2.1 Site

The site chosen for this study was the Amazon Tall Tower Observatory, ATTO (Andreae et al., 2015). This site is located in central Amazonia (02^∘08.647^′ S, 58^∘59.992^′ W), 150 km north-east of the closest large city, Manaus, Brazil. Due to the prevailing north-easterly wind direction, the influence of the Manaus plume is negligible and the measurements at this site can be considered to reflect pristine tropical forest conditions affected by air masses that have passed over about 1000 km of undisturbed rainforest. The site is equipped with a 325 m tall tower as well as two smaller towers. This study was carried out on the INSTANT tower, an 80 m walk-up tower located 600 m from the tall tower in an easterly direction. Sampling was performed on this tower below the canopy top (mean canopy height 35 m) at two different heights (12 and 24 m). For a comprehensive site description, see Andreae et al. (2015).

2.2 Air sampling

Collection of ambient air samples on adsorbent tubes, for subsequent analysis by a gas chromatography–flame ionization detector (GC-FID), was conducted with two automated cartridge samplers, described in earlier studies (Kesselmeier et al., 2002; Kuhn et al., 2002, 2005), positioned at 12 and 24 m on the INSTANT tower. The samplers consist of two main units, a cartridge magazine that holds the adsorbent-filled tubes and the control unit timing the process and recording the data. This latter unit also houses the pumps (Type N86KT, KNF Neuberger, Freiburg, Germany), pressure gauges, mass flow controllers and power supply. The cartridge magazine is equipped with solenoid valves controlling the inlet and outlet of up to 20 individual sampling adsorbent tubes. The system is a constant-flow device, with one cartridge position per loop used as a bypass for purging the system. Due to the compact weatherproof housings and the low power consumption, we were able to position one sampler at 24 m and the other one at 12 m, attached to the INSTANT tower booms with commercially available 50 mm aluminium clamps. The adsorbent tubes used for VOC sampling were filled with 130 mg of Carbograph 1 (90 m² g⁻¹) followed by 130 mg of Carbograph 5 (560 m² g⁻¹) sorbents. The size of the Carbograph particles was in the range of 20–40 mesh. Carbographs 1 and 5 were provided by L.A.R.A s.r.l. (Rome, Italy) (Kesselmeier et al., 2002). The samples were collected from 17 to 20 October 2015. Samples were taken for 30 min every hour at a flow of 200 cm³ min⁻¹ (STP), leading to a collection of 6 L of air in each cartridge using the automatic sampler. Additional sampling was performed at 24 m with a GSA SG-10-2 personal sampler pump during the years 2012–2014. These earlier samples were collected in the same type of adsorbent tubes as for the automatic sampler and were filled at 167 cm³ min⁻¹ (STP) air flow for 20 min. These additional measurements took place on 19 and 28 November 2012; 1, 3 and 4 March 2013; 11–14 June 2013; 22, 25 and 26 September 2013 and on 17 and 21 August 2014.

2.3 Instruments used for chemical analysis

2.4 Multi-Layer Canopy Chemistry Exchange Model (MLC-CHEM)

To analyse the magnitude and temporal variability of the observed monoterpene concentrations inside and above the forest canopy, we applied the Multi-Layer Canopy Chemistry Exchange Model (MLC-CHEM), driven by the observed micro-meteorology and ozone surface layer mixing ratios. The MLC-CHEM was originally developed and implemented in a single-column model. It is set up also in a global chemistry- and climate-modelling system to assess the role of canopy processes in local- to global-scale atmosphere–biosphere exchange of nitrogen oxides (Ganzeveld et al., 2002, 2008; Kuhn et al., 2010). The model's generalized representation of chemistry, dry deposition, emissions and turbulent mixing allows the role of canopy interactions in determining atmosphere–biosphere exchange fluxes and in-canopy and surface layer mixing ratios of e.g. ozone (O₃), nitrogen oxides (NO_x) and BVOCs to be studied. The BVOC emissions are calculated according to MEGAN (Guenther et al., 2006), considering the vertical distribution of biomass and direct as well as diffuse radiation to calculate leaf-scale BVOC emissions. The current implementation of canopy chemistry in the MLC-CHEM considers, in addition to standard photochemistry involving O₃, NO_x, methane (CH₄) and carbon monoxide (CO), the role of non-methane hydrocarbons including isoprene, and a selection of hydrocarbon oxidation products such as formaldehyde, higher aldehydes and acetone. Oxidation of the monoterpenes by OH, O₃ and NO₃ is taken into account, but the role of the monoterpene oxidation products in photochemistry is not considered in the current implementation of the chemistry scheme in the MLC-CHEM. For this study, we have extended the MLC-CHEM to consider, besides the compounds α-pinene and β-pinene and the observed monoterpene species α-terpinene, limonene and myrcene that are already included. The monoterpene basal leaf-scale monoterpene emission factors have been selected such that the model simulates monoterpene mixing ratios of comparable magnitude compared to the campaign-average observed mixing ratios. In the evaluation of simulated and observed mixing ratios we mainly focus on comparison of the simulated and observed temporal variability being determined by the differences in canopy processes for contrasting nocturnal and daytime conditions. For the model simulation, the basal emission factors were 0.18 µg C g⁻¹ h⁻¹ for α-pinene, 0.04 µg C g⁻¹ h⁻¹ for β-pinene, 0.11 µg C g⁻¹ h⁻¹ for α-terpinene, 0.9 µg C g⁻¹ h⁻¹ for limonene and 0.18 µg C g⁻¹ h⁻¹ for myrcene. Note the selected relative high basal emission flux for limonene is required to arrive at simulated mixing ratios comparable to the observed ones. Regarding the physical sinks, dry deposition of gases including the BVOC compounds depends on their uptake resistances calculated according to Wesely's (1989) parameterization, which estimates these uptake resistances based on the compounds' solubility and reactivity.

The simulations with the MLC-CHEM were constrained with the observed surface layer net radiation (above the canopy only), wind speed, relative humidity and O₃ mixing ratios as well as the temperatures measured above and inside the canopy (eight different heights including 12 and 24 m) from 17 to 20 October 2015, coinciding with the measurement dates. These simulations represent a set-up of the MLC-CHEM distinguishing six canopy levels with a canopy height of 30 m, implying canopy layers with a thickness of 5 m. Furthermore, we assumed a leaf area index of 5 m² m⁻² and a leaf area density profile such that about 70 % of this biomass is present in the top 15 m of the canopy, as previously observed at other tropical rainforest sites (Nölscher et al., 2016). Monoterpene emissions by vegetation were simulated using a temperature-only dependent emission flux as a function of the amount of biomass in each layer and the measured canopy temperature profiles interpolating between the 0.4 and 26 m temperature sensors. Meteorological observations for 18 October were missing and therefore the MLC-CHEM was constrained for this day by first-order estimates of the diurnal cycles in radiation, air and surface temperatures, relative humidity and wind speed comparable to the previous and subsequent days' meteorological conditions.

3.1 Time series and diel cycles

The continuous online PTR-MS measurements were compared with offline GC-FID samples over the course of 3 days in October 2015 (Fig. 1). The close agreement between the two measurement techniques provides confidence that almost all monoterpenes present in ambient air at the site were being measured. Note that in this comparison, ρ-cymene (an aromatic monoterpene) was removed from the calculations as the PTR-MS does not detect it on m∕z 137. The observed differences in the monoterpene chemodiversity in the rainforest canopy atmosphere were regarded to be driven by differences in emission, reactivity with the oxidizing species, physical removal processes and turbulent mixing conditions.

Figure 1Graph showing the speciated monoterpene mixing ratios measured hourly from 17 to 20 October 2015 for 24 m (b) and 12 m (c). The colours on the stacked bar plot indicate the different monoterpene species as they are denoted in the legend. The black line represents the PTR-MS total monoterpene mixing ratio, with a gap of data on the 19 October 2015. Temperature at 80 m is shown as the red thick line, and photosynthetically active radiation at 39 m is shown by the shaded areas (a).

Download

The total monoterpene mixing ratios were higher during the day, when temperature and solar radiation were at their maxima. Most of the observed distinct diurnal cycle in total monoterpene mixing ratios could be attributed to α-pinene, which was the dominant species during daytime (09:00 to 17:00), with mixing ratios as large as (average ± standard deviation) 0.33 ± 0.04 and 0.38 ± 0.21 ppb at 12 and 24 m respectively, and 0.15 ± 0.05 and 0.11 ± 0.06 ppb for the night (20:00 to 05:00) at 12 and 24 m. The second most abundant monoterpene species was limonene, with observed average daytime mixing ratios of 0.18 ± 0.09 and 0.19 ± 0.12 ppb at 12 and 24 m, respectively, and 0.18 ± 0.01 and 0.14 ± 0.07 ppb for the night-time at 12 and 24 m.

When comparing our results to previously published studies, we observed consistent differences with other regions of the Amazon rainforest. For instance, Kesselmeier et al. (2002) studied the seasonal monoterpene speciation in the Rondonia rainforest in southern Amazonia. Even though they found the same monoterpene species as presented in this study, their individual abundances were very different compared to the mixing ratios for the dry season at the ATTO site. α-Pinene and limonene were much higher at ATTO than in Rondonia, whereas camphene was substantially lower. In the case of β-pinene, the abundance measured at ATTO was much lower than at other Amazonian sites (Andreae et al., 2002; Karl et al., 2007). Given that emission patterns are highly dependent on species, environmental conditions and stresses, these differences underline that it cannot be assumed that the same speciation and emission rates of monoterpenes exist throughout the vast Amazon basin.

Furthermore, the difference between the 12 and 24 m height total monoterpene mixing ratios was minor given the variance of the measurements, but there was a tendency for the difference to be more pronounced during night-time (Table 1). These more pronounced differences between the measurement heights could also be due to an enhanced sensitivity of nocturnal mixing ratios to small changes in source and sink terms for the suppressed mixing conditions prevailing during the night-time.

Table 1Average mixing ratio with standard deviation in ppb at 24 and 12 m of the measured monoterpene species from 17 to 20 October 2015 as determined by the GC-FID analysis. The daytime period was chosen from 09:00 to 17:00 and the night-time period from 20:00 to 05:00 (local time). “BLD” stands for values below the detection limit. “MT sum” stands for the sum of monoterpenes.

Download Print Version | Download XLSX

Figure 2Average diel cycles for α-pinene (a), limonene (b), myrcene (c), ρ-cymene (d), β-pinene (e) and α-terpinene (f) mixing ratios for 24 m (dashed line) and 12 m (thick line). In the background, average diel cycles of isoprene mixing ratios as measured by the GC-FID are shown for 24 m (light green) and 24 m (dark green). Error bars represent the standard deviation of the averages.

Download

The continuous online measurements by the quadrupole PTR-MS indicated a clear diurnal cycle in the measured mixing ratios of the sum of monoterpenes, which has been reported previously from this site (Yáñez-Serrano et al., 2015). In order to assess the effect of each individual monoterpene species, we further investigated their diurnal cycles as obtained by the offline GC-FID samples. The measured diel cycles for the most relevant monoterpene species at the ATTO site were very similar at both heights. We also compared the measured diel cycle of isoprene as measured by the GC-FID with the observed diel cycle for the different monoterpene species for 12 and 24 m. The compounds that showed a diurnal cycle similar to isoprene were α-pinene and ρ-cymene (Fig. 2). This could be due to the emission of α-pinene and ρ-cymene being dependent on light and temperature, analogous to isoprene. However, during the night, both monoterpenes were also present, albeit at lower mixing ratios, and the nocturnal mixing ratios of the monoterpenes did not decrease as much as isoprene. This has also been noted in previous studies (Yáñez-Serrano et al., 2015).

Figure 3Pie charts representing the daytime (a, c) and night-time (b, d) average monoterpene species' abundances from 17 to 20 October 2015, with the average percentages and standard deviations at 24 (a, b) and 12 m (c, d). The day period was from 09:00 to 17:00 and the night period was from 20:00 to 05:00.

Download

Despite the higher mixing ratios of limonene compared to other monoterpene species (other than α-pinene), it was not possible to distinguish any clear diel pattern in the average data for this species (see Fig. 2). β-Pinene and α-terpinene likewise showed no obvious diel pattern in the rainforest air but were found to be above the detection limit of the GC-FID of 2 ppt.

In contrast to plant species of cooler climates, such as spruce, which emit terpenes from pools (Ghirardo et al., 2010; Lerdau et al., 1997), Amazonian plant species have been found to show an emission dependency on light and temperature (Bracho-Nunez et al., 2013; Jardine et al., 2015; Kuhn et al., 2002, 2004). This could partly explain the diurnal pattern of α-pinene mixing ratios, which exhibit some relation to a light- and temperature-dependent emission flux (Kuhn et al., 2002; Rinne et al., 2002; Williams et al., 2007). However, this behaviour was not observed for all monoterpene species. Therefore, the observed diurnal cycles of some monoterpene species might be related to a stronger temperature response.

3.2 Chemodiversity

The chemical speciation (or chemodiversity) of monoterpenes relates to the relative abundances of the different monoterpene species in the sampled air. α-Pinene, limonene, myrcene, ρ-cymene and β-pinene represented more than 85 % of the total monoterpene mixing ratio (Fig. 3). During the day (09:00 to 17:00) α-pinene had an average abundance (average±standard deviation) of 46 ± 25 and 36 ± 4 % of the total monoterpene mixing ratios at 24 and 12 m, respectively, and it was the dominant monoterpene in this study overall. However, during the night (20:00 to 05:00), its relative abundance dropped to 25 ± 13 and 25 ± 9 % at 24 and 12 m, respectively. In contrast, limonene made up 23 ± 15 and 20 ± 10 % of the monoterpenes at 24 and 12 m, respectively, by day, and increased during night-time to 33 ± 15 and 26 ± 16 % at 24 and 12 m. Thus, there was a tendency towards some differences in monoterpene species' abundances between day- and night-time. These were mainly due to the nocturnal decreases in α-pinene and the nocturnal relative increase in limonene. It is plausible that the observed decrease in α-pinene mixing ratios could be due to decreased vegetation emission, as reduced chemical destruction due to very low OH concentrations at night would lead to an increase in the nocturnal α-pinene mixing ratios.

Even though there were clear differences between the absolute and relative abundances of some monoterpene species during the day and night, there were no clear changes in the vertical gradients (e.g. for α-pinene, night-time averages were 0.15 ± 0.05 ppb for 12 m and 0.11 ± 0.06 ppb at 24 m). For the day, the apparent difference in the abundance of α-pinene was due to a single outlier data point covering 30 min at noon on 19 October 2015 at 24 m, when the α-pinene mixing ratio doubled. This increase could not be explained, although it could be related to a strong change in wind speed an hour before the measurement, when the wind was blowing from the north. In general, our observations indicate that the abundance of monoterpene species does not vary much over the heights selected (12 and 24 m) within the canopy. This is consistent with the results by Kesselmeier et al. (2000), where the monoterpene composition at the rainforest floor was comparable to the above-canopy composition at their site.

Table 2Lifetime of the different monoterpene species related to OH, O₃ and NO₃ for the OH daytime conditions at 24 and 12 m. In addition, the normalized reactivity to 1 ppb of the different monoterpene species is calculated.

Download Print Version | Download XLSX

3.3 Reactivity

The variability of the oxidants (OH, O₃ and NO₃) present in the Amazon air is important when considering the impact that monoterpenes can have on the oxidative regime in the Amazon region and Brazil in general. Hydroxyl radicals are produced mainly during the day via ozone photolysis. Low levels of OH can also be generated by the reaction of ozone with doubly bonded species (e.g. monoterpenes and sesquiterpenes), even at night. In this assessment, we considered the monoterpene contributions to OH reactivity by day only. In contrast, NO₃ is photolytically destroyed during the day but can become significant at night, so we assessed the impact of monoterpenes on NO₃ reactivity at night. Even though in the Amazon rainforest ozone levels are low (∼ 10–20 ppb) compared to other areas of the world (e.g. Williams et al., 2016), ozone is nevertheless present, and some monoterpenes are extremely reactive towards ozone. Table 2 gives an overview of the lifetime and reactivity (which is defined as the reaction rate constant (oxidant i.e. OH)*[monoterpene species]) to 1 ppb of all the investigated monoterpene species for these three oxidants. For calculating the lifetime of the different monoterpenes as presented in Table 2, typical oxidant concentrations for the Amazon rainforest conditions were used. For OH a mean value of 7×10⁵ molecules cm⁻³ was used as being representative of the site (Spivakovsky et al., 2000). For ozone reactivity calculations, 12 ppb was used, as this mixing ratio was observed during the measurement period. NO₃ mixing ratios were taken from the MLC-CHEM simulations that predicted mixing ratios of ∼ 0.4 ppt.

While α-pinene, limonene and myrcene were the most abundant species, their relative contribution to total monoterpene reactivity was not proportional to their abundances. The most abundant monoterpene, α-pinene, was not the dominant sink for the oxidants. In particular, α-terpinene dominated ozone reactivity associated with monoterpene abundance both during the day and night, as well as the nocturnal nitrate reactivity, despite the low mixing ratios measured for this compound (Table 2).

The monoterpene ozone reactivity was comparable between day ($\mathrm{1.37}\times {\mathrm{10}}^{-\mathrm{6}}$ s⁻¹) and night ($\mathrm{1.12}\times {\mathrm{10}}^{-\mathrm{6}}$ s⁻¹). α-Terpinene dominated the monoterpene–ozone chemistry, followed by myrcene and limonene. Despite the relatively high abundance of α-pinene (46 ± 25 %; average mixing ratio and standard deviation during the day was 0.34 ± 0.04 ppb at 12 m), its contribution to ozone reactivity with respect to other monoterpene species was only 11 ± 7 and 3 ± 1 % at 24 m, during the day and night, respectively, and 2 ± 1 % for both day and night at 12 m (Fig. 4). As previously noted, the differences in ozone reactivity between heights were negligible for the night and slightly higher at 24 m during the day. As ozone mixing ratios are quite similar for both heights during the day and night (11.4 ppb at 12 m and 10.4 ppb at 24 m during the night, and 16.1 ppb at 12 m and 15.6 at 24 m during the day), the higher abundance of α-pinene during the day and the lower α-terpinene mixing ratios at 24 m during the day mainly explain these changes in monoterpene–ozone reactivity. It is important to note that these results are derived from a relative abundance analysis, and unmeasured monoterpene species could change the proportions, although given the close similitude between PTR-MS and GC-FID measurements shown in Fig. 1, this is unlikely. On the other hand, very reactive species, which could dominate reactivity, may be present in very low concentrations, for which our measurements' capabilities would not allow detection.

Figure 4Pie charts representing day-time (a, e) and night-time (b, f) ozone reactivity, OH reactivity (only for day, c and g) and NO₃ reactivities from 17 to 20 October 2015, with the average percentages and standard deviations (only for night, d and h), for 12 m on the bottom and 24 m on the top. The day period was from 09:00 to 17:00 and the night period was from 20:00 to 05:00.

Download

The monoterpene reactivity towards the NO₃ radical during the night was also dominated by α-terpinene (40 ± 36 and 42 ± 27 %, for 24 and 12 m, respectively), although contributions of limonene (30 ± 13 and 25 ± 14 %, for 24 and 12 m, respectively), α-pinene (11 ± 6 and 11 ± 4 %, for 24 and 12 m, respectively), and myrcene (13 ± 11 and 16 ± 12 %, for 24 and 12 m, respectively) were also significant. No significant differences between the reactivities at different heights were observed, suggesting a rather homogeneous chemical regime regarding monoterpene chemical destruction within the canopy (from 12 to 24 m). However, note that this finding reflects the use of a single simulated NO₃ mixing ratio due to the absence of direct measurements in the Amazon rainforest, which prevents us from drawing any further conclusions. Our OH reactivity estimates demonstrate the important role of myrcene, with its higher reactivity towards OH due to its acyclic nature, especially at 12 m, where myrcene was more abundant. The total OH reactivity for the sum of monoterpenes was calculated to be 2.4 and 3.4 s⁻¹ for 24 and 12 m, respectively.

As demonstrated in this data set, chemically speciated measurements are very important for understanding how monoterpenes affect Amazon air chemistry, dependent on the time of day and season, as each monoterpene species has a different reactivity. Therefore, a lower abundance of a certain monoterpene species could not necessarily be related to a lower vegetation emission but also to a higher reactivity with atmospheric oxidants. Despite the small amount of α-terpinene present in the atmosphere, it can profoundly affect reactivity due to its fast reaction rate (its lifetime, according to the oxidant mixing ratios stated above, can be 103, 2 and 11 min to OH, O₃ and NO₃, respectively; Neeb et al., 1997). In terms of total OH reactivity accounted for by the monoterpenes, the values of this study are very low compared to the total OH reactivity measurements by Nölscher et al. (2016), with a mean total OH reactivity for the dry season of 32 s⁻¹, mostly dominated by isoprene chemistry. This suggests that the monoterpenes contributed only a small fraction to the total OH reactivity at the ATTO site during the investigated time period. This study demonstrates that the abundance does not relate to the importance in chemical reactivity, and species that are usually not considered by atmospheric chemistry models due to their modest mixing ratios might actually play a dominant role in the monoterpene atmospheric chemistry. Therefore, it is questionable to generalize the representation of terpene chemistry in models (Hallquist et al., 2009) using one or two monoterpene species only.

The gas-phase oxidation of the monoterpenes in the Amazon has numerous impacts on the environment, including the production of a multitude of new compounds that are generally longer lived than the primary emissions, increasing the lifetimes and particle production potential of certain compounds by suppressing oxidant availability. Moreover, production of OH due to the ozonolysis of monoterpenes is known to occur (Paulson et al., 1999). The production strength varies depending on the position of the double bonds, if there is more than one (Herrmann et al., 2010). Furthermore, the products of the reaction can be manifold. For instance, when α-pinene is oxidized by OH, especially at low nitrogen oxide mixing ratios, pinonaldehyde is formed in high yields (Eddingsaas et al., 2012). Chemical processing of α-pinene can also result in a further production of different monoterpenes such as the reaction of α-pinene with nitrate during the night, which can lead to the formation of ρ-cymene (Gratien et al., 2011).

Figure 5Monoterpene mixing ratio chemical speciation during the seasons of measurement. In (a), the monthly average of temperature (in red) and photosynthetically active radiation (in orange) are displayed with their standard deviations for the 80 m height. Rain, also in (a), is displayed in millimetres per month (bars). In (b), the different monoterpene species are differentiated by colours, stacked together, adding up to the sum of monoterpenes. Above each bar, a pie chart with the chemical speciation is shown for easier visualization.

Download

The implications of the measured monoterpene abundances for SOA formation at the ATTO site are difficult to quantify because the SOA formation yield is dependent on many factors. For example, it depends on the pre-existing organic aerosol mass into which these products can be absorbed (Griffin et al., 1999), and thus the SOA yield can vary between regions with similar monoterpene mixing ratios and different aerosol mass loadings. It also varies strongly between different oxidants and terpene species. For instance, α-pinene forms negligible aerosol mass under NO₃ oxidation (Fry et al., 2014), whereas there is production of organic aerosols when the oxidation of α-pinene involves O₃ (Ehn et al., 2014) and OH (Eddingsaas et al., 2012). Monoterpenes containing endocyclic double bonds (e.g. α-pinene, 3-carene) or open chains (e.g. myrcene) tend to form less aerosol mass from ozonolysis than monoterpenes with exocyclic double bonds (e.g. β-pinene, sabinene; Hatakeyama et al., 1989; Hoffmann et al., 1997). Following the equation established by Bonn et al. (2014, Eq. 5 in text), we were able to estimate the potential aerosol particle number formation rate initiated by monoterpene species only ($\mathrm{1}\times {\mathrm{10}}^{-\mathrm{5}}$ to $\mathrm{5}\times {\mathrm{10}}^{-\mathrm{5}}$ cm⁻³ s⁻¹ at 24 m) assuming steady-state conditions for radicals. Those were found to be approximately 2 orders of magnitudes smaller than the calculated potential new aerosol particle formation rate caused by oxidation products of sesquiterpenes. Our calculations assume mixing ratios of sesquiterpenes of 0.2 ppb, revealing potential formation rates of $\mathrm{1}\times {\mathrm{10}}^{-\mathrm{3}}$ and $\mathrm{4.5}\times {\mathrm{10}}^{-\mathrm{3}}$ cm⁻³ s⁻¹ at 24 m based on previous measurements in the Amazon (Jardine et al., 2011), which are remarkably smaller than observed at mid-latitude conditions (Bonn et al., 2014). Furthermore, the level of NO present (nitric oxide) also affects the potential aerosol growth (Wildt et al., 2014) and yield (Sarrafzadeh et al., 2016) at low BVOC∕NO_x ratios. As the theory assumes contributions of larger organic peroxy radicals (RO₂), which are destroyed by reactions, e.g. with NO, increasing NO_x at a constant BVOC mixing ratio will decrease the BVOC∕NO_x ratio and lead to a decline in SOA yield. Our calculations showed this effect, with a change of NO from 0.2 to 1 ppb, leading to a decrease in the formation rate at a diameter of 3 nm. This interdependence calls for a consistent consideration of the BVOC and NO_x exchange in aerosol formation and growth studies.

3.4 Seasonality

By examining GC-FID data collected in previous campaigns, an intra- and inter-annual comparison can be made. Total monoterpene averages for each season were calculated from 11:00 to 16:00 at 24 m. Based on these data, we distinguished the monoterpene mixing ratios representative of the dry season, the wet season and the wet–dry transition. The dry season conditions were represented by measurements collected in November 2012, September 2013 and August 2014, and the measurements from this study were collected in October 2015. The wet season measurements were collected in March 2013 and the wet–dry transition measurements were collected in June 2013. For the dry season conditions, the total monoterpene mixing ratios were substantially higher (1.02 ppb) compared to the observed monoterpene mixing ratios in the wet season (0.14 ppb) and the wet–dry transition season (0.18 ppb) (Fig. 5). This coincides with the occurrence of the highest radiation levels and temperatures as well as the lowest precipitation during these dry season measurement campaigns. During the wet season, the total monoterpene mixing ratios were lowest, while during the transition season in June, they were slightly higher.

For each season, an average monoterpene chemodiversity distribution is shown in Fig. 5. During the dry seasons, the chemodiversity seems relatively similar (39.4 ± 4 % for α-pinene, 20.3 ± 3 % for limonene), whereas it slightly changes during the wet season and dramatically changes during the wet–dry transition. The reason for this difference in June could be related to changes in the phenology, as demonstrated at a central Amazonian site (Alves et al., 2016; Lopes et al., 2016). Furthermore, during the dry season of 2015, a very strong El Niño event took place, leading to extremely dry conditions observed region-wide (Jardine et al., 2017).

It has been shown previously that the amounts and speciation of monoterpenes vary strongly according to plant species and leaf developmental stage. For instance, Bracho-Nunez et al. (2011) found young leaves of some Mediterranean plant species to emit more α-pinene and mature leaves to emit e-ocimene, z-ocimene and myrcene, but not α-pinene. Some species have been found to be higher emitters of α-pinene (i.e. Hevea spruceana), whereas others are higher emitters of myrcene (i.e. Quercus coccifera Bracho-Nunez et al., 2013). The leaf developmental stage is also important, as reported for flushing young leaves emitting monoterpenes, in contrast to the isoprene emission of mature leaves of the same plant species (Kuhn et al., 2004). Such a behaviour could explain the lower mixing ratios and different chemodiversity found in June. During this time of the year, leaf flushing takes place in the central Amazon region (Alves et al., 2016; Lopes et al., 2016). Under these conditions, lower α-pinene mixing ratios were found as compared to the dry season, when young leaves reach mature levels. Therefore, the seasonality in Amazon forest monoterpene emissions might depend more on the changes in aggregated canopy phenology than on the seasonality of climate drivers (Wu et al., 2016). Our study shows that chemodiversity remains relatively constant during at least the dry seasons but changed between different seasons. Therefore, the implications for the atmosphere are different for each monoterpene species. Kesselmeier et al. (2002) also showed this type of behaviour in their study, in which they did not find a strong difference in total mixing ratios, but different chemodiversity between seasons, likely expressing differences in seasonal plant developments and atmospheric reactivities, which should be accounted for in model implementations at the ATTO site.

3.5 Modelling analysis

To further support our analysis of the observed magnitude as well as temporal variability in the monoterpene mixing ratios inside the forest canopy, we used the MLC-CHEM (1) to explore how well the model represents the measured mixing ratios and (2) to assess the role of the different in-canopy processes in explaining the diel cycle of the observed monoterpene mixing ratios at the ATTO site.

From Fig. 6, which shows a comparison of the simulated (12.5 and 22.5 m) and observed (12 and 24 m) speciated monoterpene mixing ratios from 17 to 20 October 2015, it can be inferred that the simulated speciated monoterpene mixing ratios are of comparable magnitude to the measured observations. This comparison regarding the magnitude of observed and simulated mixing ratios serves mainly to assess the validity of the required selection of basal emission fluxes for the different monoterpene compounds. A more relevant result seems to be the overall quite good agreement between the simulated and observed temporal variability in monoterpene mixing ratios. Note that we also conducted a simulation in which we applied temperature- and light-dependent monoterpene emission flux. However, those simulations did not follow the observed magnitudes and temporal variability as well as the model simulations considering monoterpene emissions that only depend on temperature.

The generally quite good agreement between the simulated and observed monoterpene mixing ratios, except for an overestimation of simulated α-pinene mixing ratios for 17 October, expresses the overall result of temporally varying emissions, in-canopy chemistry, turbulent mixing and deposition. The latter also involves a potentially important role in the deposition to wet leaf surfaces (the inferred wet surface uptake resistances for the monoterpenes are ∼ 300 s m⁻¹, similar to values reported by Zhou et al. (2017)); the MLC-CHEM uses relative humidity as a proxy for the fraction of the leaf surface being wet (Lammel, 1999; Sun et al., 2016). This results in substantially smaller estimates of canopy wetness on 17 October compared to the following days, which partly explains the simulated high α-pinene mixing ratios. The simulated α-pinene mixing ratios for 18–20 October, with inferred wet surface fractions up to 1 during the night and ∼ 0.5 during daytime, are in much better agreement with the observations. Regarding the comparison of the simulated observed mixing ratios for some of the other monoterpenes, the simulated β-pinene, limonene and myrcene mixing ratios, especially at 12.5 m seem to capture the observed temporal variability quite well. Note that this result for limonene reflects the use of a high leaf basal emission factor (0.9 µg C g⁻¹ h⁻¹) required to simulate mixing ratios reaching up to 0.4 ppb. These MLC-CHEM simulations were also used to infer how much of the actual emission flux escapes the canopy, expressed by the calculated atmosphere–biosphere limonene flux divided by the canopy emission flux of limonene. This ratio reaches a maximum value of 0.5 around noon, implying that these model simulations indicate that at the middle of the day, about 50 % of the emitted limonene is removed inside the canopy by in-canopy oxidation and deposition. During night-time, this ratio reaches a minimum < 0.1, indicating simulation of very efficient in-canopy removal.

Figure 6Comparison between simulated results (solid lines) for 12.5 (orange) and 22.5 m (green) from the MLC-CHEM, with the GC-FID speciated mixing ratios measurements (in ppb) for α-pinene (a), limonene (b), myrcene (c), β-pinene (d) and α-terpinene (e) at ATTO from 17 to 20 October 2015. The error bars represent the 20 % uncertainty involved in the GC-FID measurements.

Download

These modelling results should be interpreted with caution, also given that some of the simulated processes cannot be evaluated due to missing observations of canopy wetness as well as the uptake efficiency of monoterpenes by wet surfaces. It should be considered that the simulated removal of monoterpenes by wet canopy surfaces could also compensate for a misrepresentation of other canopy processes, e.g. reduced emissions from wet canopy surfaces or an underestimation of the oxidation efficiency. Further analysis of the model-simulated process tendencies (Ganzeveld et al., 2008) indicates only small changes in the simulated source of the monoterpenes over the 4-day period. Regarding the sink of, for example, α-pinene, chemical destruction of α-pinene oxidation by O₃, OH and NO₃ appears to be a relative small term, with the overall sink being dominated by deposition to wet surfaces, showing quite large temporal variability. Consequently, the agreement between simulated and observed temporal variability in monoterpene mixing ratios that is quite reasonable indicates that deposition to wet surfaces may play an important role in monoterpene atmosphere–biosphere exchange. This should be further corroborated, calling for experiments to determine the actual efficiency (and mechanisms) of the uptake of monoterpenes by wet canopy surfaces.