A statistical approach for estimating representative emission rates of 1 biogenic volatile organic compounds and their determination for 192 plant 2 species / genera in China 3

† These authors contributed equally 9 10 Abstract. To obtain more and accurate biogenic volatile organic compound (BVOC) emission rates 11 for more plant species in China and further improve the accuracy of emission rates used in BVOC 12 emission inventories, we conducted field measurements and developed a statistical approach for 13 estimating representative emission rates. We performed field measurements of BVOC emissions from 14 50 plant species at nine locations in China using our established semi-static enclosure system. The 15 emissions of 102 VOCs, including isoprene, α-pinene, β-pinene, and other VOC species, were 16 analyzed with a custom-built online gas chromatography-mass spectrometry/flame ionization detector 17 system. From the results, broadleaf trees were the greatest potential emitters of isoprene, while 18 needle-leaf trees emitted more pinene. Shrubs had lower isoprene and pinene emission potentials, but 19 higher emission potentials for other VOCs. Methyl methacrylate, isopropylbenzene, isopentane, 20 acetone, ethane, propane, toluene, and xylene were the dominant species among other VOCs, 21 probably with high emission intensities. Therefore, their emissions should be considered in future 22 global and regional BVOC estimation studies. Next, we summarized our field measurements along 23 with reported emission rates from China and abroad. The emission intensity categories were produced 24 based on statistics, with more detailed categories, accurate emission rate intervals and representative 25 rates compared to previous studies. The results showed that the BVOC emission intensities of plants 26 displayed different categories, such as lowest, lower, low, moderate, high, higher, and highest. The 27 isoprene emission rate intervals and representative rates were: lowest, 0.08–0.11 and 0.1 μg C gdw -1 h 28

Isoprene and monoterpenes are important precursors of tropospheric ozone (O 3 ) and secondary organic aerosol (SOA) (Guenther et al., 1999;Carslaw et al., 2010;Nozière et al., 2011;Sartelet et al., 2012).Therefore, it is essential to establish accurate isoprene and monoterpene emission inventories to support accurate air quality evaluations and effective decision-making regarding air pollution control in China.
It has been reported that simulated SOA and O 3 concentrations in air quality models are usually much lower than observations, especially in polluted regions and during pollution events (Hodzic et al., 2010).For example, SOA might be underestimated by 0-75% in models over an entire year in China (Jiang et al., 2012).This limitation of air quality models might be caused by underestimated emission inventories, especially biogenic volatile organic compound (BVOC) emission inventories.In the simulation evaluations, the input China's isoprene and monoterpene emissions produced by the Model of Emissions of Gases and Aerosols from Nature (MEGAN) were 9.3-16.85and 4.49-4.9Tg C yr -1 , respectively, which resulted in underestimations of SOA, O 3 , and ambient isoprene concentrations in China (Fu et al., 2007(Fu et al., , 2012;;Geng et al., 2011;Jiang et al., 2012).Current BVOC emission inventories in China were varied, with isoprene ranging from 4.1 to 20.7 Tg C yr -1 and monoterpenes ranging from 1.8 to 4.9 Tg C yr -1 (Guenther et al., 1995;Klinger et al., 2002;Yan et al., 2005;Tie et al., 2006;Fu and Liao, 2012;Li et al., 2012;Li et al., 2013;Li and Xie, 2014;Stavrakou et al., 2014).Several studies greatly underestimated isoprene and monoterpene emissions in China.
The uncertainties arose from emission rates, leaf biomass, vegetation cover data, meteorological variables, and emission algorithms.Underestimated emission rates might have been used in these emission estimations, resulting in underestimations of isoprene and monoterpene emissions in China (Zheng et al., 2010;Situ et al., 2014).
It is both important and challenging to estimate isoprene and monoterpene emission rates accurately when developing emission inventories.The emission rate sources and estimations in previous inventories varied, resulting in different results for individual plant species among studies.In addition, emission rate measurements for specific plant species and regions in China are rare.
Different studies reported distinguished values with large uncertainties due to the use of different sampling techniques and sample sizes (Guenther et al., 1994).Uncertainty exists when selecting one emission rate from a single measurement or the mean of all measurements as the representative emission rate for the specific plant.Even if the best available observations were screened or conducted according to the criteria proposed by Niinemets et al. (2011), it would be still necessary to find a straightforward method to determine one emission rate value for a plant species when Atmos. Chem. Phys. Discuss., doi:10.5194/acp-2016-1116, 2017 Manuscript under review for journal Atmos.Chem.Phys.Published: January 2017 c Author(s) 2017.CC-BY 3.0 License.compiling an emission inventory.Therefore, instead of assigning one observation for each plant species, some studies have used emission categories to determine emission intensities and rates (Guenther et al., 1994;Simpson et al., 1999;Klinger et al., 2002;Wang et al., 2007).In this method, discrete emission categories (e.g., negligible, low, moderate, and high) were defined based on the emission intensity of vegetation, with a representative rate and a range of ±50%.For each plant species, the emission rate was determined based on the tendency of the reported emission rates to fall within certain categories.This method improved the accuracy of the final emission rates to a large extent; however, it had several limitations.For example, the process of determining emission categories, representative emission rates, and ranges was not straightforward, and lacked theoretical evidence.Furthermore, some studies used both distinct emission categories and representative values, resulting in different emission rates for specific plants.Some studies used coarse classifications of emission intensity, which might underestimate emission rates of plant species with high emission potentials.Table 1 shows the different isoprene emission categories used in several studies and the resulting emission rates for some plants with high emission potentials.Thus, it is essential to have detailed emission categories and accurate representative values and ranges to estimate emission rates accurately.Especially in China, where has a large vegetated land area and high species diversity, it becomes particularly important by using local measurements of emission rates from plants.Chinese researchers have conducted BVOC emission measurements in China, including the above-canopy emission flux and branch or leaf-level emission rates (e.g., Bai et al., 1994;Bai et al., 2012Bai et al., , 2015;;Bao et al., 2014;Guo, 2012;Zhao et al., 1996;Zhao et al., 2004).However, BVOC emission rate measurements in China are relatively uncommon and have often been conducted using static enclosure systems, which introduced large errors into the measurements (Niinemets et al., 2011).In addition, isoprene and monoterpenes have often been the only VOC species measured in China, and there is a lack of measurements for other VOCs.
In this study, firstly, to obtain more and accurate BVOC emission rates for specific plant species in China, we conducted field measurements using our established semi-static emission enclosure technology and analyzed emissions of 102 VOCs using a custom-built gas chromatography-mass spectrometry/flame ionization detector (GC-MS/FID) system.Secondly, to further improve the accuracy of the emission rates applied to the emission inventories, we developed a theoretically effective approach to estimate emission rates.We summarized our field measurements and reported emission rates from China and abroad to establish emission intensity categories to determine the emission rates of 192 plant species/genera.The emission category intervals for isoprene and monoterpene were produced separately based on statistics, and included more detailed categories, accurate emission rate intervals and representative rates than previous studies.We estimated plantspecific emission rates based on the established emission intervals, and compared our results of Atmos. Chem. Phys. Discuss., doi:10.5194/acp-2016-1116, 2017 Manuscript under review for journal Atmos.Chem.Phys.Published: January 2017 c Author(s) 2017.CC-BY 3.0 License.several tree species with previous researches.Finally, we evaluated the quality of our measurements and estimations of representative emission rates.

Enclosure system
We established a semi-static enclosure system to sample BVOC emissions from plants (Zimmerman, 1979;Lamb et al., 1985).This method improved the limitations of static enclosures, was simple to operate, and had lower environmental variations within the sampling chamber during the enclosure due to the induction of a large amount of zero air as a diluent into the enclosure within a short enclosure time (~10 min or less).Figure 1 shows a schematic of the semi-static enclosure system.To achieve a low residence time with a high enough concentration of BVOCs for sensitive chemical analysis, we used a branch enclosure.A polyvinylfluoride (PVF, Tedlar) bag with a 400-L volume (90 cm width × 160 cm length) was selected to enclose a healthy branch of a plant.The Tedlar bag was 50 μm thick, and was 90-95% transparent to photosynthetically active radiation (PAR) (Ortega et al., 2008).There was no production or adsorption of VOCs in the bag.The chamber was a rectangular bag with one open end and one port for the sampling line.The open end was sealed tightly with a Velcro strap around the trunk side of the branch, together with a temperature sensor and a zero air tube placed along the branch.The bag was carefully placed over the branch to minimize contact of the bag with foliage.After enclosing, the chamber was relatively airtight and had a certain volume of air inside the bag.The port of the bag was connected to fused-silica-lined SUMMA polished stainless steel canisters (3.2 L) with a polytetrafluoroethylene (PTFE, Teflon) tube for instantaneous sample collection.The canisters were cleaned with nitrogen (N 2 ) and evacuated to < 10 mtorr before sampling (Liu et al., 2008).The zero air was provided by the portable cylinder with synthetic air (79% N 2 and 21% O 2 ).The temperature inside the enclosure was measured using an HMP155A probe (Vaisala, Inc., Vantaa, Finland).PAR was measured with a quantum sensor (LI-190SB; LI-COR Biosciences, Inc., Lincoln, NE, USA).Owing to limitations of the field experiment, rotameters were used to monitor and control the flow rate of the zero air.Considering its relatively low accuracy, the flow rate had to be measured and calibrated with a primary air flow calibrator (Gilian Gilibrator-2; Sensidyne, Inc., St. Petersburg, FL, USA) at the beginning and end of each experiment.
To sample BVOCs using the semi-static branch enclosure system, after enclosing, ambient air was collected quickly as the background sample.Then, the bag was purged with zero air at a flow rate of 10 L min -1 for ~6 min.Next, the air flow rate was decreased to 2 L min -1 and the purge was continued for an additional ~3 min, which was expected to allow the air in the bag to be well mixed.
Finally, air in the bag was collected as the emission sample.During the experiment, the start and end times of enclosure and each air purge stage were recorded, and temperature and PAR were monitored Atmos. Chem. Phys. Discuss., doi:10.5194/acp-2016-1116, 2017 Manuscript under review for journal Atmos.Chem.Phys.Published: January 2017 c Author(s) 2017.CC-BY 3.0 License.every minute.All leaves on the enclosed branch were collected and transported to the laboratory and weighed after drying at 70°C for 48 h.We assumed that the air in the bag had been mixed well in the enclosure, and that the volume and VOC concentration were constant during the emission sample collection.After the experiments, we confirmed that the enclosure had been relatively airtight, and the total volume of the residual air in the bag at the beginning of enclosure and the purged zero air was not large enough to cause gas to leak from the bag.

Analysis
The canister samples were transported to the laboratory for analysis as soon as possible to avoid loss of BVOCs due to their high reactivities.The analyses of the sampled VOCs were performed on a custom-built online GC-MS/FID (TH-PKU 300B; Wuhan Tianhong Instruments Co., Wuhan, China) (Liu et al., 2009;Yuan et al., 2012Yuan et al., , 2013;;Li et al., 2014Li et al., , 2015)).The two-channel system used dual columns and dual detectors to analyze C2-C11 VOCs simultaneously.It combined electronic refrigeration technology and the GC-MS/FID system to concentrate and analyze the VOC samples.
The refrigeration system had an initial temperature of -160°C, which was achieved by compressing air.For each sample, 300 mL of air was concentrated when passing through the instrument at a flow rate of 60 mL min -1 after water and carbon dioxide removal.Then, the highly focused VOCs were quickly desorbed at 110°C by heating and transferred to the GC column for separation.The C2-C4 alkanes and alkenes were separated on a non-polar capillary column (PLOT-Al 2 O 3 , 15 m × 0.32 mm ID × 3 μm; J&W Scientific, Folsom, CA, USA) and quantified with FID.The C4-C10 compounds were separated on a semi-polar column (DB-624, 60 m × 0.25 mm ID × 1.4 μm; J&W Scientific) and quantified with a quadrupole MS detector.
For the quantification of C2-C4 hydrocarbons by FID, we used the external standard method.
Meanwhile, we used the internal standard method for the GC/MS quantification of VOCs, using four compounds as internal standards: bromochloromethane, 1,4-difluorobenzene, chlorobenzene-d5, and bromofluorobenzene.Table 2 summarizes the full list of the 103 VOC species identified and quantified using certified VOC mixture standards: 56 nonmethane hydrocarbons (Environmental Technology Center, Ottawa, Canada), 63 compounds of TO-15 (Linde Electronics and Specialty Gases, Stewartsville, NJ, USA), and α-and β-pinene (KIN-TEK Laboratories, Inc., La Marque, TX, USA).The calibrations were performed using five to six concentrations, ranging from 0.4 to 10 ppbv with three standard gas mixtures.The correlation coefficients for the calibration curves were mostly > 0.99 for the VOCs.The precision of the system for VOCs ranged from 0.5% to 4%.The detection limits of the GC-MS/FID system for the various tested compounds were in the range of 0.01-0.09ppbv.

Calculation of Emission rate
For each enclosure experiment, after determining the background and emission sample VOC Atmos. Chem. Phys. Discuss., doi:10.5194/acp-2016-1116, 2017 Manuscript under review for journal Atmos.Chem.Phys.Published: January 2017 c Author(s) 2017.CC-BY 3.0 License.concentrations, the emission rate was calculated with Equation (1): where ER is the emission rate (μg gdw -1 h -1 ) of the VOC species under the experimental temperature and PAR conditions; C and C 0 are the VOC species concentrations (μg m -3 ) in the emission and background samples, respectively; V is the total volume (m 3 ) of zero air purged into the enclosure bag; V 0 is the volume of dead air (i.e., residual air in the bag after enclosing but before purging zero air, m 3 ); Δt is the total enclosure time (h); and B is the dry weight (g dw) of the leaves on the enclosed branch.
In this study, we assumed that plants emitted no or negligible acetylene, so that the mass of acetylene in the background and emission sample was constant.Therefore, we estimated V 0 with Equation ( 2): where C b and C s are the acetylene concentrations (μg m -3 ) in the background and emission samples, respectively.

Sampling sites and plant species
We collected emission rate field measurements using the semi-static branch enclosure system at nine We selected sampling periods with no precipitation, low wind speed, higher visibility, and temperature and PAR close to the standard (i.e., temperature = 30°C, PAR = 1000 μmol m -2 s -1 ).All the field measurements were conducted during the daytime, usually from ~10:00 am to 15:00 pm.
Each enclosure experiment usually lasted for ~10 min.In most cases, the average temperature was 26-33°C and PAR was 600-1300 μmol photons m -1 s -1 .For each enclosure experiment, the temperature variation in the bag was typically < 2°C.

Data processing
In this study, the emission rates measured using the semi-static branch enclosure system were taken at the branch level and were derived under the experimental environmental conditions, such as temperature and radiation.Therefore, first, we normalized the results to standard conditions (i.e., temperature = 30°C; PAR = 1000 μmol photons m -2 s -1 ).Owing to a lack of definite response curves, the results were normalized with a Guenther et al. algorithm developed by Guenther et al. (1993).We assumed this to be a reasonable approximation when temperature deviated within 26-33°C and PAR deviated within 600-1300 μmol photons m -2 s -1 (Niinemets et al., 2011).Then, we converted the branch-level emission rates into leaf-level emission rates for future applications in canopy environment models to estimate BVOC emissions.The branch-level isoprene emission rates were converted into leaf-level emission rates by multiplying by 1.75 (Guenther et al., 1994;Guenther et al., 1996).Conversions were unnecessary for monoterpenes and other VOCs, due to their lower light dependence and a lack of available data.
To estimate representative emission rates, we summarized our field measurements, as well as a large number of measurements from China and abroad.The data from other studies collected under different environment conditions were normalized to standard conditions using the algorithm mentioned above.To minimize errors introduced by this extrapolation, we only included measurements conducted during the day in summer or growing season under temperature and light similar to the standards to obtain normalized emission rates of the studied plant species/genera.The reported isoprene emission rates were also measured at the branch or leaf level.Normalized leaf-level emission rates were also determined from branch-level rates using the method described above.
Before estimating the emission intervals and representative emission rates, we hypothesized that the VOC emission mechanisms were similar among species, and that their emissions were random.
All available normalized leaf-level isoprene and monoterpene emission rates from all studied trees, comprising > 400 and 300 individual values, respectively, were separately analyzed.VOCs, some of which had high emission intensities (> 10 μg C gdw -1 h -1 ).Future studies should include these VOC species when measuring BVOC emissions, and they should be considered in global and regional BVOC emission estimates.

Emission rate intervals
The frequency distributions of the measured emission rates using all available data for isoprene and monoterpenes for trees are shown in Figure 2. The emission potentials differed greatly among plant species, resulting in a wide range of emission rates.The isoprene emission rates ranged from 0 to 500 μg C gdw -1 h -1 , while those of monoterpenes fell predominantly within 0-100 μg C gdw -1 h -1 , with a sparse distribution of higher emission rates.According to the distribution, we divided the emission range of isoprene (x-axis of Fig. 2 The plant emission rates were inconsistent, but regular in distribution, falling into different intensity levels (Fig. 2).Based on these distributions, we defined seven categories for isoprene emission rates (lowest, lower, low, moderate, high, higher, and highest), and six categories for monoterpene emission rates (lowest, lower, low, moderate, high, and higher).We included more emission rate categories than those in previous studies.First, we estimated a range of emission rates for each category according to the distribution, and counted the values in each interval.improved much with enhanced rigorous measurements.

Determination of emission rates for 192 plants
The plant-specific emission rates were estimated based on the established emission intervals and representative emission rate.Emission rates were determined for each tree species based on the tendency of their measured emission rates to fall within each emission interval.Each plant was assigned the representative rate of the emission interval in which most rates fell.The normalized emission rates from a number of studies fell within a single interval for more than 70% of the 30 tree species/genera studied.When the reported rates fell within more than one category, the tree genus/species was assigned to the interval with the most rates.There were no occasions where reported rates fell equally into several intervals.
We analyzed the emission rates reported in China and abroad separately.It was a priority to apply the domestic measurements.The data from other countries would be selected when there was a lack of domestic data.When the determined emission rate obtained from domestic and foreign studies for one plant differed, and when there are much more measurements abroad than in China, such as measurements for Eucalyptus and Picea, the mean of two representative rates was considered as the emission rate.We assigned plant species with no measurements the same emission interval as that of measured species of the same family or genus.When there were no measurements for a family or genus, the plant species was assigned to the lowest emission category for isoprene and monoterpenes, with representative emission rates of 0.1 μg C gdw -1 h -1 (Klinger et al., 2002).
Table 6 lists the estimated emission rates based on measurements from domestic and foreign studies and the final values of leaf-level emission rates for the 30 dominant tree species in this study.
The final values for each plant species/genera were determined from the estimated domestic and foreign emission rates according to the principle described above.Broadleaf trees, such as Quercus, Populus, bamboo, and Eucalyptus, had higher or high isoprene emission rates.Meanwhile, needle-leaf trees, such as Pinus, Abies, and Cupressus, had higher monoterpene emission potentials.Similarly, the normalized isoprene and monoterpene emission rates of 149 shrub and grass genera and 13 crop species were determined based on the measurements from our study and those from other studies conducted in China (Wang et al., 2002;Wang et al., 2003b;Zhao et al., 2004;He et al., 2005;Bai and Baker, 2006;Xie et al., 2007;etc.),and abroad (Guenther et al., 1994;Geron et al., 2006;Simon et al., 2006; etc.) (Tables 7 and 8).In Table 7, only the average emission rates for 54 families of shrubs and grass are displayed due to space limitations.Overall, the emission potentials of crop, shrub, and grass were much lower than those of forest tree species; however, rubber had higher isoprene and monoterpene emission rates.

Comparison of representative emission rates
Previous studies of BVOC emission inventories in China applied a traditional method based on the  9 lists the isoprene emission categories and determined emission rates of some dominant tree species with higher emission potentials.In addition, previous studies used coarse emission intensity classifications (usually three to six classes) (Klinger et al., 2002;Yan et al., 2005;Wang et al., 2007), while we defined seven categories for the isoprene emission rates.Their use of fewer categories could result in underestimation of emission rates for plants with higher emission potentials.
For example, the emission rate estimates of Eucalyptus, Quercus, Populus, and bamboo were much lower in previous studies than in our study (Klinger et al., 2002;Yan et al., 2005;Wang et al., 2007), which could result in an underestimation of 4.9-7.8Tg C yr -1 for isoprene emissions in China, estimated using the methodologies described in Li et al. (2013) and Li and Xie (2014).It should be noted that different data sources contributed to differences in the determined emission rates among studies.In the future, we will perform more measurements for further improvement of the approaches and accuracy of estimating representative emission rates.

Evaluation
Accurately estimating representative BVOC emission rates is a challenging but critical step for constructing emission inventories.However, efforts have been made to develop an accurate emission rate database.Niinemets et al. (2011) presented a fairly thorough discussion on estimations of isoprenoid emission capacities from enclosure studies.They reviewed sources of uncertainties in the emission rate estimates, including measurement techniques (focusing on dynamic enclosure systems), calculations, extrapolations to standard emissions, and averaging.They first proposed a standardized protocol for the measurements and calculations and standardized the examination and screening of emission rate data from numerous reports before developing the emission rate database.This review helped the developers of emission factor databases to select and process original observations successfully.However, there might be still a large number of available varied emission rate data for one given plant species, while no data existing for some certain species.Therefore, it was necessary to determine an accurate representative value as the emission rate for application in the BVOC emission inventory estimates.Here, our established reasonable statistical method for determining speciesspecific representative emission rates, with production of more detailed emission intensity categories, accurate emission rate intervals and representative rates, should work effectively.Certainly, the foundation of our method was the evaluation and screening based on the quality of the emission rate observations and the use of reliable extrapolations, as suggested by Niinemets et al. (2011).First, they recommended that only two quality classes (quantitative measurements and semi-quantitative measurements) associated with dynamic systems could be used to construct BVOC emission Atmos. Chem. Phys. Discuss., doi:10.5194/acp-2016-1116, 2017 Manuscript under review for journal Atmos.Chem.Phys.Published: January 2017 c Author(s) 2017.CC-BY 3.0 License.
inventories.Meanwhile, non-quantitative measurements (i.e., those conducted using static enclosure systems or possibly semi-static and some dynamic systems) should not be used in BVOC modeling.
In our study, the summarized reported emission rate measurements from abroad were mainly derived from dynamic open systems.While those in China were measured using simple static systems, primarily due to a lack of other measurements, which should only be used in emission rate estimates when there are no other available observations for a region.The errors introduced due to extrapolation were minimized and within a reasonable range, as discussed in Section 2.2.
In this study, we focused on gaining a comprehensive understanding of BVOC emissions from plants in China and exploring scientific methods for the accurate estimation of plant species-specific representative BVOC emission rates based on reliable original emission rate observations.

Conclusions
We performed field measurements of BVOC emissions from 50 plant species, including 36 trees and 14 shrubs, at nine locations in China using our established semi-static enclosure system.Emission rates of 102 VOC species (i.e., isoprene, α-pinene, β-pinene, and other VOC species) from 67 experiments were determined by analyzing with a custom-built online GC-MS/FID system.Of the studied plants, broadleaf trees were the greatest potential emitters of isoprene, while needle-leaf trees emitted more pinene.Shrubs had lower isoprene and pinene emissions, but higher emissions of other VOCs.Overall, deciduous broadleaf trees had higher isoprene emission intensities than evergreen broadleaf trees.B. papyrifera was the strongest emitter of isoprene, with a normalized leaf-level emission rate of 838.62 μg C gdw -1 h -1 .One broadleaf tree species, L. formosana, had considerable αand β-pinene emission rates of 707.12 and 2542.13 μg C gdw -1 h -1 , respectively.Overall, shrubs emitted more other VOCs than isoprene and pinene.Other VOC species, including methyl methacrylate, isopropylbenzene, isopentane, acetone, ethane, propane, toluene, and xylene, were the dominant components even with high emission intensities.It was suggested that future studies should consider their emissions in global or regional estimations of BVOCs.
We established a statistical approach to estimate representative emission rates, based on a summary of our field measurements and reported emission rates from China and abroad.First, we produced isoprene and monoterpene emission intensity categories based on statistics.Tree species fell into various emission intensity categories, including lowest, lower, low, moderate, high, higher, and highest.The isoprene emission rate intervals and representative rates were: lowest, 0.08-0.11and 0.1 μg C gdw -1 h -1 ; lower, 0.9-1.3 and 1.0 μg C gdw -1 h -1 ; low, 5.2-6.5 and 5.8 μg C gdw -1 h -1 ; moderate, 13.1-15.3and 14.4 μg C gdw -1 h -1 ; high, 31.1-37.0and 33.6 μg C gdw -1 h -1 ; higher, 67.2-75.1 and 70.1 μg C gdw -1 h -1 ; and highest, 135.1-157.6 and 142.5 μg C gdw -1 h -1 .The monoterpene emission rate intervals and representative rates were: lowest, 0.08-0.11and 0.1 μg C gdw -1 h -1 ; lower, 0.17-0.22 and 0.2 μg C gdw -1 h -1 ; low, 0.5-0.7 and 0.6 μg C gdw -1 h -1 ; moderate, 1.2-1.5 and 1.4 μg C gdw - with previous studies, our emission rate categories were more detailed, and the emission rate interval and representative rates were more accurate.They would be further improved by integrating more field measurements in the future, which would be significant for reducing the uncertainty in the determination of emission rate and estimation of emissions in BVOC emission inventories.
Based on the emission intervals, we determined emission rates for 192 plant species/genera, including 30 dominant tree species, 149 shrub and grass genera, and 13 crop species.Broadleaf trees, including Quercus, Populus, bamboo, and Eucalyptus, had higher or high isoprene emission rates.
Meanwhile, needle-leaf trees, including Pinus, Abies, and Cupressus, had higher monoterpene emission potentials.The emission potentials of crops, shrubs, and grasses were much lower than those of forest plants.Of the crop species, rubber had higher isoprene and monoterpene emission rates.In our study, the isoprene emission rates of several tree species with high emission potentials were higher than those in previous studies, which could explain why China's BVOC emissions have frequently been underestimated much (Li et al., 2013;Li and Xie, 2014).
Despite our efforts to achieve reliable estimation of accurate representative BVOC emission rates to construct an emission inventory, it is still necessary to establish a measured emission rate database with minimal uncertainties.This would improve the accuracy of our estimations by integrating more quantitative measurements, as suggested by Niinemets et al. (2011), and is especially important in China.Besides, it is necessary to conduct measurements in different regions to obtain the representative emission rates of plants in the whole China.

Figure and table captions:
Table 1.Isoprene emission categories and determined emission rates of the dominant tree species in China's emission inventories (μg C gdw -1 h -1 ).
Table 2. 103 VOC species quantified with the GC-MS/FID system.
Table 3. Plant species sampled at each location, sampling time, and measured emission rates.
Table 4. Isoprene and monoterpene emission rate categories, emission rate ranges, statistics, and distributed plant species in each category (μg C gdw -1 h -1 ).
Table 7. Averaged normalized leaf-level isoprene and monoterpene emission rates for 54 shrub and grass families in China (μg C gdw -1 h -1 ).
locations in China from May to October in 2014 and 2016.The sites were located in Beijing (Peking University, Jiufeng Mountain, Yunmeng Mountain, Wuling Mountain, and Beijing Gardening Research Institute), Hubei (Yunwu Mountain and Maan Mountain, Wuhan), and Sichuan (Longquan Mountain, Chengdu) provinces.To avoid excessive oxidation and loss of reactive compounds in the canisters, most locations were set far from anthropogenic sources.Mature and healthy plants and branches were selected for the BVOC emission measurements.The age of trees was not recorded due to lack of investigation and guidance from professionals.Table3lists the measured plant species at each location, including trees, shrubs, and vines.In total, we performed 67 experiments of 50 plant species.Due to the limitation of experiment equipment and field conditions, each plant species had one or two replicates of samples for BVOC emissions.

Figure 1 .
Figure 1.Schematic of the semi-static branch enclosure system used in this study.

Figure 2 .
Figure 2. Frequency distribution of (a) isoprene and (b) monoterpene emission rates from forest trees.
Geron et al.
Table3lists by sampling location the plant species, vegetation type, and measured leaf-level emission rate normalized to standard conditions.These are the first reported emission measurements of 102 VOC species.Owing to limited space, we only included the individual emission rates of the three the strongest emitter of isoprene, with an emission rate of 838.62 μg C gdw -1 h -1 .Phyllostachys aureosulcata f. spectabilis, a species of subfamily Bambusoideae, had a high emission potential of 187.73 μg C gdw -1 h -1 .Styphnolobium japonicum, Populus tomentosa Carr., Platanus orientalis, and Quercus wutaishanica were all high isoprene emitters, producing > 60 μg C gdw -1 h -1 .However, Of the studied plants, broadleaf trees were the greate, 2017ential emitters of isoprene, while needle-leaf trees emitted more pinene.Shrubs had lower isoprene and pinene emissions, but higher emissions of other VOCs.In addition, 29 of the 39 broadleaf tree species emitted isoprene (i.e.,Atmos.Chem.Phys.Discuss., doi:10.5194/acp-2016-1116,2017Manuscriptunderreviewfor journal Atmos.Chem.Phys.Published: January 2017 c Author(s) 2017.CC-BY 3.0 License.emissionrate> 0.1 μg C gdw -1 h -1 ).In general, deciduous broadleaf trees had higher isoprene emission intensities than evergreen broadleaf trees.Broussonetia papyrifera (Linn.)L'Hér.exVent.wasvariabilis was considered to have no or little isoprene emissions.The isoprene emission rates of needle-leaf trees and shrubs were mostly < 0.3 μg C gdw -1 h -1 .Only 3 of the 14 needle-leaf trees emitted isoprene, while all of these species emitted pinene.Among them, Metasequoia glyptostroboides, Pinus tabulaeformis Carr., and Axodiaceae had especially high pinene emissions.Notably, one broadleaf tree species, Liquidambar formosana Hance, had considerable α-and β-pinene emission rates of 707.12 and 2542.13 μg C gdw -1 h -1 , respectively.In addition, the shrub Cotinus coggygria Scop.was a high pinene emitter.Owing to a large amount of compounds, the totaled emission rates of other VOCs might be high despite lower emissions of individual VOCs.For example, L. formosana was the highest emitter of other VOCs (90.44 μg C gdw -1 h -1 ), followed by Pteroceltis tatarinowii Maxim.(23.68 μg C gdw -1 h -1 ).For most shrub species, other VOCs contributed the most to their total VOC emission potentials.The other VOCs were dominated by light alkanes, alkenes, aromatics, and straight-chain alkanes, as well as some carbonyls, while halohydrocarbons were minimally emitted.For example, alkanes accounted for 43% of the total other VOC emissions in L. formosana.Meanwhile, 34% of other VOC emissions in Q. wutaishanica were attributed to alkenes.Aromatics accounted for 51% of other VOC emissions in M. glyptostroboides.In general, oxygenated compounds accounted for a greater proportion of the other VOCs in broadleaf tree species (> 50% for most trees) than in needle-leaf trees.Methyl methacrylate, isopropylbenzene, isopentane, acetone, ethane, propane, toluene, and xylene were the dominant components of the other Based on the statistics listed in Table4, the measured values displayed dispersed distributions (i.e., large SDs relative to the mean) in each emission rate range.If the mean were considered to be the representative emission rate for each emission category, large uncertainties would be introduced into the emission rate estimations of individual plants.Therefore, we considered additional statistics for all the values in each emission rate range separately, which each had a normal distribution.Using the t-test, we determined the 95% confidence interval (CI) of each range, which we considered to be the final emission rate interval for each emission category (Table5).Table5lists the statistically valid sample sizes included after the t-test.Values from the initial estimated emission rate range that fell outside the new interval were eliminated due to possible measurement errors, resulting in smaller SDs.The means of the new intervals were considered as the representative emission rates for each Table4lists the statistical frequency, mean, an, 2017dard deviation (SD) of emission rates and main plant species distributed in each category.The lowest isoprene emission category had the most isoprene emission rate measurements (42% of the total), while the higher category comprised 19% of the total measurements, and other categories each comprised only 7-9% of the total.The largest number of plants, including Tilia, Paulownia Sieb., Betula, Quercus Suber, and most needle-leaf trees, were in the lowest isoprene emission category.Future studies should perform more emission measurements of plants in the lower and high isoprene emission categories.For monoterpenes, emission rates were uniformly distributed in the six emission categories, with frequencies of 15-20%.Future studies should perform more intensive measurements of plants in the low and lowest monoterpene emission intensity categories.It should be noted that future studies integrating more emission measurements with different intensities would create more detailed categories, accurate emission rate intervals and representative rates.Thus, the accuracy in the estimation of representative emission rates for each plant would be Atmos.Chem.Phys.Discuss., doi:10.5194/acp-2016-1116,2017Manuscriptunder review for journal Atmos.Chem.Phys.Published: January 2017 c Author(s) 2017.CC-BY 3.0 License.

Table 1 .
Isoprene emission categories and determined emission rates of the dominant tree species in 693 China's emission inventories (μg C gdw -1 h -1 ).

Table 6 .
Normalized leaf-level isoprene and monoterpene emission rates for 30 dominant tree genera/species in China (μg C gdw -1