Impacts of soil moisture on de-novo monoterpene emissions from European beech, Holm oak, Scots pine, and Norway spruce

Impacts of soil moisture on de-novo monoterpene (MT) emissions from Holm oak, European beech, Scots pine, and Norway spruce were studied in laboratory experiments. The volumetric water content of the soil, Θ , was used as reference quantity to pa-rameterize the dependency of MT emissions on soil moisture and to characterize the 5 severity of the drought. When slight of were observed but with further progressing the emissions decreased to almost zero. The increases of MT emissions observed under conditions of mild drought were explainable by increases of leaf temperature due to lowered tran- 10 spirational cooling. When Θ fell below certain thresholds, MT emissions decreased simultaneously with Θ and the relationship between Θ and MT emissions was approximately linear. The thresholds of Θ (0.044–0.19 m 3 m − 3 ) were determined as well other parameters required to describe the soil moisture dependence of de-novo MT emissions for application in the Model of Emissions of Gases and Aerosols from Nature, 15 MEGAN. A factorial approach was found appropriate to describe the impacts of Θ , temperature, and light. Temperature and Θ inﬂuenced the emissions largely independent from each other, and, in a similar manner, light intensity and Θ acted independently on de-novo MT emissions. The use of Θ as reference quantity in a factorial approach was 20 tenable in predicting constitutive de-novo MT emissions when Θ changed on

the respective dependencies.
One of the first algorithms was developed by Tingey et al. (1980Tingey et al. ( , 1991, who showed that monoterpene (MT) emissions from Slash pine exponentially depend on temperature but are independent of PAR. The underlying reason is that MT emissions from conifers originate from MT diffusion out of pools, which depend on temperature and not 15 directly on PAR. In a modified syntax, Tingey's algorithm reads: In Eq. (1), Φ is the emission rate of the MT at the actual temperature T , Φ S is the standard emission rate or emission activity factor, i.e. the emission rate measured at standard temperature T S . β is the parameter describing the temperature dependence. 20 β is in the range of 0.09 K −1 . Isoprene is emitted directly after its biosynthesis. Isoprene emissions thus are denovo emissions and directly related to the rate of isoprene biosynthesis. Biosynthetic activity generally depends on T and PAR. Hence, both variables are required to describe isoprene emissions as shown in the algorithm by Guenther et al. (1993) In Eq.
(2), Φ is the emission rate at temperature T and at light intensity PAR. Φ S is the emission activity factor i.e. the emission rate measured at standard light intensity (often set to 1000 µmol m −2 s −1 ) and at standard temperature, T S (often set to 30 • C). α is the 5 parameter describing the PAR dependence of the emission, and c L is a normalization factor used to obtain Φ = Φ S when PAR is equal to the standard light intensity.
Equation (2) consists of three factors: first the emission activity factor, second the factor describing the PAR dependence and third the factor describing the temperature dependence. A factorial approach requires the individual factors to be independent of 10 each other; i.e. Guenther et al. (1993) postulated that the PAR dependence is independent of the actual temperature and vice versa, that the temperature dependence is independent of the actual PAR. Schuh et al. (1997) found interdependent impacts of temperature and PAR on MT emissions from sunflower which could not be described by a factorial approach as 15 given in Eq. (2). They ascribed the observed interdependency to the two different mechanisms of MT emissions, MT emissions from pools that are independent of the actual light intensity and de-novo emissions that depend on the actual light intensity. Depending on the species, both mechanisms can act in parallel. Schuh et al. (1997) suggested two additive terms to describe T and PAR dependence of MT emissions. One of them 20 was similar to Eq. (2) and was used to describe de-novo MT emissions. The other one was similar to Eq. (1) and describes the temperature dependence of MT emissions from pools. By using additive terms the factorial approach for de-novo emissions was retained.
In the same study, Schuh et al. (1997)  clearly shows that MT emissions also depend on other variables than T and PAR alone, and one of these other variables is soil moisture.
In the Model of Emissions of Gases and Aerosols from Nature (MEGAN, Guenther et al., 2006Guenther et al., , 2012, impacts of soil moisture are only considered for isoprene emissions. Guenther et al. (2006) implemented the empirical algorithm by Pegoraro et al. (2004) 5 who used the volumetric water content, Θ, as reference quantity for characterizing the impacts of soil moisture. Three different regimes of Θ are used in MEGAN to define the factor that describes the impact of soil moisture on isoprene emissions. Above a threshold Θ 1 , isoprene emissions are not affected by soil moisture and the factor is unity. Below Θ 1 , isoprene emissions linearly decrease with decreasing Θ until the 10 wilting point Θ W is reached. The wilting point Θ W is the soil moisture below which plants cannot extract water from the soil. At and below Θ W isoprene emissions are set to zero. In MEGAN, Θ W is taken from a database by Chen and Dudhia (2001). The difference: Θ 1 − Θ W , ∆Θ 1 , is the empirical parameter used to describe the dependence of isoprene emissions on soil moisture. Its value (∆Θ 1 = 0.06 m 3 m −3 ) is taken from 15 Pegoraro et al. (2004). Compared to isoprene there are less studies on impacts of soil moisture on MT emissions. Some studies show increasing emissions with decreasing soil moisture (Bertin and Staudt, 1996;Blanch et al., 2007;Ormeño et al., 2007), others show decreasing emissions with increasing severity of drought (Lavoir et al., 2009;Šimpraga et al., 20 2011). Besides this, different reference quantities have been used to characterize the soil moisture level. Among these are the plant water potential (Ormeño et al., 2007;Lavoir et al., 2009) and the diurnal variation of the radial stem diameter (Šimpraga et al., 2011). As these plant parameters are highly variable between individuals and influenced by a larger number of factors, they can hardly be used for up-scaling. Introduction approach is justified. Results of our experiments and the limitations of our approach are described below.

Laboratory set up
Experiments were performed at the Jülich Plant Atmosphere Chamber facility (JPAC).

5
The setup of the chambers has already been described elsewhere in detail (e.g. Schimang et al., 2006;Mentel et al., 2009Mentel et al., , 2013. The two plant chambers used for these experiments were made of borosilicate glass and operated as continuously stirred tank reactors (CSTR). To keep the temperature constant, the CSTRs were mounted in separate walk-in climate chambers. Each CSTR was equipped with a Teflon fan providing 10 homogeneous mixing and diminishing the boundary layer resistance at leaf surfaces. The chambers were equipped with several connections to introduce temperature sensors (Newport Omega HTMTSS), a light-intensity sensor (LI-COR, LI-189) and to connect the tubes for gas phase analysis and air supply. were also reduced by the adsorption dryer. The CO 2 concentration in the plant chamber was kept at levels of about 350 ppm by adding CO 2 at the inlet. The air flow through the chambers was kept constant by mass flow controllers (Brooks Instruments). Typical air flows were in the range of 20-30 L min −1 when using the small chamber and 50-100 L min −1 when using the large chamber.

5
Also the analytical equipment at JPAC is described in detail in previous publications (Schuh et al., 1997;Wildt et al., 1997;Schimang et al., 2006). O 3 concentrations were measured by UV absorption (Thermo Environmental Instruments, model 49). Concentrations of NO were measured by chemiluminescence (Tecan CLD 770 AL PPT) and for the measurements of NO 2 the analysed air was led through a photolytic converter (Tecan PLC 760). During all the measurements presented here, O 3 mixing ratios were below 1 ppb. NO x mixing ratios were about 300 ppt.
Differences in mixing ratios of H 2 O and CO 2 between chamber inlet and outlet were measured by IR absorption (Rosemount Binos 100 4P). Absolute H 2 O concentrations were determined with dew point mirrors (Walz, MTS-MK1). Mixing ratios of BVOC were 15 measured using gas chromatography -mass spectrometry (GC-MS, Agilent GC-MSDsystem HP5890 Series II + MS HP5972A, or GC-MSD-system HP6890 + HP5973, equipped with thermo desorption unit, TDSG, Gerstel, Mülheim, Germany). Calibration of the GC-MS systems was performed using a diffusion source containing pure chemicals in individual vials in combination with a dynamic dilution system. Concentra-20 tions of the compounds released from the calibration source were determined from the mass loss rates of the individual compounds and the dilution fluxes. The VOC mixing ratios were in the lower ppb to ppt range. For details on the GC systems and calibration procedure see e.g. Heiden et al. (2003).
The CSTR contained shoots and leaves of the plants. The upper part of the setup 25 was separated from the lower part containing roots and soil by PTFE sheets with a hole at the centre to accommodate the stem. The stem duct was closed by a flexible bag. The pot containing the plants' roots was positioned in a shallow dish allowing for collecting excess water from the pot. Plant, pot, and dish were mounted on a balance BGD 11,2014 Impacts of soil moisture on de-novo monoterpene emissions (Sartorius, MC1) that was used to measure the weight of the investigated plant together with the soil and the water in the soil. The flexibility of the bag that sealed the gap between stem and stem duct allowed weighing the plant/soil system on-line. The balance had a nominal resolution of 1 g but variations of the chamber's slight overpressure (5-10 mbar) imposed noise in the range of 20-30 g limiting the precision of weight 5 measurements.

Determination of flux densities
Flux densities for the compound X , Φ(X ), were calculated using the respective differences of the mixing ratios between chamber inlet and outlet and the leaf area, A leaf , as basis: In Eq. (3), F air is the air flow through the chamber (in units of mol s −1 ), [X ] i is the mixing ratio of compound X at chamber inlet and [X ] is the mixing ratio of compound X in the chamber. Flux densities are termed as MT emission rates for X = MT, transpiration rates for X = H 2 O, and net assimilation rates for X = CO 2 . According to the results 15 of several tests (e.g. Schuh et al., 1997;Heiden et al., 2003;Schimang et al., 2006) wall losses and chemical reactions were unimportant for all compounds investigated here and were therefore neglected. As a convention flux densities are positive when compounds are emitted (e.g. water, MT) and negative when compounds are taken up (e.g. CO 2 ). To allow better visual comparison in diagrams, net assimilation is multiplied 20 by −1. Stomatal conductance was calculated using Eq. (4). O] is the difference between water vapour mixing ratios in the substomatal cavity and in the chamber air, the former calculated from leaf temperature assuming saturation vapour pressure in the substomatal cavity. After finishing an experiment, about 5 % of the leaves from broadleaf species were dissected from the plant, taking care to select representative leaves. The leaves were 5 scanned for leaf area determination and evaluated by image analysis. Total leaf area was calculated by multiplying the averaged area measured per dissected leave by the total number of leaves.
Needle area for the conifers were determined as described in Shao et al. (2001). The number of needles per 10 cm of branches with green needles was counted for 10 representative branches, needles were removed from these branches, scanned and the area of dark pixels was determined. Upscaling from the measured needle area to the total needle area was performed by considering the total lengths of branches with green needles for the respective conifer.
Uncertainty in leaf/needle area determination was estimated to approximately 10 %. 15 Note that the needle areas given in Table 1 are projected needle areas and leaf areas for broadleaf species are one sided leaf areas.

Experiments with plants
Experiments were conducted with individual plants using species representative for European climate zones. European beech (Fagus sylvatica L.) and Holm oak (Quercus 20 ilex L.) represented species without storage pools for monoterpenes. Scots pine (Pinus sylvestris L.) and Norway spruce (Picea abies L.) mainly exhibited MT pool emissions but also showed de-novo MT emissions. One to two years old beech-, pine-, and spruce-seedlings were taken from the forest, potted in buckets of 15 L volume and stored outside for about a year before they were 25 used for the experiments. Seedlings of Holm oak were obtained from the forest nursery of Castelporziano Estate, Rome, central Italy. These have been successively potted in 15 L buckets and stored for several weeks in a growth room before using them in the 12994 Introduction CSTR. All plants were potted in the same soil that was a mix of peat (Einheitserde ED73 containing 2 g L −1 nutrient salts and 1 g L −1 clay) with quartz sand, volumetric ratio 5 : 1, density 0.46 ± 0.07 kg L −1 .
To control for the de-novo nature of the 1,8-cineol emission from the conifers we exposed the plants to 13 CO 2 . The chamber was flushed with synthetic air (Air Liquide, 5 Germany) and 13 CO 2 (99 % 13 C, ∼ 350 ppm) was added for ∼ 3 h in each of both experiments. Individual plants were investigated for time periods of 3-10 weeks. In most of our studies, the impact of soil moisture was investigated (Table 1, European beech, two experiments with Holm oak -experiment Holm oak 1 and Holm oak 2 -, one exper-10 iment with Norway spruce and one with Scots pine). In these experiments the plants were exposed to a diurnal rhythm of 11 h illumination, 11 h darkness and simulation of twilight by switching on or off individual lamps within 1 h in the morning and evening, respectively.
In additional studies, we varied T or PAR, respectively, in parallel to soil moisture 15 to investigate the potential limitations of the factorial approach. In the third experiment with Holm oak the chamber temperature was systematically changed during periods of illumination. Except for two days, the chamber temperature was set to 15 • C during the night and kept at 15 • C for the next 6-7 h of the following illumination period (from 3 o'clock to ∼ 10 o'clock). Then the temperature was set to 20 • C for four hours (from 20 ∼ 10 to ∼ 14 o'clock) and thereafter to 25 • C (from ∼ 14 to ∼ 18 o'clock, all data local time). The exact timing of temperature settings was adapted to the start of GC runs. The period of illumination was elongated to 15 h allowing measuring at the 3 different temperatures but at the same PAR for each day. When Θ had fallen below 0.02 m 3 m −3 , the chamber temperature was not changed for two days to follow the drought induced 25 decrease of MT emissions without changes of temperature and PAR.
In the fourth experiment with Holm oak, PAR was changed systematically during periods of illumination. Every second day PAR was set to 700 µmol m −2 s −1 for 6 h (from 4 o'clock to ∼ 10 o'clock), thereafter to 400 µmol m −2 s −1 for 5 h (from ∼ 10 to ∼ 15 BGD 11, 12985-13027, 2014 Impacts of soil moisture on de-novo monoterpene emissions Due to the radiation from the lamps, the chamber temperatures and the leaf tempera-5 tures varied with illumination. During periods of darkness leaf temperatures and chamber temperatures were similar. During periods of illumination leaf temperature was higher than chamber temperature. For well watered plants leaf temperature was about 2 • C higher than the chamber temperature. Progressing drought stress decreased transpiration and leaf temperatures increased by additional 2-3 • C relative to the chamber 10 temperature ( Fig. 1).

Determination of Θ and the Θ-dependence of MT emissions
The volumetric water content, Θ, was determined from the mass loss of water during the respective experiments (Eq. 5).
In Eq. (5), M act is the actual mass of the soil and M dry is the dry mass of the soil. V soil is the volume of the soil in the pots neglecting the volume of the roots. M act was measured online and M dry was estimated from soil samples taken from the top of the pots and oven dried at 110 • C for five days. The measured mass loss was converted to volume loss by using a water density of 1 kg L −1 .

20
The procedure of determining M dry by taking soil samples added the main uncertainty to the data we give for Θ. The error caused by this procedure for Θ is estimated to be ±15 % of the absolute data. In one case, the measured M act were lower than M dry . This led to slightly negative values for Θ which is physically impossible. However, since the deviation from zero was quite low, we left the negative values. 11,2014 Impacts of soil moisture on de-novo monoterpene emissions The dependence of MT emission on Θ was parameterized in the following manner: maximum MT emissions were reached when Θ had fallen near to the threshold below which MT emissions started to decrease (Θ 1 ). The average of the emission rates measured around the maximum emissions was set as standard emission rate, Φ S . Emission rates measured at the same light intensity and chamber temperature 5 were normalized by dividing them by Φ S . The normalised data were used to determine the relationship between MT emissions and Θ. Although the decrease of emissions with decreasing Θ appeared to be exponential, we applied a linear approximation to be comparable to other parameterization approaches. Linear fits to the normalized data allowed to determine Θ 1 as well as the volumetric water content Θ 0 below which 10 the extrapolated emissions became zero. Only data with Θ < Θ 1 and measured at the same PAR and the same chamber temperature were taken for fitting. Furthermore, all data where Φ/Φ S < 0.05 were discarded to diminish the impact of the exponential behaviour. Similar to the notation used in MEGAN, we calculated ∆Θ 1 , the difference Θ 1 − Θ 0 . This procedure of determining the Θ dependence of MT emissions was per- 15 formed for all data sets. with coefficients of determination always above R 2 > 0.95. The pattern was also not affected by the drought (Fig. 2) and therefore the effects of drought on de-novo MT emissions can be shown at the example of a single MT. For spruce and pine we choose 1,8-cineole emissions for the soil moisture dependence, because these were the only pure de-novo MT emissions from both conifers. 5 Sabinene was the dominant emission from European beech, comprising nearly 1/3 of the total emissions. Other MT emitted in substantial amounts were: γ-terpinene, α-terpinene, β-phellandrene, and α-terpinolene. We will show the response of European beech to drought at the example of sabinene. Main MT emissions from Holm oak were those of α-pinene, limonene, β-pinene, sabinene and myrcene. In two of the experiments with Holm oak α-pinene was the dominant emission (experiment 1: ∼ 48 % of total MT emissions; experiment 3: ∼ 42 % of total MT emissions), in the two other experiments the emissions of limonene were the strongest (experiment 2: limonene ∼ 45 %, α-pinene ∼ 23 % of total MT emissions; experiment 4: limonene ∼ 39 %, αpinene ∼ 25 % of total MT emissions). To demonstrate the behaviour of MT emissions 15 from Holm oak in response to drought we choose α-pinene as pivotal substance. It has to be noted that strong emissions of the acyclic ocimenes have been absent from Holm oaks. Emissions of ocimenes may show different behaviour than emissions of the cyclic MT (Staudt and Bertin, 1998). During our experiments the ocimene emissions were too low to significantly influence the sum of all MT emissions (maximum ∼ 6 % of 20 total MT emissions). Figure 3 shows the time series of sabinene emissions from European beech over a period of about three weeks. According to the de-novo nature of the emissions a sig-25 nificant diurnal variation was observed with nearly no emissions during darkness and high emissions during periods of illumination. As can be seen in the Θ-trace, the plant 12998 Introduction was watered 4 times during the experiments. Between 2.5 and 5.6 L water were added after soil moisture had reached values of Θ ∼ 0.1 m 3 m −3 (dry conditions) and Θ near to 0 m 3 m −3 (severe drought), respectively. Besides the strong light dependent variation, also impacts of soil moisture on MT emissions were observable. During severe drought (2nd period, days 5-17 in Fig. 3) 5 MT emissions decreased near to zero. After re-watering MT emissions increased again. Such increases were not instantaneous but appeared on a time scale of few days.

European beech
Transpiration (not shown in Fig. 3) and net assimilation showed similar responses to Θ as MT emissions but with different response times. As can be seen, changes in net 10 assimilation appeared earlier than changes in MT emissions, in particular during the phase of the second drought period (Fig. 3, days 5-17). Figure 4 shows the dependence of normalized sabinene emissions and net assimilation on Θ. Only data taken at PAR = 440 µmol m −2 s −1 are shown. For Θ between 0.35 and 0.1 m 3 m −3 , MT emissions showed a considerable scatter that was mainly 15 caused by different response times for the emissions to increase to pre-drought levels.
With Θ falling below 0.1 m 3 m −3 emissions decreased to minute amounts. As obvious from Fig. 4, changes of net assimilation as response to decreasing soil moisture were different from those of MT emissions. Net assimilation already decreased at higher Θ of ∼ 0.22 m 3 m −3 . 20

Holm oak
Two experiments with Holm oak were set up to investigate whether the reduction of MT emissions with decreasing Θ was determined by the soil moisture or by some other time constant of the plant itself (Table 1, experiment Holm oak 1 and Holm oak 2). Holm oak 1 was investigated at higher PAR and at higher T and showed higher 25 rates of net photosynthesis and higher transpiration (data not shown). The soil and the plants roots were also exposed to higher temperature compared to Holm oak 2. Due to the combination of higher evaporation and transpiration rates, Θ dropped from 12999 Introduction  Fig. 5a). In contrast, the decrease of MT emissions with decreasing Θ was similar for both individuals although they were investigated at different dynamics of drought 5 progression (Fig. 5b). The substantial difference observed for the temporal behaviour of the emissions (Fig. 5a) did not cause significant changes in decrease of emission rates for Θ < 0.1 m 3 m −3 (Fig. 5b). Similar as for beech, increasing MT emissions were observed for Holm oak when Θ declined from ∼ 0.4 m 3 m −3 to 0.15 m 3 m −3 . For Holm oak 1, at higher T and PAR, the emissions increased nearly 3-fold. For Holm oak 2 the 10 increase was about 40 %.
In the third experiment with Holm oak chamber temperature was changed systematically during the progressing drought (Table 1, experiment Holm oak 3). This experiment was designed to measure the impact of Θ on MT emissions at different temperatures as well as the temperature dependence at different stages of drought. 15 The temperature dependence of MT emissions was determined by linear regression analysis of ln(Φ) vs. leaf temperatures (see Eqs. 1 or 2 third factor). During the first 9 days with varying temperature at Θ > 0.1 m 3 m −3 the temperature dependence remained constant with β = 0.12 ± 0.007 K −1 (mean, 1σ standard deviation). Similar to the observations in the other experiments with Holm oak, MT emissions de-20 creased when Θ fell below 0.06 m 3 m −3 . During this measurement period the impacts of progressing drought and variations of temperature superimposed each other. Although temperature increased systematically over the day, emissions did not increase substantially. Consistent with the findings reported by Bertin and Staudt (1996), typical log-linear relationships between emissions and temperature were hardly detected. 25 We therefore had to correct the data to consider the drought induced decrease of MT emissions over the day. Approaches to determine the Θ dependence from the data obtained at the days when temperature was held constant were not successful. The small dynamic range in BGD 11,2014 Impacts of soil moisture on de-novo monoterpene emissions which Θ changed during these two days (from ∼ 0.025 to ∼ 0.013 m 3 m −3 ) prevented a reliable analysis. On the other hand, the assumed linear decrease of MT emissions with decreasing Θ is only an approximation because the decrease appeared to be exponential (compare Figs. 4 and 5b). We therefore directly used the temporal decay of the emissions during the two days when temperature and PAR were held constant 5 to develop the drought correction considering that the temporal decay of emissions reflects the decay in Θ (compare Fig. 5a and b). An exponential function was fitted to the temporal decay measured during the respective days yielding a decay rate of 0.04±0.002 h −1 , R 2 = 0.85. The correction factor was set to 1 for the time when the chamber had reached steady state conditions after 10 twilight in the morning. Correction factors were then calculated for each time when the chromatograms were taken by using the decay rate of 0.04 h −1 . The emission rates determined for the respective times were then divided by the correction factor which dropped from 1 in the morning to ∼ 0.6 in the evening. Data corrected for the drought induced decrease were used to determine the temperature coefficient β during periods 15 with low soil moisture. Figure 6 shows the values obtained for β in dependence of Θ. As the decay rate of 0.04 per hour was not applicable for Θ near to Θ 1 , data points near to Θ 1 (0.06 < Θ < 0.12 m 3 m −3 ) were discarded. Data analysis resulted in β = 0.13 ± 0.024 K −1 for the data points determined for Θ < 0.06 m 3 m −3 which is about the same as β = 0.12, 20 determined for stress free conditions within the error limits. Thus, no significant differences were found between the temperature dependency of emissions under drought stress and under stress free conditions. In the fourth experiment with Holm oak PAR was changed systematically during the progressing drought (Table 1, experiment Holm oak 4) to allow for determining the 25 dependence of the emissions on Θ at different PAR and the dependence of emissions on PAR at progressing drought.
The correction was made in the same way as described above by fitting an exponential function to the temporal decay measured during the days at constant PAR and BGD 11,2014 Impacts of soil moisture on de-novo monoterpene emissions temperature. The emission rates measured the next day under conditions of variable PAR were then corrected by dividing them by the correction factors. After correction the data were normalized using the emission rates measured at the respective day at PAR = 700 µmol m −2 s −1 . Figure 7 shows the normalised emission as a function of light intensity at the example of 3 days. Within the precision of the data the light intensity 5 dependence did not substantially or systematically change with progressing drought.

Scots pine and Norway spruce
The experiments with pine and spruce lasted for about a month each and each of the plants was exposed to 3 drought periods. At least one of these drought applications was severe with Θ < 0.01 m 3 m −3 .
Main emissions from both conifers were pool MT emissions such as α-pinene and ∆-3-carene. Besides such pool emissions the plants also emitted 1,8-cineole which is a de-novo emission (Tarvainen et al., 2005;Kleist et al., 2012). This was also tested for both individuals used in our experiments by exposing them to 13

Mild drought stress
All de-novo MT emissions from all plants investigated in our experiments were dependent on soil moisture. While mild drought (Θ > Θ 1 with Θ 1 = Θ 0 + ∆Θ 1 , see Table 2) 5 caused slight increases of MT emissions, severe drought (Θ < Θ 1 ) caused decreasing MT emissions from all investigated species.
Increasing emissions under mild drought have been reported before (Bertin and Staudt, 1996;Blanch et al., 2007;Ormeño et al., 2007). Blanch et al. (2007) studied MT emissions from Holm oak (Quercus ilex) and Aleppo pine (Pinus halepensis). They 10 report emissions to increase two-fold for Holm oak when the relative water content of the leaves decreases from ∼ 90 % to 80 %. Ormeño et al. (2007) studied the impact of water deficit on emissions from Rosemary (Rosmarinus officinalis), Aleppo pine (Pinus halepensis), Rock rose (Cistus albidus), and Kermes oak (Quercus coccifera). Using the plant water potential as a reference, they report two to three fold increases of MT 15 emissions when the water potential drops from −2 to −8 MPa.
Although these previous studies agree with our finding that MT emissions increase under mild drought, a quantitative comparison is impossible because different reference quantities are used to characterize the degree of drought stress. We assume that most of the increases in MT emissions observed during our studies were only an indi-20 rect effect of drought. In our studies leaf temperatures were reasonably well measured for the broadleaf species. Leaf temperatures increased when Θ dropped from high values (Θ > 0.3 m 3 m −3 ) to ∼ 0.1 m 3 m −3 even though chamber temperatures were kept constant. Obviously, the decline in transpiration lowered its cooling effect leading to higher leaf temperatures (Fig. 1). In the experiment with European beech and in three of the experiments with Holm oak (experiments 2, 3, and 4) MT emissions increased by less than 50 % when leaf temperature increased due to the progressing drought. Ap- 11,2014 Impacts of soil moisture on de-novo monoterpene emissions plying β ∼ 0.12 K −1 , the main fraction of increases in MT emissions from these plants was explained by the increase of leaf temperature. Only for one single Holm oak we found an increase (∼ three fold, see Fig. 5b, red circles) too high to be explained by increases in leaf temperature alone. Due to the watering-drought-watering procedure used here it is possible that parts 5 of the increases arise from recovery from a preceding drought. In particular when Θ decreases rapidly, the effects of sequential droughts may overlap. However, the effect of strongly increasing MT emissions with Θ dropping from high levels to Θ 1 was observed in one experiment only and without further information we can only speculate on this effect.

10
The reason for increases of the 1,8-cineole emissions from conifers during mild drought stress (∼ 30-50 % when Θ dropped from 0.35 to 0.1 m 3 m −3 ) also remains unproven because leaf temperatures were not reliably measurable for both conifers. Nevertheless it is reasonable to assume that also in conifer species needle temperature increase due to lower transpirational cooling caused the observed slight increases 15 in MT emissions. For modelling of isoprene emissions in MEGAN, the impact of leaf temperature is implemented by a general temperature dependence of the emissions. Such parameterization cannot differentiate between variations of leaf temperature by variations in transpiration or by variations of air temperature. Therefore the indirect effect of drought 20 does not appear in the factor for the dependence of isoprene emissions on soil moisture (Guenther et al., 2006). In analogy we suggest, to consider decreasing emissions caused by severe drought stress also for de-novo MT emissions.

Severe drought stress
Similar to previous studies (e.g. Bertin and Staudt, 1996;Llusià and Peñuelas, 1998;25 Plaza et al., 2005;Lavoir et al., 2009;Šimpraga et al., 2011;Bourtsoukidis et al., 2014) we observed substantially reduced MT emissions during periods of severe drought BGD 11,2014 Impacts of soil moisture on de-novo monoterpene emissions as well as increasing emissions after re-watering (Peñuelas et al., 2009). Llusià and Peñuelas (1998) found strong suppression of MT emissions from non-storing species such as Holm oak and a significant relation between net assimilation and MT emissions. Lavoir et al. (2009) found exponentially decreasing emissions when the leaf water potential fell below −2 MPa. Šimpraga et al. (2011) show strong attenuation of 5 MT emissions from European beech with progressing drought. The temporal shape of sabinene emissions from Beech as shown here in Fig. 3 for the second severe drought was nearly identical to the temporal shape observed by Šimpraga et al. (2011) indicating that the same general effect was found. But again, direct comparison of the results is not easy because different reference quantities are used to characterise the degree 10 of drought. Bourtsoukidis et al. (2014) use soil water availability to characterize the degree of drought. They measured BVOC emissions from Pedunculate oak (Quercus robur ) and Black cherry (Prunus serotin). Compared to the MT emissions from regularly watered plants (soil water availability ∼ 30 %) the MT emissions from plants under drought (soil 15 water availability ∼ 2.3 %) were reduced by about 65 %. This is similar to the reductions found when taking the relative water content of the soil as a reference instead of Θ.

Present state of modelling
While effects of soil moisture on de-novo MT emissions are not yet considered in the Model of Emissions of Gases and Aerosols from Nature (MEGAN, Guenther et al.,20 2006, 2012) the effects on isoprene emissions are considered. As noted above, isoprene emissions are unaffected as long as Θ is above a threshold Θ 1 , they decrease linearly between Θ 1 and the wilting point Θ W and they remain zero as long as Θ < Θ W . We found a similar behavior for the de-novo MT emissions indicating that the emission of MT on soil moisture can be modelled using the same formalism with the same basic 25 modelling parameters Θ W and ∆Θ 1 .
The wilting point Θ W is the soil moisture below the plans cannot take up water. During our measurements we always observed a recovery of the plants from the drought after 13005 Introduction

Tables Figures
Back Close

Full Screen / Esc
Printer-friendly Version

Interactive Discussion
Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | re-watering. Thus soil moisture during drought in our experiments must have been always slightly above a critical value causing permanent damage to the plant. Insofar Θ 0 listed in Table 2 may not directly reflect Θ W . However, Θ W depends on the soil and Θ W varies by up to two orders of magnitude (Chen and Dudhia, 2001). For modelling, Θ W data from Chen and Dudhia (2001) are used and such data are listed for different 5 soil types. The data we provide for Θ 0 are therefore not needed for modelling purposes and possible differences between Θ 0 and Θ W are not relevant for modelling. Basic parameter for modelling is ∆Θ 1 . Our average of ∆Θ 1 (average = 0.08 ± 0.05 m 3 m −3 , error = 1σ standard deviation, statistical weight the same for individual plants) is similar to ∆Θ 1 used to model isoprene emissions (∆Θ 1 = 0.06 m 3 m −3 ). Our data for ∆Θ 1 show substantial variability and it is unknown so far whether the high variability is caused by a different behavior of individuals or by other reasons. Within the uncertainty of the data it was also impossible to find differences between Holm oak and the other species. No conclusions can thus be drawn whether or not more drought tolerant plants such as Holm oak show different behavior than more drought sensitive 15 plants such as European beech. However, all plants behaved similar and we therefore suggest that using the average ∆Θ 1 for modeling purposes is a tenable approach. Considering the impact of soil moisture in MEGAN, results in a 7 % reduction of global isoprene emissions only. However, substantial impacts of drought isoprene emissions are expected on a regional scale (Guenther et al., 2006). Including the impacts of 20 soil moisture on de-novo MT emissions in modelling may have different effects because isoprene emitters and de-novo MT emitters are regionally differently distributed. Recent studies show that, particularly in Europe, the fraction of MT emitting species is larger than previously predicted (Oderbolz et al., 2013;Kemper Pacheco et al., 2014). The effects described here may therefore have large impacts when modelling MT emissions 25 in Europe.
Impacts of drought on MT emissions in Europe have been modelled by Lavoir et al. (2011). The authors performed model calculations to determine the impact of soil moisture on MT emissions from Holm oak in southern France. As input for their BGD 11,2014 Impacts of soil moisture on de-novo monoterpene emissions C. Wu et al.  Lavoir et al. (2009) who determined the soil moisture dependence of MT emissions from Holm oak using the leaf water potential as reference quantity. For modelling they used the relative water content of the soil (RWC) as reference considering a relationship between leaf water potential and RWC. According to Rambal et al. (2003), RWC is the ratio of current water content to water content at field 5 capacity. Using our mass based data, RWC can be calculated according to Eq. (6): In Eq. (6), M FC is the mass of the soil at field capacity, M act is the actual mass and M dry the dry mass as in Eq. (5). Field capacity is reached when the micropores of the soil are filled with water and the macropores filled with air after water is lost by gravity.
According to our procedure of waiting some hours before measuring the weight of pot and plant, the water in the macropores should have been lost. To compare our data to those of Lavoir et al. (2011) we therefore approximate M FC from the weight shortly after removing the excess water from the dishes below the pots. Lavoir et al. (2011) give a threshold of 0.7 kg kg −1 above which MT emissions 15 from Holm oak are independent of soil moisture and a linear decrease to zero at RWC = 0.4 kg kg −1 . Using our data for Holm oak we obtain RWC = 0.24 ± 0.1 kg kg −1 as the threshold. The emissions are zero at RWC ≤ 0.09 ± 0.1 kg kg −1 . Both our values are far lower than the data used by Lavoir et al. (2011) for up-scaling. Such differences are explainable by different structural properties of soils stressing the importance for 20 regional models to consider soil properties for a more realistic emission estimate. However, the differences also show that the impacts of drought on regional MT emissions may be less pronounced than modelled by Lavoir et al. (2011).

Use of Θ as reference quantity for modelling
We aimed at providing a data set to support modelling of soil moisture effects on de- an empirical relationship between de-novo MT emissions and Θ. Nevertheless, when reflecting mechanisms of the drought impact, plant physiological processes must be considered. For isoprene the decreasing emissions as result of drought are ascribed to a general decrease of the plant's performance (e.g. Brüggemann and Schnitzler, 2002). Decreased plant performance causes reduction of isoprene biosynthesis and 5 thus isoprene emissions. Basic parts of isoprene and MT biosynthesis pathways are identical suggesting that general mechanisms causing the reduction of isoprene and MT emissions are similar. We therefore suggest that, similar to isoprene emissions, also the decrease of de-novo MT emissions is caused by a general decrease of the plants performance.

10
As plant performance is coupled to the rate of transpiration and to net assimilation either of them may be regarded as reference quantity. But, as already pointed out by Peñuelas et al. (2009) and by Loreto and Schnitzler (2010), drought induced responses in transpiration or net assimilation differ from responses in MT emissions. This general behavior was also found here in all experiments and we therefore could not use either 15 of them as reference quantity.
We used Θ as reference, but, from a mechanistic point of view Θ is not a direct reference quantity. In a first step, soil moisture impacts the plants performance and in a second step the plants performance affects MT synthesis and emissions. The relationship found here between de-novo MT emissions and Θ therefore has to be 20 scrutinized. In particular the temporal behaviour of MT emissions may be skewed by the time needed for the plant to respond to changes in Θ.
Indeed there were time periods with substantial time lags between changes of Θ and the plants responses in MT emissions. Re-watering caused Θ to increase to optimum conditions for the plants on time scales of hours. For the de-novo MT emissions it took 25 some days until the same level was reached as before the drought. We assume that the time needed for the plants to recover from previous droughts was the reason for the decoupling of Θ and MT emissions during such time periods. 11,2014 Impacts of soil moisture on de-novo monoterpene emissions Quick changes of Θ also appear in nature in case of strong rainfall. If such rainfall appears after a severe drought period with Θ being far below the threshold affecting emissions, Θ and emissions are most probably also decoupled in nature impeding a correct description of the emission temporal behavior during such periods.

BGD
We tested whether or not a mechanistic model can be used to describe the impacts 5 of soil moisture on MT emissions. Niinemets et al. (2002) couple de-novo MT emissions to photosynthetic electron transport. Using the fraction of the photosynthetic electron transport necessary for MT synthesis (ε) as a surrogate for standard emissions shall lead to a more realistic description of emissions, especially under stress conditions where empiric models are supposed to fail (Niinemets et al., 2002).

10
For our tests we used the data from beech during a period of mild drought and rewatering when stomatal conductance was still reliably measurable to test the model (Fig. 3, days 0 to 11). ε was calculated as described in Niinemets et al. (2002) and plotted as a function of Θ (Fig. 8).
There was no significant change in ε when Θ changed rapidly due to re-watering, 15 indicating that a coupling of MT emissions to the fraction of electron transport may be usable for describing periods of recovery. However, ε was not constant when Θ was between 0.2 m 3 m −3 and 0.1 m 3 m −3 . This may be caused by the mild drought itself or by the changes in leaf temperature. However, impacts of soil moisture on MT emissions during desiccation cannot be described using a constant value for ε. Empirical 20 parameterizations have to be included also in this model to allow realistic description of BVOC emissions (Arneth et al., 2007). We tried to derive such empirical relationship to allow describing ε in dependence of soil moisture but these attempts were unsuccessful. With progressing drought MT emissions were still strong although transpiration and net assimilation were already 25 very low. ε is related to leaf internal CO 2 concentration which is coupled to stomatal conductance. Reliable calculations of stomatal conductance require reliable data on transpiration but our data on stomatal conductance at such low transpiration were error prone. We therefore cannot provide reliable data for ε at Θ ≤ 0.1 m 3 m −3 .
BGD 11,2014 Impacts of soil moisture on de-novo monoterpene emissions While phenomenological derived algorithms cannot describe the emissions during recovery after severe drought, they can describe impacts of soil moisture during desiccation. Desiccation appears on longer time scales than changes of soil moisture after heavy rain fall. For such longer time scales, the lag between changes of Θ and the plants' response should not carry too much weight in skewing the relationship be-5 tween emissions and Θ. This assumption was confirmed by the result from the two experiments with Holm oak at different temporal progression of drought. Although the temporal decreases appeared on different time scales the Θ dependence of MT emissions remained similar (compare Fig. 5a and b). As long as desiccation appeared on a time scale of days, the plant responses to changes in Θ were obviously quick enough 10 to allow finding the relationship between Θ and de-novo MT emissions. Although Θ is only an indirect reference quantity for plant performance, it proved to be suitable for considering the impacts of soil moisture on de-novo MT emissions in modelling.

Justification of a multiplicative approach for modelling
Modelling BVOC emissions is often performed using factorial approaches. An impor-15 tant requirement for the validity of such a factorial approach is a negligible interdependency of the individual factors. We investigated such possible interdependencies but no substantial effects were found.
As can be seen from Fig. 6, the temperature dependence of MT emissions from Holm oak was largely independent of the actual Θ. Vice versa, the soil moisture dependence 20 of the emissions was independent of the actual temperature (compare data in Table 2). Also the PAR dependence of MT emissions from this species was not substantially affected by the actual soil moisture (Fig. 7) and vice versa the Θ dependence was not substantially influenced by PAR (Table 2). Compared to the overall effect of drought on de-novo MT emissions, possible residual interdependencies were negligible. This 25 indicates that a factorial approach is justified at least for Holm oak investigated with this respect. From the similarity of basic processes leading to de-novo MT emissions we postulate that such interdependencies are also negligible for other plant species.

Conclusions
All de-novo MT emissions from all plants investigated in our experiments clearly depended on soil moisture. The investigated species are representative for European climate zones and they all are strong MT emitters. We therefore conclude that impacts of drought have to be re-considered for modelling. 5 To the best of our knowledge the direct impacts of soil moisture on constitutive denovo MT emissions are not considered in MEGAN. Severe drought reduces the emissions and therefore models neglecting impacts of soil moisture overestimate MT emissions. Depending on the climatology in the respective regions and depending on the fraction of de-novo emitted MT, modelled MT emissions may drop substantially when 10 considering soil moisture as variable influencing de-novo MT emissions.
On the other hand to the best of our knowledge a dependence of the fraction of the photosynthetic electron transport necessary for MT synthesis, ε, on soil moisture is not considered in modelling so far. According to our data ε increased already two-fold when Θ dropped to ∼ 0.2 m 3 m −3 . Hence, simulations of de-novo MT emissions assuming ε to be constant will underestimate their source strengths at conditions of moderate drought. Again, depending on the climatology of the respective regions and the fraction of de-novo emitters in the respective region, modelled de-novo MT emissions may increase substantially. Large differences of modelling results for MT emissions (Arneth et al., 2008) may therefore be reduced to some extent.

20
Although the detailed mechanisms causing the reductions in de-novo MT emissions are still unknown, one statement can be made. Estimations of future alterations of MT emissions must consider such effects because climate change will induce more and longer lasting drought periods (Dai, 2013). Future long lasting drought periods will suppress constitutive de-novo MT emissions. In combination with more intensive heat periods that also suppress de-novo MT emissions (Kleist et al., 2012), these abiotic stresses will have substantial impacts on regional and global BVOC emissions. 11,2014 Impacts of soil moisture on de-novo monoterpene emissions C. Wu et al. Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Mediterranean species. Relationship between terpene emissions and plant water potential, Chemosphere, 67, 276-284, 2007. Pegoraro, E., Rey, A., Bobich, E. G., Barron-Gafford, G., Grieve, A., Malhi, Y., and Murthy, R.:

BGD BGD
Effect of elevated CO 2 concentration and vapor pressure deficit on isoprene emission from leaves of Populus deltoides during drought, Funct. Plant Biol., 31, 1137-1147, 2004 BGD 11,2014 Impacts of soil moisture on de-novo monoterpene emissions C. Wu et al.  . Fraction of photosynthetic electron transport used for MT synthesis, ε, calculated according to Niinemets et al. (2002) in dependence of Θ. The data were calculated as daily averages. Red circles represent ε as determined for the first measurement period during mild drought (Fig. 3, days 0-5) when Θ dropped to ∼ 0.1 m 3 m −3 . Blue squares show the data after re-watering at day 6 until day 11 when Θ again had dropped to ∼ 0.1 m 3 m −3 . Experiment with European beech.