Global high-resolution emissions of soil NOx, sea salt aerosols, and biogenic volatile organic compounds

Natural emissions of air pollutants from the surface play major roles in air quality and climate change. In particular, nitrogen oxides (NOx) emitted from soils contribute ~15% of global NOx emissions, sea salt aerosols are a major player in the climate and chemistry of the marine atmosphere, and biogenic emissions are the dominant source of non-methane volatile organic compounds at the global scale. These natural emissions are often estimated using nonlinear parameterizations, which are sensitive to the horizontal resolutions of inputted meteorological and ancillary data. Here we use the HEMCO model to compute these emissions worldwide at horizontal resolutions of 0.5° lat. × 0.625° lon. for 1980–2017 and 0.25° lat. × 0.3125° lon. for 2014–2017. We further offer the respective emissions at lower resolutions, which can be used to evaluate the impacts of resolution on estimated global and regional emissions. Our long-term high-resolution emission datasets offer useful information to study natural pollution sources and their impacts on air quality, climate, and the carbon cycle. Measurement(s) nitrogen oxide • aerosol • isoprene • acetone • acetaldehyde • ethene • ethanol • propene • monoterpene • lumped monoterpenes • limonene • sesquiterpene • Emission • sea salt aerosol Technology Type(s) computational modeling technique Factor Type(s) spatial resolution • temporal resolution Sample Characteristic - Environment saline aerosol environment • soil environment • atmospheric ozone • vegetation layer • atmosphere Sample Characteristic - Location Earth (planet) Measurement(s) nitrogen oxide • aerosol • isoprene • acetone • acetaldehyde • ethene • ethanol • propene • monoterpene • lumped monoterpenes • limonene • sesquiterpene • Emission • sea salt aerosol Technology Type(s) computational modeling technique Factor Type(s) spatial resolution • temporal resolution Sample Characteristic - Environment saline aerosol environment • soil environment • atmospheric ozone • vegetation layer • atmosphere Sample Characteristic - Location Earth (planet) Machine-accessible metadata file describing the reported data: https://doi.org/10.6084/m9.figshare.12205379

Soil NO x emissions. Inside HEMCO, the algorithm for above-canopy soil NO x emissions (soil NO x ) follows Hudman, et al. 2 , with the efficiency of loss to canopy depending on vegetation type and density. Based on soil chamber and field measurements, soil NO x varies greatly with climate and edaphic conditions, and are most strongly correlated with N-availability, temperature, precipitation patterns, and fertilizer management practices 21,22 . In the Hudman, et al. 2 algorithm, soil NO x emissions flux is a complex function of biological and meteorological drivers: www.nature.com/scientificdata www.nature.com/scientificdata/ [13 01 ln 53 6] (5) dry d ry ct ′ A biome , representing the biome-dependent emission factors of N in the soil, is a function of N avail and the A w biome , coefficients. A w biome , is the wet biome-dependent emission factors updated based on estimates from Steinkamp and Lawrence 30 . Ē is the mean emission rate of fertilizer, and is treated identically to the natural pool of N.
N avail , representing the sum of fertilizer N and deposited N, is the mass of available nitrogen in the soil. F is the fertilizer application rate and τ is a decay lifetime, which is chosen as 4 months based on measurements within the top 10 cm of soil 31,32 . Although atmospheric deposition also contributes to the available nitrogen in soils (about ~5% globally, based on Hudman, et al. 2 ), this amount can only be calculated through 3-D model simulations and is thus not accounted for here. θ × f T g ( ) ( ) represents the combination of the soil temperature (T) and soil moisture dependence of soil NO x . The temperature dependence of soil NO x is an exponential dependence on temperature between 0 °C and 30 °C (constant at T > 30 °C), where 0.103 is the weighted average of temperature dependencies for several biomes. The parameterization for soil moisture is a Poisson function scaling, where θ (water-filled pore space) is defined as the ratio of the volumetric soil moisture content to the porosity 33 .
( ) P l t , dry represents the pulsed soil NO x , which occur when very dry soil is wetted resulting in a reactivation of water-stressed bacteria. The parameterization, following Yan, et al. 34 is derived from four field studies relating pulsed emissions to the length of the antecedent dry period [35][36][37][38] . The rate constant c reflects the rise/fall time of the pulse (c = 0.068 h −1 ). The value of l dry is the antecedent dry period in hours.
Emissions of sea salt aerosols. Parametrization of sea salt emissions in HEMCO is modified from Jaeglé, et al. 6 . It considers two categories of SSAs based on their radii. The radius of accumulation mode sea salt aerosol (SALA) ranges from 0.01 to 0.5 μm, while that for coarse mode sea salt aerosol (SALC) ranges from 0.5 to 8 μm.
Parametrization of sea salt aerosols emissions includes both a wind speed and a sea surface temperature (SST) dependence. The SSAs emission flux density function dE/dr 80 is formulated as follows: is based on Gong 39 . r 80 is the particle radius at RH = 80% (with r 80 ~ 2r dry ), and u 10m is the 10-meter wind speed. = .   www.nature.com/scientificdata www.nature.com/scientificdata/ γ γ γ γ γ γ = CL (8) , , ε i j , is the average emission factor of species i for vegetation type j at standard conditions (leaf temperature = 297 K; air temperature = 303 K; the photosynthetic photon flux density averaged over the past 24 h is equal to 200 μmol m −2 s −1 for sun leaves and 50 μmol m −2 s −1 for shade leaves), and χ j is the fractional grid box areal coverage for the same vegetation type. The emission factor accounts for the estimated in-canopy deposition flux so that F i represents the net above-canopy flux.
The emission activity factor (γ i ) reflects the emission response to environmental drivers. The canopy environment coefficient (C) is assigned a value that results in γ = 1 for the standard conditions and is dependent on the canopy environment model being used. A detailed description of the model parameterizations for light (γ p ), temperature (γ T ), leaf age (γ A ), soil moisture (γ S ), leaf area index (L) and CO 2 inhibition (γ C ) can be obtained from Guenther 14 .

LaI data for calculation of soil NO x and BVOCs emissions.
Vegetation composition is principal information needed to estimate BVOCs and soil NO x emissions 14,15 . The density of vegetation is represented in the parametrizations by leaf area index (LAI), which is defined as the amount of leaf area per unit surface of the ground (m 2 m −2 ). We use monthly MODIS-derived LAI with gap filling and smoothing described by Yuan, et al. 40 (here after referred to as Yuan LAI). For 2005-2017, we use year-specific Yuan LAI data. For years prior to 2005, we use the LAI values in 2005 due to lack of year-specific data. This would introduce certain uncertainty for these earlier years. We did a test to fix LAI to a certain year, and the effect on global emissions is relatively small (within 5%). Therefore, our extrapolation of LAI data before 2005 does not significantly affect the emissions time series.

Data Records
Our datasets contain 4 data records for monthly global gridded emissions. Each record contains monthly emission data for soil NO x , 2 SSAs species (SALA and SALC), and 10 BVOCs species (ISOP, ACET, ALD 2 , C 2 H 4 , EOH, PRPE, MTPA, MTPO, LIMO, and SESQ). Our data were constructed in nc file format which can be read by many tools like IDL, MatLab, and so on. Of these,

Reference
Year Method Resolution Emission (Tg C/yr)  www.nature.com/scientificdata www.nature.com/scientificdata/    Fig. 2e shows the temporal profile of the respective global total emissions. SSAs are the largest over the North Atlantic in January and over the Indian Ocean in July. Emissions are also strong over the Southern Ocean. The temporal variation of global total SSAs emissions is characterized by lower values in the 1980s and 1990s than in later years (with a difference by about 10%), and by a modest seasonality (within 15%). Figure 3 shows the spatial and temporal distributions of total BVOCs emissions (sum of ISOP, ACET, ALD 2 , C 2 H 4 , EOH, PRPE, MTPA, MTPO, LIMO, and SESQ). The total BVOCs exhibits strong seasonality and cross-hemispheric seasonal migration (Fig. 3a-d) because of changes in radiation and temperature. The highest emissions occur over the Amazon, Southeast Asia, Southeast United States, and Central Africa. The global total emission also exhibits a large seasonality, with a July to January ratio of 1.3, due to variation of LAI, especially in the Northern Hemisphere. The interannual variation is modest (within 20%) (Fig. 3e).
The parameterized nonlinear relationships between emissions and controlling factors means that the horizontal resolution of inputted meteorological and other variables has important influences on the calculated emission magnitudes and spatial distributions. Table 5 presents the global annual total emissions of soil NO x , SSAs and BVOCs derived based on GEOS-FP at different resolutions (4° lat. × 5° lon., 2° lat. × 2.5° lon., and 0.25° lat. × 0.3125° lon.) over 2014-2017. The resolution dependence of emission magnitude is evident especially for soil NO x and SSAs, that is, a higher resolution results in greater global emission totals. The global total SSAs emission increases from 3,157 Tg/yr to 3,239 Tg/ yr (by 2.6%) and to 3,860 Tg/yr (by 22.3%) as the resolution changes from the coarsest to the finest. This increase is primarily because emissions are parameterized as a function of wind speed to the 3.41-th power. For soil NO x , the global total increases from 7.1 TgN/yr to 7.5 TgN/yr (by 5.6%) and to 8.8 TgN/yr (by 23.9%) as the resolution increases. This is mainly because the parameterized NO x emission is convex functions of temperature and soil moisture. For BVOCs, the resolution dependence of the global total emission is weaker, i.e., within 5% for ISOP and within 10% for other species. The magnitude of horizontal resolution dependence for BVOCs here is similar to that of temporal resolution dependence shown by Ashworth, et al. 41 who showed that using monthly mean inputted data instead of hourly data would reduce the global ISOP emission total by 7%.  www.nature.com/scientificdata www.nature.com/scientificdata/  www.nature.com/scientificdata www.nature.com/scientificdata/ for absolute emission differences (Fig. 6a,b). The percentage difference is most evident along the coasts where a fine resolution (0.25° lat. × 0.3125° lon.) resolves the land-ocean contrast much better than coarser resolutions do (Fig. 6c,d). For sea salt emissions, the major source regions at high latitudes of both hemispheres exhibit a large resolution dependence (Fig. 7). For BVOCs, tropical regions have the largest resolution dependence in terms of absolute difference (Fig. 8a,b), while the coastal and low-emission regions exhibit the largest resolution dependence in terms of percentage difference (Fig. 8c,d). Figure 9 further shows the resolution dependence of calculated regional annual emission totals over eight major regions. Compared to results for global total emissions in Table 5, the resolution dependence of emission magnitude in some regions is more evident. For Southeast Asia, soil NO x emissions total at 0.25° lat. × 0.3125° lon. is higher than that at 4° lat. × 5° lon. by 38%. Similar results are shown for Europe (38% higher) and Australia (37% higher). For sea salt, emissions for North Hemisphere Africa and Southeast Asia increase by 38% and 30%, www.nature.com/scientificdata www.nature.com/scientificdata/ respectively, from 4° lat. × 5° lon. to 0.25° lat. × 0.3125° lon. The resolution dependence of regional emissions is smaller for BVOCs (within 10% for all regions) than for soil NO x and SSAs.

technical Validation
Uncertainty. A major source of uncertainty in our calculated emission data is the use of parameterization as an approximate of the complex processes involved in the emissions of these species 2,6,14 . Parameterization is also sensitive to errors in the inputted meteorological and ancillary data 23,26 . www.nature.com/scientificdata www.nature.com/scientificdata/ The parameterization of soil NO x emissions includes a continuous dependence on soil moisture and temperature, a representation of biogeochemistry that induces pulsing of the emissions following dry spells, and a detailed spatial and temporal representation of N-inputs both from chemical/manure fertilizer and atmospheric N-deposition (not included here). Our sensitivity test for 2017 at GEOS-FP 4° lat. × 5° lon. shows that a 1 °C increase in temperature would lead to 5.2% increase in the calculated global total emission, a 10% increase in soil moisture would lead to 15.8% decrease in emission, and a 10% increase in LAI would lead to 1% decrease in emission. Sensitivity tests for other resolutions show similar results.
For sea salt emissions, the strong power law relationship with wind speed and the polynomial relationship with SST mean that errors in wind speed and SST have a significant impact on calculated sea salt emission. Based on our sensitivity test for 2017 at GEOS-FP 4° lat. × 5° lon., a 10% increase in wind speed would lead to 38.4% increase in the calculated global emission total, and a 1 °C increase in SST would lead to 6.7% increase in   www.nature.com/scientificdata www.nature.com/scientificdata/ emission. This is consistent with the evident dependence of calculated emissions to the horizontal resolution of inputted meteorological data. By comparison, an increase in the shape parameter by 10% (from constant 30 to 33) would lead to 0.1% increase in the calculated emissions.
The parameterization of BVOCs emissions involves meteorological (temperature, solar radiation, humidity, wind speed and soil moisture), land cover data (LAI and PFT fractions) and the PFT-specific average emission factor at standard conditions. According to Guenther,et al. 14 , uncertainties associated with the global annual emissions of several compounds (isoprene, acetone and acetaldehyde) are about a factor of two while estimates of uncertainties are a factor of three or higher for other compounds here. The average emission factor is the largest contributor to the uncertainty of estimated emission. Uncertainties in land cover and meteorological variables are also important. Wang, et al. 42 showed that an average bias of about 2 °C in temperature is associated with an error in isoprene emissions by ~23% in the Pearl River Delta of China. The error in LAI and its extrapolation to 1980-2004 leads to an additional uncertainty in calculated emissions. Our sensitivity test for 2017 at GEOS-FP 4° lat. × 5° lon. shows that a 1 °C increase in temperature would lead to 12.9% increase in the calculated global total BVOCs, and a 10% increase in LAI would lead to 4.6% increase in emission.
Comparison with existing emission estimates. Comparisons with existing emission estimates are mainly for our results derived based on MERRA-2, which contain much longer data records than those based on GEOS-FP.
The existing estimates of total BVOCs range from 200 to 1,000 TgC/yr depending on the meteorological and vegetation datasets used 15,45 , and our estimate (563 TgC/yr averaged over 1980-2017 based on MERRA-2) is within this range. Table 3