Leaf-Scale Study of Biogenic Volatile Organic Compound Emissions from Willow (Salix spp.) Short Rotation Coppices Covering Two Growing Seasons

: In Europe, willow ( Salix spp.) trees have been used commercially since the 1980s at a large scale to produce renewable energy. While reducing fossil fuel needs, growing short rotation coppices (SRCs), such as poplar or willow, may have a high impact on local air quality as these species are known to produce high amounts of isoprene, which can lead to the production of tropospheric ozone (O 3 ). Here, we present a long-term leaf-scale study of biogenic volatile organic compound (BVOC) emissions from a Swedish managed willow site with the aim of providing information on the seasonal variability in BVOC emissions during two growing seasons, 2015–2016. Total BVOC emissions during these two seasons were dominated by isoprene (>96% by mass) and the monoterpene (MT) ocimene. The average standardized (STD, temperature of 30 ◦ C and photosynthetically active radiation of 1000 µ mol m − 2 s − 1 ) emission rate for isoprene was 45.2 ( ± 42.9, standard deviation (SD)) µ g g dw − 1 h − 1 . Isoprene varied through the season, mainly depending on the prevailing temperature and light, where the measured emissions peaked in July 2015 and August 2016. The average STD emission for MTs was 0.301 ( ± 0.201) µ g g dw − 1 h − 1 and the MT emissions decreased from spring to autumn. The average STD emission for sesquiterpenes (SQTs) was 0.103 ( ± 0.249) µ g g dw − 1 h − 1 , where caryophyllene was the most abundant SQT. The measured emissions of SQTs peaked in August both in 2015 and 2016. Non-terpenoid compounds were grouped as other VOCs (0.751 ± 0.159 µ g g dw − 1 h − 1 ), containing alkanes, aldehydes, ketones, and other compounds. Emissions from all the BVOC groups decreased towards the end of the growing season. The more sun-adapted leaves in the upper part of the plantation canopy emitted higher rates of isoprene, MTs, and SQTs compared with more shade-adapted leaves in the lower canopy. On the other hand, emissions of other VOCs were lower from the upper part of the canopy compared with the lower part. Light response curves showed that ocimene and α -farnesene increased with light but only for the sun-adapted leaves, since the shade-adapted leaves did not emit ocimene and α -farnesene. An infestation with Melampsora spp. likely induced high emissions of, e.g., hexanal and nonanal in August 2015. The results from this study imply that upscaling BVOC emissions with model approaches should account for seasonality and also include the canopy position of leaves as a parameter to allow for better estimates for the regional and global budgets of ecosystem emissions.


Introduction
The increase in greenhouse gases (GHGs) and the global temperature during the last few decades have resulted in mitigation strategies and climate targets, where the main goal is to keep the average global temperature increase within 1.5-2 • C [1,2] compared clay or a silty clay loam with 34-36% clay content, 45-46% silt content, and no organic content (Table 1). Willow trees growing on P1 originated from the species S. viminalis. This plantation was established in 1994 and the last time the trees were harvested was in the beginning of 2013, which means measurements were done during their third growing season in 2015. In total, P1 covered approximately 6 ha. The variety growing on P2 is called Tora and P2 was established in 2003 on an area of approximately 5 ha. The trees on P2 were harvested in spring 2013, which means they were in their fourth growing season in 2016. At both sites, all trees were planted in double rows, separated by 0.75 m and 1.25 m between each double row. The distance between the trees in the row was 0.4 m. Table 1. Information about the two plantations (P1 and P2) in the study. Size (ha), previous land use, year of establishment, canopy height (m), leaf area index (LAI, m 2 m −2 ), soil type and content, treatments, varieties, and yield (m 3 ha −1 ).  3 Taken from SGU [48]. 4 Total number of treatments during the rotation cycle. 5 Yield after four growing years.

Experimental Setup
Measurements were done with two different setups. First, we used a branch chamber (volume ca 13 L) during repeated campaigns throughout the growing seasons 2015 and 2016. The aim of this setup was to follow the seasonal pattern of BVOC emissions during the two growing seasons.
Second, a photosynthesis system with a leaf chamber (6 cm 2 leaf area) was used to provide photosynthesis information in parallel to BVOC measurements focusing on the light response.

Branch Chamber Setup
The branch chamber used was a cylindrical construction made of PFA and stainless steel covered with a thin transparent PFA film [49][50][51]. At the beginning of each campaign, the tip (ca 30 cm long) of a shoot from one Salix tree was carefully inserted into the chamber the day before measurements started, avoiding any disturbance of the leaves. The chamber was left open on one end when not in use (i.e., during night). Every day before the measurements started, the chamber lid was closed, and the chamber was flushed with purge air for one hour. After this procedure, the branch chamber was used as a dynamic chamber with a continuous purge flow of 3-4 l×min −1 , leading to a residence time of ca 4 min for the air within the chamber.
In 2015, only one branch chamber was used. Starting in June 2016, two identical sets of branch chambers were used in parallel at different canopy heights. Each campaign, two trees standing next to each other were chosen for the branch chamber measurements. Prevailing conditions of temperature (T, • C) and relative humidity (RH) were measured both inside and outside the branch chamber (CS215, Campbell Scientific, UT, USA). Pho-tosynthetically active radiation (PAR, µmol m −2 s −1 ) was measured close to the chamber and at canopy level (Li-190, LI-COR, Lincoln, NE, USA). The data were recorded on a data logger (CR1000, Campbell Scientific, Logan, UT, USA) every 10 sec and stored as 1 min averages.

Light Response Curves and Photosynthesis Rates
Besides the branch chamber, a portable photosynthesis system (LI-6400XT, LI-COR, NE, USA) was used in 2016 to provide photosynthesis information in parallel with BVOC measurements from individual Salix leaves [42,46,50,52]. The leaf chamber has a capacity to measure leaves with an area up to 6 cm 2 (2 × 3 cm 2 ) and the measurements were done on the middle part of the leaves to obtain the maximum area of the leaves. The flow rate into the leaf chamber was set to 500 µmol s −1 (ca 0.7 L min −1 ). The leaf chamber was equipped with a LED source (6400-02B). This photosynthesis system made it possible to control the PAR, T, RH, and carbon dioxide (CO 2 ) concentration within the leaf chamber. Temperature and CO 2 were set to ambient conditions (20-25 • C and 400 ppm), while PAR was set to increase between 0 and 1500 µmol m −2 s −1 in six steps (0, 150, 300, 450, 600, 1000, and 1500 µmol m −2 s −1 ) to obtain light response curves. Relative humidity was regulated to match ambient values and ranged mainly from 30% to 70%. Net assimilation rate (A, µmol CO 2 m −2 s −1 ), transpiration (Tr, mmol H 2 O m −2 s −1 ), and stomatal conductance (g s , mol H 2 O m −2 s −1 ) were measured with the Li-6400XT, and water use efficiency (WUE, mmol CO 2 mol −1 H 2 O) was then calculated by dividing A by Tr.

BVOC Sampling
BVOC sampling was based on adsorbent cartridges containing Tenax TA (a porous organic polymer) and Carbograph 1TD (graphitized black carbon (C)) (Markes International Ltd., Mid Glamorgan, UK). Prior to sampling, all cartridges were conditioned. In this process, N 2 is used as a carrier gas while the cartridges are heated up to 300 • C, which empty the cartridges from possible contamination.
Air from the dynamic chamber was pulled through an adsorbent cartridge for 20 min per sample at a flow rate of 200 mL min −1 using a flow-controlled pump (Pocket Pump, SKC Ltd., Dorset, UK), leading to a total sample volume of 4 L. In the case of the leaf chamber, the photosynthesis system had been modified with a trace-gas tube connection after the outlet of the leaf chamber [42,46,50,52]. This way, BVOC samples could be taken in parallel to the on-going photosynthesis measurements.
To provide VOC-free and O 3 -free air into the branch chamber, air entering the branch chamber was passed through a hydrocarbon trap (Alltech, Associates Inc., Chicago, IL, USA), which contained activated C and a MnO 2 -coated copper mesh. The hydrocarbon trap was also attached to the intake of the photosynthesis system. For the branch chamber, background samples of the purge air entering the chamber were taken every day, while blank samples from the empty leaf chamber were taken after the last light intensity (1500 µmol m −2 s −1 ) measurement.
During the campaigns, BVOC samples were taken during day time (8:00-18:00). In the case of the leaf chamber, measurements started one hour after the leaf was inserted to avoid potential stress-induced BVOCs from physical handling [42,52]. When a new light level was set, 30 min passed before the BVOC sampling started so the leaf had time to adapt to the new condition [52]. After sampling, the cartridges were sealed with Teflon-coated brass caps and stored at 3 • C until analysis. After each campaign was finished, the sampled leaves were harvested and dried at 75 • C for two days to measure the dry weight.
In total, 754 samples were taken on 35 trees during 15 campaigns between July 2015 and October 2016, and the campaigns were in general 2-5 days (Table 2). Due to setup and instrumental failure, 26 samples were missing in 2015 and 65 in 2016. Campaigns performed in 2015 were all on P1 except the last one in September, where the first day was on P1 (P1* in Table 2, the star means another variety) and the last two days were on P2. This change was due to a rust outbreak of a fungi (Melampsora spp.). Signs of the infestation were seen in July, and by mid-September most of the trees had shed their leaves ( Figure S2). Leaves with the least damage were chosen, but more or less all trees were equally affected. All the campaigns performed in 2016 were done at P2. No signs of Melampsora were observed on Tora but leaf-feeding beetles (Phratora vulgatissima) damaged the leaves, especially higher up in the canopy. On account of P. vulgatissima, undamaged leaves were hard to find, at least in the uppermost part of the canopy, from July onward ( Figure S3). Table 2. All campaigns done in 2015 and 2016 with branch and leaf chambers. Two different plantations (P1 and P2) were studied and new trees were studied each campaign. The star (P1*) means that another variety was measured at this site compared with the rest of the campaigns on P1. Height levels were z L (up to ca 70% of the total canopy height (TCH)), z M (up to ca 70% of the TCH but growing at the edge of the plantation), and z H (heights above z L ). z/H means the ratio between the measured leaves (z) and the canopy height (H). The heights of the measurements were divided into three categories. The lower height (z L ) included leaves that were below or in the lower part of canopy (up to ca 70% of the total canopy height (TCH)) and growing within the plantation. The higher height (z H ) included leaves above z L (>70% of TCH) and growing in a more sun-adapted environment compared with the shade-adapted leaves in z L . Trees growing at the edge that were <70% of the TCH were classified as middle (z M ) because they were expected to be more sunlit than if they grew within the plantation. The canopy height (H) was measured as the average height for the trees that were growing just next to the leaves, which the BVOC emissions were measured from, as these trees were responsible for the light conditions at the measured leaves. All measurements during 2015 were done on z H except for the last campaign. Measurements with the branch chambers in 2016 were done at the edge of P2, while the measurements with the leaf chamber were done both on the edge and within P2.

Gas Chromatography-Mass Spectrometry and Data Processing
All samples were analyzed by thermal desorption (UNITY2 thermal desorber, Markes International Ltd., Mid Glamorgan, UK) in combination with an ULTRA autosampler and gas chromatograph-mass spectrometer (GC-MS) (7890A Series GC coupled with a 5975C inert MSD/DS Performance Turbo EI System, Agilent, CA, USA) [42,53]. Separation of BVOCs was done by a HP-5 capillary column (50 m × 0.2 mm, film thickness 0.33 µm) and helium was used as a carrier gas. The oven temperature started at 40 • C for 1 min, was first raised to 210 • C with a rate of 5 • C min −1 , and was then raised to 250 • C with a rate of 20 • C min −1 . The compounds were analyzed in the program Enhanced Chem-Station (MSD ChemStation E.02.01.1177, Copyright 1989-2010 Agilent Technologies, Inc. Santa Clara, CA, USA) and identified by injected external standards (Table S1) or according to mass spectra in the NIST 8.0 database. To quantify BVOCs for which no specific standard was available, α-pinene was used for monoterpenes (MTs), humulene for sesquiterpenes (SQTs), and toluene for other VOCs. Only sample peaks that had at least twice as strong a signal as the corresponding peaks in background samples were chosen for further analysis.
Two compounds, toluene and butylated hydroxytoluene, were removed from all measurements done by LI-6400XT because they had approximately the same concentration in the background samples as the measurements from the leaves and were therefore assumed to be emitted by this instrument.
The emission rates for the BVOCs were calculated by Equation (1), according to Ortega and Helmig [54] where E (µg g dw −1 h −1 ) is the emission rate, C 2 (µg L −1 ) is the BVOC concentration taken from the chamber, C 1 (µg L −1 ) is the BVOC concentration entering the branch chamber or the BVOC concentration inside the empty leaf chamber, Q (L h −1 ) is the flow rate of the purge air, and m (g dw ) is the dried mass of the leaves contained in the chamber.
A standardization for BVOCs that were both light and temperature dependent was done by Equation (2) according to Guenther et al. [55]. The reason for this standardization was to allow for comparison with other studies, regardless of prevailing environmental conditions. The STD values for T and PAR were 303.15 K and 1000 µmol m −2 s −1 , respectively. The compounds that were STD with Equation (2) were isoprene, ocimene, and α-farnesene.
where E (µg g dw −1 h −1 ) is the actual emission at the measured chamber temperature T (K) and PAR (µmol m −2 s −1 ), E s is the STD emission, and C T and C L are dimensionless correction factors for temperature and light defined by Equations (3) and (4).
For compounds that are temperature, but not light, dependent, the STD emission can be calculated with the temperature-dependent Equation (5), which was used on all terpenoids except isoprene, ocimene, and α-farnesene.
Fitted net assimilation curves were done by using Smith's equation, Equation (6) [57] , and A max is net assimilation at the maximum PAR (1500 µmol m −2 s −1 ). The statistical tests were performed with a Kruskal-Wallis test at a significance level of 0.05 when comparing emissions between different heights since the data did not have a Gaussian distribution. The optimized parameters for the fitted light response curves were determined by minimizing the difference between measured and simulated emissions according to a root mean square procedure.

Isoprene
The emission rates for isoprene changed mostly according to prevailing T and PAR values. The average STD isoprene emission from P1 in 2015 was 59.3 (±44.5) µg g dw −1 h −1 , where the corresponding average chamber T (T C ) and chamber PAR (PAR C ) were 24.9 (± 5.4) • C and 462 (±296) µmol m −2 s −1 , respectively. Most of the measured isoprene emissions that varied between 92.0 and 153.2 µg g dw −1 h −1 during 2015 occurred in the middle of July and during the second half of August. The chamber T was high (>27.5 • C) and PAR C varied from 304 to 1654 µmol m −2 s −1 for these isoprene emissions ( Figure 1A-C). The measured emission peaked (153.2 µg g dw −1 h −1 ) in July and the STD emission peaked (177.6 µg g dw With more unsteady weather conditions and under cloudy conditions and even rain, low emissions of isoprene (1.6-7.0 µg g dw −1 h −1 ) were observed even during the peak growing season in July and August 2015 due to low PARc values (<200 µmol m −2 s −1 ). In particular, in the beginning of September when Tc was <15 • C and PARc was <160 µmol m −2 s −1 , the emission rates only reached up to 0.1 µg g dw −1 h −1 . In addition, the leaves were damaged in September due to the infestation of Melampsora on P1.
The seasonal trend for measured and STD isoprene emissions in 2016 showed an increase from spring to summer; thereafter, it decreased from late summer to mid-autumn. The average STD isoprene emission in 2016 was 30.6 (±26.7) µg g dw −1 h −1 and the average T C and PAR C were 23.2 (±7.6) • C and 521 (±444) µmol m −2 s −1 . The measured isoprene emission varied from 50.1 to 143.7 µg g dw −1 h −1 between June and August but emissions up to 46.5 µg g dw −1 h −1 were already observed in May ( Figure 2C). All measured isoprene emission rates >50 µg g dw −1 h −1 in 2016 had corresponding T C and PAR C ranging from 26.8 to 39.9 • C and from 365 to 1518 µmol m −2 s −1 . The STD isoprene emission peaked (150.0 µg g dw −1 h −1 ) in late July 2016, while the measured isoprene emission peaked (143.7 µg g dw −1 h −1 ) in August. Isoprene emissions were negligible during the last campaign in October, when T C and PAR C were low (less than 11 • C and 110 µmol m −2 s −1 , respectively).
The influence of T C on the isoprene emission rates showed a stronger exponential increase for the variety on P1 than for Tora on P2 (Figure 3). In particular, when T C reached above 23 • C, the average emission in 2015 increased faster than during 2016. The steeper response in 2015 was possibly due to the fact that this variety was a stronger emitter of isoprene, but the unexpected lower average emission when T C varied from 30 to 36 • C in 2016 also reduced the slope of the fitted curve this year. (F) STD emission rates (µg g dw −1 h −1 ) for MTs (green) and SQTs (blue). Vertical dashed lines separate the different days.
Each bar represents individual measurements.   Table S4.

Monoterpenes
The emitted MTs from P1 in 2015 were α-pinene, eucalyptol, linalool, and ocimene. The average STD MT emission was 0.365 (±0.151) µg g dw −1 h −1 . Ocimene was emitted during all campaigns and the highest measured emissions were seen during the first campaign in mid-July (0.409-1.648 µg g dw −1 h −1 ) and the fourth campaign in mid-August (0-0.932 µg g dw −1 h −1 ). Ocimene contributed with ca 93-100% to the average STD MT emission rate during all campaigns in 2015 (Table S3), and STD emissions of ocimene peaked (1.631 µg g dw −1 h −1 ) in July. α-pinene was the second most abundant MT in 2015, with the STD emission ranging from 0 to 0.076 µg g dw −1 h −1 , but α-pinene was not observed in September 2015. The total average STD MT emission decreased from 0.580 (±0.292) µg g dw −1 h −1 in July to 0.286 (±0.182) µg g dw −1 h −1 in September and a similar trend was seen for the measured average total MT emission.
More MTs were emitted from P2 in 2016. Except for the MTs mentioned above, alloocimene, limonene, myrcene, p-cymene, and 3-carene were also observed. The average STD emission of MTs was 0.281 (±0.222) µg g dw −1 h −1 . Like in 2015, ocimene dominated over the other MTs in 2016 but with higher measured (0-2.930 µg g dw −1 h −1 ) and STD emission rates (0-4.277 µg g dw −1 h −1 ). The measured emission peaked in June and the STD emission peaked in July. Instead of α-pinene, linalool was the second most abundant MT in 2016. The highest measured (0-0.206 µg g dw −1 h −1 ) and STD (0-0.335 µg g dw −1 h −1 ) emissions of linalool occurred in May, and together with ocimene they contributed with more than 97% to the STD MT emission in May. Throughout the rest of the season in 2016 and until September, they contributed with ca 95-99% to the average STD MT emission. Myrcene and allo-ocimene were only seen in June. The STD MT emission during 2016 was highest in May (0.590 ± 0.306 µg g dw −1 h −1 ) followed by June (0.488 ± 0.330 µg g dw −1 h −1 ), and lowest in October (0.022 ± 0.028 µg g dw −1 h −1 ). This pattern was also observed for the measured average total MT emission, which is in line with the MT emission trend in 2015.

Sesquiterpenes
The average STD emission of SQTs from P1 in 2015 was 0.082 (±0.060) µg g dw −1 h −1 . Caryophyllene had the highest emission rate among the SQTs in 2015 and the STD emission varied between 0 and 0.478 µg g dw −1 h −1 . The highest STD emissions occurred in September and caryophyllene was the only emitted SQT this month ( Figure 1E and Table S3). The measured SQT emission peaked in the second half of August. α-farnesene had the second highest average emission rate among the SQTs during 2015 and was emitted in July and August. Measured emissions of α-farnesene (0-0.242 µg g dw −1 h −1 ) were highest in August but STD emissions (0-0.193 µg g dw −1 h −1 ) peaked already in late July. The average STD SQT emission from P2 in 2016 was 0.012 (±0.018) µg g dw −1 h −1 and the same SQTs were observed this year as during 2015. August was the month where both the STD emission (0.054 ± 0.046 µg g dw −1 h −1 ) and the measured emission (0.084 ± 0.057 µg g dw −1 h −1 ) peaked during 2016. Caryophyllene dominated between June and August and the STD emission varied from 0 to 0.215 µg g dw −1 h −1 . It contributed with ca 64-96% to the STD SQT emission during these months. α-farnesene was emitted from July to September with STD emissions between 0 and 0.345 µg g dw −1 h −1 . It was the only SQT observed in September. Humulene was seen from June to July, where the STD emission ranged from 0 to 0.043 µg g dw The average emission of other VOCs in 2016 was 0.279 (±0.049 µg g dw −1 h −1 ). The majority of the highest other VOC emissions were observed during July and especially during mid-July. The emissions of nonanal and benzaldehyde varied from 0 to 1.060 µg g dw −1 h −1 and 0 to 0.701 µg g dw −1 h −1 in July, respectively, but during the other months they were usually less than 0.200 µg g dw −1 h −1 . Emissions of 2-ethylhexanol also peaked in July (0.974 µg g dw −1 h −1 ). Octanal peaked in July but higher emissions were seen in both May and July (0-0.441 µg g dw −1 h −1 ) compared with the other months (0-0.089 µg g dw −1 h −1 ). Tetradecane was only observed in July and August (0-0.034 µg g dw −1 h −1 ), where it peaked in August, and pentadecane was seen in July, August, and September (0-0.072 µg g dw −1 h −1 ).

Height Comparison with Branch Chamber Measurements in 2016
The average STD isoprene emission from the middle canopy height z M (33.6 ± 28.2 µg g dw −1 h −1 ) did not significantly differ from the upper canopy height z H (26.7 ± 24.1 µg g dw −1 h −1 ). The largest difference between z M and z H for the STD isoprene emission was observed in August (Tables 4 and 5). Due to P. vulgatissima, leaves at z H were more damaged than at z M in August, which could be one explanation for the different STD isoprene emissions.
Comparing MT emissions between the heights showed that z H emitted higher rates of MTs than z M . The average STD MT emission from z H was 0.509 (±0.294) µg g dw  (Tables 4 and 5). In October, it was difficult to find any intact leaves at z H ( Figure S3). Consequently, the leaves at z H were not able to emit any observable MTs in October. The number of emitted MT compounds from z H was also larger compared with z M , which did not have emissions of myrcene and allo-ocimene. Ocimene was the dominantly emitted MT from both z M (0-91.9%) and z H (0-96.8%), and linalool also contributed substantially to the MT emission for z H (0-81.7%). Limonene (0-50.0%), p-cymene (0-33.3%) and 3-carene (0-25.0%) only had a significant contribution for z M . The only MTs that were observed in October (α-pinene, 3-carene, and limonene) were emitted from z M .  The STD SQT emission from z H (0.019 ± 0.023 µg g dw −1 h −1 ) was almost three times larger than from z M (0.007 ± 0.018 µg g dw −1 h −1 ; p < 0.05). The same SQTs were observed at both heights (α-farnesene, caryophellene, and humulene). Sesquiterpenes were seen from z H in June, July, and September, but only in August from z M (Tables 4 and 5) . Caryophyllene was the dominantly emitted SQT for z H (0-90.0%) and z M (0-65.0%). The number of other VOCs from z H (20) was almost the same as from z M (22) but the average emission from z M (0.347 ± 0.061 µg g dw −1 h −1 ) was more than twice as high compared with z H (0.159 ± 0.030 µg g dw −1 h −1 ; p <0.05). The highest average emission of other VOCs from z M occurred during July. The compounds that contributed most to this emission in July were benzaldehyde, cyclopentenyl acetylene, 2-ethylhexanol, nonanal, hexanal, and octanal. The average other VOC emission from z H was highest in May followed by July. Nonanal, hexanal, octanal, and acetophenone were the dominant other VOCs from z H during May. The average emission of 2-ethylhexanol was 16 times higher from z M compared with z H . Hexanal, benzaldehyde, and octanal were 3-4 times higher from z M compared with z H . The other VOC emission from z H was considerably lower at the end of the growing season whereas z M still emitted higher amounts during September and October.
In particular, the response to PAR differed between the height levels under lowlight conditions for measured isoprene emissions ( Figure 4A). For instance, the slope between 0 and 150 µmol m −2 s −1 for z H was 0.140 µg g dw −1 h −1 /(µmol m −2 s −1 ), while it was 0.058 µg g dw −1 h −1 /(µmol m −2 s −1 ) for z L (Table S5). This fast response resulted in isoprene comprising more than 90% of the total BVOC emission already at 150 µmol m −2 s −1 for z H ( Figure 5B). When comparing the STD isoprene emissions between z L and z H , then the slope of the fitted curves differed less for low PAR values ( Figure S4 and Table S6).  Table S7. The ratio between isoprene emission and net assimilation (E/A, dimensionless) showed that z H emitted more of the synthesized C as isoprene relative to the assimilated C, compared with z M and z L ( Figure 4B). In particular, z H had the highest E/A ratio when PAR was 1000 µmol m −2 s −1 and more than 0.6% of the C was emitted as isoprene. For z M and z L , this value was less than 0.4% and 0.3%, respectively. z L was the only height level that increased the E/A ratio from 1000 µmol m −2 s −1 (ca 0.30%) to 1500 µmol m −2 s −1 (ca 0.46%).
None of the observed other VOCs from the leaf chamber measurements showed a clear relationship with light. However, the emissions of a few compounds seemed to increase or decrease for certain PAR values. The average emission of 2-methylbutane increased from 0.145 (±0.148) µg g dw −1 h −1 to 0.226 (±0.239) µg g dw −1 h −1 when PAR varied between 0 and 300 µmol m −2 s −1 and decreased thereafter ( Figure 6B). The average emission of 2-ethylhexanoic acid ranged between 0.293 and 0.641 µg g dw −1 h −1 for the PAR values 0-450 µmol m −2 s −1 , but the opposite trend was seen for 2-pentanone, the average emission of which decreased from 0.330 to 0 µg g dw −1 h −1 when PAR increased from 150 to 1000 µmol m −2 s −1 .
Transpiration at z H was higher than at z M and z L when there was light ( Figure 7B). Stomatal conductance at z H exceeded g s at z M and z L at light intensities above 600 µmol m −2 s −1 ( Figure 7C). The lower height had the lowest g s response to PAR but was mainly in between z H and z M for Tr. Water use efficiency was similar for z H and z L when PAR varied between 0 and 450 µmol m −2 s −1 ; thereafter, z H increased to ca 6 mmol CO 2 mol −1 H 2 O while z L was more or less constant and remained below 5 mmol CO 2 mol −1 H 2 O ( Figure 7D). On the other hand, WUE for z M was approximately twice as high as for z H and z L for 450-1500 µmol m −2 s −1 . Table 6. Mean values for net assimilation (A, µmol CO 2 m −2 s −1 , n = 3-6), transpiration (Tr, mmol H 2 O m −2 s −1 , n = 3-6), water use efficiency (WUE, mmol CO 2 mol −1 H 2 O, n = 3-6), and stomatal conductance (g s , mol H 2 O m −2 s −1 , n = 3-6) at 1500 µmol m −2 s −1 for different height levels. Numbers in parentheses are SD.

Isoprene Emissions, Net Assimilation, and Water Use Efficiency
As expected, isoprene was the dominant compound throughout the growing season and made up 90-99% of the total BVOC emission, with measured emission rates reaching up to 150 µg g dw −1 h −1 . The STD isoprene emission rate averaged over the whole study (45.2 µg g dw −1 h −1 ) is comparable to what has been shown in other studies [13,58,59]. However, the reported range of STD isoprene emission rates is wide [13,59]. In Morrison et al. [59], the STD isoprene emission for unknown Salix spp. varieties varied from 0.1 to 15.9 µg g dw −1 h −1 for three different sites in the United Kingdom. In their study, a similar method was used as in this study but since they included the whole year, a lower emission can be expected on account of the low temperature and PAR during the non-peak season. Moreover, the age of the measured trees was not the same in their study (first, second, and third growing season) as in this study (third and fourth growing season). Another study done close to our site reported a ca 6 times lower STD isoprene emission from clones that originated from the same species (S. viminalis) as was measured in this study, during peak summer [44]. In their study, a relaxed eddy accumulation technique scaled by foliar mass density was used, which is a different method compared with the methods used in this study and could therefore explain the disparity. In particular, this method includes measurements from the whole canopy, i.e., both sun-exposed and shaded leaves, whereas most leaves in this study were adapted to sunlit conditions. Additionally, emission rates vary within the same species [13], which can explain why the average STD isoprene emission in 2015 was twice as high as in 2016, despite similar T C and PAR C , since the plantations contained different varieties. According to Karlsson et al. [42], it has been shown that Salix varieties growing on SRC fields emit various amounts of isoprene and that the Tora-variety growing on P2 is a low emitter.
The seasonal patterns for isoprene emissions in the two Salix varieties we measured in 2015 and 2016 were similar. The highest emission rates typically occurred when weather was warm and sunny. The measurements in 2016 showed that the trees were able to emit high rates of isoprene already in the beginning of May. At this point, the weather had started to become warmer, which resulted in the trees developing small leaves with a light green color. It is well known that isoprene emission is influenced by temperature because of the linkage to the enzyme isoprene synthase [60,61]. In September, measured and STD isoprene emissions during 2015 and 2016 started to decline compared with the former months. In October 2016, isoprene emissions were negligible. The reason for the declining trend is mostly explained by changing weather (2015), and the approaching fall and leaf senescence [62,63]. In particular, all isoprene emissions were <3.0 µg g dw −1 h −1 and most of the samples contained no isoprene during October. Zero emissions of isoprene occurred when T C was ca 11 • C or lower, and PAR C <110 µmol m −2 s −1 , indicating that these values could be a threshold for isoprene emission during late season for this location and variety. These growing conditions in combination with the leaf damage caused by P. vulgatissima explain why the isoprene emissions from z H were exceptionally low in October since many leaves at z H were almost gone and more or less only the midrib was left ( Figure S3).
The Melampsora infestation on P1 likely affected the emission rates. Toome et al. [64] showed that willow trees infected with Melampsora reduced their isoprene emission rates by almost 30% and a similar result was observed for Populus [65]. Perhaps, the isoprene emissions would have been even higher in 2015 if no Melampsora outbreak had occurred.
Measured and STD isoprene emissions from the more sunlit leaves (z H ) exceeded emission rates from shaded leaves (z L ) as expected [45,66,67]. Previous studies have reported a hyperbolic curve for isoprene when PAR increases [55,68,69], which can be linked to the photosynthetic electron transport [69,70], and a similar behavior was observed in this study. Isoprene emissions from sun-adapted leaves at z H responded faster to increased light, and reached twice as high maximum emission rates. This result is in line with Sharkey et al. [66], who studied isoprene emissions from oak and aspen, and it also points out that large errors can be expected if up-scaled emissions are based on data from only one canopy height [45].
The isoprene emission increased exponentially with temperature but it is hard to state if the optimum temperature was reached in our study. In a subarctic Salix myrsinites, isoprene emission increased exponentially across the temperature range of 10-38 • C without reaching a maximum [71]. According to Niinemets et al. [72], the optimum often occurs around 40 • C, suggesting that this could be the same for the varieties in this study. Nevertheless, measurements at higher temperatures are needed to confirm this.
The net assimilation was higher for z H compared with z L, but lower than for z M . An increasing A pattern with canopy height has been reported previously [73,74]. Since also Tr from z M was below the others, this resulted in a higher WUE for z M , making these leaves more efficient in their water use than the leaves at z L and z H . The E/A ratio at z H exceeded that at z L , which means that the light-adapted leaves in the upper part of the canopy lose more C as isoprene than less light-adapted leaves. However, despite the lower isoprene emission rates from the shaded leaves at z L , they showed a strong increase (ca 50%) in their E/A ratio when PAR increased from 1000 to 1500 µmol m −2 s −1 . So, even for leaves less acclimatized to sunlight, isoprene serves as a strongly induced compound when the leaves are light-stressed. The leaves that were more sun-adapted (z H and z M ) did not show this pattern.
Measurements of A and simultaneously emitted BVOCs showed that the trees only used a minor fraction (0.4%) of photosynthetic C for BVOC emission, although compounds with less than five carbon units (i.e., methanol, acetone) were not detectable with the methods used in this study. This ratio is comparable with the results from a study conducted during summer on a spruce forest, where the assimilated C loss on average was approximately 0.3% [75]. For isoprene, the ratio between emitted C and A was similar to what had been found for some species of poplar [76], which are also known to be high isoprene emitters. Only a small percentage of the net carbon assimilation is in general released as BVOCs [77][78][79][80] but under stressed conditions, carbon loss can be 7-8% due to increased isoprene emissions [68,81].
The maximum A for the sunlit leaves is in line with results in other studies, which suggest that A for Salix trees can vary between 10 to 35 µmol CO 2 m −2 s −1 [42,82]. The WUE values for z L and z H were comparable with the values for Tora reported in Karlsson et al. [42].
The variety Tora in Karlsson et al. [42] had a ca 80% lower STD isoprene emission compared with Tora on P2 when comparing the leaf chamber measurements for sunadapted leaves. This difference is probably due to the fact that the trees were younger (first and second growing season) in Karlsson et al. [42].

Emissions of Monoterpenes, Sesquiterpenes, and Other Volatile Organic Compounds
The second most emitted compound was ocimene and together with α-pinene, they contributed with more than 95% to the MT emission throughout 2015. In 2016, ocimene dominated over the other MTs, followed by limonene, p-cymene, and linalool. Even if the emission rates of the MTs varied individually during the season, both 2015 and 2016 showed the same trend, where the measured average total emission of MTs decreased across the measurement periods, but occasional higher emissions were still possible in late summer and autumn. Emissions of MTs can be dependent on the season and willow leaves have shown a stronger capacity to emit MTs in the beginning of the growing season close to bud break [59,89]. However, we cannot verify this suggestion as we did not measure in early spring. The most abundant SQTs were caryophyllene, humulene, and α-farnesene. The emissions of these compounds peaked in the second part of the growing season between July and September. No SQTs could be observed during May and October.
Ocimene and α-farnesene were the only terpenes, besides isoprene, that were influenced by PAR. Ocimene was also found to increase with light by Karlsson et al. [42]. Many studies have shown a correlation between terpenes and light [42,[90][91][92]. This similarity with isoprene shows that emissions of some MTs and SQTs could be described by a similar light-and temperature-dependent algorithm developed by Guenther et al. [55] for short-term emissions.
Shade-adapted leaves at z L emitted less MTs and had lower emission rates than the sun-adapted leaves at z H but the emissions from z M were higher than z H in May and October. The only month when the SQT emission from the lower height (z M ) exceeded that from the higher height level (z H ) was in August. Neither ocimene nor α-farnesene were emitted from z L , which suggests that the leaves require an adaption to an environment with sufficient sunlight to be able to emit these terpenes. Monoterpenes and SQTs are very reactive and protect the plant against different abiotic and biotic stresses [49,[93][94][95][96][97][98][99][100].
Emissions of other VOCs were larger from the lower canopy height compared with the higher canopy height, but it is difficult to pinpoint the underlying reason. If the leaves were more attractive to leaf-eating insects at the upper canopy, one would expect higher emission rates from this level as a self-defending mechanism [94,106]. On the other hand, the higher emission rates from the lower canopy could be the reason why these beetles strike the upper part. However, one explanation for the difference might be related to how the leaves looked. In July and August, the leaves at the lower height were smaller with a brighter green color compared with the higher height. Further, the damage by P. vulgatissima is believed to have reduced the emission from the upper part of the canopy, especially in October.
Most studies have only focused on isoprene and a few other compounds, which usually are most abundant [43,59,107,108]. Even if isoprene is dominant and crucial for air chemistry processes, other compounds might also be of importance. Monoterpenes constitute a group responsible for particle formation and generation of SOAs [38,109]. Additionally, stress-induced compounds are usually not included (yet) in modeling processes, but biogenic non-terpenoids can contribute with the same magnitude as anthropogenic sources of non-terpenoids to SOA formation and should not be discarded [104].

Summary and Conclusions
We studied seasonal trends and canopy height differences for BVOC emissions of willow varieties growing as bioenergy crops on SRC fields. The majority of the measurements in 2015 were done on an unknown variety belonging to S. viminalis, while all measurements in 2016 were on the variety Tora. The results from this study confirm that isoprene was the most dominant BVOC during the major part of the growing season. Emissions of MTs, SQTs, and other VOCs were in general only a minor fraction of the total BVOC emission. The prevailing infestation of Melampsora (2015) was probably responsible for increased emissions of, e.g., nonanal, hexanal, and caryophyllene.
The emissions of isoprene peaked during summer (in July 2015 and in August 2016) but the leaves had high emissions already in May. The seasonal pattern for SQTs showed that these peaked during summer as well. Emissions of MTs showed a decreasing trend from May to October and the overall terpenoid emission seemed to be substantially lower by the end of the growing season as a consequence of colder weather and proceeding leaf senescence.
Our study suggests that Tora is a better choice when it comes to having a reduced risk for impaired air quality since it emitted ca 50% less isoprene, which can be a source for O 3 and peroxyacetyl nitrate. Nevertheless, Salix trees are high emitters of isoprene compared with commercial crops, and an expansion of Salix plantations would need to be strategically planned and placed at a sufficiently long distance away from pollution sources to avoid elevated photochemical production of O 3 under high NO x conditions. Leaves higher up in the canopy and acclimatized to more sunlight emitted higher rates of isoprene than the leaves growing in the lower and more shaded parts of the canopy. Emissions of MTs were also lower for the leaves in the shaded conditions and no SQT emissions were observed from the shade-adapted leaves. On the contrary, emissions of non-terpenoids were higher from the lower part of the canopy. These results point out that the location within the canopy is one important factor when it comes to emissions of different compounds, particularly those compounds that are dependent on PAR. Hence, to make better estimations of the regional and global BVOC fluxes, models need to adjust for, e.g., the vertical distribution of leaf area within the canopy layer.
Ocimene and α-farnesene were the only compounds, except isoprene, that were influenced by light, suggesting that they can be modeled with a similar algorithm as isoprene.
This study highlights that scaling-up BVOC emissions and basing models on one simplified emission potential for the whole canopy could lead to large errors.
Supplementary Materials: The following are available online at https://www.mdpi.com/article/10 .3390/atmos12111427/s1, Figure S1: Photos of the two plots. Figure S2: Photos of Melampsora-infested leaves. Figure S3: Photos of leaves in October. Table S1: Injected standards in GC-MS. Table S2: Average emission values for all detected compounds. Table S3: Average emission, T C and PAR C for each month of 2015. Table S4: Parameter values for fitted isoprene curves in Figure 3. Table S5: Slope and parameter values for fitted curves in Figure 4A. Figure S4: STD isoprene emission vs. PAR. Table S6: Parameter values for fitted isoprene curves in Figure S4. Table S7: Parameter values for fitted E/A curves in Figure 4B. Figure S5: Total average BVOC emission and fraction vs. PAR at z M . Table S8: Parameter values for fitted emission curves in Figure 6A,B. Table S9: Slope values for A in Figure 7A. Table S10: Parameter values for fitted A curves in Figure 7A.  Data Availability Statement: Data from this study is available from the corresponding author upon reasonable request.