The effects of recent control policies on trends in emissions of anthropogenic atmospheric pollutants and CO2 in China

. To examine the effects of China’s national policies of energy conservation and emission control during 2005– 2010, inter-annual emission trends of gaseous pollutants, primary aerosols, and CO 2 are estimated with a bottom-up framework. The control measures led to improved energy efﬁciency and/or increased penetration of emission control devices at power plants and other important industrial sources, yielding reduced emission factors for all evaluated species except NO x . The national emissions of anthropogenic SO 2 , CO, and total primary PM (particulate matter) in 2010 are estimated to have been 89 %, 108 %, and 87 % of those in 2005, respectively, suggesting successful emission control of those species despite fast growth of the economy and energy consumption during the period. The emissions of NO x and CO 2 , however, are estimated to have increased by 47 % and 43 %, respectively, indicating that they remain largely determined by the growth of energy use, industrial production, and vehicle populations. Based on application of a


Introduction
China suffers highly degraded air quality and related environmental impacts, mainly due to intensive fossil fuel consumption and rapid growth of the vehicle population. Based on satellite observations and chemical transport models, eastern China has been found to have the highest concentrations of airborne fine particulate matter (PM 2.5 ) (van Donkelaar et al., 2010) and vertical column densities (VCD) of tropospheric NO 2 (Richter et al., 2005) in the world. Serious air pollution has caused huge public health damages particularly in mega cities (Parrish and Zhu, 2009) and has also threatened ecosystems. The highest acidity of precipitation in the world has been observed in south and southwest China (Larssen et al., 2006). A number of analysts have estimated swift increases in anthropogenic emissions, the main cause of China's severe air pollution, during the early-2000s (Ohara et al, 2007;Zhang et al., 2007;Lu et al., 2010;Lei et al., 2011a). According to the GAINS model developed by the International Institute for Applied Systems Analysis (IIASA), China accounted for 24 %, 14 %, 25 %, and 27 % of global emissions of SO 2 , NO x , black carbon (BC), and organic carbon (OC) in 2000, respectively (Cofala et al., 2007;Klimont et al., 2009). Although considerable uncertainties exist (Zhao et al., 2011a), estimates of China's total SO 2 and PM emissions are much more than those of the US or Europe .
Under heavy pressure to improve urban air quality, reduce regional air pollution, and limit carbon emissions, China's government has implemented a comprehensive national policy strategy of energy conservation and emission reduction since 2005. Its goal is to shift the country's development mode from one dependent on intense fossil energy inputs with consequent high emissions to a more resource-efficient and environment-friendly alternative. Stringent, compulsory measures to improve energy efficiency and control emissions have been required at many major source types, targeting a range of atmospheric pollutants. These measures include: replacement of small and inefficient plants or boilers with larger, energy-efficient ones in the power sector and certain heavy industrial sectors including cement production; installation of flue gas desulfurization (FGD) systems at all newly built thermal power units; application of more stringent emission standards in cement production; and staged implementation of tighter emission standards on vehicles. Evidence of success of these measures since 2005 has been confirmed in different ways. For example, improved combustion efficiency (and thus energy efficiency) is indicated in an increasing inter-annual trend in the ratio of CO 2 to CO, because CO 2 results from complete combustion and CO from incomplete combustion of carbon fuels. The trend has been observed instrumentally in air masses representative of north China emissions at a rural site north of Beijing (Wang et al., 2010a) and also indicated in bottom-up emission inventory studies (Zhao et al., 2012a, b). Reductions in regional and national SO 2 are similarly indicated both by observations, from satellites , and by bottom-up emission trends based on fast-track energy statistics for recent years (Lu et al., 2010;2011). Some studies also assess the effects of policies on other species including NO x (Lin et al., 2011;Wang et al., 2012) and primary and secondary aerosols (Lu et al., 2011;Lin et al., 2010a).
While studies that focus on individual species and/or source types are essential to building fundamental knowledge of atmospheric processes in China, they contribute piecemeal to understanding of China's atmospheric environment as an integrated system of sources and sinks of diverse reactive species. Interaction of emission trends, however, is often as significant to environmental outcomes of interest as the trend in any one species or source category taken alone. This is critical not only to understanding physical, chemical, and biological cycles but also to evaluating and informing the development of broadly effective air quality and climate protection policy strategies. The current study meets a need for comprehensive consideration of emission trends of different atmospheric pollutants and analyses of the main drivers of these trends. It focuses on 2005 through the end of China's 11 th Five Year Plan in 2010, a discrete period of sharply heightened regulatory action in emission control.
The study analyzes the effects of recently implemented control measures on the inter-annual trends, sector and spatial distributions, and uncertainties of China's anthropogenic emissions. Incorporating the latest information from domestic field measurements and investigations, the trends of emission factors (i.e., the emission levels per unit consumption of energy or industrial production) for different kinds of pollutants from 2005 to 2010 are developed by sector and technology. Based on a bottom-up framework, provincial and national emissions are estimated for 2005-2010, indicating the effectiveness of improved energy efficiency and emission control efforts during those years. The uncertainties of emissions in 2010 are quantified statistically using Monte-Carlo simulation, which was developed and applied previously to emissions for 2005 (Zhao et al., 2011a). The causes of discrepancies between the uncertainty results for the two years are evaluated. To understand the effects of varied emissions on urban and regional air quality, available observations from ground measurements and satellites are reviewed and compared with the bottom-up emissions for corresponding time periods and locations in China. Limitations of current controls on diverse environmental impacts are analyzed based on the estimated emission trends for different species and regions, recommending a more comprehensive, multi-pollutant scope as China develops its future strategies in control of atmospheric pollutants.

The framework of the emission inventory
The methods of developing a bottom-up emission inventory are detailed in previous studies (Zhao et al., 2011a(Zhao et al., , 2012a. Figure S1 in the Supplement shows the source structure used to estimate China's anthropogenic atmospheric emissions. At the largest scale, sources fall into four main sector categories: coal-fired power plants (  reflecting the structure of available data. Species considered in this work include gaseous pollutants (SO 2 , NO x , and CO), PM according to different size classes and chemical species (Total Suspended Particles (TSP), PM 10 , PM 2.5 , BC, OC, Ca, and Mg), and the greenhouse gas CO 2 . Annual anthropogenic emissions of these pollutants for 2005-2010 are estimated both by province and sector and then aggregated to the national level, using Eq. (1): wherei,j , k, m,n and t stand for species, province, sector, fuel type, emission control technology and year, respectively; AL is the activity level, either energy consumption or industrial production; EF is the unabated emission factor; R is the penetration rate of emission control technology; and η is the removal efficiency.

Activity levels
Activity levels for 2005-2010 are compiled annually by sector from various data sources. The fossil fuel consumption and industrial production at provincial level are obtained from Chinese official energy (NBS, 2011a) and industrial economy statistics (NBS, 2011b). For some industrial sources without official statistics, such as brick making, internal production data from relevant associations have to be relied on. To avoid double counting, the fuel consumption by OIB is estimated by subtracting the fuel consumed by CEM, ISP and PRO from fuel consumed by total industry (Zhao et al., 2012a). The amount of biofuel use is taken from official statistics (NBS, 2011a). The biomass combusted in open fields is calculated as a product of grain production, waste-tograin ratio, and the percentage of residual material burned in the field. Details are described in Zhao et al. (2011a;. From 2006 to 2010, the share of primary energy consumption by coal decreased from 74 % to 70 %, indicating a shift towards cleaner energy sources. It should be noted that China's provincial and national energy statistics are often inconsistent, and such inconsistence can lead to considerable deviations in emission estimate (Guan et al., 2012). As shown in Fig. 1, the annual coal consumption levels reported officially for the entire country range 13-16 % lower than the sum of provincial consumption in each year from 2005 to 2010. On a sector basis, the differences reach 20-30 % for industry and exceed 30 % for the residential and commercial sector, while the difference for the power sector is relatively small (∼2 %). Akimoto et al. (2006) found China's provincial-level statistics to be within the uncertainty bounds of the satellite record of NO 2 over the country while the national-level statistics were not, and advised against use of the latter for emission inventories in China. Although this conclusion was drawn for 1996-2002, the differences between the national and aggregated provincial statistics have not diminished in following years. Subsequent studies of China's emissions, many including comparisons to observations by ground stations, aircraft, or satellites, have held to the same conclusion (Streets et al. 2006;Zhang et al., 2007;Zhao et al., 2012b). We likewise believe that the provincial statistics are more accurate than the national ones.
For transportation, Chinese official statistics reflect only fuel used in commercial activities, and thus cannot be applied directly. In this work, on-road vehicles are classified into light-duty gasoline vehicles (LDGV), light-duty gasoline trucks (LDGT), light-duty diesel trucks (LDDT), heavyduty gasoline vehicles (HDGV), heavy-duty diesel vehicles (HDDV), and motorcycles (MC). The oil consumption by  each vehicle type is calculated as the product of the vehicle population, annual average mileage traveled per vehicle, and average fuel economy of the corresponding type (He et al., 2005). The updated data are provided in Huo et al. (2011Huo et al. ( , 2012a and Zhao et al. (2012a). For non-road sources including railway, waterway, rural vehicles, and construction equipment, the fuel consumption in 2005 is taken from Zhang et al. (2008), while those for 2006-2010 are scaled by province according to the growth of passenger and freight traffic by rail and shipping, and the total growth of agricul-tural and construction equipment. All those data are obtained from official statistics (NBS, 2011c).

Evolution of emission factors
Driven mainly by official Chinese policies, the penetration levels of different energy efficiency technologies and emission control devices shifted considerably at the national level from 2005 to 2010, leading to strong changes in emission factors as clearly illustrated in Fig. 2. Details by source are discussed as follows. During the period of interest, the mass fractions of BC and OC in PM for industrial and transportation sectors, and those of Ca and Mg for all the related sectors are assumed unchanged, taken from Zhao et al. (2011a) and Zhu et al. (2004). It should be noted that such assumption will elevate uncertainty since the fractions of chemical compositions in PM are not always constant along with technology change for given sources. For example, increased share of BC in PM emissions has been reported for newer vehicles with improved technology (Zielinska et al., 2004). In China, however, domestic measurements are still insufficient for evaluating the clear tends of chemical fractions in PM, particularly for a relatively short period. Long-term analysis of PM emission factors with chemical species profiles are thus recommended for future.

Coal-fired power plants
Coal-fired power plants were targeted for the most stringent emission controls during 2005-2010, particularly for SO 2 . According to a unit-based dataset of coal-fired power plants over the country , the FGD penetration rate increased from 13 % of total capacity in 2005 to 86 % in 2010, and the capacity share of the units equal to or larger than 300MW rose from 51 % in 2005 to 78 % in 2010, as shown in Fig. 2a. Based on an unpublished official survey, the national average removal efficiency of FGD is set at 75 % in this study, resulting in a 61 % reduction in the SO 2 emission factor for the entire coal-fired power sector over 2005-2010. Fast growth of large power units with higher energy efficiency and advanced PM control devices like electrostatic precipitators (ESP) or fabric filters (FF) reduced the emission factors of CO and PM as well. Based on an emission factor database reported by Zhao et al. (2010Zhao et al. ( , 2012a, the emission factors of TSP, PM 10 , PM 2.5 , and CO are respectively estimated to have decreased 60 %, 55 %, 46 %, and 31 % from 2005 to 2010. (Note that increased wet-FGD also helped to reduce PM emissions due to its ancillary benefit on PM removal (Zhao et al., 2010)).
On the other hand, however, the emission factors of NO x and CO 2 for coal-fired power plants varied little from 2005 to 2010, due to the limited addition of selective catalytic/noncatalytic reduction (SCR/SNCR) and the experimental state of carbon capture and storage technologies. It should be noted that the emission factors here are expressed as the mass of emitted pollutants per unit consumption of coal. If evaluated as pollutants per unit of generated electricity, the emission factors of NO x and CO 2 declined by 12 % and 19 %, respectively, resulting mainly from the improved energy efficiency of coal-fired power generation during the years of analysis.

Cement production
Cement kiln technologies in China include shaft, precalciner, and other rotary kilns. As shown in Fig. 2b, the penetration of precalciner kilns, the most energy-efficient technology, increased from 44 % to 81 % from 2005 to 2010, while that of shaft kilns declined from 49 % to 16 %, according to official statistics. Emission factors of cement production by technology have been compiled by Lei et al. (2011b). Newly built precalciner kilns with ESP or FF lead to reduced emission factors for PM and CO. However, such benefits are accompanied by an increased NO x emission factor, because the higher operational temperatures and more automated air-flow systems of precalciner kilns increase emissions of NO x compared to shaft kilns. Combining the emission factors by  and changes in penetration of different kiln types, the emission factors (expressed as pollutants per unit of cement production) for SO 2 , TSP, PM 10 , PM 2.5 , and CO are estimated to have declined by 32 %, 72 %, 69 %, 64 %, and 55 %, respectively, during 2005-2010, while that for NO x increased by 31 %. Since CO 2 is generated mainly from the non-combustion process of carbonate calcination (Zhao et al., 2012b), the technology changes in cement production yielded little mitigation of CO 2 emissions.

Iron and steel production
The iron and steel industry employs the following processes: coking, sintering, pig iron making (in blast furnaces), steel making (nearly 90 % of which is in basic oxygen furnaces), and casting. SO 2 and NO x come mainly from the sintering process, and those emission factors are assumed unchanged during 2005-2010 given no new control requirements.
Based on national statistics (CISA, 2011), the share of coke produced in machinery coking ovens (versus modified indigenous ovens with poor technology and manual operation) increased from 82 % in 2005 to 86 % in 2010, reaching a peak of 91 % in 2007. Due to improved use of waste heat, the release ratios of machinery coke oven gas declined from 5.7 % to 1.4 % between 2005 and 2010. The release ratio of blast furnace flue gas in the making of pig iron dropped from 8.4 % to 5.0 % in the same period, and the recycled flue gas in basic oxygen furnaces increased from 60 to 79 Nm 3 /t-steel (CISA, 2011). Those improvements made the emission factors of PM, CO, and CO 2 for the combined processes (expressed as pollutants per unit of steel production) decline by 39 %, 44 %, and 18 %, respectively, as shown in Fig. 2c.

Transportation
Since 1999, staged emission standards (Stage I-IV, equivalent to Euro I-IV) for new on-road vehicles have been implemented nationwide, with earlier implementation in Beijing than in other provinces. The fleet compositions by con-  (NBS, 2011c) and the retirement of old vehicles based on assumed vehicle lifetimes in China by type. The average lifetimes of light-duty vehicles, light-duty trucks, and heavy-duty trucks are assumed to be 15, 8, and 10 yr, respectively, based on previous studies (He et al., 2005;Huo et al., 2012b).
As summarized in Tables S1 and S2 in the Supplement, prior measurements of emission factors of NO x and PM 2.5 for vehicles in China by type and control stage were thoroughly investigated in this work, including on-road tests, engine tests, carbon balance calculations, and remote sensing (see also the database for CO by Zhao et al., 2012a). Results of on-road tests with advanced measurement technologies (e.g., He et al. (2010) and Wu et al. (2012) using SEMTECH-D and Oliver (2008) using OBS-2200) are given preference to calculated vehicle emissions. If two or more studies consider the same combination of vehicle type and control stage, the emission factors used here are calculated as the average of the original data weighted by the sampling size. Due to few on-road tests of LDGT, data from roadside remote sensing (Guo et al., 2007) are applied for this vehicle type. With almost no measurements for HDGV or MC by control stage, the standard limits of stage I-II are relied upon, and typical fuel economies (2.7 L-fuel/100 km for MC (He et al., 2005) and 250 g-fuel kWh −1 for heavy-duty engines (Chen et al., 2008;MIIT, 2010)) are applied to convert the standard limits to fuel-based values. The same assumption is also applied for most non-road sources, except rural vehicles (RV), for which emission factors are taken from on-vehicle tests using SEMTECH-D by Yao et al. (2011). As shown in Fig. 2d, the nationwide emission factor levels of NO x , PM 2.5 , and CO are estimated to have declined by 44 %, 41 %, and 52 % respectively for LDGV, attributed to implementation of the staged regulations. The control effects for MC, HDDV, and rural vehicles ( Fig. 2e-g) are less, particularly for NO x , and indeed NO x emission factors for rural vehicles are estimated to have increased. The reasons include (1) slower penetrations of new MC and rural vehicle technologies than that of LDGV, and (2) higher NO x emission rates of diesel engines under recently implemented regulations.

Other industrial boilers, processes, and residential and commercial combustion
For industrial boilers and processes, the primary concern is changes of emission factors for PM, resulting from varied penetration levels of different dust collector technologies. Such information is available for 2005 from investigation by Lei et al. (2011a), but very few data can be found for subsequent years. In this work, we assume that new emission sources (reflected by the annual net growth of energy consumption or industrial production) applied the most advanced dust collectors that have already been deployed in the sector, reflecting the effect of national policies to foster energy conservation and emission reduction during 2006-2010.
For example, the PM emission factors are estimated to have decreased by over 40 % in nonferrous metallurgy, attributed to increased application of FF ( Fig. 2h). For brick making, the share of solid clay bricks in China is estimated to have declined from 86 % to 40 % from 2005 to 2010 (Xu and Wang, 2007;Zhou, 2009;Zhao et al., 2012a). Accordingly, the emission factors (expressed as pollutants per unit of brick production) declined by 43 %, 28 %, 16 % and 27 % for TSP, PM 10 , PM 2.5 and CO, respectively (Fig. 2i). Those trends of emission factors, however, should be used with caution, since there are currently very few domestic measurements available for size profiles of PM emissions from brick production, and the results from foreign countries have to be relied on (Klimont et al., 2002).
To examine the sensitivity of PM emissions to the estimated trends of dust collector penetration, we also make an additional case, in which the application rates of dust collectors for industrial boilers and processes are kept unchanged from 2005 to 2010. Comparison of emissions between the two cases is given in Sect. 4.3.
For the residential & commercial sector, very little progress in emission control is believed to have occurred during 2005-2010 in either fossil fuel consumption or biofuel/biomass burning. Emission factors for different species are mainly from the database compiled by Zhao et al. (2011a;, with most recent results from domestic tests incorporated (Shen et al., 2010(Shen et al., , 2012 4 Results

Inter-annual emission trends by sector and province
The national emissions from 2005 to 2010 for different species are illustrated by sector in Fig. 3. The SO 2 emissions are estimated to have decreased from 31.1 million metric tons (Mt) in 2005 to 27.7 Mt in 2010, with a peak value of 32.1 Mt in 2006 (Fig. 3a). It is particularly notable that the emissions from coal-fired power plants declined from 16.3 to 9.2 Mt, and the sector share of total emissions from 52 % to 33 %, resulting mainly from the swift increase of FGD systems during 2005-2010. The SO 2 emissions of other industries, however, are estimated to have increased by 32 %, accounting for over half of total national emissions in 2010; this is attributed to the dramatic growth of industrial production and energy consumption. This result confirms previous work on China's SO 2 trends (Lu et al., 2011;Zhang et al., 2012a; see also the detailed comparison in Sect. 4.2) and spotlights the industry sector as a key target for future SO 2 emission controls, not just power plants. In contrast to SO 2 , national NO x emissions increased dramatically, by 48 % to 29.0 Mt in 2010, attributed to swift growth of energy consumption and limited control measures (Fig. 3b). Specifically, emissions from cement and on-road vehicles are estimated to have increased 131 % and 61 %, respectively. Expanded application of precalciner kilns, while helping to control emissions of other pollutants, actually worsened the NO x problem. The implementation of staged emission regulation of on-road vehicles could not keep pace with rapid growth of vehicle populations in recent years, and thus failed to prevent emissions from rising overall in this subsector. Moreover, NO x emissions from residential combustion of fossil fuels also went up despite decreased coal consumption of that sector, although the share remained small. This resulted mainly from rising use of liquid and natural gas fuels, which emit much less SO 2 and particles than coal combustion but relatively more NO x . All of these facts suggest that NO x emission control will be a huge challenge for China in upcoming years. Regarding CO, a much reduced growth rate was found for 2005-2010, reflecting the benefits of improved energy efficiency in recent years (Fig. 3c). As shown in Fig. 3d-j, the emissions of PM of different sizes and chemical species are estimated to have declined to varied extent, and Ca is the species with biggest emission abatement, by 25 % from 2005 to 2010. Attributed to the penetration of improved production technologies and dust collectors, national emissions of TSP are estimated to have decreased from 33.2 Mt in 2005 to 28.7 Mt in 2010, of PM 10 from 18.9 to 17.0 Mt, and of PM 2.5 from 13.0 to 12.2 Mt. Emission control in cement production is found to have been highly effective, with PM emissions of different sizes and chemical species reduced around 50 %. The cement-making share of total emissions decreased from 22 % to 13 % for TSP and from 52 % to 34 % for Ca. In contrast, emissions of different PM categories from iron and steel plants are estimated to have increased 24 %-39 % from 2005 to 2010, attributed mainly to huge growth of steel production. The annual variations of national emissions of primary carbonaceous aerosols ranged less than 10 %, and the source contributions were relatively stable. The combustion of fossil fuel and biofuel/biomass in residential and commercial activities  accounted respectively for around 19 % and 30 % for BC emissions, and 29 % and 50 % for OC emissions. CO 2 emissions are estimated to have increased from 7126 Mt in 2005 to 10 174 Mt in 2010, with an annual growth rate of 7.4 % (Fig. 3k). These totals include emissions from biofuel/ biomass burning, which are omitted in many CO 2 inventories. Regarding sector distributions, emissions from power plants, industry, and transportation increased by 44 %, 60 %, and 64 % during 2005-2010, and the three sectors accounted for 32 %, 41 %, and 8 % of national total emissions in 2010, respectively. Emissions from residential and commercial activities declined slightly. Table 1 are the emissions of different species at the provincial level in 2010. The developed regions of east, north-central, and south-central China (as defined in Table 1 and illustrated in Fig. S2 in the Supplement) are estimated to account together for around 70 % of total national emissions of all concerned species in 2010. Notably, the SO 2 emissions in north-central and east China declined by 12 % and 20 % respectively during 2005-2010, indicating considerable achievements of emission control in these heavily polluted regions. The analogous reductions for northeast and southwest China, however, are merely 2 %. In the northeast, where coals with low sulfur content (such as lignite) are widely used, the SO 2 emissions from coal combustion were relatively small and thus there is less need for FGD at existing power units than in other areas. In contrast, the sulfur content of coals in the southwest are extremely high and many power units had FGD already installed by 2005, leading to limited potential for further reduction of emissions after 2005. Among all the regions, the northwest had the fastest growth of NO x and CO 2 emissions, indicating a relatively rapid increase of economic activity and energy consumption in that less-developed area, although its shares of total national emissions remained small. For SO 2 and PM of different size classes, combining results of other studies and this work shows that the growth of emissions in the early-2000s have been gradually reversed in recent years. Generally consistent trends of SO 2 emissions from 2005 to 2010 were indicated by Lu et al. (2011) and this work, although the current estimates are consistently lower. The discrepancy is likely attributable mainly to applications of different FGD removal efficiencies in the two studies. Lacking estimates in more recent years for comparison, the emissions of TSP, PM 10 , and PM 2.5 for 2005 in Lei et al. (2011a) and this work are close. Regarding chemical species, however, the different studies are relatively inconsistent. For example, decreasing trends of carbonaceous aerosol emissions were given by Klimont et al. (2009) Lu et al. (2011), but much lower OC. The carbonaceous aerosol emissions estimated by this work are close to or higher than those of Klimont et al. (2009) and Lei et al. (2011a), but clearly lower than those of Lu et al. (2011). Although Lu et al. (2011) included emissions from burning of forest and savanna that is omitted by this work, the contribution of that source was very small and cannot fully explain the differences. The divergent results indicate large uncertainties for estimates of emissions of chemical species of PM, particularly from the burning of fossil fuel and biofuel/biomass in residential activities, helping to motivate the uncertainty analysis of bottom-up emissions described in Sect. 4.3.

Comparisons with other studies
Compared to gaseous and PM pollutants, the disparities in CO 2 emissions between different studies are smaller, as shown in Fig. 3k. Note the results of other studies and this work for total CO 2 emissions cannot be directly compared because the emissions estimated by the US Carbon Dioxide Information Analysis Center (CDIAC, http://cdiac.ornl.gov/ftp/trends/emissions/), the PBL Netherlands Environmental Assessment Agency (PBL, http://www. pbl.nl/sites/default/files/cms/publicaties/500212001.pdf), the US Energy Information Administration (USEIA, http://www.eia.gov/cfapps/ipdbproject/IEDIndex3.cfm?tid= 90&pid=44&aid=8), and the International Energy Agency (IEA, http://www.iea.org/publications/freepublications/ publication/name,4010,en.html) include only those from fossil fuel combustion and sometimes cement production, while this work also includes emissions from other non-combustion industrial processes and biofuel/biomass burning (see details in Zhao et al., 2012b). (Note that USEIA and IEA do not report cement process emissions, and the estimates by CDIAC are added to USEIA and IEA fossil fuel emissions in Fig. 3k to facilitate some comparison of the studies by equivalent inclusion of source types.) Even excluding the emissions from biomass/biofuel, discrepancy remains between this work and other studies. The CO 2 emissions for 2005-2010 estimated by us are generally higher than those by most of other studies, attributed mainly to the application of a domestic CO 2 emission factor database (Zhao et al., 2012b), and the use of provincial-level energy data in this work, as confirmed by Guan et al. (2012). Moreover, the non-combustion CO 2 emissions from industrial processes, such as the emissions from primary aluminum production due to the consumption of carbon anodes in the reaction to convert aluminum oxide to aluminum metal, are also included in this work, although the amount from those sources is relatively small. From 2000 to 2005, the annual growth rates of CO 2 emissions from fossil fuel combustion plus cement production are estimated to have ranged from 10.2 % by PBL to 13.9 % by USEIA, while growth in annual emissions declined to 8.5 % from 2005 to 2010 based on the current study. This suggests the effects both of slowed economic development and of improved energy efficiency for the country during 2005-2010.

Uncertainty analysis of emissions in 2010
The uncertainties of emissions for different species in 2010 are estimated using a Monte-Carlo framework developed by Zhao et al. (2011a). The principles of determining the uncertainties of all the parameters, expressed as the probability distribution function (PDF), were described in detail by Zhao et al. (2011aZhao et al. ( , 2012a. With updated PDFs for 2010, a total of 10000 simulations are performed and the uncertainties of emissions, expressed as 95 % confidence intervals  Mg f PM2.5, pulverized boiler (24 %) F Mg, iron and steel production (60 %) -EF TSP, coal stove (80 %) AL coal (9 %) R iron an steel production cyclone (12 %) -AL coal (11 %) (CIs) around the central estimates, are generated by sector and species, as shown in Table 2. The parameters most significant in determining the uncertainties of emissions, judged by their contribution to the variance, are also identified by the Monte-Carlo simulations and are shown in Table 3.
The uncertainties of China's anthropogenic emissions of gaseous pollutants SO 2 , NO x , and CO in 2010 are estimated to be -15 % to +26 %, −15 % to +35 %, and −18 % to +42 %, respectively; those for primary aerosols TSP, PM 10 , PM 2.5 , BC, OC, Ca, and Mg are −22 % to +54 %, −15 % to +54 %, −15 % to +63 %, −28 % to +126 %, −42 % to +114 %, −75 % to +77 %, and −46 % to +152 %, respectively; and that of the greenhouse gas CO 2 is −10 % to +9 %. In general the results of the uncertainty analyses are similar to those for 2005 (Zhao et al., 2011a(Zhao et al., , 2012a in that: (1) the uncertainties of emissions of gaseous pollutants are smaller than those of primary aerosols; (2) among sectors, the uncertainties associated with residential and commercial activities are the largest; and (3) in most cases, parameters related with emission factors contribute most to the uncertainties of   atmospheric pollutants, while activity levels are more significant to uncertainties of CO 2 , except for the industry sector ( Table 3).   The uncertainty of total national SO 2 emissions is accordingly larger in 2010, at −15 % to +26 %, than that in 2005, at −14 % to +13 %. This results mainly from the swift penetration of FGD systems, of which the SO 2 removal efficiency may vary nationally and is poorly quantified. Although designed FGD removal efficiencies can reach 95 %, the installed FGD systems are not believed to have achieved such high values because they were not always operated fully (Xu et al., 2009;Xu, 2011). FGD systems operated less consistently than expected reduces the anticipated benefits of SO 2 emission control on power plants, and increases the uncertainty of the SO 2 emission inventory for recent years as well. As shown in Table 3, the removal efficiency of FGD contributed 72 % to the variance of SO 2 emissions from power plants, and more investigations are thus necessary to better quantify typical removal efficiencies of FGD systems. Second, the uncertainties of NO x and PM emissions from transportation declined from 2005 to 2010. As staged emissions regulations of vehicles have been implemented since 2005, more measurements on the emission factors of onroad and rural vehicles designed for different control standards have been conducted and reported (e.g., He et al., 2010;Wu et al., 2012). The increased sampling sizes and improved measurement methods of those studies have helped considerably to reduce the uncertainties of vehicle emission factors. Third, the uncertainties of PM emissions of different particle size classes have increased, particularly for TSP. This is attributed mainly to penetration levels of dust collectors for given industrial processes during 2005-2010 that must be assumed, without sufficient field data. As shown in Table 3, the penetration level of dust collectors for lime production is the most significant parameter contributing to the uncertainty of TSP emissions from the industry sector for 2010, while the analogous parameters for 2005 are the emission factors for non-combustion cement processing and grate boiler combustion (Zhao et al., 2011a). Compared to fine particles, the uncertainty of TSP emissions increased more significantly, reaching −22 % to +54 % in 2010, exceeding that of PM 10 . If no technology improvement of dust collectors is assumed for industrial processes, as described in Sect. 3.5, the national TSP emissions are reestimated at 39.6 Mt, i.e., 38 % higher than the original estimate. The analogous values for PM 10 and PM 2.5 would be 16 % and 12 % higher, respectively, much smaller than that of TSP; and those for BC and OC less than 5 %. These results indicate that the uncertainty of dust collector penetrations in industrial sources have fewer impacts on emission estimates for fine particles and carbonaceous aerosols than for TSP. The benefits of those technologies on PM control are particularly from certain industrial sources with relatively large emission fractions of coarse particles including lime and brick production, as indicated in Klimont et al. (2002).

Comparisons with ground observation
The inter-annual trends of emissions are compared with those derived from observations. During the study period, SO 2 , NO 2 , and PM 10 were criteria pollutants for which concentrations were reported for 113 "key" cities by the Ministry of Environmental Protection (MEP) of China, based on their large populations, developed economies, and/or heavy pollution levels. Figure 4 shows comparisons of relative interannual variations between the estimated emissions and the reported concentrations of monitored cities (compiled from the datasets available at: http://datacenter.mep.gov.cn/) by region from 2005 to 2010, normalized to 2005 levels. The maps illustrate the changes in emissions for provinces and concentrations for cities, while the panels around the maps illustrate the trends in emissions for broader regions (as defined in Table 1) and concentrations for cities in those regions.
The inter-annual trends of emissions and observed concentrations are generally consistent with each other, although there are also some discrepancies for given species and regions. For SO 2 , as shown in Fig. 4a, similar declining trends for emissions and concentrations are found for north-central, northeast, east, and south-central China between 2005 and 2010, confirming the effects of national policy on SO 2 control. However, increased SO 2 concentrations are found for several cities in east China while decreased provincial emissions are estimated. This is possibly because (1) some local sources like small coal stoves which are missed in the energy statistics and thus omitted in the emission inventory framework grew in recent years; and/or (2) the operation of FGD was actually poorer than expected in some specific areas. In contrast, decreased concentrations were observed with increased emissions between 2005 and 2010 for some provinces in southwest China. As discussed in Sect. 4.1, the limited potential expansion of FGD for power plants in the region could not significantly reduce SO 2 . The declining concentrations thus imply that some local measures of SO 2 control, such as coal washing, may not be well characterized by this work. The disagreements of emission and concentration trends in certain areas indicate the pressing needs of more detailed local investigations on emission characteristics and control measures.
As shown in Fig. 4b, NO x emissions are estimated to have increased much faster than NO 2 concentrations from 2005 to 2010 in all regions. Indeed decreased urban average NO 2 concentrations were reported in north-central and northeast China, shown in the line plots for those regions. The relatively large gaps between regional emissions and observed urban concentrations likely reflect that NO x (the estimated emission species) includes NO as well as NO 2 (the observed concentration species). Complex local NO x sources contributed higher levels of NO in urban areas than rural or remote ones, though it is omitted in measurements of urban NO 2 concentrations. Satellite observation at larger spatial scales will be used to further examine the inter-annual trends of estimated emissions, as described in Sect. 4.5.
The inter-annual trends of emissions and concentrations for PM 10 match well for most regions, as shown in Fig. 4c. In north-central China, faster reduction of urban PM 10 concentrations is found than that of estimated provincial emissions. This is probably attributable to gradually implemented control measures (e.g., road paving and afforestation) on fugi-tive dust from construction sites, unpaved roads, or natural sources. These are not included in the current emission inventory. Similar to SO 2 , there are some cities in east China with increased observed PM 10 concentrations despite estimated declines in regional emissions from 2005 to 2010, suggesting that some local small industrial and/or residential sources, which generate PM emissions from coal or biomass combustion, probably increased though they are not well accounted for in recent years.
The government has not systematically reported urban concentrations of other species including BC, OC, CO, or CO 2 . As summarized in Table 4, however, a number of independent studies have been conducted in different periods and locations in China to observe ambient BC, OC, CO, and CO 2 levels. Correlation slopes between some of these observed species, e.g., dBC/dCO, dOC/dBC, and dCO 2 /dCO, have been estimated to approximate emissions. To test the accuracy of the bottom-up emissions in this work, the ratios of BC to CO emissions (µg m −3 ppbv −1 , note that 1 µg m −3 ppbv −1 = 1.25 kt kt −1 ) and those of OC to BC (µg m −3 /µg m −3 ) are calculated for the corresponding periods and locations in which the observations were conducted (see also an analysis of CO 2 /CO ratios in Zhao et al., 2012a). Monthly variations of emissions are generated following the methods of Zhang et al. (2009), and are used in the calculation of emission ratios for periods matching the observations. In one case, Andreae et al. (2008), the observations were conducted outside of the study period of this work, in October-November 2004, so emissions for the corresponding months in 2005 are instead applied here for comparison. The ratios of emissions are determined at two spatial scales: a local scale, based on provincial emissions, and a multi-province regional scale, as defined in Table 1. One exception to this definition is for the comparison with Kondo et al. (2011), which reported the results of dBC/dCO observed at Cape Hedo on Okinawa Island, Japan, representing emissions exported from east China. In this case the local emission ratio is calculated based on emissions in east China while the regional one is based on those for the whole country.
As shown in Table 4, in most cases the BC/CO ratios from estimated emissions are higher than the correlation slopes from observations in China. However, once the influence of wet deposition and atmospheric chemistry processes are excluded Wang et al., 2011), the ratios are much closer to each other, indicating consistency of observations and bottom-up emissions. In all cases, the BC/CO emission ratios at regional levels are larger than those at local levels. Since the observations were mainly conducted in or close to developed mega cities, on-road gasoline vehicles which generate elevated CO contribute more to local emissions than they do at regional scale. The regional emissions include more combustion of residential solid fuels in less developed rural areas, which generate higher BC emissions. Zhou et al. (2009) conducted observations in Beijing (in north-central China) and Taicang (in Jiangsu province, east China) during the same period and found a much higher dBC/dCO in Taicang than Beijing. However, this big diversity is not well indicated by the bottom-up emission inventory, with just a slightly higher local BC/CO emission ratio for Jiangsu province than Beijing. Although the large population of gasoline vehicles in Beijing and the diesel use by shipping in east China helped generate higher BC/CO emission ratio for Jiangsu than Beijing, the more consumption of residential coals with elevated BC/CO in Beijing than Jiangsu partly compensated the difference. Further studies are thus recommended on differentiated emission characteristics by region and sector. The comparisons of OC/BC ratios are shown in Table 4 as well. The minimum OC/BC slopes from observation, if available, are used in the comparisons to eliminate the effects of secondary organic aerosols as much as possible. The ratios from estimated emissions are generally close to observed slopes, enhancing confidence in the bottom-up emission inventory for primary carbonaceous aerosols. However, the estimated emissions fail to capture the very high OC/BC ratios in rural Inner Mongolia in fall and winter (Han et al., 2008), suggesting that the current inventory might miss or underestimate emissions from some sources that generate large amounts of OC, such as biofuel use and biomass open burning. Besides short-term observations, relatively long-term observations (2005)(2006)(2007)(2008) of BC and OC have also been conducted in both urban and rural Beijing (Yang et al., 2011). That the ratios from regional (north-central China) emissions are closer to the observed slopes than local (Beijing) emissions indicates that sources outside Beijing contribute substantially to carbonaceous aerosol levels. Although BC emissions in Beijing are estimated to have declined by 22 % from 2005 to 2008, BC concentrations in urban Beijing did not decline much, and observed BC levels in rural Beijing in fact increased by 21 % (Yang et al., 2011), resulting mainly from increased BC emissions in nearby provinces.    Imaging Absorption Spectrometer for Atmospheric CHar-tographY (SCIAMACHY), and the monthly level-3 product with spatial resolution of 0.25 • × 0.25 • from Support to Aviation Control Service (SACS) is used (data source: http://sacs.aeronomie.be/archive/month/index VCD month.php). To approximate the effects of anthropogenic activities, it is assumed that the SO 2 is in the lowest 2 km above the surface, i.e., that SO 2 is found only in the plan-etary boundary layer. The VCDs of tropospheric NO 2 are from the Ozone Monitoring Instrument (OMI), retrieved by the Royal Netherlands Meteorological Institute (Boersma et al., 2007;2011), and the monthly data with spatial resolution of 0.125 o ×0.125 o are used (data source: http://www.temis.nl/ airpollution/no2col/no2regioomimonth v2.php). The annual means of SO 2 and NO 2 VCDs over mainland China for  Figure 5 illustrates the trends of monthly VCDs of tropospheric SO 2 and NO 2 from satellite observations and of annual emissions of SO 2 and NO x , from 2005 to 2010 by region. To eliminate seasonal variations, the satellite observations are presented as 12-month moving averages, calculated as the means of the data for the previous and subsequent six months. All values are normalized to 2005 levels, to reflect relative changes during the study period. Generally, the growth of tropospheric NO 2 is consistent with the trends of bottom-up annual emissions, confirming increasing NO x pollution in mainland China. Specifically, the emissions and observations match well for developed regions including north-central, east, and south-central China. In the west of the country, however, bigger discrepancy is found between the growth trends of emissions and VCDs (Fig. 5e-f). This is partly because the random retrieval errors of satellite observation can be significant over western regions with relatively clean environment and low NO 2 values. Emissions from natural sources (such as soil and lighting), which had not grown as much as anthropogenic emissions, could contribute more to the VCDs than they do in the developed eastern regions (Lin et al., 2010b). Moreover, the estimated emissions fail to fully capture the drop of NO 2 during late 2008-early 2009, attributed to limits on economic activities and energy consumption to improve air quality for the Beijing Olympics (Wang et al., 2010b) and/or economic downturn in the country (Lin et al., 2011). This discrepancy reveals the limits of emission inventories at annual temporal resolution to reflect responses to short-term variations of economic activity and control policies at local or regional scales.

Comparisons with satellite observation
The comparisons for SO 2 are perplexing. As shown in On the other hand, the exaggerated growth of SO 2 columns particularly for 2010 (Fig. S4f) could also be from the highly uncertain satellite data retrieval due to variations of atmospheric conditions including the varied aerosol absorption and conversion efficiency of SO 2 to sulfate, as suggested by Lu et al. (2011). The discrepancy between emissions and observation has been realized by SACS and reduced uncertainty of data retrieval is expected through reprocessing of SO 2 VCDs from SCIAMACHY (personal communication with SACS).

The effects of policy on emission abatement
During 2005-2010, substantial efforts were undertaken in China to achieve national targets in both energy conservation and emission reduction, particularly in sectors of power generation, cement production, and iron and steel production. Figure 6 illustrates the inter-annual trends of production, coal consumption, and emissions for the three sectors from 2005 to 2010.
For the power sector, as shown in Fig. 6a, the electricity generated from coal-fired plants increased 62 % during 2005-2010, while coal consumption by the power sector increased 46 %, reflecting the progress of energy conservation in the sector. This is mainly due to the replacement of small and old power units with more energy-efficient large units (e.g., super-critical and ultra super-critical units), reducing the coal consumption per unit electricity generation by 10 %, from 370 to 335 grams of coal equivalent per unit kilowatthour (gce kWh −1 ) (updated from previous work by . The higher penetration of large units raised as well the application of advanced emission control technologies for PM (e.g., ESP and FF) and SO 2 (e.g., wet-FGD systems), resulting in strong emission abatement of the two pollutants, by 46 % and 42 % respectively from 2005 to 2010. The annual emissions of NO x and CO 2 , however, had similar growth trends as that of coal consumption, indicating that current control of these two species depends significantly on growth of energy consumption. Although the penetration of SCR technology reached 10 % in the power sector in 2010, the actual effects on NO x control cannot yet be verified (Zhao et al., 2010), in the same manner that FGD has not taken full control effect on SO 2 in recent years. Since coal will continue to dominate the energy structure of China's electricity generation in the near future, the improvement of SCR use, in terms not only of penetration in the sector but also of operational performance and removal efficiency, is likely the most effective way to constrain growth of NO x emissions from coal-fired power plants.
As shown in Fig. 6b, cement production has increased by 76 % from 2005 to 2010, and the actual production of 1.9 billion metric tons in 2010 already exceeded the prediction for 2020 of an analyses using a computable general equilibrium (CGE) economic model . The dramatic growth of cement production, together with that of steel (described later), reflects the unexpectedly swift development of infrastructure facilities in the past few years in China. Meanwhile, the increased penetration of precalciner kilns has improved sector-wide combustion efficiency and expanded the use of emission control devices like FF, leading to considerable reduction of CO and PM. However, NO x emissions increased by 130 % during the study period, resulting both from the swift growth of cement production and the higher NO x emission levels produced by precalciner kilns compared to other kilns. This tension between improved technology and increased NO x emissions in the cement industry indicates that current policies are far from sufficient to reduce associated NO x emissions, and suggests that SCR/SNCR systems need to be promoted in the sector in the future.
During 2005-2010, iron and steel production increased by 116 %, while the coal consumption of smelting and pressing of ferrous metals increased by only 45 % (Fig. 6c). This big achievement in energy saving resulted mainly from the retirement of small steel production plants and the increased use of recycled gas in coke ovens, blast furnaces, and basic oxygen furnaces (Zhao et al., 2012a). However, those improvements had very limited effects on the sintering process, leading to continued growth of SO 2 and NO x emissions. Moreover, PM emissions from the iron and steel industry are estimated to have increased as well, resulting mainly from the fugitive emissions from coking, pig iron production, and casting processes. Among all the species, CO 2 emissions are estimated to have increased fastest, by 87 % from 2005 to 2010, although this growth rate is still lower than that of production. Generally, technology improvement in the sector have had some emission control effect but not to the extent of reversing growth of emissions of any species. A way to further abate emissions would be to increase the penetration of electric arc furnaces in steel making, which employs a short flow process (i.e., reuses waste steel in the material flow) and thus has much higher energy efficiency and lower emission factors than basic oxygen furnaces.
As a summary, although emission control was implemented in some key sectors, the emission trends of given species (e.g., NO x and CO 2 ) are still largely driven by the underlying activity levels, i.e., energy consumption or industrial production. Because of the economic downturn from late 2008 to early 2009, there was a clear leveling off of energy and industrial production at that time, leading to slowed growth in NO x and CO 2 emissions. However, starting in 2009, emissions accelerated again corresponding to the policy to stimulate the economy including enormous investments in infrastructure construction. This strong dependence of emissions on the economy and energy implies that there are still major challenges in emission abatement in China as the economy continuously develops.

Implications of emission trends of different aerosol species
From 2005 to 2010, China's emissions of TSP, PM 10 and PM 2.5 are estimated to have decreased by 14 %, 10 % and 6 %, respectively. The lesser abatement of fine particles indicates more difficulty in emission control than the coarse fraction. In recent years, although penetration of dust collectors into industrial process sources has grown, many of them are cyclones or wet scrubbers, with much lower removal efficiencies for fine particles than TSP. FF systems, which are considerably more effective at PM 2.5 control, are still applied at limited sources including power, cement, and iron & steel plants. Similarly the emissions of BC and OC were less reduced than TSP, since the main sources of those species are residential small stoves burning solid fuels and open biomass burning, with very few technology improvements successfully deployed during recent years. Since fine particles and carbonaceous primary aerosols are much more closely associated with public health and radiative forcing than TSP, there is an urgent need for control measures targeting those aerosol species, particularly for industrial processes and residential fuel combustion. Reduced emissions of PM and thereby alkaline base cations with acid-neutralizing effects may increase the   Fig. 7. The emission reduction rates of 2005-2010 are faster than those of a longer-term trend projected by the authors on the basis of current policy commitments, which suggest that China's base cation and SO 2 emissions will decline 49 % and 22 % from 2005 to 2020, respectively (Zhao et al., 2011b). In contrast, the US national emissions of SO 2 and PM 10 (as a surrogate for base cations, since no emission of base cation was reported) are officially reported to have declined 36 % and 24 %, respectively, from 1990 to 2005, the 15 years following enactment of the 1990 amendments to the Clean Air Act (USEPA, 2012). Even with this aggressive level of SO 2 abatement in the US, it was estimated that the amendments would not be adequate to protect surface waters and forest soils of the northeastern US against further anthropogenic acidification based on long-term observation of a forest catchment (Likens et al., 1996). From 2005 to 2010, another 50 % of US SO 2 emissions were reduced while PM 10 emissions kept relatively stable (Fig. 7). Comparing the situations of the two countries, the much smaller percentage decline of SO 2 and much larger decline of base cations in China indicate that recovery of acidification in the country may be more difficult under current control policies than the US experienced in 1990-2010. Recently, a long-term monitoring study found an association between increased acidity of precipitation and decreased PM concentrations at many sites across China, which cannot be explained by changes in natural sources (Tang et al., 2009). The observation confirmed increased acidification risks due to decreased anthropogenic base cations over the country. Since PM control efforts will doubtlessly continue in China to achieve important benefits of reduced aerosol pollution and avoided damages to public health, little other choice is available to alleviate acidification but to pursue even more stringent SO 2 controls.

The spread of air pollution challenges from urban centers to less developed areas
While China's mega cities have been suffering from poor air quality for a long time (Parrish and Zhu, 2009), satellite observations suggest that even faster growth of air pollutants such as NO 2 is now seen in less densely developed regions compared to mega cities . As shown in Fig. S3 in the Supplement, very limited increase in tropospheric NO 2 VCDs was found in the developed Yangtze River Delta from 2005 to 2010; indeed there was even a small reduction in NO 2 VCD in the mega city of Shanghai. In contrast, much larger growth (exceeding 20 % during 2005-2010) was found for the less developed areas adjacent to Shanghai and the Delta, such as north Jiangsu and Anhui provinces. Similarly, the NO 2 increase during the five years in the mega city of Beijing (around 20 %) was much smaller than that in nearby provinces including Tianjin and Hebei (over 40 %) during 2005-2010. These trends indicate that China's air pollution challenges have been expanding from developed urban areas to nearby regions, explained partly by changes in anthropogenic activities and thereby emissions. In recent years, China's rapid urbanization of relatively small cities and development policies targeting interior regions have spread economic growth to less developed areas, resulting in increased industrial production and energy consumption. Meanwhile, tightened emission controls in the most highly developed, heavily polluted urban areas has lead to relocation of major emission sources from urban to rural regions. Based on county-level economic data provided by the China Data Center, University of Michigan (http://chinadataonline.org/), the fraction of national GDP generated by "secondary industry" (including mining, manufacturing, and construction) in China's urban areas is estimated to have declined from 54 % to 48 % during 2005-2010. The fraction of capacity of coal-fired power plants in urban areas decreased from 56 % to 47 %, according to the updated unit-level database of power plants by . Moreover, the emissions of on-road transportation, which has contributed an increasing share of urban pollution, have been gradually controlled through the implementation of staged emission standards, while those of small industrial boilers and residential stoves, which dominate rural emissions, have not. At the provincial level, Table 5 Table 5 confirm that China's economic activities, energy consumption, and emissions have been shifting proportionately to poorer areas. Thus national air pollution control strategies will increasingly need also to address conditions in those areas in the future.

Conclusions
Under pressures of enormous energy consumption and severe atmospheric pollution, China has been implementing a series of policies in energy conservation and emission reduction in recent years. These include the retirement of small and inefficient power and industrial plants, deployment and operation of FGD and SCR systems in the power sector, and implementation of staged emission control regulations for on-road vehicles. The measures have had varied impacts on the inter-annual trends of emissions of different atmospheric species. The emissions of SO 2 and primary PM have been gradually reduced, although uncertainties around these emission estimates have increased from 2005 to 2010, mainly because of the weakly understood operational conditions of the swiftly increased FGD systems and the unclear penetration levels of dust collectors in key industrial sectors. Emissions of NO x and CO 2 are estimated to have continued increasing, with average annual growth rates of 8.0 % and 7.4 % during 2005-2010, respectively, indicating the limited progress of current measures and ongoing major challenges in emission control of these two species. Although emission control policies greatly reduced TSP, fewer benefits were achieved for fine particles and carbonaceous aerosols, which contribute more to human health damages and climate forcing. Moreover, the estimated swift decline of alkaline base cations in primary PM compared to SO 2 suggests rising acidification risks to ecosystems, as also indicated by long-term obser-vations at multiple sites across the country. There is thus a great future need for a comprehensive, multi-pollutant control strategy, consisting not of separately developed piecemeal policies targeting single atmospheric species but rather conceived to redress a complex of emissions and diverse environmental impacts. In this work, emissions estimated bottom-up are compared with available observations from ground sites and satellites for different species, locations, and periods. While observations generally reflect inter-annual trends of emissions in most cases, clear discrepancies exist for given regions, seasons, or years. These discrepancies result from uncertainties both of observations and of bottom-up emissions. Data limits prevented full realization of regional and seasonal differentiations in activity levels, technology distributions, and emission factors for some sectors, particularly industrial processes and residential fuel combustion. Considering the rapidly changing complex of emission sources across China, further investigation of spatial and time distributions of contributing factors are needed to reduce uncertainties and to generate more accurate trends of emissions.