m −2 s −1 and 3.3-23 µmol m −2 s −1 for Amazonian and Mediterranean plants, respectively.Such rates have been reported for shade leaves of tropical C3-plants (Da Matta et al., 2001;Larcher, 2003) and for Mediterranean plants in spring or autumn (Llusia and Peñuelas, 2000).Mostly, the transpiration rates were < 6 mmol m −2 s −1 except for Cistus monspeliensis, Rosmarinus officinalis, Hevea spruceana (igapó) and Pouteria glomerata, showing higher values (Table 7) and consequently higher stomatal conductances.The other species exhibited transpiration rates and stomatal conductances in the range as reported before for tropical and temperate plant species (Parolin et al., 2001;Bazzaz, 1980).Monitoring the net CO 2 exchange data did not only enable a review of the physiological activities but also provided a data set to be compared with the carbon lost as VOC.Relating the mean values of photosynthetically driven net CO 2 uptake to the mean values of carbon emitted as VOC species identified so far, we estimated a carbon loss between 0-4 % of the net amount of carbon assimilated by the plants (Table 7).

General remarks
Using phylogenetic reconstructions to explore the trait of isoprene emission, large gaps in our knowledge have been recently demonstrated (Monson et al., 2013;Sharkey, 2013), indicating the need to screen more plants even in the case of the most intensively investigated isoprene emission.The issue becomes much more complex by taking into account other VOC species.Furthermore, the mixture of emitted VOC species is highly dependent on multiple stress factors (Holopainen and Gershenzon, 2010) and is triggered by intra-and interspecific plant competition, economizing the loss of carbon in view of defense strategies (Kigathi et al., 2013).Therefore, a robust assessment of regional and global VOC emissions is needed in order to understand their effects on atmospheric chemistry/physics and the carbon budget.For this purpose, we need more information at the plant species level to deal with processes and to contribute to the understanding on the ecosystem level, especially as flux studies covering larger terrains miss substantial amounts of emitted VOC because of their high reactivity.This study contributes to the description of plant-specific VOC emissions from plant species of two such special ecosystems, i.e. the Amazonian compared to the Mediterranean area.Investigation of Mediterranean plant species are numerous and demonstrate the domination of monoterpene-emitting species.In contrast, screening of tropical plant species for the identification of VOC on the species level is still rare, though compounds such as isoprene, monoterpenes, methanol and acetone have previously been detected in high concentrations above Amazonian forests (Karl et al., 2004;Eerdekens et al., 2009).For our screening study we chose plant species at random and the small numbers cannot lead to a final view, but the results do indicate a trend and will improve data bases.

Tropical plants
Investigations of Amazonian plant species are fragmentary in terms of areas, plant species and analytical techniques (Kesselmeier et al., 2009).Several screening projects of VOC emissions from tropical African (Harley et al., 2003), Asian (Klinger et al., 2002;Geron et al., 2006b;Oku et al., 2008;Llusia et al., 2013), or South American including Amazonia (Lerdau and Keller, 1997;Lerdau andThroop, 1999, 2000;Harley et al., 2004;Pegoraro et al., 2006) have been performed reporting on isoprene emissions.A few plant species were identified as monoterpene emitters, among them Hevea brasiliensis (Klinger et al., 2002;Baker et al. 2005;Geron et al., 2006b;Wang et al., 2007).Additionally, Wilske et al. (2007) identified six out of eight SE Asia tropical tree species as monoterpene emitters, though at low rates.For the Amazon region one more species (Apeiba tibourbou) was identified by Kuhn et al. (2002bKuhn et al. ( , 2004) ) as a light-dependent monoterpene emitter.Within our investigation, the most substantial isoprene emissions of Amazonian plants were found for Vatairea guianensis from várzea and igapó, reaching values similar to those measured for several Mediterranean Quercus species (Kesselmeier and Staudt, 1999).Other isopreneemitting species emissions were found to be consistent with emission rates reported for tropical tree species (Harley et al., 2004;Padhy and Varshney, 2005).Monoterpene emission was found to dominate in the group of Hevea species.The rubber tree Hevea brasiliensis showed a normalized monoterpene E s of 10.9 to 38.1 µg g −1 s −1 , even exceeding the high emissions reported earlier (Klinger et al., 2002;Baker et al., 2005;Geron et al., 2006b;Wang et al., 2007).In addition to Hevea brasiliensis, we also identified Hevea spruceana and Hevea guianensis as strong monoterpeneemitters with some differences related to the origin of the species (igapó or várzea).All monoterpene-emitting tropical trees of this work emitted α-pinene followed by limonene and sabinene, contrasting the emission pattern reported previously for Hevea brasiliensis (Wang et al., 2007).Further investigation is needed for a better understanding of these differences.Nevertheless, it is of special interest to note that all Hevea species investigated so far belong to the group of monoterpene emitters.In addition to the Hevea species, the terra firme species Scleronema micranthum was found to be a low monoterpene-emitting species (< 1 µg g −1 s −1 ) with an emission pattern similar to Hevea brasiliensis and Hevea spruceana (Table 5).No bibliography data were found about the VOC emission pattern of this important tree of the Central Amazonian forest.

Mediterranean plants
In contrast to the set of Amazonian plants, Mediterranean plant species were found to be dominated by monoterpene emitters.These findings are in close accordance with earlier investigations of VOC emissions from Mediterranean ecosystems (Owen et al., 2001).Nevertheless, we identified several species such as Ficus carica, Spartium junceum, Chamaerops humilis and Brachypodium retusum to emit substantial amounts of isoprene in contrast to earlier studies (Pio et al., 1993;Benjamin et al., 1996;Owen et al., 2001).
www.biogeosciences.net/10/5855/2013/Biogeosciences, 10, 5855-5873, 2013 These differences could be due to the different origins of the plants within different Mediterranean areas, suggesting genetic differences.However, uncertain species identification, missing normalization of the data, and other technical aspects in previous studies cannot be ignored.In some cases an emission was not reported but assigned based on genus average as in the case of the isoprene emission from Ficus carica and Chamaerops humilis reported (Benjamin et al., 1996) or for some tropical species (Harley et al., 2004).Observed monoterpene emission rates were quantitatively comparable to literature data in a few cases only, e.g. for Ceratonia siliqua, Pinus halepensis, Quercus coccifera and Rosmarinus officinalis (Llusia and Peñuelas, 2000;Owen et al., 2001;Ormeno et al., 2007aOrmeno et al., , b, c, 2009;;Staudt and Lhoutellier, 2011).Quercus suber has been identified previously as a non-isoprenoid-emitter tree (Seufert et al., 1997;Steinbrecher et al., 1997).But we estimated an emission factor of 35.6 ± 5.5 µg g −1 h −1 , confirming more recent literature showing typical summertime values of 10-43 µg g −1 h −1 (Staudt et al., 2004;Pio et al., 2005;Staudt et al., 2008).
A further oak species, Quercus afares, was found to emit high quantities of monoterpenes and low amounts of isoprene (see also Welter et al., 2012).Furthermore, we classified Ficus carica and Spartium junceum as non-monoterpene emitters contrasting earlier reports (Benjamin and Winer, 1998;Pio et al., 1993).Monoterpene emissions from Chamaerops humilis, Coronilla valentina and Buxus sempervirens were reported for the first time, though at low rates.Monoterpene emission rates found for Olea europea here were higher (< 1 µg g −1 h −1 ) than previously reported (Arey et al., 1991;Benjamin and Winer, 1998;Llusia et al., 2002;Owen et al., 2001;Pio et al., 1993;Winer et al., 1983Winer et al., , 1992)).Such variability of monoterpene emissions may be understood to result from a variety of influences, such as different sampling and analytical methods, plant origin, plant developmental stages and environmental conditions, biotic and abiotic stresses or even questionable plant identification, making it difficult to use such data for emission models (Niinemets et al., 2011).

Monoterpene species heterogeneity
Monoterpene species emission patterns as found for the tropical plants species were rather homogeneous, contrasting the high variability of monoterpene species in the case of the Mediterranean plants.The species pattern found here for Ceratonia siliqua (myrcene, Z-ocimene, E-ocimene) was completely different from that reported by Llusia et al. (1998), who reported an emission of α-phellandrene only, as well as by Owen et al. (2001) or by Llusia et al. (2002), who identified α-pinene and limonene or only α-pinene, respectively.Our results with Cistus monspeliensis, i.e. emissions of α-pinene, camphene and ß-pinene, coincided with Owen et al. (2002).Furthermore, we could additionally identify myrcene and E-ocimene in our study.For Olea europaea we found Z-ocimene and E-ocimene, the latter com-pound also reported by Arey et al. (1991).But our results completely disagreed with the emission pattern reported in several other reports (Winer, 1983;Pio et al., 1993;Benjamin and Winer, 1998;Owen et al., 2001;Llusia et al., 2002).Pinus halepensis is one of the most intensively investigated conifers of the Mediterranean area with a large variety of emission patterns (Corchnoy et al., 1992;Benjamin and Winer, 1998;Llusia andPeñuelas, 1998, 2000;Owen et al., 2001Owen et al., 2001;;Simon et al., 2005;Blanch, 2007;Ormeno et al. 2007a, b, c, d).In our current study we detected Z-ocimene, E-ocimene, myrcene and at low concentration terpinen-4-ol, whereas α-pinene, ß-pinene, limonene, sabinene, and others were not found, a result which indicates the high emission variability of this plant species.In the case of Quercus suber, emissions of 15 different monoterpene species were found in our study, contrasting with other reports with only about 4 to 7 monoterpene species (Pio et al., 1993(Pio et al., , 2005;;Staudt et al., 2004;Staudt et al., 2008).Measurements of Rosmarinus officinalis confirmed the high variability of emissions as reported earlier (Hansen, et al., 1997;Seufert et al., 1997;Owen et al., 2002;Ormeno et al. 2007aOrmeno et al. , b, c, 2009;;Olivier et al., 2011).

Sesquiterpenes
Sesquiterpene emissions were detected only in the case of Mediterranean vegetation and not for tropical plants.Due to their high reactivity and relatively low vapor pressure and emission rates, these compounds are easily missed or underestimated (Duhl et al., 2008).Special analytical adaptations can facilitate their quantification (Merfort, 2002;Helmig et al., 2004;Tholl et al., 2006).Bracho-Nunez et al. (2011) reported about a higher sensitivity by offline gas chromatographic methods as compared to online PTR-MS methods, the latter also being less specific.Other studies have confirmed the difficulty of PTR-MS for the measurements of sesquiterpenes, reporting high fragmentation patterns (Tani et al., 2003;Demarcke et al., 2009).The characterization of sesquiterpene emissions in the case of the Mediterranean area is sparse and only a few oaks, birches and pines typical for this region are reported to emit sesquiterpenes (Llusia and Peñuelas, 1998;Ciccioli et al., 1999;Hansen and Seufert, 1999;Ormeno et al., 2007b, d;Staudt and Lhoutellier, 2007;Staudt et al.;2008).In the course of our work on Mediterranean vegetation we found a high variability for these compounds in close accordance with other reports (Duhl et al., 2008;Llusia and Peñuelas, 1998;Ormeno et al., 2007aOrmeno et al., , b, c, d, 2009)).Sesquiterpene emissions from Buxus sempervirens, Ceratonia siliqua and Olea europaea could be reported for the first time here.Sesquiterpene emissions from Quercus coccifera were quantitatively but not qualitatively comparable.We found only (−)E-caryophyllene emissions, whereas a variety of sesquiterpene species were reported in other studies (Ormeno et al., 2007a(Ormeno et al., , c, d, 2009;;Staudt and Lhoutellier, 2011).It is important to note that for Cistus albidus most of Biogeosciences, 10, 5855-5873, 2013 the sesquiterpene species identified here were partially in accordance with other studies performed in the Mediterranean region (Llusia and Peñuelas, 1998;Ormeno et al., 2007a, b, c, d) except for emissions of α-and ß-cubebene.Sesquiterpene emissions of Cistus albidus as low as 0.63 ± 0.32 µg g −1 h −1 were in concordance with Ormeno et al. (2007a, d) but contrasted significantly higher emissions found in other studies (Llusia and Peñuelas, 1998;Ormeno et al., 2007b, c).We did not find any sesquiterpene emissions from Pinus halepensis and Rosmarinus officinalis, in contrast to observations reported by Ormeno et al. (2007aOrmeno et al. ( -d, 2009)).The emission of (−)E−caryophyllene by Cistus monspeliensis disagrees with the findings by Llusia and Peñuelas (1998).Such variation of sesquiterpene emissions may be due to differences in analysis protocols and sampling and plant enclosure techniques.But other factors such as seasons, measuring conditions and overall stress effects may also contribute.Sesquiterpene emissions are known to be induced by a variety of stresses, including heat and oxidative stress (Loreto and Schnitzler, 2010;Holopainen and Gershenzon, 2010).
For example in a recent study on Quercus coccifera, Staudt and Lhoutellier (2011) observed that sesquiterpene emissions became boosted during exposure to heat and high radiation, while at moderate temperatures sesquiterpene emissions were low or even undetectable in this species.In line with this, field studies by Ormeno et al. (2007dOrmeno et al. ( , 2009) ) observed the highest standard emissions of Rosmarinus officinalis in June, whereas lower emission rates were found in March or January, suggesting a seasonal trend possibly associated with heat and oxidative stress occurring under Mediterranean summer conditions.Similar trends were reported for Cistus albidus and Pinus halepensis.It has to be noted that the variability of sesquiterpene emissions is even more striking than that for monoterpenes.We have to take into account, however, potential interactions of oxidative stress and herbivory (Holopainen and Gershenzon, 2010), a complex interaction which will alter VOC emission qualities and quantities.

Oxygenated VOC
Despite their importance in plant physiology, ecology, and also in air chemistry (Fall, 2003), our information on oxygenated VOC compounds emission from vegetation is sparse, and their contribution to the global VOC budget has been poorly described (Eerdekens et al., 2009).Regular environmental triggers such as inundation of large areas in the Amazon region may have substantial impact on oxygenated VOC species (ethanol, acetaldehyde) emitted (Bracho-Nunez et al., 2012).The technological improvement of using a PTR-MS has facilitated the study of such short-chain oxygenated VOC.The exchange of the oxygenated VOC like methanol and acetone has been detected for a variety of plant species (Isidorov et al., 1985;MacDonald and Fall, 1993a, b;Nemecek-Marshall et al., 1995;Kirstine et al., 1998;De Gouw et al., 1999;Holzinger et al., 2000;Hüve et al., 2007;Seco et al., 2007).Methanol is a product of the demethylation of pectin during cell wall formation and is produced by leaf growth, for example (Nemecek-Marshall et al., 1995;Galbally and Kirstine, 2002).The leaves we measured were mature but probably still in a growing period and a significant emission of methanol was found.However, the physiological background for methanol emission is still a matter of discussion (Folkers et al., 2008).In the case of acetone, several studies have reported leaf level acetone emissions from vegetation (MacDonald and Fall, 1993b;Holzinger et al., 2000;Janson and de Serves, 2001;Kreuzwieser et al., 2002;Villanueva-Fierro et al., 2004;Cojocariu et al., 2005;Geron et al., 2006a, b;Grabmer et al., 2006) as well as high concentrations of this compound in the troposphere above forested areas (Helmig et al., 1998;Pöschl et al., 2001;Geron et al., 2002;Karl et al., 2003;Müller et al., 2006).In our study, emissions of methanol and acetone were demonstrated for the chosen plant species for the first time, except for the Mediterranean tree Pinus halepensis, which has been reported to emit methanol and acetone (Filella et al., 2009).Though the tropical tree Hevea brasiliensis has been described as an acetone emitter by Geron et al. (2006b), we did not find acetone emissions during our studies.Interestingly, emissions of methanol by the Mediterranean species investigated in this study were significantly lower (p < 0.05) than those detected for our tropical vegetation species.We may understand that difference as being due to different developmental stages of the measured leaves, although in both experiments fully expanded mature leaves were measured.On the other hand, the light-dependent acetone emissions found in this study are supposed to be direct emissions from the leaves.Only one tropical plant species released acetone, whereas seven Mediterranean plants were found to be acetone emitters.The metabolic pathway of acetone has not yet been proven.It was hypothesized that acetone is produced in spruce needles by the decarboxylation of acetoacetate (MacDonald and Fall, 1993b), whereas a cyanohydrin-lyase catalyzed reaction was found in cyanogenic plant species (Fall, 2003), leading to acetone release.Hevea spruceana, Hevea guianensis and Hevea brasiliensis are known as cyanogenic plants containing the cyanogenic ß-glucoside linamarin (Lieberei, 1986), but acetone emissions were found only with Hevea spruceana.Furthermore, a cyanogenic pathway causing acetone emission was also proposed for Olea europaea, Coronilla glauca and Prunus persica by Bracho-Nunez et al. (2011).On the other hand, it is known that in the presence of NO x the atmospheric OH-oxidation of several monoterpenes, like αand ß-pinene, appears to be a potentially relevant source of acetone (Wisthaler et al., 2001).NO x data are not available but we cannot exclude a presence of this compound.Therefore, a contribution of this oxidation pathway to the measured acetone concentrations in the plant cuvette could not be excluded, particularly in the case of all tropical plant species and the Mediterranean species Cistus monspeliensis,

Unknown VOC species
Diurnal emissions of an unknown compound with mass 73 have been detected from the tropical plant species Garcinia macrophylla.This mass might be considered to be protonated methyl ethyl ketone (MEK), also known as 2-butanone.Not much is known about the biosynthetic pathways leading to the formation of MEK and its emission mechanisms, but it has been reported to be emitted by a variety of plants and grasses (Isidorov et al., 1985;Kirstine et al., 1998;De Gouw et al., 1999;Jardine et al., 2010).Furthermore, a contribution to this mass by the volatile compound 2-methyl propanol cannot be excluded (Baraldi et al., 1999;Jardine et al., 2010).
On the other hand, Holzinger et al. (2007) identified mass 73 as representing protonated methylglyoxal, a secondary oxidation product of isoprene (Lee et al., 2006).Since Garcinia macrophylla is a high isoprene emitter, formation of such an oxidation product could be plausible (Jardine et al., 2011).But then emissions of MVK and/or methacrolein (primary oxidation products of isoprene) would have been expected, which was not the case.The detection of an unknown compound demonstrates the need for further investigation in order to identify compounds that probably have been overseen until now due to technical limitations.This is of special importance, as we seem to have a gap of knowledge in understanding the contributions of unmeasured and possibly unknown primary biogenic VOC species, which may contribute significantly to total OH reactivity in the atmosphere (Nölscher et al., 2013).

Conclusions
This study provides a categorization of VOC emissions from common tropical and Mediterranean plant species to further the understanding on interactions between vegetation and the atmosphere.Screening should be continued in the future, since the characterization of VOC emissions from representative species will support our understanding of emission processes, regulation and development.In particular more attention should be given to sesquiterpenes and oxygenated compounds, whose emissions are highly variable and uncertain due perhaps to methodological limitations and their inherent association with stress and phenology (Bracho-Nunez et al., 2011, 2012).Such studies would contribute to our understanding of VOC, representing a substantial carbon loss for vegetation.The average loss rates below 4 %, as found in our study, are in close agreement with Kesselmeier et al. (2002) and indicate a normal unstressed plant behavior.The scientific community will not be able to screen all tree species.However, we should note that the role of biodiversity of a tropical forest may be overestimated.These communi-ties are sometimes even regarded as not qualitatively different from temperate forest with a few dominating tree species (Montero et al., 2012;Pitmann et al., 2001;Ter Steege et al., 2003, 2006;Wittmann et al., 2002Wittmann et al., , 2010Wittmann et al., , 2013)).Taking into account individuals with < 10 cm diameter at breast height (dbh) the number of tree species/ha can be estimated at 250 (terra firme), 172 (Várzea) and 79 (Igapó).In view of these numbers we screened more than 1 % of trees.This is definitely not enough but such estimates may reflect local features.Furthermore, without screening on a leaf area level we will not learn about the release of reactive chemical species and emission regulation processes.An enhanced understanding of VOC emission behavior from screening additional plant species will increase confidence in predictive models, although future investigations of possible sources of observed variability are warranted, including examining the role of biodiversity and plant competition, as well as comparing different measurement and analytical techniques.
Five of them could be classified as high emitters (E s

Table 1 .
Plant species, family, functional type, ecosystem and distribution of the 12 investigated tropical plant species.

Table 2 .
Plant species, family, functional type and occurrence of the 16 investigated Mediterranean plant species.

Bracho-Nunez et al.: Leaf level emissions of volatile organic compounds which
store VOC.It has been shown that mechanical stress can cause large bursts of VOC from these plants, leading to large overestimations of VOC emission rates normalized as standard emissions (E s ) at 30 • C and 1000 µmol m −2 s −1 (PAR) www.biogeosciences.net/10/5855/2013/Biogeosciences,10, 5855-5873, 2013A.

Table 4 .
VOC emitted a from Mediterranean and tropical plant species measured with PTR-MS (Except for sesquiterpene emissions from Buxus sempervirens, Ceratonia siliqua and Olea europaea that were only detected with GC-FID).All averages ± standard errors are calculated from < 100 samples (tropical) or 18-83 samples (Mediterranean) from n tree individuals.

Table 6 .
Sesquiterpene species emitted a from Mediterranean and tropical plant species measured with GC-FID.Averages ± standard errors are calculated from n number of replicates.All averages ± standard errors are calculated from < 100 samples (tropical) or 18-83 samples (Mediterranean) from n tree individuals.

Table 7 .
Specific leaf weight (SLW), net photosynthetic rate (NPR), transpiration (E), and stomatal conductance (g s ) for Mediterranean and tropical plant species.Carbon loss (VOC carbon in relation net CO 2 assimilation) is estimated from the mean values of NPS (this table) and the sum of VOC carbon derived from the species listed in Table4.Except for SLW, all averages ± standard errors are calculated from < 100 samples (tropical) or 5-125 samples (Mediterranean) from n tree individuals.