Impact of a new emission inventory on CAM 5 simulations of aerosols and aerosol radiative effects in eastern China

Emissions of aerosols and gas precursors in China have increased significantly over the past three decades with the rapid economic growth. These increases might have a large climate effect. However, global aerosol-climate models often show large biases in aerosol distribution and radiative forcing in China, and these biases are often attributed to uncertainties 15 and biases associated with the emission inventory used to drive the models. In this study, an energy-statics and technologybased emission inventory, Multi-scale Emission Inventory for China (MEIC), was compiled and used to drive the Community Atmosphere Model Version 5 (CAM5) to evaluate aerosol distribution and radiative effects in China against observations, compared with the model simulations with the widely-used IPCC AR5 emission inventory. We found that the new MEIC emission improves the annual mean AOD simulations in eastern China by 12.9% compared with MODIS 20 observations and 14.7% compared with MISR observations, and explains 22%-28% of the AOD low bias simulated with the AR5 emission. Seasonal variation of the MEIC emission leads to a better agreement with the observed surface concentrations of primary aerosols (i.e., primary organic carbon and black carbon) than the AR5 emission, while the seasonal variation of secondary aerosols (i.e., sulfate and secondary organic aerosol) depends less on the emission. The new emission inventory estimates the annual averaged aerosol direct radiative effect at TOA, surface, and atmosphere to be -0.50, 25 -12.76, and 12.26 W m respectively over eastern China, which are enhanced by -0.19, -2.42, and 2.23 W m compared with the AR5 emission. Due to higher winter BC emission in MEIC, the atmospheric warming effect and the surface cooling of BC are twice as much as those using the AR5 emission. This study highlights the importance of improving the aerosol and gas precursor emissions in modeling the atmospheric aerosols and their radiative effects. 30


Introduction
China has been experiencing three decades of rapid economic growth that brings emissions of atmospheric pollutants that are very different from the past and other parts of the world (Streets et al., 2008;Zhang et al., 2009;Klimont et al., 2009;Lu et al., 2010;Lei et al., 2011;Wang et al., 2012).In 2000, the emissions of sulfur dioxide (SO 2 ), black carbon (BC), and organic carbon (OC) in eastern China accounted for 18.8%, 17.3% and 8.7% of global total emissions, respectively (Liao et al., 2015).China's relative contribution to the global radiative forcing due to emission of sulfate, nitrate, BC, and OC are estimated to be 28%, 24%, 14%, and 5%, respectively (Li et al., 2016).
Emissions of aerosols and their precursor gases in East Asia remain highly uncertain (Zhao et al., 2011;Fu et al., 2012;Wang et al., 2014;Chang et al., 2015;Zhang et al., 2015).The uncertainties (defined as 95% confidence intervals around the central estimates from Monte-Carlo experiments) of a bottom-up Chinese emission are -14% to 13% for SO 2 , -13% to 37% for NO x , -25% to 136% for BC and -40% to 121% for OC (Zhao et al., 2011).Using a different definition of the uncertainty (standard deviation divided by the mean value across different inventories), Chang et al. (2015) estimate that the uncertainties of the surface concentrations of sulfate, nitrate, BC, primary organic matter (POM), and secondary organic aerosol (SOA) due to emissions are 3.9%, 40.0%, 11.1%, 16.7%, and 15.4% over eastern China, respectively.
The uncertainties of emission inventories can affect the model simulations of the aerosols and their radiative effects.Shindell et al. (2013) shows that aerosol optical depth (AOD) are underestimated in East Asia by the multi-model mean of 10 Climate Model Intercomparison Program phase 5 (CMIP5) (Taylor et al., 2012) models compared with satellite observations.Eight out of the 10 global climate models underestimate the AOD by about -36% to -58% relative to ground-based Aerosol Robotic Network (AERONET) observations.The AOD biases are substantially larger than those in North America and Europe.For example, the Community Atmospheric Model version 5 (CAM5) significantly underestimates AOD in East Asia (Liu et al., 2012).A moderate emission adjustment (idealized 20%-30% regional increases in SO 2 , OC, and BC emissions) largely improves the simulations of precursor gases and primary aerosols from CAM5 (He and Zhang, 2014).To examine the impact of the emission inventory on the simulation of aerosols and their radiative effects in CAM5, it is straightforward to replace the default emission inventory in China with a new one that is supposed to improve the representation of the current status of Chinese emissions of aerosols and gas precursors.
The default emission inventory of CAM5 was prepared following the protocol of the Intergovernmental Panel on Climate Change (IPCC) Fifth Assessment Report (AR5) experiments.In order to provide a consistent emission inventory for the CMIP5 models participating the IPCC AR5, a global historical (1850~2000) dataset for anthropogenic and biomass burning emissions of aerosols and reactive gases was developed (Lamarque et al., 2010).In East Asia the AR5 emission inventory incorporates the Regional Emission inventory in ASia (REAS) dataset (Ohara et al., 2007).The IPCC AR5 emission inventory has been widely used for global and regional climate studies (Giorgi et al., 2009;Jones et al., 2011;Shindell et al., 2013).However, the AR5 emission inventory has disadvantages in modelling the rapid change of aerosols in China.For example, data are updated every 10 years since annual data are available only in recent past and for limited species.Seasonal variations are ignored for all emissions except biomass burning, soil NO x , ship and aircraft emissions due to limited data availability.
To characterize the rapid change of China's emissions, an improved technology-based Multi-scale Emission Inventory for China (MEIC) is developed at Tshinghua University, China (http://www.meicmodel.org/index.html).MEIC covers 10 major atmospheric pollutants and over 700 anthropogenic emission sources from 1990 to 2012.It has the following advantages compared with the AR5 emission inventory: (1) adoption of a detailed technology-based approach, (2) application of a dynamic methodology of rapid technology renewal, (3) re-examination of China's energy statistics, and (4) monthly emissions to represent species that have strong seasonal variations for the model results to be better compared with timespecific measurements (Zhang et al., 2009).The MEIC emission inventory is verified to produce consistent aerosol precursor loadings with satellite observations (Li et al., 2010;Wang et al., 2010Wang et al., , 2012;;Zhang et al., 2012;Liu et al., 2016).It has widely been used to study the trend of aerosol concentrations in China (Wang et al., 2013), the Asian air pollution outflow (Zhang et al., 2008;Chen et al., 2009), relative contribution of emission and meteorology to the aerosol variability (Xing et al., 2011) and sensitivity of air quality to precursor emissions (Liu et al., 2010).
Here we examine the impacts on aerosol concentrations and their radiative effects simulated from CAM5 due to the change of emission datasets from the default AR5 to the new MEIC emission inventory.With the inclusion of emission seasonality in MEIC, we will be able to isolate the seasonal variation of aerosol concentrations due to the impact of meteorological factors (e.g., temperature, precipitation, wind speed, etc.) from the impact of emission seasonality.For example, monsoon precipitation induces seasonal shift of the spatial distribution of AOD over eastern China without including the seasonal variation of the aerosol emissions (Jiang et al., 2015).
We rely on climate models to estimate the historical and future change of aerosol direct radiative effect (ADRE).However, the uncertainty of aerosol emissions used in the climate models could add another dimension of uncertainty in simulating the change of aerosol radiative effect.The emission structure changed in China during the last decade.The emission of SO 2 decreased by 9.2% from 34.0 Tg in 2006 to 30.8 Tg in 2010 (-2.4% annual growth rate) (Lu et al., 2011).The NH 3 emission decreased slightly from 10.5 Tg in 2006 to 9.7 Tg in 2012 with an annual rate of -1.4% (Kang et al., 2016).Meanwhile, some researches show that the emissions of BC, OC, and NO x have been increasing since 2006.The BC/OC emissions increase from 1.6 Tg/3.6Tg in 2004 to 1.9 Tg/4.0 Tg in 2010, with an annual growth rate of 2.8%/1.7%.The MEIC developer team estimates that the emission of BC/OC decreased from 2006 to 2010 due to reduced emissions in the domestic and transportation sectors.The NO x emission grew by 113.9% from 12.2 Tg in 2000 to 26.1 Tg in 2010, with an annual growth rate of 7.9% (Zhao et al., 2010).To examine the change in the ADRE as the change in the emission structure, we also carry out a simulation using MEIC emission from 2002 to 2012.This paper is organized as follows.Sect. 2 describes the model configurations and how MEIC is incorporated in the model.
Sect. 3 shows the results of aerosol properties and their radiative effects simulated from the model using the new MEIC emission compared to the AR5 emission.Sect. 4 discusses the change of aerosol radiative effects over eastern China as the change of emission uncertainty in the last decade.Conclusions are presented in Sect. 5.

Model configuration
We run the simulations with the National Center for Atmospheric Research (NCAR) CAM5 with the 3-mode Modal Aerosol Model (MAM3), which represents aerosol size distribution and mixing state between aerosol components in the Aitken, accumulation, and coarse modes (Liu et al., 2012).The aerosol species include sulfate, BC, POM, SOA, dust, and sea salt.
MAM3 assumes that the aerosol species are internally mixed within modes and externally mixed among modes.The mass mixing ratio of each aerosol species in the mode and the number mixing ratio of the mode are predicted.The emission, transport, gas-and aqueous-phase chemical production, dry and wet deposition, microphysics (nucleation, condensation and coagulation), and physical, chemical and optical properties of aerosol are simulated in a physically based manner.The transported gas species are SO 2 , H 2 O 2 , dimethyl-sulfide (DMS), H 2 SO 4 and a lumped semi-volatile organic gas species (SOAG).The aerosol radiative effects are calculated in CAM5 with the Rapid Radiative Transfer Model for GCMs (RRTMG) (Iacono et al., 2008).The refractive indices for most aerosol components are taken from OPAC (Hess et al., 1998), but for BC the value (1.95, 0.79i) from Bond and Bergstrom (2006) is used.
We run CAM5 in a "constrained meteorology" mode where the model meteorological fields (winds, temperatures, etc.) are nudged towards ERA-Interim (Dee et al., 2011) with 6 h relaxation timescale (Ma et al., 2013(Ma et al., , 2014;;Zhang et al., 2014).The horizontal resolution is 0.9 o ×1.25 o and vertically there are 30 layers from surface to 2.25 hPa, with the lowest 4 layers inside the boundary layer.We focus our analysis of model results over eastern China (22 o -44 o N, 100 o -124 o E, the red frame in Fig. 1) where the active anthropogenic emissions are located.

Mapping the MEIC emission inventory for CAM5
In addition to running CAM5 with the default AR5 emission inventory, we implement the new emission inventory MEIC into the model.We replace the default AR5 emission with the MEIC emission in China and keep the same as the AR5 emission elsewhere.We process the MEIC emission in the same way as the AR5 emission is processed for the CAM5 simulations.We map the MEIC aerosols and precursor gases in different sectors to those required by the CAM5 chemistry and aerosol modules (see Table S1 in the supplementary materials).MEIC includes anthropogenic emissions in sectors named by power, industry, residential, and transportation, which correspond to the AR5 emission sectors named by energy, industry, domestic, and transportation, respectively.Emissions due to shipping, agricultural waste burning and waste treatment as well as natural sources such as forest fires, grass fires and continuous volcanoes are not specified in the MEIC emission, so we keep them the same as those in the AR5 emission.
MEIC provides the emissions of aerosols and precursor gases including SO 2 , BC, OC, and non-methane volatile organic compounds (NMVOCs).Additional work is done to obtain a full set of species required by CAM5-MAM3 aerosol simulations, which includes SO 2 , primary sulfate aerosol, BC, POM, SOAG, DMS, and number concentration in the accumulation and Aitken modes.These aerosols and precursor gases are emitted as surface or elevated sources.2.5% (by molar) of the SO 2 emission is regarded as the primary sulfate aerosol emitted directly from sources following the Aerosol Comparisons between Observations and Models (AeroCom) protocol (Dentener et al., 2006) and the rest are emitted as SO 2 .
The energy and industrial SO 2 and primary sulfate are treated as elevated sources at 3 levels between 100 and 300 m, and their sources due to forest fires and grass fires are emitted in six vertical levels at 0 to 6 km (Dentener et al., 2006).SO 2 and primary sulfate from agriculture, domestic, transportation, waste, and shipping sectors are emitted at surface.Primary sulfate aerosols from domestic and transportation are put in the Aitken mode and those from other sectors are put in the accumulation mode (Liu et al., 2012).
The POM emission is assumed to be 1.4 times the OC emission in order to include the aerosol mass of other elements (i.e., oxygen, hydrogen, and nitrogen) (Seinfeld and Pandis, 1998).POM and BC from forest fires and grass fires are treated as elevated sources, while POM and BC from other sectors are treated as surface sources.All POM and BC aerosols are put in the accumulation mode.The SOAG species in CAM5-MAM3 are lumped semi-volatile organic gas-phase species that can condense onto pre-existing aerosols to form SOA. Since the IPCC AR5 dataset does not provide biogenic volatile organic compound (VOC) emission, to simulate SOA in CAM5-MAM3, the SOAG emission is derived from the emission fluxes of five primary VOC categories (isoprene, monoterpenes, big alkanes, big alkenes, toluene) that are prescribed from the Model for OZone And Related chemical Tracers version 4 (MOZART-4) dataset (Emmons et al., 2010) In MOZART-4 the biogenic emissions of isoprene and monoterpenes are based on the Model for Emissions of Gases and Aerosols Emissions from Nature (MEGAN) (Guenther et al., 2006).The MEIC emission provides anthropogenic sources of the five VOC categories and the mapping table for lumping the MEIC VOC species to MOZART is provided by Li et al. (2014).Since the MEIC emission inventory does not provide biogenic sources of isoprene and monoterpenes, which are much larger than their anthropogenic sources, we make the total emissions from anthropogenic and natural sources of these two species the same as those in the AR5 emission.To derive the emissions of the SOAG species, the MEIC emissions of the VOCs are multiplied by the assumed mass yields (Liu et al., 2012).We note that the emissions of SOAGs are increased by a factor of 1.5 in CAM5 to account for the large uncertainty in SOA formation.
The number emission fluxes in both AR5 and MEIC are calculated from the mass fluxes in a consistent way.The mass to number conversion is based on , where D v is the volume-mean emitted diameter and ρ is the aerosol particle density (Liu et al., 2012).Since the MEIC emission does not provide mass emissions from agricultural waste burning, waste treatment, forest fire, grass fire and continuous volcanoes, we use the number fluxes from the AR5 emission for these sectors.
Since the reanalysis data used for our offline meteorology is for year 2009, we obtain the emissions in 2009 by the linear interpolation between year 2008 and 2010 for the MEIC emission and between year 2005 and 2010 for the AR5 emission.
To examine the change in the ADRE as the change in the emission structure, we carry out a simulation using MEIC emission from 2002 to 2012.We choose these 11 years because China's economy recovered from a depression in 2002, and since then the SO 2 emission has started to grow dramatically and decreased after 2006 due to the application of desulfurization equipment.After 2012 the annual emission rates did not change as dramatically as the previous years.

Aerosol observational data
AOD retrievals from MODIS and MISR in 2009 are used to compare against the model results.This study uses the monthly mean AOD from MODIS Terra collection 6 (MOD08_M3), which were obtained from NASA Level1 and Atmosphere Archive and Distribution System (LAADS, https://ladsweb.nascom.nasa.gov/).We use the combined MODIS AOD retrievals of the Dark Target (Levy et al., 2010) and the Deep Blue (Hsu et al., 2004) algorithms.
The AERONET AOD and single scattering albedo (SSA) retrievals in 12 sites in Mainland China, Hongkong, Taiwan, Japan, and Korea are used to compare with the model results and the locations are shown in Fig. 1.We calculate the monthly averaged AOD and SSA in 2009 from the daily averages and exclude the months with less than 3 daily values.
Chemical observation data of surface concentrations in China are collected from literatures (see Table S2 and the references in the supplementary materials).The chemical compositions of particulate matter with diameter smaller than 2.5 µm (PM 2.5 ) are analyzed for sulfate, OC, BC, and SOA in these studies.The measured OC concentrations are multiplied by a factor of 1.4 for calculation of the total organic mass (Seinfeld and Pandis, 1998).To characterize the seasonal cycle of a full year, we

Aerosol direct radiative effect calculation and observational data
We estimate the ADRE due to instantaneous impact of aerosol scattering and absorption on the Earth's energy budget in 2009 following the method by Ghan (2013).The ADRE is calculated by the difference between the "all-sky" radiative flux in the standard model simulation and a diagnostic call to the model radiation code from the same simulation but neglecting the aerosol scattering and absorption.This method considers the impact of cloud presence on the aerosol ADRE, which is different than the conventional method that considers the clear-sky flux changes only (references in Ghan, 2013).The advantage is that this method takes into account the radiative warming enhanced by absorbing aerosol above clouds.We calculate the shortwave ADREs of all aerosols and each individual aerosol species (sulfate, BC, POM, SOA) at the top of the atmosphere (TOA), at the surface, and in the atmosphere.
The ADREs have been estimated based on ground-based and satellite observations at different locations in China (Li et al., 2016).Table S3 in the supplementary materials lists the observations used in this study.Most of the data are from the Chinese Sun Hazemeter Network (CSHNET) (Xin et al., 2007;Li et al., 2010).The ADREs are consistently defined as difference of the irradiance at TOA, surface, and in the atmosphere with and without the presents of aerosols.The ADREs are either calculated by radiative transfer models using the measured or retrieved aerosol properties (AOD, single scattering albedo (SSA), phase function, Ångström exponent (AE), and size distribution) and surface reflectance (Xia et al., 2007a;Li et al., 2010;Liu et al., 2011;Zhuang et al., 2014), or derived from the fitting equation of irradiance measurements as a function of AOD (Xia et al., 2007b, c).Since the MEIC emission inventory is for anthropogenic aerosols, we only compare with observations that are free of the impact of dust aerosol near deserts.For the same reason, the shortwave radiation is discussed since anthropogenic aerosols are mostly fine particles, the impact of which in the long-wave radiation can be ignored.All data analyses are performed after cloud screening to ensure clear-sky conditions.Since the solar irradiance depends on solar zenith angle (i.e., the time of the day), we compare with the measurements that are 24 h averaged.If both the TOA and the surface ADREs are provided, we calculated the ADRE in the atmosphere by subtracting the ADRE at TOA by ADRE at the surface.
The cloud-screening (clear-sky) method in observations neglects the effects of aerosol above and below clouds.Absorption of reflected solar radiation by absorbing aerosol above clouds should result in larger radiative warming effect.The radiative cooling by scattering aerosol below clouds is not as strong as in the clear sky due to less sunlight penetrating the clouds.
Both of the two factors will results in lower (more negative at TOA) radiative effect in clear-sky than that in all-sky.
Therefore, the observation-derived ADRE in clear-sky provides an upper-limit of the model-estimated all-sky ADRE at TOA.

Comparing MEIC and AR5 emission inventories
Figure 2 compares the MEIC and AR5 emission inventories in China.The spatial distributions of the two emission inventories are generally consistent.The MEIC emission rates in China are 14.59 Tg Sulfur (S)/year, 1.78 Tg Carbon (C)/year, 5.16 Tg C /year, 3.50 Tg C /year for SO 2 (including primary sulfate), BC, POM, and SOAG, respectively (Table S1) and are 10.11%, 11.59%, 11.44%, and 36.34%higher than the AR5 emissions, respectively.The emissions are mostly concentrated in eastern China where the MEIC emission rates are 13.60 Tg S/year, 1.59 Tg C/year, 4.38 Tg C/year, 2.86 C Tg/year for SO 2 (including primary sulfate), BC, POM, and SOAG, respectively, which are 12.57%, 13.35%, 12.04%, and 46.88% higher than the AR5 emissions, respectively.SO 2 , BC and POM emissions in MEIC are generally higher in northern China and the Sichuan Basin than those in AR5, whereas they are lower in southern China.The SOAG are higher in MEIC than AR5 in most part of eastern China, especially in southern China.Both emission inventories highlight the emission "hotspots" in the Jing-Jin-Ji region, Henan, Shangdong, Jiangsu Provinces, and the Sichuan Basin.SO 2 industrial emission is the major contributor to the "hotspots".Industrial and domestic emissions result in the high POM and BC emission in these "hotspots".Dust emission is identical between the two runs since the same constrained surface wind speed drives the dust emission.
In addition to differences in the magnitudes and spatial patterns of the total emissions in 2009, there are large differences in the seasonality of two emission inventories (Fig. 3).The AR5 SO 2 and BC emissions do not have seasonal variations, and the POM emission features a weak variation due to biomass burning with a slightly lower emission in the winter.The MEIC emission is characterized by monthly variations for SO 2 , BC, and POM that peak in the winter.SOAG shows a consistent seasonal variation that peaks in the summer in both the MEIC and the AR5 emissions because the emissions of biogenic VOCs (isoprene and monoterpenes) dominate the SOAG emission and peak in the summer.

Aerosol optical depth
Figure 4 shows the spatial distributions of the annual averaged AOD over China simulated by CAM5-MAM3 using the MEIC and the AR5 emissions in 2009.They are compared with the MODIS and MISR satellite AOD retrievals.The modeled AOD (including dust aerosol) averaged over eastern China using the MEIC emission is 0.25, which is 30.38%higher than the AOD with the AR5 emission (0.19) (Table 1).The impact of anthropogenic emissions on the modeled dust AOD is small (< 1.0% difference) due to slightly different removal rates of dust resulting from the internal mixing of dust with anthropogenic aerosols (e.g., sulfate).The modeled AOD with the MEIC emission is 45.10% lower than the MODIS AOD (0.46) and 37.16% lower than the MISR AOD (0.40).The modeled AOD with the AR5 emission is 57.95% lower than the MODIS AOD and 51.87% lower than the MISR AOD.The MEIC emission improves the AOD simulations by 12.85% relative to MODIS and 14.71% relative to MISR compared with the AR5 emission.This suggests that the emission uncertainty (bias) could account for 22.18%-28.37% of the underestimation of AOD simulated by CAM5 with the AR5 emission in eastern China.
The dominant contributor to the total AOD is the sulfate aerosol followed by dust, POM, BC, and SOA in both emission runs (Table 1).The MEIC emission produces 17.40%-70.35%higher AODs of different anthropogenic aerosol species than the AR5 emission.The sensitivities of modeled AOD to the emission change between the two inventories are quite different for each aerosol species due to different aerosol refractive indexes.12.04% of POM emission difference results in 70.35% of the AOD difference.In contrast, 46.88% of the SOAG emission difference leads to only 17.40% of the AOD difference of SOA.
Do the emission inventories produce the reasonable AOD spatial distribution in spite of the underestimated magnitudes?By comparing the spatial distribution of the emissions (Fig. 2) with the AOD distribution (Fig. 4), we found that the AOD distribution basically agree with the emission patterns of sulfate, dust, POM and dust aerosols, which contributes to about 85% of the total AOD.Fig. 4e and Fig. 4f shows the scaled MODIS and MISR AOD by one-half and two-thirds, respectively, which are approximately the ratios between the modeled AOD with the MEIC emission and observed AODs averaged over eastern China.It is found that the model generally reproduces the spatial distribution of MODIS and MISR retrieved AOD.The Jing-Jin-Ji Region, Sichuan Basin, Shandong, Henan, Aihui, Hunan, Hubei Provinces are characterized by higher AODs than other parts of China, which is consistent with the higher anthropogenic emissions in these regions.the SO 2 emission maximum in early summer around 30 o N (Figure S2).Therefore, it is likely that this AOD maximum is due to sulfate aerosol formation and it is heavily underestimated by the model.Since the uncertainty of SO 2 emission is relative low (±12%), this underestimation cannot be explained by emission alone.Other causes (e.g., missing nitrate, particle size distribution, aerosol hygroscopic growth, etc.) in the model are more responsible.
The satellite retrievals show another two AOD maximums around 22-25 o N in March and September.The model simulates the first maximum by capturing the POM maximum at this time and location.We note that there is no strong POM emission and therefore the maximum possibly results from the reduced wet scavenging or low wind speeds.The model results do not show the second maximum or the peak is too weak.Figure 5 indicates that the MEIC emission improves the model performance in reproducing the AOD spatiotemporal distribution.However, there are still processes other than emission that need to be improved in order to reduce the discrepancy between model and observations.Among the possible processes, wet scavenging should be improved since it significantly affects the model results.The CAM5-MAM3 does not include the treatment of nitrate aerosol, which can be an important aerosol component in East Asia (Gao et al., 2014).The modeling of dust aerosol is of particular importance in northern China by contributing to the AOD maximum in autumn and providing the surface area for sulfate condensation.Xianghe, and Xinglong.The satellite retrieved AOD shows a minor peak in June but smaller than that in April.This implies that the timing of the dust AOD maximums over northern China is not well simulated in the model.In addition, sulfate and other anthropogenic aerosols that contribute to the high AOD in the summer are underestimated in the model with both emission inventories.This comparison highlights the importance for better modeling the dust aerosol in northern China.
The Taihu site is in the southern China where the modeled AODs are underestimated at all seasons compared with observations.The model shows that the dominant aerosol species is sulfate in all the year, followed by POM in winter and SOA in the late summer.Sulfate aerosol peaks in September and June and SOA peaks in September.These peaks are observed in MODIS and MISR but are a factor of 2-3 larger than the modeled ones.
The impact of using the new emission inventory in China is also reflected in the modeled AODs in the downwind regions.
AODs at the sites in 20-30 o N (Taiwan and Hong Kong sites) are featured by the summer minimums in both observations and model results due to the scavenging of aerosols by the summer monsoon precipitation.The model captures the seasonality but underestimates the AOD by a factor of 2-3.The MEIC emission has a significant impact on AOD in Hong_Kong_PolyU site in all seasons and only has a small impact in winter in Taiwan sites (NCU_Taiwan, EPA_Taiwan, and Chen-Kung_Univ).AODs at the sites north of 30 o N (Gwangju_GIST, Osaka, and Shirahama) are characterized by the AOD maximums in spring and June as observed by MODIS, MISR and AERONET.Model results indicate that sulfate aerosol is the main contributor to the AOD.The model captures the summer maximum at Gwangju_GIST site and the modeled AOD with the MEIC emission is higher than that with the AR5.The difference between the two emission inventories is not evident at Osaka and Shirahama.
Figure 7 shows the modeled SSA using the MEIC and AR5 emissions and the comparison to the observations by AERONET.The modeled SSA at Beijing, Xianghe, and Xinglong agree with the AERONET retrievals in the strong seasonal variation of lower SSA in the winter and higher SSA in the summer, indicating higher fractions of absorbing aerosols in the winter.However, the modeled SSAs are systematically lower than the AERONET retrievals.This indicates the significant underestimation of scattering aerosols (e.g., sulfate and POM), consistent with the AOD underestimation at these sites.The SSA simulated with the MEIC emission is lower than that using the AR5 emission by up to 0.05 in winter, which is consistent with the higher BC emission in the MEIC emission.The SSAs simulated with the MEIC emission in Taihu are slightly higher than that with the AR5 emission throughout the year, which is consistent with the higher MEIC sulfate emission.Outside Mainland China the modeled SSAs agree with AERONET retrievals reasonably well at the Hong Kong, Taiwan and Japanese sites, although underestimations can be still in some months.

Aerosol concentrations
Figure 8 compares the modeled surface concentrations of sulfate, BC, POM, and SOA with the chemical observations.The surface concentrations of these aerosol species are generally underestimated in the model, which is consistent with the AOD underestimation.The simulated surface concentrations of these species are improved by using the MEIC emission compared with those modeled by using the AR5 emission.The POM and BC surface concentrations are significantly improved by the MEIC emission due to higher emission rates especially in the winter.The sulfate aerosol concentration is underestimated by about a factor of 3 (the linear regression slope of 0.35) using the MEIC emission but is improved compared with about a factor of 5 (the linear regression slope of 0.18) using the AR5 emission.The root mean square errors (RMSEs) using the MEIC emission are 10.01, 14.63, 3.32, 6.58 µg m -3 for sulfate, POM, BC, and SOA, respectively, which are smaller than RMSEs of 13.38, 19.21, 3.97, 8.38 µg m -3 if using the AR5 emission.The correlation coefficients (R 2 ) between model simulations and observations of all these species are also improved.Considering that most observations are carried out at single points and at altitudes close to the surface, the underestimation could be partly due to the coarse model horizontal and vertical resolutions.The model with a coarse horizontal resolution does not account for the subgrid variability of aerosols (Qian et al., 2010).With the coarse vertical resolution, the aerosol species are assumed to be well mixed in the bottom model layer with a thickness of about 120 m.
Figure 9 shows the seasonal variations of surface concentrations of sulfate, POM, and BC simulated with the MEIC and AR5 emissions and the comparisons with the chemical composition observations of PM 2.5 at 10 locations in China.Note that CAM5-MAM3 simulates the aerosols of diameter less than 3.65 µm so the values shown here represent the upper limit of the simulated concentrations.The MEIC emission improves the modeled aerosol concentrations compared with the AR5 emission for all the three aerosol species.The seasonal variations of surface concentrations of POM and BC are mainly determined by their emissions.As shown in Fig. 3, there are no seasonal variations of POM and BC in the AR5 emissions.
By introducing the seasonally varied MEIC emissions of POM and BC, the seasonal changes of POM and BC surface concentrations are better simulated with higher concentrations in the winter and lower concentrations in the summer, which are consistent with the observations.The impact of emission on the seasonal cycle of sulfate surface concentrations is less obvious since sulfate aerosol production depends on meteorology (i.e., temperature and precipitation).The sulfate concentrations in northern China (Harbin, Chengde, Shangdianzi, Beijing, Tianjin, Shijiazhuang, Zhengzhou) are characterized by the summer maximums in both model results and observations.The sulfate concentrations in the southern cities (Xiamen and Guangzhou) do not have the summer maximum due to the Asian summer monsoon with strong winds and precipitation.
The emission of SO 2 peaks in the winter in northern China (Harbin, Chengde, Shangdianzi, Beijing, Tianjin, Shijiazhuang, and Zhengzhou) due to heating in the domestic section.However, the modelled sulfate aerosol concentrations show their minimum in the winter (Figure 9), which is commonly seen for many climate models (personal communication with Prof. Hong Liao from Nanjing University of Information Science & Technology).Obviously, the surface concentrations cannot be explained by emissions alone, and atmospheric physical and chemical processes are more responsible for the low bias.We examine the processes that determine the sulfate aerosol concentrations in the model, including gas-phase and aqueous-phase production, the dry removal and wet scavenging, as well as the controlling meteorological variables (Figure S3).The simulated seasonal variations of surface concentrations of sulfate aerosol are controlled by the gas-phase and aqueous-phase production processes, as oppose to the emission of SO 2 .The gas-phase chemistry is most active in the summer due to the temperature-dependence of the reaction rate in the photochemical oxidation of SO 2 by OH.The aqueous-phase formation of sulfate aerosol also peaks in the summer due to high relative humidity.Although the MEIC SO 2 emissions peak in the winter, both gas-phase and aqueous-phase oxidations are inefficient in the winter, which results in lower concentrations of sulfate aerosol than in the summer.
The seasonal variations of modeled sulfate aerosol concentrations are verified by some observations (Zhao et al., 2013;Zhang et al., 2013;Geng et al., 2015).However, observations from CAWNET (Zhang et al., 2012) show that sulfate aerosol concentrations in the northern Chinese cities (Gucheng and Zhengzhou in Figure S4) peak in the winter as opposed to summer in spite of a minor maximum in the summer, which is consistent with our general recognition.The contrasting result in observations could be due to different location and time.The difference between the simulated and the observed seasonal variations by the CAWNET may reveal that some mechanisms of sulfate aerosol formation in the winter over China are missing in the model.The importance of sulfate production through heterogeneous reactions of SO 2 on deliquescence preexisting particles catalyzed by transition mental ions, which can increase PM2.5 concentrations and the mass fractions of secondary inorganic aerosols in the wintertime hazes of northern China (Wang et al. 2014;Huang et al., 2014;Zheng et al, 2015;Chen et al., 2016;Dong et al., 2016).The coexistence of NO 2 and SO 2 also promotes sulfate production (He et al., 2014;Wang et al., 2014).The aqueous-phase oxidation of SO 2 by NO 2 (Wang et al., 2016) or O 3 (Palout et al., 2016) is efficient to form sulfate aerosol under high relative humidity and NH 3 neutralization conditions.One thing to notice is that aqueous-phase oxidation occurs only in cloud droplets in the model.The aqueous-phase oxidation of SO 2 by NO 2 on preexisting aerosols is not modeled by CAM5-MAM3, which is important in China (Wang et al., 2016).
Having the same "constrained" meteorology for the two runs with different emission inventories provides us with an opportunity to examine the impact of emission on the seasonality of aerosols versus the impact of meteorology.The longitudinal averaged BC burden in the MEIC emission run shows a strong seasonal variation with higher burden between 25 and 40 o N in the winter (Fig. 10a), which is correspondent with the seasonal variation of BC emission (Fig. 10b).We note that there is no seasonal variation for BC aerosol in the AR5 emission (Fig. 10d), and thus the seasonal variation of BC concentration can only be due to the impact of meteorology in the AR5 emission run (Fig. 10c).Fig. 10 indicates that seasonal variations of both the emission and meteorology play important roles in determining the seasonal variation of BC concentrations.
The impacts of emission on seasonal variations of primary (e.g., BC) and secondary aerosols (e.g., sulfate) are different.
Figure 11 shows that the seasonal variation of the differences in the longitudinally averaged BC burden between the two emission runs resembles closely the pattern of difference in the BC emission.However, the dependence of seasonal variation of sulfate burden on the SO 2 emission is less evident (Fig. 11c and Fig. 11d).The large difference of sulfate burden between the two emission runs is obviously aligned with high temperatures between 30 and 40 o N in the summer (May to July) (Fig. 11e) where evident SO 2 emission difference exists there in the summer.It is because H 2 SO 4 gas production to form sulfate aerosol are favored at higher temperatures.In contrast, although there is a significant difference in the SO 2 emission between 34 and 40 o N in the winter (November to March) the sulfate burden difference is not as evident due to the fact that low temperatures inhibit the sulfate production.Wet scavenging by clouds and precipitation helps to reduce the concentrations and the absolute concentration differences in southern China during the spring and summer (Fig. 11f and Fig. 11g).Stagnant wind field that propagates from 22 o N to 30 o N in the winter to 30 o N to 40 o N in the spring (Fig. 11h) also contributes to the large differences of BC and sulfate burdens between the MEIC emission and AR5 emission in corresponding seasons and regions.Diffusion of sulfate aerosol due to higher wind speeds north of 35 o N in the winter also helps to explain the smaller sulfate burden difference in spite of the evident SO 2 emission difference there.Thus, compared to primary aerosols, the spatiotemporal patterns of secondary aerosol burdens follow less closely to their emissions.

Aerosol direct radiative effect
Figure 12 shows the spatial distribution of annual averaged shortwave aerosol DREs in China simulated using the MEIC and the AR5 emissions due to all aerosol species (sulfate, POM, BC, SOA, dust and sea salt) at TOA, surface, and in the atmosphere.The TOA radiative cooling effect is evident in eastern China due to anthropogenic aerosols.At some parts of the southwestern China the ADRE at TOA is positive due to strong BC absorption in the atmosphere.The most pronounced surface cooling and atmospheric warming are located in the northern China and the Sichuan Basin, which is consistent with the spatial patterns of the emissions.In these locations the surface and atmospheric differences of the ADREs between the MEIC and the AR5 emissions are also significant.
As shown in Table 2 the annual averaged cooling effect at TOA is reduced (more negative) by -0.19 W m -2 (63.32%) by all aerosols using the MEIC emission (-0.50 W m -2 ) than the AR5 emission (-0.31W m -2 ).The small TOA ADRE is the sum of the large surface cooling and atmospheric warming effects.At the surface there is a strong cooling effect of -12.76 W m -2 using the MEIC emission, which is reduced (more negative) by -2.42 Wm -2 (23.42%) compared with that using the AR5 emission (-10.34W m -2 ).The atmospheric warming effect of all aerosols using the MEIC emission is estimated to be 12.26 W m -2 , which is 2.23 W m -2 (22.20%) stronger than the estimation made by the AR5 emission (10.03 W m -2 ) over eastern China.
Table 2 also shows the annually averaged ADREs over eastern China by individual aerosol species (sulfate, BC, and POM).
The ADREs of SOA are not shown due to its large emission uncertainty.Due to larger AODs simulated with the MEIC emission, the ADREs by every individual aerosols are larger than the ADREs using the AR5 emission by 29.91% to 66.03% at TOA.The ADRE percentage differences are similar to the AOD percentage differences as listed in Table 1.Tables 1 and   2 show that 12.04% to 46.88% difference of the anthropogenic emission rates of various aerosol species results in 30.38% difference of the total AOD of all species (including anthropogenic and natural aerosols), about 22% difference of the surface and atmosphere ADREs, and about 63% of the ADRE at TOA of all aerosol species over eastern China.The impacts of the emission on AOD and ADRE are significant.
The normalized radiative effect (NRE) represents the radiative effect efficiency per unit aerosol optical depth (Schultz et al., 2006).The light scattering aerosols (sulfate and POM) have very similar negative NREs (-14.34 and -14.71W m -2 τ aer -1 with the MEIC emission hereafter).The light absorbing BC aerosol shows a much higher positive NRE (122.40W m -2 τ aer -1 ) which is comparable to the mean NREs of the AeroCom models (153 W m -2 τ aer -1 ) considering the wide range of the estimates among the models (28 to 270 W m -2 τ aer -1 ) (Schulz et al., 2006).This indicates that the ADREs are much more sensitive to BC aerosol burden than the other aerosol species.This highlights the importance of the BC emission and concentration to correctly represent the ADRE in the model.BC also makes the largest contributor to the ADRE in the atmosphere and at the surface.The NREs of BC are much higher than the other aerosols species, especially the warming in the atmosphere (281.22Wm -2 τ aer -1 ).We note that the ADREs of light scattering aerosols (sulfate and POM) in the atmosphere are also warming effects.The explanation is that coating of these scattering aerosols on BC increases the absorption capability of the internally mixed aerosol particles (i.e., particles in the same aerosol mode with BC) (Chung et al., 2011).
The impact of emission on the spatial distributions of ADRE of BC is more significant compared with the meteorological effects.Since there is no seasonal variation of BC emission in the AR5 dataset, the resultant different spatial distributions of BC aerosol and the ADRE between summer and winter are due to the monsoon circulation and precipitation (Jiang et al., 2015).The averaged ADREs over eastern China in the winter (-3.37 W m -2 at surface and 5.40 W m -2 in the atmosphere) are close to the ADREs in the summer (-3.49W m -2 at surface and 5.75 W m -2 in the atmosphere) using the AR5 emission.Due to higher MEIC BC emission in the winter, the cooling effect of BC at the surface is much more significant using the MEIC emission (-5.86 W m -2 ) than the AR5 emission averaged over eastern China (Fig. 13).Likewise the warming effect of BC in the atmosphere is 9.31 W m -2 , which is nearly twice as much as that using the AR5 emission (Fig. 14).Driven by the same constrained meteorology, the MEIC emission results in stronger seasonal variation of ADRE of BC than the AR5 emission.
Figure 15 shows the comparison between the measured and modeled ADREs at TOA, surface, and in the atmosphere in China.Observations from 25 nationwide stations shows that clear-sky ADREs are characterized by a strong radiative heating in the atmosphere, which implies a substantial warming in the atmosphere and cooling at the surface (Li et al., 2007;Li et al., 2010).This implies the enhanced haze pollution in winter due to the strong atmospheric inversion (Wang et al., 2015;Ding et al., 2016).Model simulations show small ADREs (-4 to 1 W m -2 ) at TOA with both the MEIC and AR5 emission inventories, while the measurements gives a much larger range of ADREs at TOA (-14 to 2 W m -2 ).At the surface and in the atmosphere, the modeled ADREs using the MEIC emission inventory at most locations are within a factor of 2 compared with observations.The MEIC emission inventory produces better agreement with the observations than the AR5 emission inventory.Overall, the model underestimates the ADREs at the surface and in the atmosphere compared with observations due to underestimated aerosol loading, which results in a smaller range of the ADREs at TOA than the observations.

Summary and Conclusions
Anthropogenic aerosols in East Asia may have substantial effects on regional air quality and climate.However, CMIP5 models generally have low biases in anthropogenic aerosol burdens in this region (Shindell et al., 2013), and thus the aerosol effects may be underestimated.The reasons behind the low biases are unclear, but may include the bias in aerosol emissions, missing of some aerosol processes, coarse model resolutions, etc.In this study, we simulated the aerosol concentrations, optical depth, and the radiative effects in eastern China using the Community Atmospheric Model version 5 with the 3-mode model aerosol module (CAM5-MAM3).The "constrained" meteorology is applied to avoid the ambiguity from model predicted meteorological bias.A technology-based emission inventory, Multi-scale Emission Inventory for China (MEIC), was implemented into CAM5-MAM3 and results were compared with the default IPCC AR5 emission inventory.
We found that the MEIC emission improves the annual mean AOD simulations in eastern China by 12.85% compared with the MODIS observations and 14.71% compared with the MISR observations, which explains 22.18%-28.37% of the AOD underestimation simulated with the AR5 emission.The MEIC emission generally reproduces the AOD spatial distribution although AOD is still underestimated compared with the MODIS and MISR satellite retrievals.The MEIC emission run captures the AOD maximums around 30 o N in the spring and the autumn better than the AR5 emission run.However, both emission runs underestimate the AOD maximum around 30 o N in the summer, which coincides with the modeled summer monsoon precipitation.Wet scavenging by summer monsoon precipitation should be reasonably represented since it significantly affects the model AODs.The modeling of dust aerosols is also of particular importance in northern China.
The simulated surface concentrations of anthropogenic aerosol species are improved by using the MEIC emission compared with that modeled by the AR5 emission.Seasonal variation of the MEIC emission leads to better agreement with the observed surface concentrations of primary aerosols (i.e., POM and BC) than the AR5 emission, while the seasonal variations of secondary aerosol species (i.e., sulfate) depend less on the emission.The difference of the primary aerosol burdens in the two emission runs agree with the difference of the emission patterns, while the secondary aerosol concentrations depend on both the emission and the meteorological factors (e.g., temperature, precipitation, wind).The emission of SO 2 in the northern cities peaks in the winter due to domestic heating.Since both the gas-phase and in-cloud aqueous-phase formation of sulphate aerosol are simulated to peak in the summer, the surface concentrations of sulphate aerosol peak in the summer.The seasonal variation is different from observations by some observations (e.g., CAWNET), which cannot be explained by the emission alone.This may reveal that some mechanisms, such as production through heterogeneous reactions of SO 2 on aerosols, which is not currently included in the model, can be important sources of sulfate aerosol in the winter.
Different emissions also have substantial effects on the aerosol radiative effects.By using the MEIC emission, the annual averaged ADREs at TOA and at the surface over eastern China are reduced (more negative) by -0.19 W m -2 and -2.42 W m -2 , respectively, while the warming in the atmosphere is increased by 2.23 W m -2 .The ADREs between the MEIC and AR5 emissions with all aerosols species (including natural dust) increases by 63.32%, 23.42%, and 22.20% at TOA, surface and in the atmosphere, respectively.The ADRE is much more sensitive to BC aerosol burden than the other aerosol species.Due to the higher MEIC BC emission in the winter, the warming effect of BC in the atmosphere and the cooling effect at the surface are much higher than those using the AR5 emission.Over eastern China about 12% to 47% difference of the emission rates of different aerosol species results in 30.38% difference of the total AOD, about 22% difference of the surface and atmosphere ADREs, and about 63% of the TOA ADRE.The impacts of the emission on AOD and ADRE are significant.
By examine the change of ADRE from 2002 to 2012 using the estimation of emissions made by the MEIC developer team, we find that the change of ADRE at TOA in eastern China due to the change of emission structure is not significantly different than that estimated by the MEIC and AR5 emissions in 2009.However, the decadal changes of ADRE at surface and in the atmosphere are smaller than the uncertainties by using the two different emissions.This indicates that there is an urgent need to constrain the emission inventories of aerosol and precursors by in-situ and satellite observations.This research highlights the critical importance of improving aerosol and precursor emissions for the modeling of aerosols and aerosol radiative effects in East Asia.We note that modeled aerosol AOD is still underestimated in CAM5 even with the MEIC emission.The CAM5 model with a horizontal resolution of 0.9 o ×1.25 o may miss the subgrid aerosol variability (Qian et al., 2010) as well as not able to capture the collocation between aerosols and clouds important for aerosol wet scavenging (Ma et al., 2014).CAM5-MAM3 may also miss some important aerosol species (e.g., nitrate) which can have similar mass burdens as sulfate in eastern China (Gao et al., 2014).Current work is under the way to improve the model resolution and to implement nitrate aerosol in CAM5-MAM3.The impacts of these new developments on aerosols in East Asia will then be evaluated.
Fundamental Research Funds for the Central Universities (Grant No. 310400090).We thank the AERONET PI investigators and their staff for establishing and maintaining the 12 sites used in this investigation.
technique facilitates the model-observation comparison of aerosols and gas species.When simulating the decadal change from 2002 to 2012, we use the reanalysis data in 2009 cyclically to nudge the model meteorological fields.In this way, the change of aerosol concentration is controlled by the change of emission alone since the meteorological fields are identical among the years.
extend our time selection of the observations from 2009 to 2010.Many of the studies collected the samples continuously during April, July, October, and January to represent the concentrations in spring, summer, autumn, and winter.The observation in Xiamen was carried out for a full year of sample collection.For the observations that only lasted for several days within a month, we average the daily data to represent the month.Since we do not find the SOA measurements in 2009, we use the data in other years and are aware of the uncertainties due to the time difference.The geographical locations of these chemical observations are shown in Fig. 1.

Figure 5
Figure5shows the seasonal variation of longitudinal (100 o E to 124 o E) averaged AOD over eastern China.The CAM5 simulation using the MEIC emission captures the observed AOD maximums around 30 o N in the spring (February to May) and in the autumn (August to October) but the magnitudes are lower than the observations.The model results using the AR5 emission fail to capture the first maximum and underestimate the second one even more than the MEIC emission.By examining the model AOD components by species (FigureS1), the first maximum is mostly due to sulfate aerosol and to a less extend due to POM aerosol, and the second maximum is mostly due to sulfate.Sulfate AOD peaks in the summer because the production of sulfuric acid gas (H 2 SO 4 ) is more efficient at higher temperatures.Wet scavenging due to the strong East Asian summer monsoon precipitation reduces the aerosol concentrations.This is evident in both the model results and the observations with minimal AODs in the summer between 20 o N to 30 o N.However, the modeled precipitation (mostly large-scale precipitation) extends further north in the summer compared with the Global Precipitation Climatology Project (GPCP) observations (not shown here).This causes the model to underestimate the AOD maximum around 30 o N in the summer in spite of evident sulfate production at this time and location.The model simulates a maximum between 35 o N and 40 o N in early summer (from May to July), which is to the north of the observed AOD maximum..This model maximum is mostly due to dust aerosol transported from the west, while the satellite retrievals do not show such strong dust emission and transport.Since the dust emissions are the same in the simulations using the MEIC and AR5 emissions, the difference of the modeled AOD maximums between 35 o N and 40 o N is mainly due to sulfate condensed on dust aerosol.We notice that the maximum AOD in the satellite retrieval occurs around 30 o N in early summer (May to July), as oppose to 35 o N to 40 o N as simulated by the model.The observed AOD maximum complies with

Figure 6
Figure6shows the seasonal variation of monthly averaged AOD in 2009 at 12 AERONET sites in and around China.The model simulations using both emissions generally underestimate AOD compared with AERONET and MODIS observations.The model results in Beijing, Xianghe and Xinglong agree better with MISR observations in spring and winter.The modeled AODs in the MEIC emission run are higher and agree better with the observations than the AR5 emission run in Beijing, Xianghe, Taihu, and Hong Kong, where anthropogenic emissions are significant.In Beijing, Xianghe, and Xinglong the MODIS and MISR AODs are high in February, which are better simulated with the MEIC emission that has the seasonal variations of BC and POM emissions, whereas the AR5 run does not capture this feature because of a lack of seasonal variations of the emissions.In the summer, the observed seasonality in Beijing, Xianghe, and Xinglong features a maximum in July and a lower AOD in June, where the modeled AOD peaks in June.Model results show that dust aerosol dominates the modeled AOD maximum in June and both the dust and sulfate aerosol dominate in July at these sites.The dust aerosol maximum in June is also simulated at the upwind SACOL site.Dust aerosol is usually transported from SACOL to Beijing,

Figure 16
Figure 16 shows the MEIC's SO 2 , BC, and OC emission trends from 2002 to 2012 in eastern China.Since spatially-gridded MEIC emission data are only available for 2008, 2010, and 2012, we scale the spatial distribution and seasonal variation of other years during the period to the 2008 gridded emission with the annual emissions.The annual emission rates of each species (SO 2 , BC, OC) are estimated by the MEIC developer team.Each species in different sectors (power, energy, residential, and transportation) has a different scaling factor.The annual trends are consistent with other researches(Lu et al., 2011;Lei et al., 2009) although the absolute values are different.We use the MEIC estimations because their fuel usage is based on China Energy Statistical Yearbook (CESY) as this dataset is based on the same algorithm as the gridded MEIC emission data that we used for 2009.

Figure 17
Figure 17 shows that the change of the ADREs from 2002 to 2012 relative to those in 2002 are 0.01~0.16W/m 2 , -0.44~0.39W/m 2 , and -0.23~0.51W/m 2 at TOA, surface and in the atmosphere, respectively.The decadal trend of ADRE agrees with the trend of emissions as shown in Figure 16.The warming in the atmosphere and the cooling at the surface enhanced with the increase of emissions of SO 2 , BC, and POM from 2003 to 2006.The ADRE at TOA increased slightly indicating more energy retained in the atmosphere-earth system.From 2006 to 2009, the changes of ADREs were not significant due to the stabilized emission of BC.Since 2010, the warming in the atmosphere and the cooling at the surface increased due to the increased emission of SO 2 and BC.The changes of ADREs at surface and in the atmosphere from 2002 to 2003 may reflect

Figure 1 .
Figure 1.Geographical locations of the AERONET sites and chemical composition sites where the observational data are used in this study.The provinces and regions mentioned in the context are marked.The red frame denotes eastern China (22-44 o N, 100-124 o E).

Figure 3 .
Figure 3. Seasonal variation of sulfate, BC, POM, and SOAG in the MEIC emission and the AR5 emission in eastern China.

Figure 4 .
Figure 4. Spatial distributions of annual averaged AOD at 550 nm over China in 2009 simulated by CAM5-MAM3 using (a) the MEIC emission, (b) the AR5 emission, observed by (c) MODIS and (d) MISR satellites, (e) One-half of the AOD observed by MODIS, and (f) two thirds of the AOD observed by MISR.

Figure 5 .
Figure 5.The seasonal variation of longitudinal averaged (100 o E-124 o E) AOD at 550 nm over eastern China simulated by CAM5-MAM3 using (a) the MEIC emission, (b) the AR5 emission, observed by (c) MODIS, and (d) MISR satellites in 2009.

Figure 6 .
Figure 6.Monthly averaged AOD simulated by CAM5-MAM3 using the MEIC emission and the AR5 emission compared with the AERONET, MODIS and MISR observations at 12 locations in and around China.The error bars represent one standard deviation of the daily AERONET observations within the month.

Figure 7 .
Figure 7.The seasonal variation of SSAs simulated by CAM5-MAM3 using the MEIC emission and the AR5 emission and observed by AERONET (red dots) at 12 AERONET sites in and around China.Error bars stand for one standard deviations of the observations.

Figure 8 .
Figure 8.The monthly averaged surface concentrations of sulfate, POM, BC, and SOA using the MEIC emission and the AR5 emission compared with observations.The solid lines are linear regressions between the model results and observations.The red dashed lines represents the 1:2, 1:1, and 2:1 lines.The regression functions and the correlation coefficients (R 2 ) are also shown.

Figure 9 .
Figure 9.The seasonal variation of monthly averaged surface concentrations of sulfate, POM, and BC modeled by CAM5-MAM3 using MEIC and AR5 emissions compared with the observations.Error bars stand for one standard deviation

Figure 13 .
Figure 13.Spatial distributions of ADREs of BC in the summer (June, July, August) and the winter (December, January, February) at the surface (SFC).

Figure 14 .
Figure 14.Same as Figure 13 but for the ADRES of BC in the atmosphere (ATM).

Figure 15 .
Figure 15.ADREs at TOA, surface, and atmosphere modeled by CAM5-MAM3 using the MEIC and the AR5 emissions compared with ADRE observations in China.

Figure 16 .
Figure 16.The change of emission rates of SO 2 , BC, and POM from year 2002 to 2012 in eastern China.

Figure 17 .
Figure 17.The change of ADREs at TOA, surface and in the atmosphere relative to year 2002 with MEIC emission in 850 eastern China.