Estimation of biogenic volatile organic compound ( BVOC ) 1 emissions from the terrestrial ecosystem in China using real-time 2 remote sensing data

8 Because of the high emission rate and reactivity, biogenic volatile organic 9 compounds (BVOCs) play a significant role in the terrestrial ecosystems, human health, 10 secondary pollution, global climate change and the global carbon cycle. Past 11 estimations of BVOC emissions in China on the national scale were based on outdated 12 empirical algorithms suggested around 10 years ago and coarsely-resolved 13 meteorological data, and there have been significant aging of the land surface 14 parameters in dynamic meteorological models and BVOC estimation models, leading 15 to large inaccuracies in the estimated results. To refine BVOC emission estimations for 16 China, we used the latest algorithms of MEGAN (Model of Emissions of Gases and 17 Aerosols from Nature), with modified MM5 (the Fifth-Generation Mesoscale Model) 18 providing highly resolved meteorological data, to estimate the biogenic emissions of 19 VOCs for China in 2006. Real-time MODIS (Moderate Resolution Imaging 20 Spectroradiometer) land-use and vegetative cover data were introduced in MM5 to 21 replace the land surface parameters and to improve the simulation performance of 22 MM5. Highly-resolved 8-day MODIS leaf area index (LAI) data were also used to 23 determine the influence of LAI and leaf age deviation from standard conditions. In this 24 study, the annual BVOC emissions for the whole country totaled 13.02 Tg C, higher 25 than the recent estimation of Tie et al. (2006) by 19.9%, which might be attributed to 26 the aged land-surface data and meteorological input, as indicated by several case 27 studies, and higher than Klinger et al. (2002) by 72.9%. Therein, the most important 28 individual contributor was isoprene (9.39 Tg C yr -1 ), followed by α-pinene (1.24 Tg C 29 yr -1 ) and β-pinene (0.81 Tg C yr -1 ). Spatially, isoprene emission was concentrated in 30 South China, which is covered by large areas of broadleaf forests and shrubs. While 31 Southeast China was the top-ranking contributor of monoterpenes, in which the 32 dominant vegetation genera consist of coniferous forests. In the main southern cities 33 (Fujian, Guangxi, Hainan, Hunan, Jiangxi, and Yunnan), Shaanxi and Inner Mongolia, 34 BVOC emissions predominated over anthropogenic NMVOC emissions. Temporally, 35 BVOC emissions primarily occurred in July and August, with daily emissions peaking 36 at about 13:00~14:00 hours (Beijing Time, BJT). 37


Introduction
Large quantities of non-methane volatile organic compounds (NMVOCs) are emitted from various anthropogenic and natural sources, such as vegetation, marine algae (McKay et al., 1996) and microbiological decomposition (Kuzma et al., 1995;Zemankova and Brechler, 2010).On the global scale, natural emissions of NMVOCs equal or exceed anthropogenic emissions, nearly by an order of magnitude (Guenther et al., 1995;Benkovitz et al., 2004;Monks et al., 2009).On the regional scale, although anthropogenic sources usually dominate in urban areas, in many cases, BVOCs made significant contributions to the overall VOC inventories of both urban and rural areas (Benjamin et al., 1997).
The importance of BVOCs was first acknowledged about 50 yr ago (Went, 1960) and, compared with other abundant NMVOCs, is increased because of its key role in tropospheric physics and chemistry.First, BVOC emissions strongly influence the composition of the troposphere (Guenther et al., 1999;Monks et al., 2009).In particular, isoprene and monoterpenes are thought to account for a major proportion of the total BVOC emissions (Kesselmeier and Staudt, 1999;Goldstein and Galbally, 2009;Monks et al., 2009).Second, many BVOC species are emitted in copious quantities and have extremely high reactivities with tropospheric oxidants.Through their oxidation, BVOCs can significantly affect the concentrations of ozone (O 3 ), hydroxyl radicals (OH), nitrogen oxides (NO x ) and so forth, consequently exerting a controlling influence on the oxidizing capacity of the atmosphere, and thus affecting local and global air quality (Ryerson et al., 2001;Wiedinmyer et al., 2004;Arneth et al., 2008;Goldstein and Galbally, 2009;Hallquist et al., 2009;Monks et al., 2009).Third, BVOCs, through complex oxidation processes, have also been implicated as key precursors to biogenic secondary organic aerosol (BSOA), thereby providing an additional source/burden of aerosol in the atmosphere and further exerting a strong influence on global climate-related issues (Griffin et al., 1999;Atkinson and Arey, 2003;Kanakidou et al., 2005;Szidat et al., 2006;Goldstein and Galbally, 2009;Hallquist et al., 2009;Pacifico et al., 2009;Introduction Conclusions References Tables Figures
To further explore the roles of BVOCs, it is essential to accurately estimate time-and space-resolved BVOC emissions.The most important emitter of BVOCs is vegetation, especially forest ecosystems (Zemankova and Brechler, 2010).China is a country that encompasses large areas, which are covered by a variety of plants, leading to remarkable temporal and spatial variations in BVOC emissions.In recent decades, studies have been conducted in China focused on the development of national or regional BVOC emission inventories using various models or approaches, which reported an annual emission value of 4.06∼15.00Tg C for isoprene and 3.16 ∼ 4.30 Tg C for monoterpenes on the national level (Guenther et al., 1995;Klinger et al., 2002;Wang et al., 2003;Tie et al., 2006;Wang et al., 2007;Leung et al., 2010;Zheng et al., 2010;Wang et al., 2011).
However, past estimations were deficient with regard to estimation algorithms and input data.In terms of methodology, the estimation algorithms applied in these past studies did not fully consider the factors controlling BVOC emissions, such as LAI and leaf age.Some studies even unreasonably assumed the emissions of monoterpenes and other reactive VOCs to be solely temperature dependent, failing to recognize their dependence on light (Wang et al., 2003;Zheng et al., 2010).In addition, some estimates used outdated USGS (United States Geological Survey) land-cover data to identify vegetation distributions and/or to drive the dynamic meteorological model (Tie et al., 2006;Wang et al., 2011).The use of USGS data resulted in the misidentification of plant function types (PFTs) and large uncertainties in the simulated meteorological results, and potentially led to disagreements between the land surface parameters of the dynamic meteorological model and the BVOC estimation model.Many studies have explored the importance of land surface parameters in meteorological simulation (Molders, 2001;Atkinson, 2003) and have highlighted the potential of satellite data to Introduction

Conclusions References
Tables Figures

Back Close
Full improve simulation performance in dynamical meteorological models (Gutman and Ignatov, 1998;Crawford et al., 2001;Kurkowski et al., 2003;de Foy et al., 2006;Yucel, 2006;Ge et al., 2008;Meng et al., 2009).Thirdly, many studies were based on coarsely resolved meteorological data interpolated from daily or monthly weather datasets, leading to coarse resolutions of emission inventories (for example 0.5 • × 0.5 • in studies by Guenther et al., 1995) and failures to capture maximum emission values (Ashworth et al., 2010).Consequently, all the factors above jointly led to large uncertainties and great variations among the results of different studies.
In this work, we aimed to estimate the amounts, spatial distributions and temporal variations of BVOC emissions in China in 2006 using MODIS-MM5-MEGAN, with MM5 providing hourly meteorological outputs to drive the estimations.Real-time MODIS data were introduced into MM5 to update the land surface parameters and improve the simulation performance, and were also used to provide a highly resolved input database for the calibration and estimation of BVOC emissions.Two primary classes of BVOCs, isoprene (C 5 H 8 ) and monoterpenes (C 10 H 16 ), including a group of α-pinene, β-pinene, limonene, myrcene, sabinene, 3-carene and ocimene, were considered in this study because of their substantial contributions to total BVOC emissions and their key role in atmospheric processes.

Estimating BVOC emissions
We estimated BVOC emissions based on the parameterized canopy environment emission activity (PCEEA) algorithms in MEGAN, described by Guenther et al. (2006) and Sakulyanontvittaya et al. (2008), as shown schematically in Fig. 1.
The net BVOC emission fluxes (mg m −2 h −1 ) into the above-canopy atmosphere were empirically specified according to Eq. ( 1)

Conclusions References
Tables Figures

Back Close
Full where EF (mg m −2 h −1 ) is a standard canopy-scale emission factor, which represents the BVOC emission rates under standard conditions.Global gridded EFs for isoprene and seven monoterpene species, which have a base resolution of 30 s for the year 2000 based on measured tree species emission factors and global PFT distributions, were downloaded from the CDP (Community Data Portal) website (available at http: //cdp.ucar.edu/).Changes in BVOC emissions due to deviations from standard conditions were modified through a set of dimensionless emission activity factors (γ T , γ P , γ age , γ LAI , γ SM and ρ).The light dependence of the BVOC emission processes was considered using the light-dependent function (LDF).Detailed information and calculation processes for all correction terms can be found in the published reports of Guenther et al. (2006) and Sakulyanontvittaya et al. (2008), so no further details are provided here.
In our estimations, the influences of soil moisture and detailed canopy information were neglected; thus, γ SM and ρ, which represent the influence of soil moisture and production and loss of BVOCs within the canopy, respectively, were set to 1.
Highly resolved meteorological outputs of air temperatures at 2 m and solar shortwave radiations from MM5 were used to estimate the light and temperature dependencies (γ T and γ P ) of BVOC emissions.The horizontal resolution of the modeling domain was 12 km × 12 km, centered at (37.40 • N, 102.52 • E) with 440 × 380 cells in the horizontal direction and 35 layers in the vertical direction.MM5 was run for the entire year of 2006, and each run covered 3.5 days with 12-h spin-up time.
The 8-day MODIS LAI data (MCD15A2), which have a 1 km resolution for the year 2006 and provides information regarding the seasonal evolution of vegetation characteristics (Justice et al., 2002), were used to estimate the influences of LAI (γ LAI ) and leaf age (γ age ) on BVOC emission capacities.Introduction

Conclusions References
Tables Figures

Back Close
Full

Update of real-time MODIS data in MM5
Accurate simulations of meteorological fields are important for the estimation of BVOC emissions.It is now widely recognized that several key land surface parameters, including land cover, vegetation fraction (VGF), soil temperature and moisture, significantly affect land-atmosphere interactions and are thus important in weather simulations (Wittich and Hansing, 1995;Crawford et al., 2001;de Foy et al., 2006;Yucel, 2006).By default, MM5 uses the USGS global 1 km land-use data derived from the AVHRR (Advanced Very High Resolution Radiometer) observation, which is based on 1-yr data from April 1992 to March 1993.The VGF data in MM5 are derived from monthly AVHRR NDVI (Normalized Difference Vegetation Index) data at a coarse resolution of 10 min (de Foy et al., 2006;Meng et al., 2009).
However, over the last decade, global terrestrial ecosystems underwent great changes, such as urbanization, desertification and deforestation, causing the existing land-use data in MM5 to be outdated and, thus, likely to produce errors in weather simulations (Pielke et al., 2002;Yucel, 2006).By far, many studies have been conducted to update land surface data in meteorological models with satellite data (Crawford et al., 2001;Kurkowski et al., 2003;Tian et al., 2004;de Foy et al., 2006).Studies also reasoned that the global coverage, enhanced resolutions and accurate calibrations for retrievals of land surface parameters of MODIS enabled regional and global studies of biogeochemical cycles, land cover changes and so forth, for which AVHRR-based studies have limitations (Justice et al., 2002).
We introduced the latest MODIS land-use data and VGF data to replace the landsurface parameters in MM5.MODIS land-use data (MCD12Q1) for the year 2006 and water mask data (MOD44W) for the year 2000, which both have a resolution of 500 m (Justice et al., 2002), were used to obtain a new land-use map by mapping the 17 MODIS land-use categories defined by the International Geosphere-Biosphere Program = (IGBP) onto the existing 24 USGS categories, as shown in Table 1.To validate the accuracy of the MODIS land-use data, the 1:100, 0000 plant distribution map Introduction

Conclusions References
Tables Figures

Back Close
Full (Fig. 2c), which is based on local field surveys from the Plant Research Institute of the Chinese Academy of Sciences, was used for comparison.The land-use distribution in China changed dramatically over the past decade, as shown by Fig. 2. For example, according to the USGS data, South China was mainly covered by large areas of crops (Fig. 2a), while, according to MODIS observations, South China contained mixed forest, grass and shrubs (Fig. 2b).For plant distributions obtained from field investigations, it demonstrated similar distribution patterns with those of the MODIS data.Therefore, we concluded that the MODIS land-use data better reflects the present land cover characteristics of China and may help to improve the simulations.
3 Results and discussions

BVOC emission budgets
Hourly emissions of isoprene and seven monoterpene species in China were calculated for the year 2006, with a spatial resolution of 12 km × 12 km.In the following section, all the results were measured as carbon weights of the constituent compounds, unless stated otherwise.Seasonal and annual total emission budgets determined by this study are listed in Table 2.As shown, the annual BVOC emissions totaled 12.97 Tg C (which equals Introduction

Conclusions References
Tables Figures

Back Close
Full 14.70 Tg compound).Regarding the relative contributions of individual species, the dominant contributor was isoprene, which had an annual emission budget of 9.36 Tg C, accounting for approximately 71.6 % of the total BVOC budget.The next most predominant contributor was α-pinene (1.24 Tg C yr −1 ), which was responsible for 9.8 % of the total BVOC emissions and 34.3 % of the total monoterpene emissions, followed by β-pinene (0.81 Tg C yr −1 ) and 3-carene (0.67 Tg C yr −1 ).The other four monoterpene species were less significant, and the annual emission budgets on the national scale were as follows: ocimene (0.32 Tg C), limonene (0.28 Tg C), sabinene (0.18 Tg C) and myrcene (0.11 Tg C).Overall, the annual emission budget of monoterpenes was 3.61 Tg C, contributing approximately 28.4 % to estimated BVOC emissions.According to our estimations, the annual total emission budget of isoprene was approximately 2.5 times that of the monoterpenes.
Because of the distinct reactivities of BVOC species, the better understanding of the relative contributions of individual BVOC species may be essential for the further exploration of secondary products of BVOCs and the determination of appropriate regulatory oxidant control strategies (Hoffmann et al., 1997;Wiedinmyer et al., 2004;Hallquist et al., 2009).

Spatial distributions of BVOC emission fluxes
Because of the complex plant distributions (Fig. 2) and the wide range of meteorological variations across China, BVOC emissions showed considerable spatial variations.As described in Sect.2.1, the spatial distribution of BVOC emission fluxes (expressed as the total emission of BVOCs per unit area per unit time) depended on the distribution of standard EFs, LAI as well as meteorological conditions.
The field measurements of previous studies indicated that broadleaf forests (especially Quercus, Populus and Eucalyptus) and shrubs were of high isoprene emission capacities, while an intense emission of monoterpenes generally corresponded to the dense distribution of coniferous forests (especially Pinus and Picea).soever (Kesselmeier and Staudt, 1999;Wang et al., 2003;Guenther et al., 2006;Wang et al., 2007Wang et al., , 2011;;Sakulyanontvittaya et al., 2008;Karl et al., 2009;Zheng et al., 2010).Thus, the types and magnitudes of BVOC emission fluxes were controlled by the amounts and compositions of plant species.
As shown in Fig. 2, forests and shrubs were mainly distributed in Northeast and South China.According to the survey results from the Plant Research Institute, Northeast China was primarily covered by deciduous coniferous forests (mainly Larix gmelini) and deciduous broadleaf forests (mainly Quercus mongolica, Tilia Mongolia and Betula platyphylla).By comparison, the distribution pattern of tree species in South China was more complex.Large areas of evergreen coniferous forests (mainly Pinus massoniana and Cunninghamia lanceolata) and shrubs were found in Southeast China, while the main plant genera in Southwest China were larger groups of evergreen tree species, including evergreen broadleaf forests (e.g.Quercus aquifolioides), evergreen coniferous forests (Picea likiangensis var.balfouriana,Pinus yunnanensis and Keteleeria evelyniana Mast) and shrubs.Notably, large areas of tropical rain forests were concentrated in the southeast of Tibet and south of Yunnan Province.Additionally, high fractional cover of deciduous broadleaf forests (mainly Quercus variabilis and Quercus liaotungensis) was found south of Shaanxi.Correspondingly, the spatial distributions of standard emission factors for isoprene and monoterpenes, presented in Fig. 3 (in which blank areas represent districts that had no plant cover), were consistent with the plant distributions in China (Fig. 2).On the whole, BVOC emission hotspots mainly occurred in the northeast and south of China.Specifically, plants in the northeast and south of China as well as in the south of Shaanxi exhibited high isoprene emission capacities (Fig. 3a).High monoterpene emission capacities were found in the northeast and southeast of China as well as the Sichuan-Tibet area (Fig. 3b).Introduction

Conclusions References
Tables Figures

Back Close
Full As illustrated in Fig. 4, the distribution patterns of biogenic emissions modified by real conditions agreed well with those under standard conditions (Fig. 3).The high EFs as well as the low latitude and resulting high temperature in the south of China led to strong isoprene emission there (Fig. 4a).High isoprene emission flux values were also found in the south of Shaanxi.Although the emission capacities of the plants in the northeast of China were extremely high (Fig. 3a), the real emission flux values were relatively low (Fig. 4a).This result is because the high latitude of Northeast China led to a relative low temperature throughout the year.Additionally, as mentioned above, Northeast China was mainly covered by deciduous forests that cease to grow and nearly emit no isoprene in winter.High monoterpene emission flux values were centered in the south (especially the southeast) of China, where a high density of coniferous forests were concentrated (Fig. 2c).In Southwest China, especially the Sichuan-Tibet area, the high altitude and resulting low temperature of the Tibetan Plateau led to relatively low emission flux.
The lowest BVOC emissions occurred in Northwest China (Fig. 4), which is primarily covered by barren land and low emitting grass (Fig. 2).In North China and the north of East China region, which is mainly covered by crops (Fig. 2), BVOC emissions were also low (Fig. 4).

Province-specific emission
Each province's relative emission contribution depended on plant distribution, forested area, climatic conditions, and so forth.Detailed information depicting plant cover by province, listed in Table S1, was based on tree species data from the Chinese Academy of Sciences.
According to the estimated seasonal and annual BVOC emission budgets for 33 provinces (Table 3) respectively contributed 9.5 %, 7.9 % and 7.0 % to the total isoprene emission.The three provinces all had dense plant cover.For example, Yunnan Province had approximately 11.0 × 10 4 km 2 of forests, accounting for 28.7 % of its land area (Table S1).
Subtropical evergreen broadleaf forests were also common in Yunnan, accounting for approximately 28.2 % (3.1 × 10 4 km 2 ) of the forested area.Moreover, Yunnan was also covered by large areas of shrubs (8.6 × 10 4 km 2 ) (Table S1).The high isoprene emission capacities of broadleaf forests and shrubs led to the strong emission of isoprene in Yunnan.However, our results also depicted a contradiction between the extremely high forest and shrub coverage (30.1 × 10 3 ,km 2 of broadleaf forests and 108.4 × 10 3 km 2 of shrubs) and the relatively low isoprene contribution in Sichuan Province.As mentioned above, this phenomenon could result from the low temperatures (due to terrain height) in the west of Sichuan and the low solar radiation in the east of Sichuan, combined with the enclosed terrain and the abundant cloudiness in the atmosphere (Fig. 5).
The emission rates of monoterpenes were in general agreement with the distribution of coniferous forests.Specifically, Guangxi Province was the highest emitter, with an annual emission budget of 0.36 Tg C (contributing 9.9 % to the annual emission of monoterpenes), followed by Yunnan (0.35 Tg C yr −1 , 9.8 %) and Hunan (0.29 Tg C yr −1 , 8.1 %) (Table 3).Coniferous forests were extensively distributed in these regions.For example, Guangxi was covered by 5.0 × 10 4 km 2 of coniferous forests, mainly Pinus massoniana, accounting for 72.5 % of the forested area of Guangxi (6.9 × 10 4 km 2 ).
In determining regional environmental strategies, estimating province-specific BVOC emissions can provide constructive information for local administrations to make effective environmental decisions associated with oxidant control strategies, large-scale tree planting and urban greening programs (Schell et al., 2001;Liao et al., 2007;Leung et al., 2010).

Temporal variations in BVOC emissions
There is increasing evidence that the seasonal changes of meteorological conditions and plant phenology greatly affect fluctuations in BVOC emission.Thus, seasonal Introduction

Conclusions References
Tables Figures

Back Close
Full variabilities must be considered to achieve an accurate estimation of vegetative BVOC emissions.
Corresponding to the seasonal variations of LAI, temperature and solar radiation (Fig. 5), the distributions of isoprene (Fig. 6) and monoterpene (Fig. 7) emission fluxes showed considerable seasonal variations.
Isoprene emissions were highest throughout the country in the summer and there were fewer spatial differences in emission distributions, consistent with the high temperatures, the high solar radiation, and the vigorous growth of forests, resulting in the high LAI values across China (Fig. 5).In winter, the LAI values decreased dramatically with the loss of leaves in deciduous forests, particularly in Northeast China, where temperate deciduous forest dominated.By comparison, in the southeast and southwest of China, the LAI values remained high throughout the year because of the prevalence of tropical and subtropical evergreen forests (Fig. 5a).However, the low temperatures and low solar radiation in winter (Fig. 5) still led to low BVOC emissions across China (Fig. 6d).In spring and autumn, the distribution pattern of isoprene emissions were similar and concentrated in the south of China (Fig. 6a and c), partly because of the relatively dense vegetation cover in South China.The distribution pattern of temperature, which decreasing from the south to the north and from the east to the west, and decreasing solar radiation from the northwest to the southeast (Fig. 5b and c) could also lead to the dense distribution of isoprene in South China.
The spatial distribution patterns and seasonal variations of monoterpene emissions (Fig. 7) were similar to those of isoprene emissions, but the magnitudes of the former were much smaller and the spatial variations were larger.In addition, the distribution of monoterpene emissions were more centered in Southeast China (Fig. 7), where the density of coniferous forests was higher (Fig. 3).
Figure 8 shows the bell-shaped monthly evolution patterns of the total BVOC on a national level as well as the monthly changes of LAI, temperature and solar radiation.Marked differences in BVOC emissions were observed from month to month, which generally agreed with LAI, temperature and solar radiation trends.BVOC emissions Introduction

Conclusions References
Tables Figures

Back Close
Full concentrated from April to October and peaked at 3.17 Tg C in July and August (Table 2).As the season transitioned with accompanying decreases in temperature, radiation and vegetation, BVOC emission intensities declined dramatically, reaching a lowest value of 0.11 Tg C in January (Table 2).Thus, monthly isoprene emissions fluctuated between a maximum (2.41 Tg C) in July and August and a minimum (0.05 Tg C) in January.The same pattern was observed for monoterpene emissions, for which a peak emission of 0.76 Tg C occurred in July and the lowest emission (0.05 Tg C) occurred in February.Notably, the seasonal variability of monoterpenes was not as obvious as that of isoprene.The ratio of monoterpenes emitted in summer (2.04 Tg C) to that in winter (0.17 Tg C) was approximately 12.0, much smaller than the value (32.9) for isoprene (6.25 Tg C in summer and 0.19 Tg C in winter), an observation that is attributed to the lack of light dependence of monoterpene emission and the strong temperature dependence of isoprene emission (Tie et al., 2006).
In addition to seasonal changes, the temporal patterns of BVOC emissions displayed obvious diurnal cycles, which were primarily determined by diurnal variations in temperature and solar radiation.Biogenic isoprene emissions peaked at approximately 13:00 (BJT) and were lowest at night.During the night, because of the absence of solar radiation and isoprene's strong light-dependence, emissions of isoprene nearly ceased.In general, the diurnal pattern of monoterpene emissions agreed with that of isoprene, with the strongest emission of the former occurring at about 14:00 (BJT).
By comparison, the diurnal variation observed for monoterpene emissions was much smaller than isoprene emissions.In addition, the monoterpene emissions maintained high levels during the night.According to previous studies of the oxidation mechanisms of BVOCs, isoprene oxidation is dominated by reactions with hydroxyl ( . OH) during the day and reactions with nitrate radicals (NO 3 ) during the night, while ozone (O 3 ) plays a minor role at all times (Atkinson and Arey, 2003;Monks et al., 2009).Thus, the strong diurnal cycles of BVOCs have important effects on the formations and diurnal variations of ozone and SOA.Introduction

Conclusions References
Tables Figures

Back Close
Full

Comparisons to anthropogenic emissions
Recent studies proposed a link between BVOCs and AVOCs (anthropogenic volatile organic compounds), implying an interaction between them in the formation of SOA (Goldstein and Galbally, 2009).More important, because of the significant differences in the composition, reactivity and oxidation products, the regional variation of relative weights for BVOCs compared with AVOCs could affect local air quality and pollution controlling strategies (Lane and Pandis, 2007).Therefore, to further explore the role of BVOCs in atmospheric chemical processes, it is essential to compare BVOC emissions to anthropogenic emissions.The inventory of AVOC emissions in 2006 was taken from the study by Zhang et al. (2009), which gave AVOC emissions for all the major anthropogenic sources (power, industry, residential and transportation) in China.In total, approximately 23.2 Tg NMVOCs (measured as full molecular weights of the constituent compounds) were emitted from China in 2006 (Zhang et al., 2009).This value is approximately 1.6 times our estimate of natural emissions (14.7 Tg compound yr −1 ).Thus, the total emission of NMVOCs was 37.94 Tg on the annual scale for the base year 2006, of which Guangdong made the largest contribution with an annual emission of 2.62 Tg, followed by Sichuan (2.32 Tg yr −1 ) and Shandong (2.20 Tg yr −1 ).However, because of the complex plant distribution and economic structure across China, the relative ratio of BVOCs/AVOCs varied greatly by region.
The total amount of VOC emissions from natural and anthropogenic sources for each province is listed in Table 3.The distribution of anthropogenic NMVOC emissions was closely related to energy-consuming activities and population density, and reached a high value in the eastern area of China, while the distribution of BVOCs was closely dependent on plant distribution and centered in the south and northeast of China.Of the 32 provinces for which complete data were listed (the anthropogenic NMVOCs emission data of Taiwan was not available), BVOC emissions were comparable to anthropogenic emissions (the difference of the two sources was less than 0.02 Tg Introduction

Conclusions References
Tables Figures

Back Close
Full natural or anthropogenic origins in western provinces, which were undeveloped and sparsely populated (e.g., Qinghai and Tibet), all maintained a negligible value.On the other hand, in the other eight provinces (including Fujian, Guangxi, Hainan, Hunan, Inner Mongolia, Jiangxi, Shaanxi, and Yunnan), where the plant cover fractions were high, BVOC emissions predominated over anthropogenic NMVOC emissions (Table 3), making it likely that BVOCs would play a more important role in local photochemistry processes and should not be ignored.

Comparisons with other studies
Because the algorithms and data applied in this work represent the latest findings on BVOC emissions, our results may differ greatly from previous published studies and it's important to compare the results for further improvement.
The results of this study were comparable with that of previous studies (Table 4), hence, our results were assumed to be reasonable.As shown in Table 4, the annual emission budgets estimated by the present study fall between the values given by Klinger et al. (2002) and Guenther et al. (1995).In addition, the spatial results of Klinger et al. (2002) indicate a higher BVOC emission in northern China than our value.This discrepancy is likely due to differences in model input values for the BVOC emission factors and estimation algorithms, because the two studies were both based on the outdated G95 algorithms (Guenther et al., 1995).
The results of this study are comparable (though slightly higher) with the recent study of Tie et al. (2006).Compared with the results of Tie et al. (2006), which are based on the algorithms of Guenther et al. (2000), our annual isoprene and monoterpene emissions were higher by 22 % and 14 %, respectively.However, Tie et al. (2006) used USGS data to represent the land-cover distribution and did not account for the influences of LAI and leaf age, which may certainly explain the difference between our study and theirs.The differences between the two works may also be explained, in part, Introduction

Conclusions References
Tables Figures

Back Close
Full by the different meteorological simulation outputs given by WRF (Weather Research and Forecasting) in Tie et al. (2006) and by MM5 in our study.
To further evaluate China's BVOC emission capacity, we calculated the average BVOC emission flux from the total emission budget and land area.Comparing our estimated emission fluxes with previous regional estimations for Beijing (Guenther et al., 1995;Klinger et al., 2002;Wang et al., 2003), PRD (Pearl River Delta) (Zheng et al., 2010;Wang et al., 2011) and Hong Kong (Tsui et al., 2009;Leung et al., 2010), it is clear that the estimated average emission flux for China is lower than the average regional emission fluxes obtained for Hong Kong and the PRD, largely due to the vast areas of barren land in Northwest China.Notably, the results of different studies vary dramatically by province, mainly because of differences in input data and algorithms.
Compared to the global emission estimations of Guenther et al. (1995), the annual emissions of isoprene and monoterpenes in China accounted for about 2.0 % of the total global amount (629.9Tg C).Therein, isoprene was responsible for 1.8 % of global emissions (502.9Tg C), and monoterpenes accounted for 2.8 % of global emissions (127.0Tg C).Thus, although China covered approximately 6 % of the global land area, it only contributed 2 % to global BVOC emissions.Currently, China is undergoing rapid land-cover changes and is now the world's leading nation in existing plantation area (24 % of the global total) (Geron et al., 2006) and implying a greater future impact on BVOC emission contributions.Introduction

Conclusions References
Tables Figures

Back Close
Full The uncertainties in our estimations of BVOC emissions are substantial because of a number of factors.It is difficult to quantify the uncertainties in BVOC emissions studies for lack of data.Here, we merely try to identify the main sources of uncertainties to better understand the results and improve future research directions.

Emission algorithms
Experimental evidence over the past two decades implicated a number of physical and biological factors in modifying the capacity of a leaf to emit BVOCs.Hence, the uncertainties in the emission algorithms used to describe the dependence of emissions have been a significant source of the overall uncertainty.
1. To date, Guenther's algorithms represent the most advanced approach for reliable determinations of BVOC emissions and are widely used.However, like all empirical modeling approaches, Guenther's algorithms suffer from uncertainties related to first, whether empirical relationships based on short-term and limited measurements can be generalized to describe long-term changes, and second the lack of process-understanding regarding isoprene production and emission (Pacifico et al., 2009).
2. The algorithms of light and temperature dependence by Guenther et al. (2006) require a set of input parameters, and as demonstrated herein, even minor errors in the parameterized coefficients substantially affect daily and maximum fluxes (Guenther et al., 1993;Arneth et al., 2008).Observations in boreal or subarctic environments have suggested that BVOC emissions may increase more steeply than indicated by the standard parameter values (Hakola et al., 1998;Rinne et al., 2000;Janson and de Serves, 2001;Arneth et al., 2008), posing great challenges to the algorithm relationships and coefficients in MEGAN.Introduction

Conclusions References
Tables Figures

Back Close
Full 3. Additionally, the algorithms in this study also simplified the spatial distributions of solar radiation, leaf temperature and LAI within the canopy, neglecting the extinction of solar radiation as a function of the LAI, the distribution of LAI values inside the canopy, and the fractions of shaded and sunlit parts at each vertical layer (Karl et al., 2009).However, previous studies have found that the interception of solar radiation by a vegetation canopy generally yields an exponential decrease in radiation intensity as the canopy is descended, and whether a simple or a detailed canopy environment is considered might greatly influence the estimation results (Lamb et al., 1993(Lamb et al., , 1996;;Muller et al., 2008;Karl et al., 2009).Hence, the detailed canopy environmental model, provided as an option with MEGANv2.1, may provide a more reliable result.
4. Finally, although the parameterized algorithms adopted in this study consider the main controlling factors in BVOC emissions, there are still many other influencing factors, such as humidity, wind speed, CO 2 concentration, and environmental stresses, that require further consideration (and whose lack of adequate consideration to this point is largely due to insufficient data and their obscure relationships) (Guenther et al., 2006).In particular, several studies have focused on the influence of CO 2 and have shown that elevated CO 2 concentrations inhibited isoprene emissions, while low CO 2 concentrations enhanced isoprene emission.These results indicate that great changes in BVOC emissions might be projected for future climate warming scenarios (Arneth et al., 2007a;Arneth et al., 2007b;Calfapietra et al., 2008;Heald et al., 2009;Wilkinson et al., 2009).
The most recent study introducing an empirical factor in MEGAN to account for the direct effects of CO 2 on isoprene emissions was described in detail in Heald et al. (2009) and may provide a promising new direction for accurate estimations of BVOCs.Introduction

Conclusions References
Tables Figures

Back Close
Full

Emission factors
The emission factor dataset in this study was based on species-specific emission factors and land-cover information derived from ground measurement inventories, such as regional tree inventories, satellite-based inventories and eco-region descriptions (Guenther et al., 2006).Thus, the uncertainties associated with emission factors predominantly arose from plant misidentifications and errors in BVOC emission measurements.Errors from misidentification are likely quite low because professional botanical experts were employed for plant identification (Klinger et al., 2002) and the validity of the satellite data was verified in other studies (Justice et al., 2002;Wang et al., 2011).Hence, uncertainties in the BVOC emissions determined for a given species should predominantly result from inaccurate determinations of BVOC emissions and from discrepancies between assigned versus actual EFs.Species-level EFs were compiled by Wiedinmyer et al. ( 2004) from published measurements.The determination of emission capacity were based on direct enclosure flux measurements (Wiedinmyer et al., 2004), above-canopy isoprene flux measurements (Guenther et al., 1996;Guenther and Hills, 1998) or inverse modeling and gradient approaches (Greenberg et al., 1999;Wiedinmyer et al., 2004) conducted abroad, and the results are associated with substantial uncertainties due to the measurement techniques and our limited understanding of the complex conversion process within the canopy (Guenther et al., 2006).
Additionally, previous studies indicated that the emission capacities differed considerably within plants, and there were no obvious or consistent taxonomic relationships in terms of non-emitters or emitters, even within the same genus (Pacifico et al., 2009).Hence, it may not be appropriate to directly extrapolate the measured EFs from certain plant species to similar unmeasured family members.In particular, in comparison with trees, there are relatively few measurements of emission factors for shrubs, grasses and other plant species.Because of the scarcity of measurements relative to the large variability of plants on Earth and the inevitable lumping of a large number Introduction

Conclusions References
Tables Figures

Back Close
Full of plant species into functional groups, there may be great variations between BVOC estimation values for/in China that are obtained locally and those obtained via foreign measurements.

Other input data
In our study, highly resolved MM5 outputs were used to drive the estimation.In addition, we updated the land-use and vegetation data in MM5 to improve the simulation performance, but other land surface parameters, e.g., albedo, surface emissivity and surface temperature, still required further updates (de Foy et al., 2006;Meng et al., 2009).Seasonal changes in vegetation and physiological activities, as reflected in LAI values calculated from the remote sensing input, influenced emissions via variations in the amount of emitting leaf biomass.Uncertainties in the determination of vegetation abundance (LAI) arise from satellite measurements and retrieval algorithm errors (Justice et al., 2002;Tian et al., 2002).
Although the present study has made obvious improvements in the methodology and input data, there is still a considerable degree of uncertainty, and there is much more to investigate, including creating a typical and representative map of local BVOC emission factors and improving the estimation algorithms to reduce uncertainties.

Conclusions
Using MODIS-MM5-MEGAN, we estimated the total emission and spatial-temporal dis- (1.24 Tg C, 9.8 %) and β-pinene (0.81 Tg C, 6.4 %).Spatially, isoprene emission centered in Northeast and South China, with Yunnan contributing the largest (9.5 %), followed by Hunan (7.9 %) and Sichuan (7.0 %).While monoterpene emissions centered in Southeast China, where coniferous forests were extensively distributed and the three top-ranking emitters were Guangxi (9.9 %), Yunnan (9.8 %) and Hunan (8.1 %).Temporally, the seasonal and diurnal cycles of solar radiation and temperature as well as plant growth jointly led to corresponding strong variabilities in BVOC emissions.Generally, BVOC emission rates peaked in July and August, with daily maximum values occurring at about 13:00∼14:00 (BJT) and lowest values occurring at night.The strong temporal cycle of BVOC emissions could have an important effect on the formation of secondary pollutants.
The BVOC emission estimations presented in this study initiated an attempt to provide a systematic and real-time update of high-resolution BVOC emissions in China.On the basis of this study, we will further investigate the role of BVOCs in SOA formation and global climate change using a three-dimensional atmospheric chemical and transport model, with efforts to quantitatively evaluate the role of BVOCs in the troposphere and provide information that is useful for local administrative decisions.a Land area of China and provinces were calculated using modeling grids in this study.b Land area of Beijing area was calculated using data provided by Klinger et al. (2002).c PRD referred to the region located at the center of Guangdong Province in South China, including Guangzhou, Dongguan, Foshan, Shenzhen, Zhuhai, Zhongshan, Jiangmen, Huizhou and Zhaoqing.d Land area of PRD was calculated using data provided by Wang et al. (2011).e Land area of PRD was calculated using 3 km × 3 km grids in the study by Zheng et al. (2010).

Supplementary material related to this article is available online at
f Land area of Hong Kong was calculated using data provided by Tsui et al. (2009).
g Land area of Hong Kong was calculated using data provided by Leung et al. (2010).Introduction

Conclusions References
Tables Figures

Back Close
Full Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Crops and grass were considered to have low isoprene emitting capacities or to emit no isoprene what-Discussion Paper | Discussion Paper | Discussion Paper | Screen / Esc Printer-friendly Version Interactive Discussion Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | , BVOC emissions varied considerably by province, with the highest emissions concentrated in the southern provinces of China.The top-ranking isoprene contributor was Yunnan, with approximately 0.86 Tg C emitted yearly, followed by Hunan (0.72 Tg C yr −1 ) and Sichuan (0.64 Tg C yr −1 ).The three provinces Introduction Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | compound yr −1 ) in Chongqing, Guizhou and Heilongjiang.Emissions of NMVOCs from Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | tributions of BVOC emissions from terrestrial ecosystems in China for the year 2006.The introduction of MODIS data and comprehensive consideration of influence factors may potentially improve the estimation results.The annual total emission budget of BVOCs was roughly estimated to be 12.97 Tg C. Isoprene, with an annual emission budget of 9.36 Tg C, was the most abundant species (71.6 %), followed by α-pinene Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper |

Table 1 .
Mapping of the land-use classes of MODIS to USGS classifications.

Table 4 .
Comparisons of the estimated BVOC budgets (Tg C yr −1 ) and average emission fluxes (ton C km −2 yr −1 ) with previous studies.