Simulating precipitation decline under a Mediterranean deciduous Oak forest : effects on isoprene seasonal emissions and predictions under climatic scenarios

Seasonal variations of Q. pubescens physiology and isoprene emission rates (ER) were studied from June 2012 to June 2013 at the O3HP site (French Mediterranean) under natural (ND) and amplified (+30%, AD) drought. While AD significantly reduced the stomatal conductance to water vapour over the season excepting August, it did not significantly limit CO2 net assimilation, which was the lowest in summer. ER followed a 20 significant seasonal pattern, whatever the drought intensity, with mean ER maxima of 78.5 and 104.8 μgC gDM -1 h -1 in July (ND) and August (AD) respectively. Isoprene emission factor increased significantly by a factor of 2 in August and September under AD (137.8 and 74.3 μgC gDM -1 h -1 ) compared to ND (75.3 and 40.21 μgC gDM -1 h -1 ), but no changes occurred on ER. An isoprene algorithm (G14) was developed using an optimised artificial neural network 25 trained on our experimental dataset (ER + O3HP climatic and edaphic parameters cumulated over 0 to 21 days before measurements). G14 assessed more than 80% of the observed ER seasonal variations, whatever the drought intensity. In contrast, ER was poorly assessed under water stress by MEGAN empirical isoprene model, in particular under AD. Soil water (SW) content was the dominant parameter to account for the observed ER variations, regardless the 30 water stress treatment. ER was more sensitive to higher frequency environmental changes under AD (0 to -7 days) compared to ND (7 days). Using IPCC RCP2.6 and RCP8.5 climate scenarios, SW and temperature calculated by the ORCHIDEE land surface model, and G14, an annual 3 fold ER relative increase was found between present (2000-2010) and future (2090-2100) for RCP8.5 scenario compared to a 70% increase for RCP2.6. Future ER 35 Biogeosciences Discuss., doi:10.5194/bg-2017-17, 2017 Manuscript under review for journal Biogeosciences Discussion started: 9 February 2017 c © Author(s) 2017. CC-BY 3.0 License.

significant seasonal pattern, whatever the drought intensity, with mean ER maxima of 78.5 and 104.8 µgC g DM -1 h -1 in July (ND) and August (AD) respectively.Isoprene emission factor increased significantly by a factor of 2 in August and September under AD (137.8 and 74.3 µgC g DM -1 h -1 ) compared to ND (75.3 and 40.21 µgC g DM -1 h -1 ), but no changes occurred on ER.An isoprene algorithm (G14) was developed using an optimised artificial neural network 25 trained on our experimental dataset (ER + O 3 HP climatic and edaphic parameters cumulated over 0 to 21 days before measurements).G14 assessed more than 80% of the observed ER seasonal variations, whatever the drought intensity.In contrast, ER was poorly assessed under water stress by MEGAN empirical isoprene model, in particular under AD.Soil water (SW) content was the dominant parameter to account for the observed ER variations, regardless the 30 water stress treatment.ER was more sensitive to higher frequency environmental changes under AD (0 to -7 days) compared to ND (7 days).Using IPCC RCP2.6 and RCP8.5 climate scenarios, SW and temperature calculated by the ORCHIDEE land surface model, and G14, an annual 3 fold ER relative increase was found between present (2000)(2001)(2002)(2003)(2004)(2005)(2006)(2007)(2008)(2009)(2010) and future (2090-2100) for RCP8.5 scenario compared to a 70% increase for RCP2.6.Future ER remained mainly sensitive to SW (both scenarios) and became dependent to higher frequency environmental changes under RCP8.5.

Introduction
A large number of Mediterranean deciduous and evergreen tree species produce and release isoprene (2-methyl-1,3-butadiene, C 5 H 8 ).Under non stress conditions only 1-2% of the 5 carbon recently assimilated is emitted as isoprene, while, under stress conditions such as water scarcity, this value can reach up to 20-30% (Q.pubescens, Genard-Zielinski et al., 2014).Although the role of isoprene is still under discussion, it seems likely that C 5 H 8 helps plants to optimise CO 2 assimilation during temporary and mild stresses, especially during the growing and warmer periods (Loreto & Fineschi, 2015).The major role of isoprene in plant 10 defence probably explains its large annual global emissions (440-660 TgC.y -1 , Guenther et al., 2006), forming the largest quantity of all Biogenic Volatile Organic Compounds (BVOC) emitted.Although present in the atmosphere at the ppb or ppt level, isoprene has a broad impact on atmospheric chemistry, both in the gaz phase (especially in the O 3 budget of some urbanised areas) and in the particulate-phase (secondary organic aerosols formation) 15 (Goldstein & Steiner, 2007), hence on the biosphere-atmosphere feedbacks.
It has extensively been shown that, under non-stressful conditions, isoprene synthesis and emission are tightly linked, and mainly dependent on the mere instantaneous light and temperature conditions (Guenther et al., 1991(Guenther et al., , 1993)).By contrast, isoprene emission and synthesis are uncoupled under stress environmental conditions which influence isoprene 20 emissions in a way that is not completely understood and still under debate (Affek & Yakir, 2003;Peñuelas & Staudt, 2010).Whilst some authors highlighted an isoprene emission increase under mild water stress (Sharkey & Loreto, 1993;Funk et al., 2004;Pegoraro et al., 2004;Genard-Zielinski et al., 2014), other studies report the opposite (Bruggemann & Schnitzler, 2002;Rodriguez-Calcerrada et al., 2013;Tani et al., 2011).Furthermore, isoprene 25 emission intensity is also closely related to plant internal factors such as leaf ontogeny in deciduous species.The capacity of young leaves to release isoprene only occurs when a cumulated temperature threshold (degree day) is reached after bud breaking (Grinspoon et al., 1991).Depending on the emitter and the location, this threshold was observed to range from 120 to 210 °C (Salix phylicipholia, Hakola et al., 1998) to 400 °C (Populus tremuloides, 30 Monson et al., 1994).Moreover, after the emission onset, the seasonal variations of the leaf capacity to emit isoprene which can range over several orders of magnitude along the season Biogeosciences Discuss., doi :10.5194/bg-2017-17, 2017 Manuscript under review for journal Biogeosciences Discussion started: 9 February 2017 c Author(s) 2017.CC-BY 3.0 License.(Monson et al., 1994, Geron et al., 2000;Boissard et al., 2001;Petron et al., 2001;Hakola et al., 1998).
Two main approaches have been considered so far to model isoprene emission variations: empirically-based parameterisations to represent the observed emission variations due to environmental drivers, and process-based relationships built on our understanding of the 5 ongoing biological regulation (Ashworth et al., 2013).Although both types of model are adapted for present day global/regional modelling, only the former ones are commonly used for atmospheric applications, especially for air quality exercises for which mechanistic models are far too complex.Although Grote et al. (2014) showed the ability of such models to account fairly well for mild stress effects on seasonal isoprene variations of Quercus.ilex., 10 the high degree of descripting parameters required still represent an obstacle for their broad use in air quality and global scale emission exercises (Ashworth et al., 2013).On the other hand, because the more widely used empirical models (see MEGAN: Model of Emissions of Gases and Aerosols from Nature, Guenther et al., 2006) were developed using measurements made under 'optimum' growing conditions, it is still necessary to broaden their ability to 15 assess isoprene emissions over a larger range of environmental conditions (including stress conditions) and variation frequencies (from high or instantaneous to low or seasonal).Such improvements imply ongoing updates as research studies go along and thereby require more advanced algorithms.In particular, MEGAN model is still struggling in considering the water availability (from precipitation or present in the soil) effects on isoprene emissions.Such a 20 weakness can be particularly detrimental to isoprene emission inventories made over areas covered with a large quantity of high isoprene emitters and subject to frequent drought episodes, like the Mediterranean region.Besides, climate models predict over this area an amplification of the natural drought during summers due to a precipitation reduction which could, locally, reach up to 30% by the year 2100 (Giorgi & Lionello, 2008;IPCC 2013;25 Polade et al. 2014).Due to the strong interactions existing between air pollution over the large Mediterranean urban areas and strong BVOC emissions by local vegetation, the assessement of how global changes could impact isoprene emissions is a growing environmental issue to be adressed (Atkinson & Arey, 1998;Calfapietra et al., 2009;Chameides et al., 1988;Pacifico et al. 2009).In that context, a recent study underlines the 30 importance of long-term monitorings of, both, isoprene emissions and soil moisture in water limited ecosystems (Zheng et al., 2015).Since Quercus pubescens Willd. is the second isoprene emitter in Europe and the main one in the Mediterranean zone (Keenan et al. 2009), Biogeosciences Discuss., doi:10.5194/bg-2017-17, 2017 Manuscript under review for journal Biogeosciences Discussion started: 9 February 2017 c Author(s) 2017.CC-BY 3.0 License.
it represents an ideal model species to further investigate the isoprene emission variability under drought conditions.
The objectives of this study were to (i) compare observed seasonal impacts of ND vs AD on Q. pubescens gaz exchanges (CO 2 , H 2 O) and isoprene emission rates (ER) at the O 3 HP site, (ii) develop a specific isoprene emission algorithm taking into account drought (ND and AD) 5 impacts on Q. pubescens ER, and, (iii) make some projections on the impacts of an AD of 30% on Downy oak ER over the Mediterranean area by the year 2100, under IPCC RCP2.6 (moderate) and RCP8.5 (extreme) scenarios.

Experimental site O 3 HP 10
Experimental data was obtained at the O 3 HP site (Oak Observatory at the Observatoire de Haute Provence, 5°42'44" E, 43°55'54" N).This site is part of the French national network SOERE F-ORE-T (System of Observation and Experimentation, in the long term, for Environmental Research) dedicate to the forest ecosystem functioning.The O 3 HP site (680 m above mean sea level) is located 60 km north of Marseille and consists of a homogeneous 70 15 year-old forest dominated by Q. pubescens (5 m height, LAI = 2.2) which accounts for  90% of the biomass and 75% of the trees.O 3 HP facilities, in particular the rain exclusion system dynamically reducing precipitation by a deployable roof above the canopy, enabled to study this ecosystem under natural and amplified drought (averaging -30% of the annual cumulated precipitation), hereafter named 'ND' and 'AD' plot respectively.Ambient and soil 20 environmental parameters were continuously monitored with a dense network of sensors (for details see section "COOPERATE environmental data base").Access to the canopy was at two levels:  0.8 and 4 m (top canopy branches) above ground level, with the higher level being the focus of this study.Further description can be found in (Santonja et al. 2015).

Seasonal sampling strategy 25
Isoprene emission rate measurements were performed at least during one week, once a month, from June 2012 to June 2013, except from November 2012 to March 2013 when Q. pubescent is fully senescent with leaves remaining on the tree (marcescent species).This calendar allowed us to capture isoprene emissions during leaf maturity but also during bud break April 2013 when tree-to-tree and within-canopy variability was assessed.One ND branch was sampled throughout all intensive campaigns, while the 5 other ND and AD branches were 5 alternatively sampled during 1 to 2 days (Genard-Zielinski et al., 2014).Isoprene samples were collected on cartridges packed with adsorbents, except in April 2013 where on-line isoprene measurements were performed using a PTR-MS directly connected to the enclosure via a 50 m length 1/4" PTFE line.When cartridges were used, samples were taken from sunrise to sunset, roughly every 2 hours.PTR-MS measurements allowed a higher sampling 10 frequency (between 120 and 390 s -1 ).
Branch enclosures were mostly installed on the previous day before the first emission rate measurement took place, and, at least, 2 h before.

COOPERATE environmental database
Ambient and edaphic parameters used for the artificial neural network (ANN) optimization 15 were obtained from the COOPERATE database (https://cooperate.obs-hp.fr/db)and daily averaged for each day of our study.Ambient PAR (µmol m -2 s -1 ) measured above the canopy at 6.5 m (Licor Li-190®; Lincoln, NE, USA) in the ND plot was used as the PAR reaching all the top canopy branches studied.Ambient air temperature (T, °C) measured at 6.15 m (multisensor Vaisalla) in the ND and AD plot was used for both sets of branches.Since some 20 precipitation (P, mm) values were missing from the COOPORATE database during our data processing, P values from the nearby (< 10 km) Forcalquier meteorological station were used.
Soil water content (SW, L L -1 ) and temperature (ST, °C) at -0.1 m (Hydra Probe II SDI-12, Stevens, Water Monitoring Systems Inc., OR, USA) specific for each of the sampled trees were selected and extracted from the COOPORATE database.When data were missing, they 25 were extrapolated from the nearest equivalent data point measurements.Daily mean PAR, T, P, SW and ST were cumulated over a time period ranging from 1 to 21 days before the Sampling was carried out using two identical dynamic branch enclosures (detailed description in Genard-Zielinski et al., 2014).Briefly, the device consists of a  60 L PTFE (PolyTetraFluoroEthylene) frame closed by a sealed 50 µm thick PTFE film to which ambient air was introduced at Q 0 ranging between 11-14 L min -1 using a PTFE pump (KNF N840.1.2FT.18®,Germany).Gas flow rates were controlled by mass flow controllers 5 (Bronkhorst) and all tubing lines were PTFE-made.A PTFE propeller ensured a rapid mixing of air inside the chamber.Microclimate (PAR, T, relative humidity) inside the chamber was continuously monitored (relative humidity and temperature probe LI-COR 1400-04®, and quantum sensor LI-COR, PAR-SA 190®, Lincoln, NE, USA) and recorded (Licor 1400®; Lincoln, NE, USA).CO 2 /H 2 O exchanges from the enclosed branches were also continuously 10 measured using infrared gas analysers (IRGA 840A®, Licor) in order to assess the net assimilation P n (in µmol CO2 m -2 s -1 ) and the stomatal conductance to water vapour G w (mol H2O m -2 s -1 ) using the equations from Von Caemmerer & Farquhar (1981) as detailed in Genard-Zielinski et al. (2015).
Total dry biomass matter (DM) was calculated by manually scanning every leaf of each 15 sampled branch enclosed in the chamber and applying a dry leaf mass per area conversion factor (LMA) extrapolated from concomitant measurements made on the same site.Mean (range) DM was 0.16 (0.01 -0.45) g DM , and mean (range) AF was 13.17 (0.82 -36.67) Isoprene emission rates (ER) were calculated as: 20 where ER is expressed in µgC g DM -1 h -1 , Q 0 is the flow rate of the air introduced into the chamber (L h -1 ), C out and C in are the concentrations in the inflowing and outflowing air (µgC L -1 ) and DM is the sampled dry biomass matter (g DM ).
All over the seasonal cycle, except in April, isoprene was collected using packed cartridges 25 (glass and stainless-steel) prefilled with Tenax TA and/or Carbotrap.Isoprene was then analysed in the laboratory according to a gas chromatography-mass spectrometry procedure detailed in (Genard-Zielinski et al., 2014), with a level of analytical precision better than 7.5%.
In April 2013, additionally to cartridges, two types of PTR-MS were used for on-line isoprene 30 sampling and analysis.A quadrupole PTR-MS (HS-PTR-MS, Ionicon Analytik GmbH, Innsbruck Austria), connected to the ND branch enclosure, was operated at 2.2 mbar pressure, 60 °C temperature and a 500 V voltage in order to achieve an E/N ratio of  115 Td (E: Biogeosciences Discuss., doi :10.5194/bg-2017-17, 2017 Manuscript under review for journal Biogeosciences Discussion started: 9 February 2017 c Author(s) 2017.CC-BY 3.0 License.electric field strength (V cm -1 ); N: buffer gas number density (molecule cm -3 ); 1Td=10 -17 V cm 2 .The primary H 3 O + ion count assessed at m/z 21 was 3 10 7 cps, with a typically < 10% contribution monitored from the first water cluster (m/z 37) and < 5% contribution from the O 2 + (m/z 32).Measurements were operated in scan mode (m/z 21 to m/z 210) every 380 s.
After 15 to 20 min of sampling of incoming air, the outgoing air was sampled for 30 to 60 5 min.A high resolution (m/Δm ≈ 4000) time of flight PTR-MS (PTR-ToF-MS-8000, Ionicon Analytik GmbH, Innsbruck Austria) connected to the second enclosure used in our study enabled to discriminate compounds when their masses differ at the tenth part.The main experimental characteristics were similar to the PTR-MS-Quad, but a 550 V voltage was used in order to reach an E/N ratio of  125 Td.The H 3 O + ion count assessed at m/z 21 was 10 1.1 10 6 cps with a similar < 10% contribution monitored from the first water cluster (m/Z 37) and < 2.5% contribution from the O 2 + (m/z 32).The signal at m/z 69 corresponding to protonated isoprene was converted into mixing ratio by using a proton transfer rate constant k of 1.96.10 -9 cm 3 .s−1 (Cappellin et al., 2012), the reaction time in the drift tube, and the experimentally determined ion transmission efficiency.The relative ion transmission 15 efficiencies of both instruments were assessed using a standard gas calibration mixture (TO-14A Aromatic Mix, Restek Corporation, Bellefonte, USA; 100 ± 10ppb in nitrogen).
Assuming an uncertainty of ±15 % in the k-rate constants and in the mass transmission efficiency, the overall uncertainty of the concentration measurement is estimated to be of the order of ±20 %. Background signal was obtained by passing air through a platinum catalytic 20 converter heated at 300 °C.Detection limits defined as three times the standard deviation on the background signal were 10 and 50 ppt with the PTR-ToF-MS and the HS-PTR-MS respectively.
The overall uncertainty (sampling + analysis) on ER assessment was between 15 and 20%.

Statistics 25
All statistics were performed on STATGRAPHICS® centurion XV by Statpoint, Inc.
Isoprene emission factors (I s ) under each plot were calculated as the slope of the linear regression between C L × C T (factors accounting for light and temperature variability of isoprene emissions respectively, as in Guenther et al., 1993) and measured ER, with the exception of July when no such correlation was found (cf.Results).C L × C T were calculated 30 using PAR and T recorded in the enclosure.Differences in P n , G w , ER and I s between the ND and the AD plot were tested using U-Mann & Whitney tests).Seasonal changes in these Biogeosciences Discuss., doi :10.5194/bg-2017-17, 2017 Manuscript under review for journal Biogeosciences Discussion started: 9 February 2017 c Author(s) 2017.CC-BY 3.0 License.ecophysiological parameters were tested using Krusal-wallis test (K) and the analysis was performed separately on trees from the ND and AD plot.Comparisons between COOPORATE environmental data were made using a Wilcoxon test when data was not lognormal and a t-test when log-normal.

Artificial Neural Network (ANN) 5
The Artificial Neural Network (ANN) developed in this study was based on a commercial version of the Netral NeuroOne software v.6.0 (http://www.netral.com,France) used as a Multi-Layer Perceptron (MLP) in order to calculate multiple non-linear regressions between a set of input regressors x i (the environmental variables measured at the O 3 HP) and the output data y meas (the measured isoprene ER).The assessed ER (y calc ) was calculated as follows: 10 where w 0 is the connecting weight between the bias and the output, N the number of neurons N j , f the transfer function, w 0,j the connecting weight between the bias and the neuron N j , w i the connecting weight between the input and the neuron N j , and x i the n input regressors.The MLP optimisation of the weights w was achieved according to Boissard et al., (2008).Every 15 input regressor x i was centrally-normalised. Two sub-datasets were considered, for the ND and AD plot respectively.For each sub-dataset, 80% of our data were used for training and optimising the MLP, and the overall 20% were used for blind validation based on root mean square error (RMSE).Training/validation splitting was made using a Kullback-Liebler distance function available in NeuroOne v 6.0.Only the nonlinear hyperbolic tangent (tanh) 20 function was tested as transfer function f.Up to N=7 neurons were tested for every ANN setting.Overtraining phenomenon (a too large number of neurons vs the number of input parameters) was checked against the RMSE training /RMSE validation evolution vs the number N of neurons tested.

Future climate data over the Mediterranean area and ORCHIDEE model 25 description
For the present-day climate, T, P and PAR were assessed as the 2000-2010 daily averages and derived from the ISI-MIP (Inter-Sectoral Impact Model Intercomparison Project) climate data set (Warszawski et al., 2014)  Mediterranean region were performed with the ORCHIDEE model at 0.5 × 0.5° spatial resolution, using the soil parameters (clay, silt and sand fractions) from Zobler (1986) and since this study focuses on isoprene emissions from Q. pubescens we fixed the vegetation 20 with the corresponding PFT 'temperate broad-leaf summer green tree'.The above-described ISI-MIP historical forcings and the ISI-MIP future projections were used as climate data.
Equilibrium was reached by running ORCHIDEE on the first decade of the climate forcing ) repeated in a loop, and the value of atmospheric CO 2 corresponding to the year 1961.Among the two different hydrology schemes available in ORCHIDEE the rather 25 complex physically based 11-layer scheme was used (Guimberteau et al., 2013).

Environmental conditions observed at the O 3 HP
Mean daily ambient air temperature T varied between -3 and 26 °C (January 2013 and August 2012 respectively, Fig. 1a).Seasonal PAR variations were in line with T variations, with daily 30 means peaking at 900 µmol m -2 s -1 in July (Fig. 1b).In 2012, the amplification of the ND was

Physiology and isoprene seasonal variations
G w and P n showed similar seasonal patterns in both plots (Figs.2a, 2b), with the lowest values 15 in July-September (10-20 mol H2O m -2 s -1 and 1 µmol CO2 m -2 s -1 respectively) and the highest in June (80-170 mol H2O m 2 s -1 and 9 µmol CO2 m -2 s -1 respectively).Respiration dominated over gross CO 2 assimilation in April, resulting in negative net assimilation (P n  -1 µmol CO2 m -2 s -1 ) in both plots.By contrast, G w and P n were not similarly influenced by water stress.
Whereas G w was significantly reduced under AD since July 2012, P n remained stable, expect 20 in June 2013 when unexplained high P n values twice higher under AD than ND were observed.This higher value was however attributed to only one of the AD branches and was, moreover, unlikely to be due to AD since rain exclusion started in 2013 after June.
Water stress impacted the ER seasonal pattern during summer alone (Fig. 2c).Maximum ER was one month delayed in the AD plot (104.8 µgC g DM -1 h -1 in August) compared to the ND 25 plot (78.5 µgC g DM -1 h -1 in July).ER was the lowest in October ( 6 µgC g DM -1 h -1 in both plots).During April bud-break and isoprene emission onset, ER was as low as 0.5 and 1 µgC g DM -1 h -1 in the ND and AD plot respectively.
Although I s was calculated every month as the slope of ER vs C L  C T (as in Guenther et al., 1993), this correlation was not significant in July, especially for AD branches (P>0.05,R² = 30 0.06 and 0.01 for ND and AD respectively).As a result, I s was assessed in July by averaging ER measured under environmental conditions close to 1000±100 µmol m -2 s -1 and 30±1 °C.In general, AD branches showed poorer ER vs C L  C T correlations than branches growing in the ND plot (data not shown).I s was significantly higher by a factor of 2 in August (P=3.9)and in September (P=3.9) for the AD branches compared to the ND (Fig. 2d).As for ER, I s maximum was reached in August (137.8µgC g DM -1 h -1 ) in the AD plot, while the maximum in the ND plot occurred in July (74.3 µgC g DM -1 h -1 ).The general high variability observed 5 during the isoprene emission onset in April was as large as the AD-ND variability, and, thus, could not be attributed only to the water stress treatment.The annual I s difference between ND and AD relatively to ND was +45%.

Modeling the isoprene seasonal variations
Because we were aiming at testing an isoprene emission model widely employed for air 10 quality applications, the empirical MEGAN model (Model of Emissions of Gases and Aerosols from Nature, Guenther et al., 2006) was tested to assess our observed seasonal and drought variability of ER.A value of 53 µgC g DM -1 h -1 was used for the emission factor I s as given by Simpson et al. (1999) for Q. pubescens, together with a wilting point w of 0.138 m 3 m -3 (Chen & Dudhia, 2001).Some agreement between the calculated and the measured ER in 15 the ND plot was found in June 2012 & 2013, October and April measurements (Fig. 3a), although June 2013 ER were strongly (more than a factor of 5) underestimated by MEGAN.
In contrast, under AD, ER was correctly assessed by MEGAN only in June 2012.During the driest months (July, August, September), under ND and AD, SW recorded at the O 3 HP (0.05 m 3 m -3 , see Fig. 1d) being lower than the w used in MEGAN (0.138 m 3 m -3 ) calculated ER 20 were set to zero.When SW effect was not selected in MEGAN, an overall understimation of 60 and 70% was still found in the ND and AD plots respectively, and no more than 50% of the seasonal variations were captured (data not shown).
Among the different ANN settings tested, the G14 optimised architecture (lowest RMSE, no overtraining, best correlation between measured and calculated ER over the whole range of 25 value, see Boissard et al., 2008) was found for N=3 and a set of 16 x i with their corresponding connecting weights w i (Appendices 1).More than 80% of the ER seasonal variations were assessed by G14, whatever the water stress (ND or AD) and the month, except in July (Fig. 3b) when ER were always poorly represented whatever the different ANN settings considered.30 Biogeosciences Discuss., doi:10.5194/bg-2017-17,2017 Manuscript under review for journal Biogeosciences Discussion started: 9 February 2017 c Author(s) 2017.CC-BY 3.0 License.
Among the environmental regressors used in G14, SW was foundwhatever the frequency considered -to be the dominant parameter to explain the observed ER seasonal variations, especially when Q. pubescens was growing under AD (Fig. 4a).When water stress became significant in July, ER seasonal variations were found to be more sensitive to higher frequency changes in the environmental conditions (0 to 7 days before the measurement), 5 whatever the regressors used, compared to April and June when ER were more sensitive to lower frequencies (14 to 21 days) for both ND and AD (Fig. 4b).On the overall, T and L, whatever their frequencies but in particular when considered instantaneously, did not account for more than 15% in the seasonal ER pattern.Mediterranean 10 Relative changes between present and future climates were averaged on a monthly basis for the climate parameters used in G14 (SW, ST, T, PAR and P) and ER, according to the RCP2.6 and RCP8.5 scenarios (Fig. 5).PAR changes between present and future climates being negligible for both scenarios they are therefore not presented.Among the different G14 parameters, the highest monthly relative changes was found for P (+80% in March, RCP2.6) 15 and ST (+105 and +120% in January and December respectively; RCP8.5, Fig. 5a).

2100 projections over the European
The annual P cum was found to be 1401 and 758 mm according to the RCP2.6 and RCP 8.5 projections respectively (data not shown), leading to an annual P cum relative change of +6% and -24% for RCP2.6 and RCP8.5 respectively (Fig. 5a).Major P relative changes occur in late winter for RCP2.6 (a 50% increase) and during the summer for RCP8.5 (a strong 20 decrease down to -90% in August).SW relative change profiles were found to be in line with

P.
T relative changes were found to be always positive according to both scenarios, with an annual relative increase of 11% (+1.5 °C) and 41% (+5.4 °C) for the RCP2.6 and RCP 8.5 respectively (Fig. 5a).For both scenarios the strongest monthly changes were observed during 25 the winter time (+50% and +87% in December for RCP2.6 and RCP 8.5 respectively), while the summer T/T was of the order of +10 and +40% respectively.The ST relative change profiles were found to be in line with T.
Whatever the projections, dER G14 /ER G14 was found to be relatively more impacted by water availability W (calculated as the sum of relative impact of SW and P) than by temperature 5 effect (calculated as the sum of relative impact of ST and T): 67 and 63% for RCP2.6 and RCP8.5 respectively (Fig. 5d).By contract, temperature contribution became higher (more than 60%) in October and June under RCP2.6 and RCP8.5 respectively.
Whatever the G14 parameters used, the seasonal variations of dER G14 /ER G14 were mainly affected by 14-d frequencies (36.8 %, Fig. 5e), especially during early spring (65%) under the 10 RCP2.6 scenario, while instantaneous variations of the G14 parameters became predominant (30.3 %) for the RCP8.5, and especially during summer (50 %).The lowest frequency (21-d) impact was similar for both scenarios and was close to 10% over the season

Water stress impacts on the seasonal gas exchange of Q. pubescens 15
Despite a significant G w reduction in summer 2012 due to the AD, Q. pubescens maintained a positive P n during all summer, regardless of the water stress (ND or AD).Such behaviour enables trees to limit the evapotranspiration under water stress and, as a drought-acclimated species, to ensure enough carbohydrates accumulation for winter (Chaves et al., 2002).Such strategy was also observed in a study conducted on the same species but under greenhouse 20 conditions (Genard-Zielinski et al., 2014).The seasonal regulation/conservation of P n and G w enabled isoprene emissions to be maintained even during the summer water stress (ND and AD).The observed significant increase (a factor of 2) of the tree capacity to emit isoprene under AD (August and September) illustrates how isoprene is likely to be important for shortterm Q. pubescens drought-resistance, in particular through the ability of isoprene to stabilise 25 the thylakoids membrane, under, for example, a thermal or oxidative stress as shown for Populous species (Velikova et al., 2012).Previous studies highlighted the possibility for a plant growing under a water stress to synthesise isoprene using an alternative carbon source (extra-chloroplastic carbohydrates, Lichtenthaler et al., 1997;Funk et al., 2004).Interestingly, the maximum I s value was very similar for both treatments ( 140 µg.gDM -1 .h - ) but was 30 reached one month later under the AD (August compared to July).These observations The strong uncoupling between ER and C L × C T reported for July measurements occurred 10 when soil water content conditions significantly decreased at the O 3 HP, in both plots, down to their seasonal minimum values (0.05 and 0.03 m 3 m -3 ).Such an uncoupling was also observed for some other strong isoprene emitters under water stress (Quercus serrata Murray and Blume, Quercus crispula, Tani et al., 2011).It confirms the assumption of these authors that extra-chloroplastic isoprene precursors supply the carbon basis for isoprene biosynthesis (and 15 not only from CO 2 fixed instantaneous in the chloroplast) when water stress occurs, explaining why isoprene emissions become less dependent of the classical abiotic factors PAR and T considered in the empirical MEGAN model, or the other environmental abiotic regressors tested in this study.However, our statistical approach used in this study would require a larger set of data to further investigate this particular point.20

Towards what isoprene emission modeling improvement?
Since most of the empirical isoprene emission algorithms of MEGAN have been developed from measurements carried out under 'optimal' (i.e.'none stressing') conditions, they still show difficulties in capturing the impacts of various stress conditions, such as the water deficit, for instance.Hence, ER measured during this study, was observed to be rarely 25 dependent on T and PAR only: indeed, except in June (e.g.just before the ND and AD onset), MEGAN was unable to predict isoprene emission variations.In particular in July, the strong uncoupling between ER and C L  C T illustrates the extent to which other additional parameters than the abiotic parameters T and PAR should be taken into account in isoprene MEGAN algorithm in order to better capture the shift from carbon allocation to growth 30 towards secondary metabolites synthesis during reduced water availability conditions.Biogeosciences Discuss., doi:10.5194/bg-2017-17,2017 Manuscript under review for journal Biogeosciences Discussion started: 9 February 2017 c Author(s) 2017.CC-BY 3.0 License.
Our approach in developing a specific algorithm for Q. pubescens (G14) confirmed the strong impact of the soil water content in the Mediterranean area, and not only under stress conditions.However, the SW effect as considered in the MEGAN isoprene model was found to be not adapted to account for Q. pubescens capacity to resist to water stress: isoprene emissions did not decrease nor cease under ND or AD conditions as expected in summer by 5 MEGAN; on the contrary, they were observed to be maintained, and even increased, during the maximum AD period when SW became lower than the wilting point w.Since the SW availability modulation (1 to 0) used in MEGAN was based on the single observation made by Pegoraro et al. (2004) between Populus deltoids photosynthesis and stomatal conductance to water vapour, it is obviously not appropriate for water stress-resistant species.Such a 10 discrepancy under conditions other than Mediterranean was also noticed by Potosnak et al., 2014) during a seasonal study over a mixed broad-leaf forest mainly composed of Q. alba L. and Q. velutina Lam.(Missouri, USA).Although they found that MEGAN robustly captured 90% of the observed variance during most of the annual cycle, it was unable to reproduce the time-dependent response of isoprene emission to water stress (w of 0.084 m 3 m -3 ).Improved 15 isoprene empirical models should be able to more realistically account for the drought adaptation by the emitter.One way could be, for MEGAN, to add an emitter-dependent parameter in the general isoprene algorithm.Moreover, Guenther et al. (2013) suggested that including soil moisture averaged over longer time periods (such as the previous month and not only the mean over the previous 240h) may help to yield better predictions during drought 20 periods.Note that MEGAN performed better on our experimental data when its SW modulation was set to 1 (no SW effect), since almost half of the isoprene variations were then assessed, whatever the water stress intensity; however a general underestimation of a factor of 2 remained (data not shown).
Our work also highlighted the seasonal change occurring in the regulation frequency of 25 isoprene emissions, especially when stressed environmental conditions appeared.Indeed, Q. pubescens isoprene emissions became more highly sensitive to rapid environmental changes as drought intensity increased: in June 2012, ER G14 variations were mainly controlled by a 14-day frequency (under ND and AD) while they became later in the season mainly 7day, and even 0-to 7-day dependent as soon as the ND and AD, respectively, increased (Fig. 30 4).Moreover, only a small fraction of ER G14 variations were due to instantaneous environmental variations changes (as mostly accounted for in MEGAN).The highlighting of a dynamical regulation over oak seasonal leaf development and of a lower than instantaneous frequency dependency of their isoprene emission was also made for Q. alba and Q.
Biogeosciences Discuss., doi:10.5194/bg-2017-17,2017 Manuscript under review for journal Biogeosciences Discussion started: 9 February 2017 c Author(s) 2017.CC-BY 3.0 License.macrocarpa (Geron et al., 2000;Petron et al., 2001 respectively): isoprene onset was observed to also strongly correlate with ambient temperature cumulated over 2 weeks ( 200to 300 degree day, d.d., °C) while maximum ER was observed at 600-700 d.d.°C.If part of this dynamical regulation is already included in the leaf age emission activity  A,i of MEGAN isoprene emission algorithm, the part due to the impacts of stressing conditions is not yet 5 considered, and not only when the plant is experiencing a drought period.For instance, Wiberley et al. (2005) observed that the onset of kudzu isoprene emissions were shortened by one week under elevated temperature compared to cold growth.
On one side, and as suggested by Zheng et al. (2015), ad-hoc long-term direct measurements of isoprene emission as the one performed at the O 3 HP, are still essential to provide further 10 information for empirical model improvement on how isoprene emissions are seasonally affected by water availability.However, on the other side, representation of soil moisture in land-surface and climate models is currently poor (Köstner et al., 2008).Consequently, further actions need to be undertaken in order to provide a better description of soil moisture at the surface (e.g. the NASA Soil Moisture Active Passive instrument) and thus improve the 15 representation of the effective water available for plants in emission models.

What future impacts of amplified drought on isoprene emissions?
Depending on the future scenario tested, changes in precipitation (thus in soil water content) and/or in temperature will differently affect the Mediterranean area tested in our study.
A net P cum reduction was predicted only under the extreme CO 2 trajectories RCP8.5, with an 20 annual decrease of 24%, similar to the AD applied at the O 3 HP site during our study (-30% of precipitation), together with strong temperature increase (+5.4 °C).On the contrary, the only main change predicted under the RCP2.6 scenario was a moderate temperature increase (+1.5 °C), whereas the annual predicted P cum remained more or less stable (a slight +6% increase, due to heavier rain falls in winter).In terms of AD, the O 3 HP experimental strategy 25 followed during this work illustrated the upper limit of the drought intensity that Q. pubescens should undergo on the 2100 horizon in the Mediterranean area.
The mere RCP2.6 T increase had a similar predicted impact on ER G14 relative change (+70%) than the mere precipitation reduction had on our observed ER (+45%).If an isoprene emission increase is generally predicted and observed in relation with future temperature 30 enhancement (Peñuelas & Staudt, 2010), such a response seems not so clear under Mediterranean water deficit conditions (Llusià et al., 2008(Llusià et al., , 2009)).In our case, when the Biogeosciences Discuss., doi :10.5194/bg-2017-17, 2017 Manuscript under review for journal Biogeosciences Discussion started: 9 February 2017 c Author(s) 2017.CC-BY 3.0 License.combined effects of temperature and drought increase were considered (RCP8.5 scenario), the predicted ER G14 relative changes were enhanced compared to the sum of the individual relative effects of drought and temperature: on the average, ER G14 was multiplied by a factor of 3 (RCP8.5,Fig. 5c).During summer, this enhancement was particularly pronounced when the drought became maximum (August) with ER G14 relative changes nearly 10 times higher 5 under the RCP8.5 than the RCP2.6 (+430 and +46% respectively).Yet, the highest predicted enhancement occurred when the drought period ended, in October, with ER G14 increased by a factor of 8 under RCP8.5.In addition to expected high dER G14 /ER G14 values during the drought period, the main relative change in isoprene emission would actually occur in autumn.This prediction is in agreement with observations made on different emitters in the 10 Mediterranean area where isoprene emissions were observed in autumn to reach levels as high as in spring as soon as the water stress is declining and the net assimilation, hence the isoprene emissions, are favored again (e.g.Owen et al., 1998).Moreover, the relative effect of T on dER G14 was predicted to increase and became even higher (up to 64%, RCP2.6) or similar (RCP8.5,Fig. 5d) to the W relative effect in October, as soon as drought intensity 15 lowered: ER and T variations became, again, highly coupled as in the current isoprene algorithms (e.g.G95, Guenther et al., 1995).However, in June, the relative higher T effect on dER G14 compared to W (60 vs 40%, RCP8.5, Fig. 5d) could illustrate a higher sensitivity of Q. pubescens isoprene emissions to temperature stress as the drought is setting in.Such a coeffect was also observed by Genard-Zielinski (2014) on Q. pubescens branches that were 20 growing in the AD plot.Except in June and October, dER G14 /ER G14 seasonal variations were predicted to become mainly driven by the water budget (more than 65%) compared to temperature (less than 35%), whatever the scenario used.These 2100 projections were made considering an unchanged Q. pubescens biomass, i.e. not affected by long term acclimation to T and drought increase.However, one can question, on 25 the one hand, if Q. pubescens could maintain such a high allocation of its primary assimilated carbon (primary plant metabolites, PPMs) to isoprene emissions (secondary plant metabolites, PSMs).Indeed, for a constant assimilation, the PSMs/PPMs ratio could, on the overall, be multiplied by up to a factor of 3, and could reach a 7 fold increase during the highest drought periods (Saunier et al., under review). Genard-Zielinski et al. (2014) showed that, under 30 moderate and severe drought, Q. pubescens aerial and foliar growth was negatively affected.Furthermore, on a long term, such a cost could impact the overall energy budget and speed up the plant senescence (Loreto & Schnitzler, 2010).Our predicted isoprene emission increases could then be offset, or even reversed.On the other hand, one should also consider the Biogeosciences Discuss., doi:10.5194/bg-2017-17,2017 Manuscript under review for journal Biogeosciences Discussion started: 9 February 2017 c Author(s) 2017.CC-BY 3.0 License.additional co-effects of the CO 2 increase expected a year by 2100.Bytnerowicz et al. (2007) reported that, if the temperature increase would have little effect, an elevated CO 2 would favor both growth and water use efficiency of plants, and account for a 15-20% increase on forest NPP.When CO 2 enhancement was considered, the leaf mass (g) per square meter of the PFT ground (broad leaf temperate) tested in ORCHIDEE during this work was predicted to, 5 relatively, increase by 35 and 100 % under RCP2.6 and RCP8.5 respectively.Tognetti et al. (1998) observed a similar positive effect on the assimilation rate of both Q. pubescens and Q. ilex during a long term CO 2 enhancement study, and measured a net increase in the diurnal course of isoprene emissions.
Understandably, the G14 algorithm should be validated on a longer period of measurements 10 in order to assess how Q. pubescens acclimate over a longer period of drought, and confirm or deny these projections.In that context, since June 2013, measurements have been continued at the O 3 HP on the same branches as the ones studied in this study (Saunier et al., under review).However, the major impact of the future climate changes (higher drought, temperature and CO 2 ) on isoprene emissions could, eventually, be related to the expected 15 general land cover change, with a shift of the actual Mediterranean species to more favorable conditions.Such impacts can only be assessed with -improved -isoprene emission models coupled with global dynamic vegetation models.Table A2.The selected input regressors x i as follows : x 1 L0 x 2 L-1 x 3 T0 x 4 T-1 x 5 T M -T m x 6 T-14 x 7 T-21 x 8 SW-1 x 9 SW-7 x 10 SW-14 x 11 SW-21 x 12 ST-7 x 13 ST-14 x 14 P-7 x 15 P-14 Biogeosciences Discuss., doi:10.5194/bg-2017-17,2017 Manuscript under review for journal Biogeosciences Discussion started: 9 February 2017 c Author(s) 2017.CC-BY 3.0 License.(April 2013) and leaf senescence (October 2012).Three trees were studied in each plot along the whole seasonal cycle, with a single branch at the top of the canopy being mostly sampled for each tree.More intensive measurements were carried out in June 2012 (3 weeks) and
over the Mediterranean area which contains the bias-corrected Biogeosciences Discuss., doi:10.5194/bg-2017-17,2017 Manuscript under review for journal Biogeosciences Discussion started: 9 February 2017 c Author(s) 2017.CC-BY 3.0 License.daily simulation output of the earth system model HadGEM2-ES.The corresponding data for the 2090-2100 period were derived from two ISI-MIP future projections forced along two Representative Concentration Pathways (RCPs): the so-called 'peak-and-decline' greenhouse gas concentration scenario RCP2.6 (optimistic scenario) and the 'rising' greenhouse gas concentration scenario RCP8.5 (extreme scenario).All T, P and PAR data were extracted for 5 the entire Mediterranean region from the global ISI-MIP data set and subsequently averaged over the area.SW and ST were assessed by running the global land-surface-model ORCHIDEE (ORganizing Carbon and Hydrology In Dynamic EcosystEms) over the European part of the Mediterranean region.Calculated SW and ST were averaged over this area.ORCHIDEE is a 10 spatially explicit process-based model calculating the CO 2 , H 2 O, and heat fluxes exchanged between the land surface and the atmosphere.The processes in the model are represented at the time step of ½ hour basis, but the variations of water and carbon pools are calculated on a daily basis.Vegetation is described using 12 Plant Functional Types (PFT).Each PFT follows the same set of governing equations but takes different parameter values, except for the leafy 15 season onset and offset, which are defined by PFT specific equations.A detailed description of the model is given in Krinner et al. (2005).Simulations over the European part of the Biogeosciences Discuss., doi:10.5194/bg-2017-17,2017   Manuscript under review for journal Biogeosciences Discussion started: 9 February 2017 c Author(s) 2017.CC-BY 3.0 License.adjusted from May to reach its maximum (32%) in July and maintained until November when rain exclusion was stopped (Fig.1c).The annual cumumaled P (P cum ) in the AD plot was lower by 273 mm than in the ND plot at the end of 2012 (782 compared to 509 mm).In 2013 the AD started only at the end of June, simulating a later amplification; our April and June 2013 measurements were thus not impacted by AD.From August to October 2012, SW was 5 50 to 90% lower in the AD plot than in the ND plot ( 0.02 and to 0.05 L H2O L soil -1 respectively in August, Fig.1d).The AD plot soil water deficit remained significant until the end of the experiment (Mann & Whitney, P<0.05 in June 2012, P<0.001 from July 2012 to June 2013), although the rain exclusion system was not activated between December 2012 and June 2013.10 No significant difference was noticed for monthly PAR and T means between the ND and the AD plot, except in September 2012 when branches sampled on the ND plot received significantly more PAR than branches on the AD plot (Mann & Whitney, P<0.001).
Biogeosciences Discuss., doi:10.5194/bg-2017-17,2017   Manuscript under review for journal Biogeosciences Discussion started: 9 February 2017 c Author(s) 2017.CC-BY 3.0 License.strengthen the hypothesis that, under environmental stresses (like water stress), some plants favour the allocation of carbon to secondary metabolites production (such as isoprene) rather than allocation to growth.Maximum I s was very close to previously measured values obtained for the same species under Mediterranean conditions during greenhouse and in-situ experiment (114.3 and 134.7 µg.g DM -1 .h - ,Genard-Zielinski et al., 2014; Simon et al., 2005  5   respectively).The difference observed in April 2013 between I s in the ND and AD plot cannot be attributed only to the AD effect, since the rain exclusion was started only at the end of June 2013; apart from a possible 'memory effect', it is likely due to the high natural variability in bud breaking and isoprene emission onset at this period of the year.

Figure 2 .:
Figure 2.: Seasonal variations of monthly Q. pubescens gas exchanges observed at O 3 HP (June 2012 to 2013) under ND (blue) and AD (red) (mean  SD).(a) Stomatal conductance to water vapour G w .(b) Net photosynthetic assimilation P n .(c) Measured branch isoprene 10 emission rate ER.(d) Isoprene emission factor (I s ) calculated according to Guenther et al. (1993) using in situ ER vs C L  C T correlations, except in July where mean ER measured under enclosure conditions close to 1000 µmol m -2 s -1 and 30 °C was used.Differences between ND and AD using Mann-Whitney tests are denoted using lower case letters (a>b).Differences among months using Kruskal-wallis tests are denoted by asterisks (*: P<0.05; **: 15 P<0.01; ***: P<0.001).

Figure 4 .:
Figure 4.: Seasonal variations of the relative contribution on isoprene emission rates (ER) of (a) the environmental parameters used in G14 (the received incident PAR L, air temperatureT, soil water content SW, soil temperature ST and precipitation P) whatever their frequencies 25 considered in G14; and, of (b) the different frequencies in G14 (0, 7, 14, 21 days before the measurement), whatever the environmental parameters in G14.

Table A1 .
The optimised weights w as follows 10