Modeling Intra‐ and Interannual Variability of BVOC Emissions From Maize, Oil‐Seed Rape, and Ryegrass

Air chemistry is affected by the emission of biogenic volatile organic compounds (BVOCs), which originate from almost all plants in varying qualities and quantities. They also vary widely among different crops, an aspect that has been largely neglected in emission inventories. In particular, bioenergy‐related species can emit mixtures of highly reactive compounds that have received little attention so far. For such species, long‐term field observations of BVOC exchange from relevant crops covering different phenological phases are scarcely available. Therefore, we measured and modeled the emission of three prominent European bioenergy crops (maize, ryegrass, and oil‐seed rape) for full rotations in north‐eastern Germany. Using a proton transfer reaction–mass spectrometer combined with automatically moving large canopy chambers, we were able to quantify the characteristic seasonal BVOC flux dynamics of each crop species. The measured BVOC fluxes were used to parameterize and evaluate the BVOC emission module (JJv) of the physiology‐oriented LandscapeDNDC model, which was enhanced to cover de novo emissions as well as those from plant storage pools. Parameters are defined for each compound individually. The model is used for simulating total compound‐specific reactivity over several years and also to evaluate the importance of these emissions for air chemistry. We can demonstrate substantial differences between the investigated crops with oil‐seed rape having 37‐fold higher total annual emissions than maize. However, due to a higher chemical reactivity of the emitted blend in maize, potential impacts on atmospheric OH‐chemistry are only 6‐fold higher.

The magnitude and compound composition of BVOC emissions from a particular plant or vegetation type depends on its genetical properties (i.e., emission capacity), while the actual emission rate and its dynamics depend on environmental conditions such as temperature, absorbed photosynthetic active radiation (PAR), ambient CO 2 concentration, as well as various biotic and abiotic stressors and the plant phenological state (Brilli et al., 2019;Grote, 2019;Manco et al., 2021;Niinemets & Monson, 2013).In particular, it is known that BVOC emissions from different species are highly divers in their chemical composition (Courtois et al., 2009;Monson et al., 2013).This variety is important to consider (Heald & Kroll, 2020) but not fully reflected in regional emission inventories that often pool various species of mixed ecosystems into one emission type (e.g., tropical forests, Nichol & Wong, 2011) or use the same emission assumptions for various (monospecific) land covers that have a similar management type (e.g., continental agricultural areas, Karl et al., 2009).These assumptions are challenged by findings that indicate a large regional variability in emission types within apparently similar forests (Gu et al., 2017) as well as different properties of crops, in particular those introduced for bioenergy production (Szogs et al., 2017).An additional aspect is that agricultural regions are subject to relatively fast changes in species selection and crop rotations.For example, the agricultural area used for maize plants in Germany has steadily increased during the last 30 years from 1.6 Mha (14%) in 1990 to 2.7 Mha (23%) in 2020 (https:// www.destatis.de, Statistisches Bundesamt, 2021, last access: 30.06.2021).A similar development can be seen for oil-seed rape which increased from 0.7 Mha (6%) in 1990 to almost 1.5 Mha in 2010 (but decreased thereafter to about 1 Mha in 2020; https://www.destatis.de, Statistisches Bundesamt, 2021, last access: 30.06.2021).Considering that very different BVOC emission rates have been depicted for maize and oil-seed rape compared for example, with wheat (Gonzaga Gomez et al., 2019), it seems unreasonable to use a lumped emission factor for crops, in particular when longer periods should be investigated.
The change in species composition is partly driven by an increasing demand for fodder and bioenergy, favoring species such as maize, oil-seed rape and grass, which is a development projected to continue (Bentsen & Felby, 2012;Gelfand et al., 2020).However, knowledge about the composition of BVOC emissions and on their environmental drivers is still scarce for many of these plants.Particularly full seasonal BVOC observations that cover the whole range of phenological phases are hardly available, although the linkage of emissions to physiological processes and anatomical developments is decisive for simulations that aim to dynamically respond to various combinations of environmental conditions.It should also be considered that available information is often derived at regions where these species have been grown in the past, which have not been necessarily the same conditions than they experience under a central European climate.Therefore, we are presenting several weeks of observations in each of three different crop types (Zea mays, Brassica napus, Lolium multiflorum) that are already planted in Central Europe and/or are supposed to increase in importance.Since a detailed analysis of BVOC fluxes from maize have been presented before (Wiβ et al., 2017), we now concentrate on the comparison of the different crops, evaluating their potential impact on air quality.We therefore elaborate and parameterize a physiology-oriented BVOC emission model based on measurement results.Simulating not only developments during the observation period but also throughout a full three-year period, we present a comprehensive comparison of intensity and quality of BVOC emissions for the three investigated species.

Field Observations
Three intensive measurement campaigns were carried out throughout 2015-2017 at an agricultural field site close to the research station Dedelow, located in north-eastern Germany (N 53.3793, E13.7856).In each of the 3 years, a different crop (maize (Zea maize), oil-seed rape (Brassica napus), and ryegrass (Lolium multiflorum)) has been investigated (see Table 1 for management and observation details).The meteorological conditions during the three years were similar with respect to temperature and radiation conditions (average 9.6, 9.8, and 9.5°C for 2015, 2016, and 2017 respectively).Only minor interannual variations could be detected such as a relatively cool early April but a warm August in 2015, warm periods in January/February and September in 2016, as well as a cool late April but warm October in 2017 (see meteorological developments in the supplements Figure S1-S3 in Supporting Information S1).However, precipitation was different in the three years (annual rainfall of 475, 392, and 755 mm for 2015, 2016, and 2017 respectively) with 2017 showing exceptional high rainfall rates particularly during June and July.In each year 160/0 (maize), 80/90 (oil-seed rape), or 0/160 (grass) kg N ha −1 was added to the site as urea and calcium-ammonium nitrate, respectively.Fertilization as well as application of herbicides, fungicides and insecticides (for maize and oil-seed rape) was carried out well before the start of the measurements.The agricultural site is embedded within the CarboZALF project (Sommer et al., 2016) which provides meteorological information, as well as the facilities for automatically controlled closed-chamber measurements.
In the following we briefly present the setup and execution of measurements as well as calculations, while a more detailed description can be found in Wiβ et al. (2017).All gas exchanges were derived from concentration measurements within large polycarbonate automatic chambers of 1.5 m × 1.5 m × 2.5 m size each.The chambers are mounted onto a steel frame and lifted or lowered by a cable winch (Hoffmann et al., 2017).An aliquote of the air was sampled (and re-added) while the chamber was closed (otherwise the chambers were kept open), thereby maintaining a flow-through nonsteady state system, and fluxes are calculated from the difference in concentration during the closure, which is less than 15 min in order to minimize the influence on environmental conditions.CO 2 and water concentrations were determined using an infrared LI-840 gas analyzer (LICOR Bioscience, Lincoln, Nebraska, USA) and a Campbell 500 data logger (see also Hoffmann et al., 2015).Air was drawn from the chamber at an airflow rate of 10 dm 3 min −1 and redirected into it except for a minor fraction (70 cm 3 min −1 ) that was used for BVOC detection using a PTR-QMS 500 (Ionicon Analytik GmbH, Innsbruck, Austria).Overall, 18, 25, and 32 compounds were measured for maize, oil-seed rape, and ryegrass, respectively, covering m/z values between 21 and 225.All compounds with less than mean flux rates of 0.01 nmol m −2 s −1 were neglected in further analyses.For compound identification, additional air samples were collected and analyzed by gas chromatography mass spectrometry (GC-MS (GC type: 7890A; MS type: 5975C; both from Agilent Technologies, Palo Alto, CA, USA).Please note that isoprene and monoterpene fragments can be omitted from the analysis as their emission rates are based on measurements of the parent ions and quantified using certified standards.BVOC fluxes were calculated as the slope of a linear regression (ordinary least squares), fitted to the concentration measurement points over time obtained during 12 min of sampling.All fluxes from nonsignificant linear regressions (H0: slope of regression line is zero; p > 0.05; that is, r 2 < 0.60 for a two-tailed t distribution test with 12 data points) were set to zero.From all measurements within an hour, the mean fluxes and the standard emission factors (SEFs, emissions that can be expected at 1000 μmol m −2 s −1 PAR and 30 o C) are derived by curve fitting via the Levenberg-Marquardt algorithm with nonlinear least squares using the Python SciPy library (https://sourceforge.net/projects/scipy/files/scipy/0.14.0/, last access: 30.06.2021).

Parameter Derivation of BVOC Modeling
Light-dependent BVOC emissions are represented as a fraction of the electron transport rate as proposed by Niinemets et al. (1999); a theoretical approach that is based not on the transport rate as such but on the excess of electrons (JJv model) has been suggested by Morfopoulos et al. (2014) and elaborated by Grote et al. (2014).
Emissions that are assumed to origin from storages and therefore are temperature-dependent only, are simulated with an exponential relation suggested by Guenther et al. (1993).Since the electron transport dependencies in the JJv model only replace the modifiers for light and temperature in the Guenther model, it can directly be used with SEF parameters that have been derived from measurements and fitted with the Guenther equations.
BVOC Emissions (E BVOC , calculated in ng g −1 DW aboveground biomass h −1 ) can potentially origin from two different sources, namely from a direct production (de novo emissions) that is considering radiation and temperature as simultaneous influences, as well as from storages that only depend on (tissue) temperature T (well described with a simple exponential relationship suggested by Guenther et al. (1993)).Therefore, we derived SEF and the slope parameter β, combining both approaches and linking them using the parameter LDF (light dependent fraction, Equation 1), which in turn is determined by iterating from 0 to 1 (in steps of 0.1) selecting the ratio that best fitted emissions (Table 2). ] (1) with γ ph , γ ph_norm as well as γ en, γ en_norm being the photosynthetic (Equation 2) and enzymatic (Equation 3) emission potentials at current and standard environmental conditions as described in Grote et al. (2014).Parameters T S (standard temperature 303.15 K), c 0 (32 J mol −1 ), AE E (83129 J mol −1 ), DAE E (284600 J mol −1 ), and ΔS (887.5 J mol −1 K −1 ) are taken from Niinemets et al. (1999).The values for c 1 (0.1765e −3 ), c 2 (0.0028e −3 ) and ΔJ SAT (34 μmol m −2 s −1 ) are taken from Grote et al. (2014) and R G is the universal gas constant (8.314J mol −1 K −1 ).
The electron transport rate J (μmol m −2 s −1 ), the electron flux required to support Rubisco-limited carbon assimilation J v (μmol m −2 s −1 ), the intracellular CO 2 concentration C i (μmol mol −1 ) and the CO 2 compensation point Γ*(μmol mol −1 ) are supplied by the photosynthesis model (under either actual or standard conditions (T = T S , radiation = 1000 μmol m −2 s −1 photosynthetically active quantum flux density).

Biogeochemical Modeling
We employed the physiology-oriented BVOC emission model JJv according to Grote et al. (2014) coupled to the ecosystem module PlaMo x (Kraus et al., 2016;Petersen et al., 2021) that runs on an hourly temporal resolution within the LandscapeDNDC model framework (Haas et al., 2013).The model framework provides dynamic crop growth that is calculated from photosynthesis (Farquhar et al., 1980;Leuning, 1995) and ecosystem respiration (Thornley & Cannell, 2000) in dependence on environmental conditions (including nitrogen supply and water availability).Allocation of carbon is distributed into different compartments according to development stage using species-specific parameters.The height of the crop depends on the biomass in the stalk compartment.
Water balance is calculated based on the original DNDC model (Holst et al., 2010;Li et al., 1992) and soil water contents below a certain threshold directly reduce photosynthesis.
All plant carbon exchanges (CO 2 as well as BVOCs) were calculated on an aboveground-biomass-basis according to the environmental conditions that are estimated within a specific canopy layer.Layers have a flexible thickness that depends on crop height, and leaf biomass is equally distributed between them.Foliage develops dynamically based on the calculation of growing degree day sums.Leaf area is directly related to leaf biomass by species-specific conversion factors (specific leaf area for maize, oil-seed rape and ryegrass are 16, 39.6, and 24.Photosynthesis parameters, that is, those that influence the substrate provision for BVOC emissions, have been derived from literature for each of the investigated species (Table 3).BVOC simulation was carried out with the JJv model according to Equation 1 for all compound emissions that depend on light (Grote et al., 2014;Morfopoulos et al., 2014), using SEF, β and LDF values separately determined for all compounds as presented in Table 2. Thereby, compounds that are emitted from specific storages and thus showed no light dependence (LDF = 0) were solely calculated based on traditional storage-based emission schemes.
In order to derive the reactivity of the emission, we weighted the emission with the species-specific average blend for the whole season and multiplied with the depicted OH reactivity (Table 4).It should be noted that we excluded negative emissions, that is, potential deposition processes, from the analysis although these were sometimes substantial, in particular for oxygenated compounds when air humidity was high, and condensation on surfaces may have occurred.In these cases, deposition processes that are enhanced through solution into water can be expected.
We ran the coupled model with field-measured hourly weather data (temperature, radiation, precipitation) over the entire three years (2015)(2016)(2017).The soil was initialized with measured values of soil structural properties down to 2 m depth (more than 50% sand and 10%-20% clay) from which pore space and water holding capacity were estimated.Simulations were carried out with the respective crop of the year starting with sowing date, considering a fertilization event before sowing.In addition, runs were set up for each year with every other crop so that crop development and BVOC emission could be compared under equal environmental conditions.
Finally, we estimated the importance of BVOC emissions for air chemistry by multiplying the emission of a specific compound i with its reaction rate coefficient with OH (Table 4).Using the simulations, we scaled the emissions up to a full growth period of each crop and compared them using the total BVOC-OH reactivity of maize as standard.Hu et al., 2018;Sarkar et al., 2020), assuming that emissions are linearly proportional to BVOC concentration changes within the boundary layer (i.e., assuming a similar temperature response of BVOC reactivity with other air chemistry components).Another constrain is that the potential BVOC-OH reactivity could only be considered for those compounds which were identified by the GC-MS analysis, meaning that about 9.9% and 1.3% of emissions for maize and oil-seed rape, respectively, were neglected.

Gross Primary Production and Biomass Development
The three species each showed specific growth patterns that were characterized by a relatively long growing period from May to October with growth starting to decline in about the middle of the period (maize), a dormancy during winter after a first growth and a relatively steep biomass increase between March and the end of June (oil-seed rape), and a continuous growth from March to August that is intermitted by cutting events in May, June, and early August (ryegrass; see Figure 1).Maize reached about 1.5 times the biomass of oil-seed rape (1.66 ± 0.1   The LandscapeDNDC model could well represent photosynthesis, as well as leaf and biomass development, particularly in the early developmental phases (Figure 1).In the late phases, carbon uptake seems to be slightly under-(maize) or overestimated (oil-seed rape), which might be related to underestimated drought (maize) and disregarded seasonal decline of enzymes (oil-seed rape).

BVOC Emissions
Daily mean BVOC fluxes can be accumulated for different periods as well as the whole observation (growing) period, which differ in length and flux magnitude (see Table 5).Highest mean values over the entire investigation period were generally observed for the green leaf volatiles (GLVs, mainly maize) and other (oxygenated) compounds with relatively low reactivity (i.e., methanol, ethanol, acetone, acetaldehyde, and acetic acid), which together contribute 73%, 99%, and 94% of all observed emissions for maize, oil-seed rape and ryegrass respectively.Methanol was particular dominant in oil-seed rape and ryegrass, contributing more than 80% of the whole year emissions on a molecular basis while it was still one of the two main compounds in maize.Terpenoids are emitted by all plants but contributed to major degrees only in maize.
Whereas emissions correlated well with temperature and leaf area throughout the season in general, BVOC emissions were significantly less sensitive in some periods than others.For maize, these different sensitivities have been previously attributed to the different phenological stages (Wiβ et al., 2017).Now, this sensitivity seems to be compound-specific since for example, acetic acid increases its SEF during growth in oil-seed rape while the SEFs of acetaldehyde decrease.However, the impacts are not easily distinguished, since meteorological conditions and growth stages vary simultaneously, for example, the onset of flowering in oil-seed rape occurs in parallel with a substantial temperature and PAR increase.
Using compound-specific parameterization differentiated by plant species, BVOC emissions could generally be well reproduced (except in cases where large depositions occurred).Examples for all species are presented in Figure 2.These also demonstrate the impact of the before-mentioned sensitivity change between seasons or developmental stages, resulting for example, in an overestimation of monoterpene emissions from maize in late August 2015.
From the analysis of all different compounds, three emission groups were derived that differ in their reactivity (see Table 4).All simulated and measured compounds within one group were summed up for each species (Figures 3-5).Note.The k-rates are given for 298.15K in units of cm 3 molecule -1 s -1 .The compounds are grouped by mass to charge ratio (m/z) and sorted by reactivity, distinguishing 3 different groups: high reactivity with OH k-rates > 1.0 x 10 -10 (dark grey), medium high reactive VOCs (ORVOCs also including some oxygenated monoterpenes (oMTs), medium grey) with OH k-rates between 1.0 (0.988 in case of camphor) and 10 x 10 -11 , and low reactive other VOCs (OVOCs, white background) with OH k-rates < 1.0 x 10 -11 .If not separately indicated, the values were taken from the Master Chemical Mechanism, MCM v3.2 (Saunders et al., 2003)

Rotation Cycle Comparisons
Simulations for three consecutive years have been carried out to support conclusions about average BVOC emissions and their dependence on inter-annual meteorological differences during a growing season (Figure 6).Regarding biomass and GPP development, slightly warmer spring periods in 2016 (intensive) and 2017 (moderate) led to an earlier onset of plant growth and also to an earlier growth cessation of about 1-2 weeks compared to 2015.It can be seen that drought limited crop growth only in 2016, where a severe precipitation deficit occurred in July, which particularly affected ryegrass development.Overall, the impact of interannual variation on growth is relatively small.
Regarding BVOC emissions, however, the differences in weather led to considerably higher variations.Generally, simulated emissions of each crop were the highest in 2015, which is mainly caused by the high temperature in August of +3°C above the long-term mean.In contrast, temperatures in 2016 were relatively high in June/July and September, periods in which the leaf area and thus the potential for BVOC emissions was reduced for maize (due to not fully developed foliage), oil-seed rape (due to early cutting before September) and ryegrass (due to previous cuttings and less growth in autumn).Table 6 presents simulated mean annual BVOC emissions rates and variabilities for all 3 years.
Generally, with 91.4 ± 8.0 mmol m −2 , cultivation of oil-seed rape lead to the highest amount of total BVOCs emission per year on a mole basis followed by ryegrass (15.7 ± 0.6 mmol m −2 ) and maize (2.5 ± 0.1 mmol m −2 ).
Besides the variability and seasonality, the different composition of the blend is decisive for the potential impacts of BVOCs on air chemistry because all compounds have a different reactivity (see Table 4).The results of the weighting procedure indicate that in contrast to the huge variability of total annual BVOC emission fluxes (37-fold between oil-seed rape and maize, see supplement Figure S10-S12 in Supporting Information S1), the difference between the potential annual impact on air chemistry is relatively small (6-fold between oil-seed rape and ryegrass, see Figure 7).The highest impact on atmospheric chemistry is expected from oil-seed rape BVOC   emissions (factor 2.4 larger than maize), which is followed by maize and ryegrass (only 40% that of maize).As the composition of BVOCs emitted from oil-seed rape is similar to that of ryegrass, the difference between their total emissions (5.8-fold) is similar to the difference of their relative impact on atmospheric reactivity (6.0-fold).
In contrast, maize, emits relatively large amounts of terpenes (limonene, α -humulene) as well as the homoterpene DMNT, which are highly reactive.Thus, despite the relatively small amount of annual BVOC emissions compared to oil-seed rape and ryegrass (approximately 3% and 16% of the oil-seed rape and ryegrass emissions), the potential impact on air chemistry of maize is about 2.5-fold higher than that of ryegrass and only about 2.4-fold less than oil-seed rape.

Comparison of Simulations With Field Measurements
From the comparison of simulations with measurements a couple of important points can be derived.First, it seems that particularly high reactive compounds tend to be underestimated at least during specific development phases (e.g., maize in the flowering phase (Period 1), Figure 3).This clearly indicates that a period-specific parametrization would be superior to the use of a whole-year emission parameter, particularly with respect to light-dependent and generally more reactive compounds than emissions from storages.
Second, some uncertainties regarding the absolute amount as well as the temporal variation of emissions may be related to BVOCs formed in or deposited to the soil.BVOC emissions from soil have been reported in many studies before (Peñuelas et al., 2014) but also deposition processes seem to be common and sometimes relevant in magnitude (Bachy et al., 2020;Spielmann et al., 2017).Emissions as well as consumption has been attributed to soil microorganisms, resulting in microbial activity and diversity being the main driver (Abis et al., 2020).These effects are influenced by agricultural management, which alters BVOC exchange through lagged effects (e.g., the previously planted crop) or the management of pests and diseases (Malone et al., 2020).Note.Period 1 and period 2 refer to flowering (24 days) and ripening (28 days) in maize, to inflorescence emergence (14 days) and flowering (4 days) in oil-seed rape (Osr), and to heading (19 days) and flowering (20 days) in ryegrass, respectively.
Finally, for some specific compounds, that is, ethanol, methanol and (for ryegrass) acetic acid, the fit between simulations and measurements is very poor (Figure S7 in Supporting Information S1).These discrepancies are mostly explained by the water-solubility of the mentioned compounds, because during wet periods they are dissolved in the moisture at leaf or soil surfaces, which leads to reduced net emission rates or even negative emissions.The importance of this deposition process has thus been demonstrated with our measurements but is not yet considered in the model.
The results also demonstrate that at least for oil-seed rape and ryegrass, which can grow from late spring until early autumn, measurement periods did not cover the whole range of potential importance.

Comparison With Other Emission Measurements
Several measurements of BVOC emissions have been presented in the literature for maize, some for oil-seed rape and only a few for ryegrass (see Table 7).For maize, the data indicate a particularly large variation of what compounds are emitted in which quantity, with the majority of publications indicating much larger net emissions of methanol (Das et al., 2013;Gonzaga Gomez et al., 2019;Graus et al., 2013;Mozaffar et al., 2018) than recorded in our field survey (Wiβ et al., 2017).In contrast, our measurements demonstrated larger fluxes of monoterpenes (see also Wiβ et al., 2017) than presented in most other investigations (Bachy et al., 2016;Graus et al., 2013;Street et al., 1997).Moreover, we herein demonstrate the emission of sesquiterpenes and oxygenated monoterpenes that have not been described in any of the field measurements before (although they have been recognized in laboratory experiments).In contrast to maize, our findings for oil-seed rape are rather at the upper range of previous field observations regarding methanol (Acton et al., 2018;Gonzaga Gomez et al., 2019) as well as GLV emissions (König et al., 1995).Notably, considerable net emissions of other oxygenated hydrocarbons such as of acetone, acetaldehyde, and ethanol were found that were only seldom reported before (Bsaibes et al., 2020;Gonzaga Gomez et al., 2019;McEwan & MacFarlane Smith, 1998).Terpenoid emissions in oil-seed rape were either higher (Bsaibes et al., 2020;Himanen et al., 2009;McEwan & MacFarlane Smith, 1998) or lower (König et al., 1995;Müller et al., 2002) than indicated by our observations, with one investigation that could not  4) for maize 2015.
find any monoterpene emission at all (Morrison et al., 2016).Regarding ryegrass, we present the first comprehensive estimate of the BVOC emission spectrum that has been derived from field measurements.The few other investigations concentrated on OVOCs and indicated either no (Custer & Schade, 2007) or moderate net emission of methanol and acetaldehyde for Lolium multiflorum (Kirstine et al., 1998).Laboratory investigations indicated furthermore high emissions of GLVs and sesquiterpenes as well as emissions of monoterpenes and oxygenated monoterpenes (Pańka et al., 2013), which however, seem to play only a minor or no role under field conditions.
We hypothesize that these differences may at least partly be due to too short measurement periods that do not include deposition (because measurements took place under favorable weather conditions).Estimations based on such measurements would thus be prone to overestimate in particular emission of compounds that are water-soluble such as methanol and ethanol (Das et al., 2013;Gonzaga Gomez et al., 2019).In some occasions, plants may have been in early or late developmental stages that were not representative for whole-season emissions.Also, it should be noted that measurements at single leaves or small plants may be considerably biased if ecosystem fluxes shall be derived (Graus et al., 2013).Likely, some of the differences measured herein may also be due to the use of different cultivars or the presences of undetected stressors.For example, insect infestation has increased GLVs (Christensen et al., 2013), fungal infections are known to induce alcohols (Usseglio et al., 2017), while various stressors were connected to elevated terpenoid emissions (Block et al., 2019;M. Chiriboga. et al., 2018).It should also be noted that by using large, canopy chambers, all plant-atmosphere exchanges including BVOC emissions are derived at the whole ecosystem level, including trace gas exchange from the soil.

Importance for Atmospheric Chemistry
Cropland emission inventories have been built with and without differentiation of crop types and species using quite different SEFs (see Table 7).In most cases, croplands are not considered at all (pooled into grassland species), while sometimes fixed SEFs for cropland (usually non-irrigated) are assumed that are scaled with a fixed (average; Zheng et al., 2010) or regionally differentiated, for example, based on satellite images, biomass value per area (Gulden & Yang, 2006;Li et al., 2018).The fixed SEFs and leaf biomass densities, however, lump together crop species with very different emission pattern, for example, emitting or not emitting isoprene.While the average emission parameters may be correct under a specific condition, such inventories cannot account for changes in crop abundance, as for example, expanding cannabis plantations (Wang et al., 2019) or the continuous increase in maize covered area in Europe (see introduction).In contrast, inventories based on species-specific parameterization of BVOC emissions as well as foliage biomass density can be easily updated with agricultural statistics.For example, the cropland parameterization from the latest inventory of Europe (Karl et al., 2009) largely overestimates emissions of all compound classes when compared with the three species investigated  here.The difference would be even more severe for other low emitting crop species which have been found in various genera of energy crops such as Miscanthus or Panicum (Copeland et al., 2012;Hu et al., 2018;Morrison et al., 2016), while planting strong isoprene emitters such as giant cane (Arundo donax; Porter et al., 2012) or bamboo (Phyllostachys spp; Okumura et al., 2018) would result in an underestimation of BVOC emission fluxes.
Overall, we show that total as well as compound-specific BVOC emissions from bioenergy crops can well be represented using the principal mechanisms of the JJv model that integrates phenological and physiological responses on environmental conditions.The model enables to represent leaf expansion and the development of specific ripening stages which results in large and realistic seasonal variations in BVOC emission rates and patterns.Thus, a highly spatial and temporal resolved inventory might be feasible provided that emission parameters for the major species are available.Nevertheless, some deviations between simulations and observations (mean ± standard deviation of NSE of 0.48 ± 0.09, 0.84 ± 0.08, and 0.47 ± 0.18, for maize, oil-seed rape, and ryegrass respectively for all simulated BVOC emissions) indicate that modeling would still profit from a dynamic seasonality effect that accounts for plant developmental stages such as flowering.How to consider such an effect is still an open discussion.A simple temperature and light dependency that had been developed particularly for woody species (Guenther et al., 2006) has been applied also for crops (Karl et al., 2009) but its suitability for species that shift into different development stages remains to be shown.In conjunction with a seasonally realistic agricultural rotation management (i.e., sowing and harvesting) also spatial variation can be reasonably covered.
Finally, the conversion of BVOC emissions into reactivity has shown that not only the emission strength of individual compounds but also the emission pattern and the individual (or group) reactivity of the compounds need to be considered.This has recently also been postulated for a short-term experiment at an oil-seed rape field in France (Bsaibes et al., 2020).In particular compounds with very high OH-reactivities can determine air chemistry in specific periods.The ability of emitting such reactive compounds, however, is very differently expressed in different agricultural species, but this species-dependence is generally not differentiated in regional air chemistry calculations (Chatani et al., 2015;Kim et al., 2017;Poupkou et al., 2009).For example, significant emissions of limonene, DMNT and sesquiterpenes have been found solely in maize, while in ryegrass only small amounts of isoprene were emitted, rendering this species as low-impact to air chemistry despite its relatively large emissions of other oxygenated compounds.In this regard, it should be noted that within the same chemical class, reactivities can differ by almost two orders of magnitude (see Table 4).This leads to the fact that a low xylene emission in maize approximately compensates the OH reactivity of the very high acetic acid emissions in oil-seed rape.The presented research demonstrates that compounds with small quantitative contributions do significantly affect air chemistry.Even if the molar amount of emissions is one order of magnitude different between each reactivity group (oil-seed rape and ryegrass), the importance for air chemistry is in the same range since also the reactivity difference is about one order of magnitude.In order to derive more general and responsive simulations that are able to potentially cover interannual as well as decadal trends and are applicable on a regional scale, species-specific emission strength as well as compositions should therefore be considered.Finally, the full year simulations reveal that the variability of emissions during the year is high for all crops and that single days or short periods can have disproportional large impacts on air chemistry.Note.In order to provide this comparison, calculated emission obtained in this study were scaled to a photosynthetic photon flux density (PPFD) of 1000 μmol m −2 s −1 and 30°C in units of μg compound g −1 DW h −1 .The dry weight of leaf or plant biomass density used for scaling is also given in g m −2 .a based on Kirstine et al. (1998).
kgDW m −2 and 1.11 ± 0.09 kgDW m −2 , respectively), which was, however, not much different from the the overall biomass production of grass for the whole year (1.49 ± 0.05 kgDW m −2 ).Different development stages have been recognized for different species such as reproductive growth, emerging inflorescences and flowering, fruit development and ripening.Only for ryegrass, development did continue beyond flowering as the cuts interrupted the post-flowering development.Gross primary production (GPP) followed temperature and PAR (see supplement FigureS4-S6 in Supporting Information S1).Drought stress cannot fully be excluded in late summer 2015 for maize, as well as for a short period in July 2016, but is unlikely for 2017 which was characterized by high precipitation.

Figure 1 .
Figure 1.Gross primary production simulated and derived from measurements (a, c, e) as well as biomass development (b, d, F) for the three investigated crops: maize (a), (b), oil-seed rape (c), (d), and ryegrass (e), (f).Simulations are shown with red lines, observations are represented with black dots (Error bars indicate ± the standard error of the sum).

Figure 2 .
Figure 2. Examples of observed and simulated emissions for the specific crops (see complete results in the supplements Figure S7-S9 in Supporting Information S1).Top: oxygenated monoterpenes for maize (oMTs, mainly 1,8-Cineole); Middle: acetic acid for oil-seed rape; Bottom: methyl ethyl keton (MEK) for ryegrass.Simulations are shown with red lines, observations are represented with black dots (Error bars indicate ± the standard error of the sum).

Figure 6 .
Figure 6.Simulated Gross Primary Production (a, c, e) and biomass development (b, d, f) for the three investigated crops maize (a), (b), oil-seed rape (c), (d), and ryegrass (e), (f) for all three investigated years.

Figure 7 .
Figure 7. Potential mean annual BVOC-OH reactivity from maize, oil-seed rape, and ryegrass based on the 3 years of the investigation period.All values are scaled to the reactivity of maize.Only compounds with contribution >0.1 are considered.

Table 1
Plant Species, Date of Seeding and Harvest, Measurement Period, Number of Chambers

Table 2 BVOC
5, respectively).Since maximum carboxylation rate and electron transport rate are reduced proportionally to leaf 10.1029/2021MS002683 5 of 22 a n.s. a n.s. a n.s. a n.s. a n.s. a a n.s. a n.s. a n.s. a n.s. a n.s. a Emission Parameter for Maize, Oil-Seed Rape, and Ryegrass Fitted to the Joint JJv-Pool Emission Model 10.1029/2021MS002683 6 of 22 development and leaf senescence, also photosynthesis is decreased and precursor supply for BVOCs is reduced during this time.There are no other indirect seasonal effects on emission calculations.

Table 3
Species-specific Photosynthesis Parameters Based on Literature

Table 4 BVOC
-OH Reaction Rate Coefficients (K-Rates) for Those Compounds Detected During the Field Campaigns Throughout 2015-2017

Table 5
24-Hour Mean Molar BVOC Emissions for the Experimental Period in Pmol m −2 Ground and s −1 (Negative Fluxes Set to Zero) Including the Fraction of Each Compound to the Total BVOC Emission

Table 6
Simulated Mean Annual BVOC Emission Rates (μmol m −2 a −1 ) From Maize, Oil-Seed Rape, and Ryegrass Measured From 2015 to 2017 at the CarboZALF Field Site in Dedelow (SD Denotes the Standard Deviation Relative to the Number of Measurements Indicated in Table 1)

Table 7
Overview of Terpenoid BVOC Standard Emission Factors (SEF) FromMaize, Oil-Seed Rape, and Ryegrass Compared With Standard Cropland Emissions Used in  Regional ModelsThis study is incorporated into the Carbo-ZALF project providing the automatic chambers equipped with instruments for meteorological observations and gas exchange under the lead of Jürgen Augustin from ZALF.We are especially grateful to Gernot Verch, head of the ZALF research station in Dedelow and to Marten Schmidt for maintenance and setting up most of the measuring infrastructure.We also want to thank Steffen Klatt for his IT support with Land-scapeDNDC, Christof Lorenz for his help with the data repository, as well as two anonymous reviewers for their valuable comments.The project has been financed by the Federal Ministry of Food and Agri- culture (FKZ 22006414 and 12NR257) through the Agency for Renewable Resources (FNR).We also acknowledge support by the KIT-Publication Fund of the Karlsruhe Institute of Technology for publishing open-access and by a Fellowship in the Helmholtz Research School "Mechanisms and Interactions of Climate Change in Mountain Regions" (MICMoR) through KIT/IMK-IFU.Open access funding enabled and organized by Projekt DEAL.