Interaction between isoprene and ozone fluxes in a poplar plantation and its impact on air quality at the European level

The emission of isoprene and other biogenic volatile organic compounds from vegetation plays an important role in tropospheric ozone (O3) formation. The potentially large expansion of isoprene emitting species (e.g., poplars) for bioenergy production might, therefore, impact tropospheric O3 formation. Using the eddy covariance technique we have simultaneously measured fluxes isoprene, O3 and of CO2 from a poplar (Populus) plantation grown for bioenergy production. We used the chemistry transport model LOTOS-EUROS to scale-up the isoprene emissions associated with the existing poplar plantations in Europe, and we assessed the impact of isoprene fluxes on ground level O3 concentrations. Our findings suggest that isoprene emissions from existing poplar-for-bioenergy plantations do not significantly affect the ground level of O3 concentration. Indeed the overall land in Europe covered with poplar plantations has not significantly changed over the last two decades despite policy incentives to produce bioenergy crops. The current surface area of isoprene emitting poplars-for-bioenergy remains too limited to significantly enhance O3 concentrations and thus to be considered a potential threat for air quality and human health.

ecosystem-level measurements of isoprene and O 3 fluxes by using the eddy covariance technique and investigated the influence of [O 3 ] on both isoprene emission and total O 3 uptake.
In addition, we have scaled-up the measured isoprene emissions through the chemistry transport model (CTM) LOTOS-EUROS associated with the current surface area planted with poplar-for-bioenergy in Europe, as reported by the Food and Agriculture Organization of the United Nations 17 , and by the European Biomass Association 18 . This allowed us to quantify the potential impact of the isoprene fluxes on ground-level [O 3 ].

Results
Experimental observations. Measured isoprene fluxes showed a well-defined daily cycle with maximum emission rates during the afternoon, from midday till 16.00 CEST (Fig. 1b), and corresponding with maximum values of incoming short-wave radiation (Rg) (Fig. 1a), while surface temperature (S t ) peaked about four hours later (Fig. 1a). Total O 3 uptake (Fig. 1b) also showed a diurnal trend that mirrored the isoprene fluxes and also occurred in parallel with increasing [O 3 ] (Fig. 1c). Uptake of O 3 was dominated by the stomatal component (Supplementary information Fig. S3) that represented 70% of the total O 3 uptake. When looking at the entire season, isoprene fluxes were characterized by a remarkable peak emission occurring for a few days in August with the highest values reaching 38.6 nmol m −2 s −1 on 18 August 2012 and 38.0 nmol m −2 s −1 on 19 August 2012 (Fig. 2d). During these days (Fig. 2, shaded area), an increment of about 10 °C (around midday) in both S t and Air t (Fig. 2a), an increase in the fluxes of latent heat (LE) (Fig. 2b) and of stomatal O 3 uptake (Fig. 2d), as well as a simultaneous decrease in the fluxes of sensible heat (H) (Fig. 2b) were observed. Despite these changes, Rg (Fig. 2a), gross primary production (GPP) and net ecosystem exchange (NEE) (Fig. 2c)    Average daily variation measured during August 2012 of (a) net radiation, air temperature and surface temperature; (b) energy exchange as latent heat (LE) and sensible heat (H); (c) CO 2 flux as net ecosystem exchange (NEE) and gross primary production (GPP); and (d) isoprene emission and total O 3 uptake. The shaded area represents the days characterized by a peak of isoprene emission.  Isoprene fluxes obtained using the basic configuration of the LOTOS-EUROS chemistry transport model with a resolution of 0.5 × 0.25 degrees and the land use database CORINE 2006 (ID run named "basic" in SI Table 2) showed that the largest emissions were associated with isoprene emitting forests (Fig. 3a). The north-south gradient in isoprene emission rates reflected the temperature gradient across Europe, as southern Mediterranean areas feature the highest temperatures with respect to Northern ones. After incorporating into the LOTOS-EUROS model the areas covered by poplar plantations as reported by the FAO for the different countries in Europe (Table S3) and selecting the vegetation-specific isoprene emission factor for poplar plants (ID run "poplar" in SI Table 2) isoprene fluxes increased from 0.01% to 11.6% compared to the "basic" run. However, when a vegetation-specific isoprene emission factor obtained from the present experimental observations was applied to LOTUS-EUROS (ID run named "poplar emission factor" in SI Table 2), total isoprene emissions were 26% lower compared to the LOTOS-EUROS basic configuration and a maximum increase of 2.4% of the isoprene emissions was obtained (Fig. 3b). Overall, the simulated isoprene fluxes increased from 3.4 to 3.5 Tg C for the period of April-October 2012 when the present surface area of poplar plantations was included in LOTOS-EUROS model by using the vegetation-specific isoprene emission factor obtained in our observations.

Variable
In this study, the isoprene concentration pattern (Fig. 4a) mapped the actual distribution of isoprene emissions in Europe: indeed very low concentrations (< 1 ppb) were displayed in north-western Europe whereas in southern Europe concentrations

Discussion
The highest isoprene fluxes observed in this study occurred in mid-August 2012 (Fig. 1), when isoprene synthase activity was maximized by the increase of the air and surface temperatures under high light intensities in a fully developed canopy where strong isoprene emitting adult leaves outweighed the young ones 19 . As the poplars were not water stressed 20 during the whole growing season, they kept their stomata open and increased the fluxes of evapotranspiration in an attempt to reduce the leaf temperature, although the cooling effect due to isoprene emission was negligible (as it accounted for only 0.1-0.3% of the cooling effect of water). Because the amount of Rg did not change (during the measurements period), fluxes of H decreased simultaneously to increased LE, causing the observed reduction of the Bowen ratio. However, GPP and NEE were not affected by this thermal variation; therefore poplar plants emitted isoprene at high rates in the field when the photosynthetic machinery resulted undamaged by stress. This evidence further suggests that isoprene makes the chloroplasts membranes more resistant and more performing under high temperatures as observed in the past 12 , also in laboratory trials 21 . Indeed, experiments employing Populus grown in pots 22 have evidenced the role of isoprene in unstressed plants, as the chloroplasts of isoprene emitting plants dissipated less energy as heat than the chloroplasts of non-emitting plants, when exposed to physiologically high temperatures (29-38 °C) that did not damage the photosynthetic apparatus. While the process of isoprene emission is crucial for tropospheric O 3   a laboratory experiment 15 , which showed that the emission of isoprene and oxygenated six-carbon (C6) volatiles were inhibited when plants were exposed to an [O 3 ] of 80 ppb for two weeks. However, the impact of [O 3 ] on isoprene emission has been investigated in several other laboratory studies with conflicting results: some studies highlighted that high doses of O 3 enhanced isoprene emission 15,23,24 , whereas others [25][26][27]  The maximum effect of isoprene emission on [O 3 ] was calculated using the highest emission factor (60 μ g gDM −1 hr −1 ) regarding isoprene emission from poplar trees reported in the literature, where it was found to range between 45 and 60 μ g gDM −1 hr −1 18,29,30 . However, both the isoprene emission factor and leafy biomass derived from our experimental observations provides a 26% lower total isoprene emission than that resulted from LOTOS-EUROS model in the basic configuration. A misrepresentation of the seasonal variability of isoprene emissions in LOTOS-EUROS caused an underestimation of observed isoprene fluxes in September. This bias influenced O 3 production in the model, but since peaks of [O 3 ] mainly occured before September 2012, the impact on our conclusions is expected to be small. The observations presented here are representative for the northern temperate climate zone of Europe, which contains the countries with the highest area of poplar plantations (SI Table 3). Since the relationship used in LOTOS-EUROS model to simulate isoprene emissions from poplar plantations on the basis of the observations is temperature-dependent 30 , LOTUS-EUROS produced reliable results for the rest of Europe. On the other hand, the isoprene emission factor representing the agricultural land employed here to run the LOTOS-EUROS model is lower than that used by both Ashworth et al. 31 and Lathière et al. 32 . This means that the impact of poplar plantations on the [O 3 ] simulated in our study may be relatively high considering that isoprene emission factor and leafy biomass quantities of poplars used in LOTOS-EUROS are on the high end of the values reported in literature. However, total isoprene emission resulting from our study was lower than that reported in e.g. Ashworth et al. 6 which accounted for 11.5 Tg C yr −1 while in Beltman et al. 33 it was estimated to be 5 Tg C for the period April-September (2012). The differences in the estimates of isoprene emission between past studies and our present work are caused by differences in domain (in Ashworth et al. 6 plantations expanded much further east) and by a diverse time period of analysis (only summertime has been considered in Beltman et al. 33 , whereas the whole year in Ashworth et al. 6 ). So far, simulations of isoprene fluxes from poplar plantations have mainly used future scenarios [31][32][33] , where the impacts of isoprene emissions on [O 3 ] levels were driven by the assumption of a very large increment in both the area of poplar plantations and in the extent of the emission factor chosen to perform the simulations. Therefore, results from past simulations suggested that isoprene fluxes emitted from the foreseen increasing biomass production in expanding poplar-for-bioenergy plantations could potentially increase the risk of O 3 damage to human health, crops and ecosystems. As a consequence, to counteract the increasing level of tropospheric [O 3 ] might require a large investment for the abatement of NO x 33 . The impact of domestic biomass production for fuel and power generation needed to meet EU targets for 2020 regarding the [O 3 ] is small, but significant (Ashworth et al. 6 ). In particular, the study of Ashworth et al. (2012) 6 assumed that 72 Mha of poplar were to be planted to meet the projected use of biomass as a renewable energy source, leading to a 39% increase in isoprene emissions. Moreover other authors 34 suggested that, especially in eastern Europe, a large proportion of agricultural land might have been converted into biomass forest.
However, despite policies encouraging the use of biomass for energy are already in place 35 , data from the literature confirmed the trend that the area covered by poplar plantations did not show any significant changes over the last 20 years, at least in the countries considered in our study (ref. 17, Table S3).The current area of poplar SRC plantations present in Europe (Table S4) is far too small to create a threat for O 3 formation, leading to only 1% extra isoprene emissions, as demonstrated by our simulation. If anthropogenic emissions of non-methane volatile organic compounds (NMVOC) are underestimated in our model study, which is not unlikely as these emissions are quite uncertain, the impact of isoprene emissions on O 3 formation calculated here could be overestimated. Hence, our study indicates that increasing the production of bioenergy from poplar plantations in Europe would not reduce the air quality in the short-term; therefore policymakers should not be concerned in supporting policies that encourage the planting of bioenergy crops.

Material and Methods
The research site was a poplar (Populus) bioenergy plantation located in Lochristi, East-Flanders (Belgium; 51°06′ 44″ N, 3°51′ 02″ E) at an elevation of 6 m above sea level. The plantation was established in April 2010 with 12 selected clones of Populus deltoides, P. maximowiczii, P. nigra, and P. trichocarpa, at a density of 8000 plants ha −1 on a surface of 18.4 ha. The main environmental conditions and stand characteristics are reported in Table S1. The research site was equipped with two different eddy covariance (EC) systems. One system was used to monitor the CO 2 , latent heat (LE), sensible heat (H) and O 3 fluxes between the ecosystem and the atmosphere and comprised a three-dimensional sonic anemometer (model CSAT3, Campbell Scientific, Logan, UT, USA) to measure the wind speed components, a fast response LOZ-3F O 3 analyzer (Drummond Technology Inc., Ontario, Canada) and a fast response CO 2 /H 2 O infrared analyzer (LI-7000, LI-COR, Lincoln, NE, USA). The LOZ-3F O 3 analyzer is based on chemi-luminescence with Eosin-Y dye circulated continuously through a peristaltic pump in the sample cell, and it measures the O 3 mixing ratio at a 10 Hz sampling frequency. A slow response O 3 analyzer (API 400E, Teledyne Instruments, CA, USA) was used to continuously monitor [O 3 ], which was then compared with the concentrations measured by the LOZ-3F, to determine whether any drifting had occurred in the signal of the LOZ-3F. The API instrument was calibrated every six months by the Flemish Environment Agency (www. vmm.be). The second EC system monitored the fluxes of biogenic volatile organic compounds (BVOCs) and included a sonic anemometer (model USA1, Metek GmbH, Elmshorn, Germany) coupled with a proton transfer reaction "Time-of-Flight" mass spectrometer (PTR-TOF-MS) (Ionicon, Innsbruck, Austria) that measured the volume mixing ratios (VMRs) of BVOCs. The data streams of the anemometer and the PTR-TOF-MS were acquired independently by two different computers and synchronized with dedicated software (NTP, Network Time Protocol, University of Delaware, DE, USA) to an independent external clock through the internet, with an accuracy of< 20 ms. The two EC systems were completely independent, although they were installed very close to each other with a 1 m spatial separation between the inlets of the two sampling lines.
Fluxes of O 3 were calculated using EddyPro (version 5.2.0): the mean lateral and vertical velocity components were forced to zero by a two-component rotation. The maximum cross covariance function (within the 30-min averaging time) was used to determine the lag time for O 3 , which was about 2 s. A frequency response correction was performed according to Moncrieff et al. 36 .
The fluxes of CO 2 and of water vapor were calculated using the EdiRe software (R. Clement, University of Edinburgh, UK; www.geos.ed.ac.uk/abs/research/micromet/EdiRe/).
In order to reduce the burden of PTR-TOF-MS data analysis, the overall raw dataset collected during the measuring campaign was post-processed by the routine programs of Müller et al. 37 to screen for the presence of emitted/deposited fluxes of the most common protonated ions unambiguously found to be related to a BVOC. After post-processing, PTR-TOF-MS data were background corrected by subtracting ambient VOC-free air generated by a gas calibration unit (GCU) (Ionimed, Innsbruck, Austria) and calibrated regularly with the same gas standard (Apel Riemer, USA) during the length of the field campaign to quantify the VMRs of the selected BVOCs. This latter technique has been previously described in detail 11 . In order to standardize the computation of isoprene fluxes with the EC method the EddyPro software was modified into a new customized version that was named EddyVoc. Similarly to EddyPro, the processing routine programmed in EddyVoc masked the raw data with a quality flag to exclude individual spikes, values out-of-range and background calibration periods of the PTR-TOF-MS. Further details about the EC data processing were provided previously (Brilli et al. 11 ).

Estimation of stomatal O 3 fluxes.
Stomatal resistance to water vapor (R sto , Eq. 1) was calculated from the EC measured evapotranspiration using the evaporative/resistance method 38 : where λ is the latent heat of vaporization in air (J kg −1 ), γ is the psychrometric constant (0.065 kPa K −1 ), E L is the transpiration rate (kg H 2 O m −2 s −1 ) after subtracting the evaporative contribution from the soil, c p is the specific heat of air (J kg −1 K −1 ), ρ (kg m −3 ) is the density of dry air, VPD is the vapor pressure deficit at leaf level (kPa), R a and R b (s m −1 ) are the aerodynamic, respectively, laminar sublayer resistances.  3 ], Rg, and Air t and a regression model with only Rg, and Air t . The difference in r 2 between these two models represents the fraction of the variance in the outcome (either isoprene emission or O 3 uptake) that can only be attributed to [O 3 ]. Linear regression models were fitted using the lm() function in the statistical software package R, version 3.1.2.
The CTM LOTOS-EUROS model. The LOTOS-EUROS v1.10 is a 3-D regional chemistry transport model that simulates air pollution in the lower troposphere. For a more detailed description of the model reference is made to Schaap et al. 39 . The model uses a normal longitude-latitude projection and allows the user to specify the resolution and domain within its master domain that encompasses Europe and its periphery. For this work, runs were performed at a 0.5 × 0.25 degree resolution for the European domain (15° E -35° W, 30-60° N) and at 1/8 × 1/16 degree for a domain containing Belgium. The model ceiling is at 3.5 km above sea level and consists of three dynamical layers: a mixing layer and two reservoir layers above. The height of the mixing layer at each time and position is extracted from the ECMWF meteorological data used to drive the model. The height of the reservoir layers is set to the difference between the ceiling and mixing layer height. Both layers are equally thick with a minimum of 50 m. If the mixing layer is near or above 3500 m high, the top of the model exceeds 3500 m. A surface layer with a fixed depth of 25 m is included in the model to monitor ground-level concentrations.
Advection in all directions is handled with the monotonic advection scheme developed by Walcek 40 . Isoprene chemistry and hydrolysis of N 2 O 5 were described by Adelman 41 and by Schaap et al. 42 , respectively. Stomatal resistance is described by the parameterization of Emberson et al. 43 and the aerodynamic resistance is calculated for all land-use types separately. Isoprene emissions were calculated based on detailed information about tree types in Europe: a land use dataset 44 was combined with the distributions of 115 tree species over Europe 29 . During each simulation time step, isoprene emissions, for each grid cell, were calculated as a function of the biomass density and standard emission factor of the species or land-use class, taking into account the growing season of deciduous trees and agricultural crops. Moreover, the role of the local temperature and photosynthetically active radiation were taken into account in the biogenic emissions by following the empirically designed algorithms described by Guenther et al. 30 . Anthropogenic emissions were taken from the TNO-MACC database 45 which is a widely used emission database for air quality modelling. Emission totals and trends reported in the TNO-MACC database are in line with what is reported in other emission databases for Europe 46,47 . The model set-up used here did not contain secondary organic aerosol formation or a volatility basis set approach. The different model runs were analyzed and compared in terms of isoprene emissions, isoprene and O 3 concentrations. Modeled O 3 concentrations were validated using measured concentrations from the European Monitoring and Evaluation Programme (EMEP) monitoring network (www.emep.int). Only background stations situated in rural areas below 700 m height were considered.