Spatial distribution and interannual variation of surface PM 10 concentrations over eighty-six Chinese cities

The spatial distribution of the aerosols over 86 Chinese cities was reconstructed from air pollution index (API) records for summer 2000 to winter 2006. PM 10 (particulate matter≤10 μm) mass concentrations were calculated for days when PM10 was the principal pollutant, these accounted for 91.6% of the total 150 428 recorded days. The 83 cities in mid-eastern China (100 ◦ E to 130 E) were separated into three latitudinal zones using natural landscape features as boundaries. Areas with high PM 10 level in northern China (127 to 192 μg m−3) included Urumchi, Lanzhou-Xining, Weinan-Xi’an, Taiyuan-Datong-Yangquan-Changzhi, Pingdingshan-Kaifeng, Beijing-Tianjin-Shijiazhuang, Jinan, and Shenyang-Anshan-Fushun; in the middle zone, high PM10 (119–147 μg m−3) occurred at Chongqing-ChengduLuzhou, Changsha-Wuhan, and Nanjing-Hangzhou; in the southern zone, only four cities (Qujing, Guiyang, Guangzhou and Shaoguan) showed PM 10 concentration >80 μg m−3. The median PM10 concentration decreased from 108 μg m−3 for the northern cities to 95 μg m −3 and 55 μg m−3 for the middle and southern zones, respectively. PM10 concentration and the APIs both exhibited wintertime maxima, summertime minima, and the second highest values in spring. PM10 showed evidence for a decreasing trend for the northern cities while in the other zones urban PM 10 levels fluctuated, but showed no obvious change over time. The spatial distribution of PM10 was compared with the emissions, and the relationship between the surface PM 10 concentration and the aerosol optical depth (AOD) was also discussed. Correspondence to: W. J. Qu (quwj@ouc.edu.cn)


Introduction
The atmospheric aerosol plays an important role in visibility impairment (Watson, 2002), and studies of potential aerosol effects on climate have become of increasing interest following the pioneering work of Twomey (1977) and Charlson et al. (1992).Concerns over adverse influences of particulate matter (PM) on human health (Pope et al., 1995;Tie et al., 2009b) have added to the scientific and public interest in aerosol particles (hereinafter, simply aerosols).
Defining a representative aerosol distribution is essential for understanding the aerosols' effects on climate, especially over regional scales.However, this is challenging because the atmospheric aerosol shows marked spatial and temporal variations owing to the relatively short residence times of the particles and the heterogeneity of their sources (Kaufman et al., 2002).Aerosol distributions over the continent are generally regional because of the diversity in emissions (especially anthropogenic sources), unevenness in population density, and variations in terrain, which can concentrate air pollutants or enhance their dispersion.In this context, gradients in aerosol concentrations obviously exist, and these result in differences in the aerosols' effects.For example, data from an aircraft study over the Los Angeles Basin revealed horizontal gradients in aerosol concentrations that could result in variability of more than 50% within a 5×5 km computational grid cell (Collins et al., 2000).Related studies over the tropical Indian Ocean indicated that gradients in aerosols could lead to an inter-hemispheric difference in the solar heating (Rajeev and Ramanathan, 2001).Along these same lines, a gradient in the aerosol optical depth (AOD) could lead to difference in the reduction of the noontime solar flux from as much as −38 W m −2 in the Arabian Sea to as little Published by Copernicus Publications on behalf of the European Geosciences Union.
as −2 W m −2 south of the Intertropical Convergence Zone, thus producing a strong north to south gradient in the climate forcing (Meywerk and Ramanathan, 1999).Recent research has confirmed a strong latitudinal gradient in the total aerosol and black carbon (BC) over the Bay of Bengal with both decreasing rapidly from north to south (Nair et al., 2008).In addition to these latitudinal aerosol gradients, a strong west-to-east gradient of BC has also been observed during research cruises along latitudes of 30 • N and 35 • N in the central Pacific Ocean; this was attributed to outflow from Asia (Kaneyasu and Murayama, 2000).
As for the health impacts of air pollution, the sources of air pollutants are mostly restricted to the earth's surface; the main exception to this is aircraft emission in the upper troposphere (Highwood and Kinnersley, 2006).Meanwhile, approximately 47% of the global population lives and works in urban areas, which are the most polluted parts of the planet.Investigations of surface air quality over urban areas and strategies for its improvement are therefore compelling.
China, a developing country with the world's largest population, has undergone rapid economic growth since economic reforms began in 1978.Along with urbanization and industrialization has come an increase in energy consumption, and this has brought with it growing concerns over air pollution.Investigations of the spatial distribution and interannual variations in the atmospheric aerosol and air pollutants are necessary to understand the aerosol's potential effects on atmospheric chemical process and climate, these are also essential for the development and implementation of effective air pollution control strategies.
There have been numerous ground-based studies of the aerosol in China, including measurements of the physical and/or chemical properties at remote sites (Tang et al., 1999;Li et al., 2000;Qu et al., 2008), regionally representative rural sites (Xu et al., 2002;Wang et al., 2004a;Zhang et al., 2005) and urban sites in large and mega cities (Fang et al., 1999;Davis and Guo, 2000;He et al., 2001;Zhang et al., 2002;Cao et al., 2003;Ye et al., 2003;Ta et al., 2004;Duan et al., 2005Duan et al., , 2006;;Wu et al., 2005;Zheng et al., 2005;Feng et al., 2006;Li et al., 2006Li et al., , 2008;;Meng et al., 2007;Chu et al., 2008;Deng et al., 2008aDeng et al., , 2008b)).However, even taken together, the existing studies do not provide a systematic picture of the spatial distribution of the aerosol across China.This is in part because of the limited and discontinuous coverage of most observations and also because of difficulties in inter-comparing results among research groups who use different sampling instruments and analytical methods.As a result, there is limited information on the spatial distribution of the surface aerosol across the country.
On the other hand, information on the spatial distribution of aerosol over China has been obtained from radiation/optical measurement networks and satellite observations (Luo et al., 2000(Luo et al., , 2001;;Li et al., 2003;Wang et al., 2008).However, the columnar AOD or aerosol optical thickness (AOT) derived from these types of observations is influenced by many factors including the vertical structure and height of the atmospheric mixing layer; therefore, the characteristics and variations of the AOD/AOT will differ from those of the surface aerosol as will be discussed in Sect.3.7 below.Chan and Yao (2008) and Fang et al (2009) have reviewed the air pollution situations and air quality management practices in Chinese mega cities.Another review (Shao et al., 2006) indicated high levels of airborne PM in megacities such as Beijing, Shanghai, and Guangzhou as well as regional aerosol pollution in the vast region extending from the North China Plain (NCP) to the Yangtze River Delta region (YRD) and the heavily urbanized Pearl River Delta region (PRD).Tie et al. (2006a) characterized air pollution in Eastern China and the Eastern US to assess the differences between photochemical conditions in these two regions.Tie and Cao (2009a) discussed several crucial issues regarding aerosol pollution in the highly populated regions over China.A recent study (Song et al., 2009) characterized spatial and seasonal variations of PM 10 concentrations (for 47 groundbased air quality monitoring sites) and the Moderate Resolution Imaging Spectroradiometer (MODIS) AOD over China during the period 2003-2005, but interannual variations were not considered.
As inhalable particulate pollutant, PM 10 (PM with diameters ≤10 µm) is of primary concern for most Chinese cities, and PM 10 is often reported as the principal pollutant (abbreviation as P prin hereinafter, see below for definition) for urban areas that are monitored by the Environmental Protection Agency of China (EPA-China).In this paper we make use of 137845 daily air pollution index (API) records for eighty-six major cities (Fig. 1, Table 1) to estimate the concentrations of PM 10 over urban areas in China.The results are then used to study the spatial variability of PM 10 and to evaluate its seasonal and interannual variations.
The paper is organized as follows.In Section 2 we give an overview of methods and data.Section 3.1 presents occurrences of the air quality categories for the cities.Seasonal variations of the APIs and PM 10 are presented in Section 3.2, and the influence of Asian dust on the springtime APIs is presented in Section 3.3.Section 3.4 and 3.5 describe spatial distribution of urban PM 10 concentration versus emissions and precipitation as well as the broad pattern of latitudinal and longitudinal gradients in PM 10 level for the cities; Section 3.6 presents interannual variation trends in urban PM 10 for northern, middle and southern China.Section 3.7 presents comparison of PM 10 variations with AOD/AOT results, and the factors that influence the relationships between PM 10 and AOD/AOT are also discussed.Section 4 finally gives conclusions.
Fig. 1.Locations of the eighty-six Chinese cities for which data are available.Full names of these cities are listed in Table 1.Red frames labeled with G-1 through to G-14 include the cities assigned to the specific group as in Table 1.

Air Pollution Index (API) data
The Chinese air pollution index (API) is a semi-quantitative measure for uniformly reporting air quality, and it is based on a set of atmospheric constituents that have implications for human health.The API for each city, which is reported daily, converts the concentrations of the pollutants of interest into a dimensionless number from 0 to 500.To calculate the APIs, (1) five pollutants (PM 10 , SO 2 , NO 2 , CO and O 3 ) are continuously measured at several monitoring stations in different districts or types of areas in a city (commercial, cultural, downtown, residential, traffic, industrial, etc.) and at a clean background station; (2) the daily average concentrations of these pollutants (except CO and O 3 which are reported as hourly means) are then calculated from the measurements made at all of the stations; (3) a sub-pollution index (sPI) is calculated for each pollutant by linear interpolation of the average concentration of the specific pollutant between the grading limits for each air quality classification listed in Table 2; (4) the maximum sPI is reported as the API for the city and day, and the pollutant with the maximum sPI is reported as the P prin .The APIs thus provide a broad measure of the air quality in the entire city, including polluted areas as well as a clean site.Air quality classifications corresponding to the APIs, air quality management recommendations, and pollutant concentrations are summarized in Table 2. Further information about the API may be found in Zhang et al. (2003), Chu et al. (2008) 1).

PM 10 Data
For the days when PM 10 was reported as P prin , daily PM 10 concentrations were derived from the APIs by using the following equation: where C is the concentration of PM 10 , I is the API reported.I low and I high represent API grading limits that are lower and larger than I , respectively; C high and C low denote the PM 10 concentrations corresponding to I high and I low (Table 2), respectively.The method for calculating PM 10 concentrations from APIs has been described in detail by Zhang et al. (2003).A total of 137 845 PM 10 concentrations (on days when PM 10 was P prin or no P prin was reported) were calculated using this approach, and this accounts for 91.6% of the API records.The other 8.1% and 0.3% of the API records correspond to days when SO 2 or NO 2 was reported as P prin .Characteristics of the deduced PM 10 concentrations are the main focus of this paper, and the days when the oxides of sulfur or nitrogen were the P prin are not considered further.
In China, PM 10 concentrations are most often measured with the use of Tapered Element Oscillating Microbalance analyzers (TEOMs, model 1400a, Rupprecht & Patashnick, USA) deployed at the stations by the EPA-China environment monitoring system.The uncertainty of the daily PM 10 measurement is typically less than 1% (Xia et al., 2006).However, for some cities and stations, β ray particulate monitors (model BAM-1020, Met One, USA) were used to monitor PM 10 .The principles of the operation of the TEOM analyzers and β ray particulate monitors as well as their field applications are summarized elsewhere (Bari et al., 2003; http://www.metone.com/documents/BAM-10206-08.pdf).
It is worth noting that for the days when PM 10 was P prin and the API was reported as 500, the PM 10 mass loadings actually could have exceeded the upper limit for the air quality classification of "high-level pollution" (600 µg m −3 ).Such high paticulate loading can occur during severe dust storms in some northern cities.However, in this study, we set all of the API records of 500 to PM 10 loadings of 600 µg m −3 .This approximation does introduce some uncertainty in the analysis, but the instances when this happened were infrequent.For the full dataset, only twenty-five cities had maximum API records of API = 500.Furthermore, only four cities recorded more than ten days when API = 500; these were Lanzhou (47 days), Urumchi (36 days), Xining (24 days) and Beijing (12 days), and the days when API = 500 accounted for small percentages, about 1.9%, 1.5%, 0.98% and 0.49%, of the days with API records (2451 days).Taken together, the other twenty-one cities had 57 records with a maximum API (API = 500).Therefore, the impact of this uncertainty should be quite limited.
Another consideration is that no P prin was reported on days when the API was less than 50 (the air quality classified as "clean").In this regard, an inspection and comparison of the daily sPI records for PM 10 , SO 2 and NO 2 in selected cities has shown that when no P prin was reported, PM 10 most often had the highest sPI, and therefore the APIs were most often a reflection of the PM 10 loadings.Accordingly, for the days when no P prin reported, we assumed that PM 10 was the P prin that day and deduced PM 10 concentration according to Eq. ( 1).This is another source of uncertainty, but again the impact of this assumption is likely to be small.
As the data quality (the validity of this PM 10 data set) is of the foremost concern, a comparison was conducted between this study and related previous work.The specific concern here is that the use of TEOM-based analyzers (which are commonly used in the EPA-China's air quality monitoring network) can result in underestimation of the aerosol mass because the elevated temperatures of the TEOMs (50 • C) can cause the loss of some material via volatilization.King (2000) has reported that the results obtained with the TEOMbased analyzers are similar to the results of other instruments, but the TEOM results tend to be lower by about 30% because of the volatilization effect.Sciare et al. ( 2007) has indicated that the mass contributions from semi-volatile material (SVM) and liquid water in Beijing aerosols can cause large errors in TEOM results.These authors showed losses of SVM as high as 140 µg m −3 during high relative humidity (RH, >60%) periods in summer 2004.
It is worth noting that the error of TEOM measurements (heating at 50 • C) due to volatilization of SVM is systematic in the sense that it is always negative.This potential artifact in PM mass measurements could therefore result in a systematic underestimation of PM 10 .On the other hand, literature reported PM 10 mass concentrations based on filter sampling and gravimetry are subject to both positive and negative artifacts.That is, reactive gases can accumulate on the filters, but there also can be losses of SVM -such as nitrate and some organic material -via volatilization (Turpin et al., 2000).
These comparisons show that any artifacts in the PM 10 dataset would likely have a limited influence on the characteristics of spatial distribution and temporal variations in PM 10 as revealed here.That is, the PM 10 values calculated from the APIs were generally comparable with the aerosol concentrations based on other techniques such as filter sampling and gravimetry (supplementary material 1).Linear regression (Fig. 2) shows that the literature reported PM 10 concentrations were about 10% higher than our API-deduced PM 10 , and the correlation coefficient was 0.84 (p<0.0001significance), thus validating the API-derived PM 10 concentrations in our study.Furthermore, because of differences in site selection -as noted above, a clean background site is included in the stations used to calculate APIs -it is possible that the literature reported aerosol concentrations, which mostly measured at residential, commercial or heavily trafficked site, were in fact higher than the API-deduced PM 10 values.
Other researchers (Okuda et al., 2004(Okuda et al., , 2008;;Bi et al., 2007) have compared PM estimated from API records and measured PM, and they found similar results confirming the validity of the API dataset.For example, studies have shown that the monthly mean PM 10 concentrations reported by EPA-China in Nanning city during 2006 were close to those from a network operated by the China Meteorological Administration, and they showed the same variations with time (Mo et al., 2008).These comparisons all attest to the validity of the API-based PM 10 data used in our study.

Grouping the cities according to APIs
A fuzzy clustering procedure was used to determine the similarities between the APIs of the cities (supplementary material 2).In addition, the relationships between among the APIs and the geographical locations of these cities were taken into account in the grouping procedure.For example, the cities in a specific topographical basin were mostly assigned to a same group; thus, Huhehot and Datong which are both located on the northeastern margin of the Great Bend of the Yellow River were assigned to group G-3.On the other hand, cities located on different sides of a mountain were assigned to different groups, e.g.Shijiazhuang and Taiyuan which are located on the eastern and the western sides of the Taihang Mountains are assigned to groups G-6 and G-3, respectively.Meanwhile, cities in the same administrative prefecture (province) and those adjacent geographically were in most cases assigned to the same group for the convenience of discussion.Through these procedures, eighty-three cities in mideastern China (longitude 100 • E to 130 • E) are divided into fourteen groups (Fig. 1, Table 1).These groups generally correspond to provincial regions.Three cities, Kelamayi (KLY), Urumchi (UMQ) and Lhasa (LS) located in western China were not included in these groups.To better understand spatial patterns in the air quality categories, the fourteen groups of the cities (Fig. 1, Table 1) were further partitioned into northern, middle and southern zones, with the Qinlin Mountain -the Huaihe River and the Yunnan-Guizhou Plateau -the Jiangnan Hill -the Wuyi Mountain as boundaries between zones (marked with black dash lines, Fig. 3).
The northern zone (north of 33 • N, including thirty-eight cities, i.e.G-1 to G-8) extended westward to the northeastern margin of the Tibetan Plateau, northward to the Chinese boundary with Mongolia and Russia, eastward to the Japan Sea, the Bohai sea and the Yellow sea.The middle zone (between 28 • N to 33 • N, including twenty-six cities, i.e.G-9 to G-11) extended westward to the eastern margin of the Tibetan Plateau, eastward to the Yellow sea and the East China Sea.While the southern zone (south of 28 • N, including nineteen cities, i.e.G-12 to G-14) extended westward to the southeastern margin of the Tibetan Plateau, eastward to the East China Sea and Taiwan Island, and southward to the South China Sea.The northern and middle zones actually contain two basins oriented east/west while the southern zone is mainly a hilly area near the South China Sea and the East China Sea.

Results and discussion
An inspection of the Chinese API records suggests that PM 10 pollution is likely a greater concern than SO 2 and NO 2 in major cities because PM 10 is the most common P prin (Table 3).One can see that for the northern and middle zones of the country, PM 10 was P prin on more than 80% of the days; while for the southern zone, about half of the days had PM 10 as P prin (and ∼40% of the days with no P prin reported).The percentages of days with SO 2 as P prin ranged from 6.6% to 9.3% for the three latitudinal zones.In comparison, the numbers of days with NO 2 as P prin were quite small but with relatively larger proportions for the southern zone (0.87%) compared with the northern and the southern zones (0.11% and 0.12%, respectively).This pattern in NO 2 as P prin is probably due to more emissions from motor vehicle and more favorable conditions for the photochemical production of NO 2 in the south.

Occurrences of the air quality categories
The percent occurrences of days with an air quality category of "clean" (API≤50) were greater than 17.4% for most of the cities in the southern zone (hereinafter referred to as the southern cities) but less than that for most of the middle and northern cities (Fig. 3a).Another feature of the data, which was to be expected, was that the occurrences of "clean" air quality were generally greater for the coastal cities compared with the inland cities (Fig. 3a).Fig. 3b illustrates the percent occurrences of the days with air quality classification as "clean" or "good" (API≤100, that is, days on which the air quality met the Chinese National Grade II Ambient Air Quality Standard, CNAAQS-GII).One can see that in the southern zone, fifteen cities had air quality that met this standard on more than 95% of the days.The exceptions to this were Guiyang, Liuzhou, Shaoguan and Guangzhou; in these four cities, less than 95%, but more than 76.6%, of the days had air quality that met the standard.
Meanwhile in the northern zone, there were only six cities (four coastal cities: Dalian, Qinhuangdao, Yantai, Rizhao and two inland cities: Tai'an, Weifang) that had more than 95% of the days when the air quality met CNAAQS-GII (Fig. 3b).The cities that most often met the standard (occurrences of "clean" or "good" air quality >76.6%, Fig. 3b) were restricted to the coastal area (six cities), the Shandong Peninsula (SDP, five inland cities), and the remote Heilongjiang and Jilin Provinces (three and one cities, respectively).In the remaining large area of the northern zone, eighteen of the twenty-three cities, less than 76.6% (but more than 50.3%) of the days had air quality that met the standard (Fig. 3b).
For the cities in the middle zone, the percentages of the days on which air quality met CNAAQS-GII were intermediate between the northern and the southern zones (Fig. 3b).In contrast to the coastal cities in the other two zones, the occurrences of "clean" or "good" air quality for the coastal cities in the middle zone were less than 95% (Fig. 3b).This was most likely due to heavy anthropogenic emissions in YRD, which is on the coast of the middle zone.
The percent occurrences of days with air quality categorized as "low-level pollution" (100<API≤200) were larger for the northern cities than for those in the middle zone; and the probabilities for this classification were lowest for the cities in the southern zone (less than 10.6% except Liuzhou, Fig. 3c).Similar patterns can be seen for the occurrences of days with air quality categories of "mid-or high-level pollution" (API >200, Fig. 3d).Except for Guiyang and Guangzhou, which had relatively low occurrences of "midor high-level pollution" (0.041% and 0.12%, respectively), these classifications never occurred in the southern zone.

Seasonal variations of the APIs and PM 10
High APIs most often occur during winter (December, January and February) and spring (March, April and May) while the lowest APIs are recorded during summer (June, July and August, Fig. 4).This is true for the individual cities and more generally for the three zones.The seasonality in PM 10 for the three latitudinal zones was similar to that of the APIs, i.e. with wintertime maxima and summertime minima (Table 4).In addition, the seasonal variability of PM 10 in the southern zone was not as large as in the other zones.
The wintertime PM 10 maxima were associated with the combustion of fossil fuels for domestic heating (Li et al., 2008) and with more frequent occurrences of stagnant weather and intensive temperature inversion during the colder months.These can result in the accumulation of atmospheric particles and lead to high PM episodes (He et al., 2001;Xia, et al., 2006;Chan and Yao, 2008).Meanwhile, high springtime PM 10 concentrations always followed the wintertime PM 10 maxima (Table 4), and this is probably due to dust events (Wang et al., 2004b).The exception is slightly lower PM 10 levels in spring than autumn in the southern zone where the influence from dust events is weakest.In contrast, compared with other seasons, the PM 10 minima in summer were typically associated with (1) the absence a Proportion in percentage (%).b The three latitudinal zones are same as those illustrated in Fig. 3.
of anthropogenic emissions from domestic heating, (2) more efficient diffusion-dilution of the pollutants due to enhanced convection in a higher atmospheric mixing layer (Xia, et al., 2006), (3) more frequent and abundant precipitation due to the summer monsoon which increases wet scavenging and (4) the inflow of comparatively clean maritime air in the summertime prevailing wind pattern, especially in coastal areas.
For the cities in the northern zone, the maximum APIs were commonly 500 during winter and spring (Fig. 4a), but they rarely (four records) reached 500 for the cities in the middle zone (Fig. 4b).Continuing this trend, the maximum APIs for cities in the southern zone only reached 300 for a few days in autumn 2001 and 2003 (Fig. 4c).A possible explanation for this is that strong wintertime emissions from heating as well as mineral aerosol from springtime dust storms impact the northern zone, and variations in sources such as these can undoubtedly lead to strong seasonal changes in the APIs.In comparison, in southern China, there is less domestic heating in winter, and there are weaker influences of dust storms in spring.c The three latitudinal zones are same as those illustrated in Fig. 3.
Changes in types of pollutants emitted over decadal timescales could impact the seasonality of air pollution.For example, Guinot et al. (2007) found that the winter "heating season" pollution appears to be of lesser importance in Beijing than previously, whereas automobile traffic is likely to dominate downtown anthropogenic emissions in the future.In addition, problems associated with photochemical processes and the formation of fine secondary particles also tend to occur in Beijing summer when temperatures and relative humidities are high (Song et al., 2002).However, the air pollution situations are complicated as they vary largely between different cities.
Accompanied with the implementation of the air quality improvement measures (such as substituting natural gas for coal, controlling emissions from coal combustionimproving combustion technology and gas desulfurization technology, etc.) in China, a dramatic increase in the number of motor vehicles has occurred at the same time.More stringent standards (such as the Euro-III standards since 2006 in Beijing and the Euro-II standards since 2003 in Shanghai) were also adopted to reduce vehicular emissions of air pollutants (Chan and Yao, 2008).In their review, Chan and Yao (2008) indicated that due to effective control measures, NO 2 and CO concentrations have not increased in China's mega cities (including Beijing, Shanghai as well as Guangzhou, Shenzhen and Hong Kong in PRD) although the number of vehicles has increased by about 10% per year; SO 2 emissions were successfully controlled in Beijing, but not in Shanghai or the PRD; meanwhile, PM pollution is still severe and is the major air pollution problem in the mega cities.

Influence of Asian Dust on the APIs during Spring
The number of springtime occurrences of the maximum API (= 500) for the northern cities was about 30 for 2001 and 2002; this decreased to about 5 for 2003 through 2005, but increased steeply again to 22 in spring 2006 (top of Fig. 4a).Note that the occurrences as tabulated here reflect the total number of days with API = 500 for all cities in the northern zone, that is, if the API records reached 500 for more than one (e.g.n) city on a given day, the number of occurrences was tabulated as n for that day.The trend in the numbers of the days with API>300 for the northern cities during spring were similar to the occurrences of the API maxima, from 82 and 69 in 2001 and 2002, respectively, to about 12 for 2003 through 2005, then increasing again to 60 in spring 2006 (top of Fig. 4a).
The high APIs for the northern cities co-varied with the occurrences of dust events, and this reflects the fact that the springtime air quality in northern China can be degraded by Asian dust storms.The dust data used for this comparison are the frequencies of springtime dust storms in northern China that were extracted from satellite images by Zhang et al. (2008).Their study showed that there were 11, 8, 2, 3, 3 and 10 dust events in sequence from 2001 to 2006; and this measure of dust storm activity matches the interannual trend in the occurrence of the maximum APIs for the northern cities during spring (top of Fig. 4a).Li et al. (2007) also reported similar trends in the frequencies of springtime dust events in Shijiazhuang city (2, 0, 1, 2 and 3, respectively for 2002 to 2006).A similar pattern in dust event frequencies also has been reported for Xi'an (Ning et al., 2005).More to the point, Wang et al. (2006) showed that dust events contributed to variations in PM 10 over fourteen northern cities in China.The PM 10 levels in the cities in the southern zone were generally lower than those in the northern zone and the middle zone, and there were no areas with arithmetic mean PM 10 concentration >109 µg m −3 (Fig. 5a).However, the arithmetic mean PM 10 concentrations (in µg m −3 ) exceeded 80 in four cities: Qujing (91), Guiyang (80.1),Guangzhou (85) and Shaoguan (87), and these were much higher than the total average PM 10 level of 58 µg m −3 for the southern zone.

Spatial distribution of PM
In general, the PM 10 hotspots were in cities with dense populations and intensive industrial activities.Moreover, cities located within geographical basins tended to have higher PM 10 loadings than those outside the basins.Relatively high PM 10 concentrations were documented in areas such as central-eastern China, NCP, SCB, YRD (Nanjing and Hangzhou), and PRD (Guangzhou and Shaoguan, PM 10 concentration >80 µg m −3 , Fig. 5a).These regions also have been identified as high AOD/AOT areas by the radiation/optical measurements and satellite observations (Luo et al., 2000(Luo et al., , 2001;;Li et al., 2003;Wang et al., 2008).At the same time, one should note that during a pollution episode, the aerosol spatial distribution could be quite different from the average conditions as illustrated in Fig. 5a.More information about the relationships between PM 10 and optical measurements can be found in Sect.3.7 below.
The spatial distribution of PM 10 concentration is not presented here as an interpolated contour plot because the dataset is biased toward sites that are generally the most polluted "points" on a regional scale.That is, the cities typically are high-value centers in the field of PM 10 , with low-value areas interspersed between these high PM 10 loci.Although these results do not represent the full spatial distribution of the aerosol as evident in satellite observations, they do provide important insights into the spatial distributions of the aerosol and the problems of air pollution based on independent measurements of PM in surface air.
Table 4 tabulates the arithmetic mean PM 10 concentrations for the fourteen groups described in Section 2.3 above.Note these values tend to be less than the individual PM 10 levels in mega cities because some cleaner small and satellite cities are included in the groups.From Table 4 one can see that in the northern zone, G-4 (Heilongjiang and Jilin Provinces) and G-7 (SDP) exhibited lower PM 10 levels compared with other groups during spring, summer and autumn; while during winter, G-1 (Qinghai, Gansu and Ningxia Provinces), G-2 (Shaanxi Province), G-3 (Inner Mongolian and Shanxi Provinces) and G-5 (Liaoning Province) had higher PM 10 levels than other groups.For the three groups in the middle zone, G-9 (YRD) showed the lowest PM 10 concentration in all four seasons.In the southern zone, PM 10 levels were comparable for the three groups.For a more detailed description of the seasonal patterns of PM 10 spatial distribution, see supplementary material 3.
Although the spatial distribution of PM 10 is complex, an overall north to south decrease in our API-derived PM 10 data for the Chinese cities is distinct and can be seen in Fig. 5a; this will be discussed in detail in Sect.3.5.

Comparison of PM 10 with emissions and precipitation
Accurate emissions data are scarce for China, and research in this area is complicated by differences in the combustion technologies used throughout the country and the limited www.atmos-chem-phys.net/10/5641/2010/Atmos.Chem.Phys., 10, 5641-5662, 2010   availability of information regarding energy consumption, industrial output, etc.An inventory of anthropogenic air pollutant emissions in Asia was developed for the year 2006 for the INTEX-B project (the Intercontinental Chemical Transport Experiment-Phase B, Zhang et al., 2009).We compare the PM 10 distributions we derived with the emissions of both BC and PM 10 from this inventory.Note BC and PM 10 distributions from this inventory show similar patterns with results from other reseach on the anthropogenic nitric oxide (NO) emissions (see Fig. 7 in Tie et al., 2006b), indicating they are representative of emissions from anthropogenic activities.
PM 10 loadings in mid-eastern China (Fig. 5a) were generally similar to the emission patterns of anthropogenic BC (Fig. 5b).Cities with high PM 10 (>109 µg m −3 , indicated by pink-magenta-red symbols with concentrations from low to high, Fig 5a) coincided with the areas with strong BC emissions (>500 tonnes/year/0.5 • cell, filled with yelloworange-red with emissions from less to large, Fig. 5b) such as the Beijing-Tianjin municipalities and southern Hebei Province, mid-eastern Henan Province, and western-central Liaoning Province in northern China.The high PM 10 area of Chongqing-Chengdu-Luzhou (Fig. 5a) coincided with the intensive BC emission areas of SCB and northwestern Guizhou Province (Fig. 5b), while the high PM 10 area of Wuhan-Changsha (Fig. 5a) coincided with the intensive BC emission areas of mid-eastern Hubei Province and mid-eastern Hunan Province (Fig. 5b).
In Shanxi Province, on the other hand, the high PM 10 loadings for Taiyuan-Datong-Yangquan-Changzhi (ranging from 132 to 179 µg m −3 , Fig. 5a) were seemingly at odds with the generally moderate BC emissions in that province (Fig. 5b).However, point sources of BC (>500 tonnes/year/0.5 • cell, filled with yellow, orange or red) as shown in Fig. 5b coincided quite well with the locations of those four cities in Fig. 5a.The same was true for Weinan, Xi'an and Baoji in Shaanxi Province, for Lanzhou in Gansu Province, for Xining in Qinghai Province, and for Urumchi in the Xinjiang Uyghur Autonomous Region.
In contrast, the distributions of PM 10 concentration (Fig. 5a) along the east coast of China apparently do not follow BC emissions (Fig. 5b).For example, although BC emissions in SDP and YRD are high (Fig. 5b), the average PM 10 concentrations for the cities in SDP during 2000 to 2006 were not particularly high, less than 109 µg m −3 except for Jinan (135 µg m −3 , Fig. 5a).PM 10 concentrations for the cities in YRD also were <109 µg m −3 except for Nanjing (125 µg m −3 ) and Hangzhou (119 µg m −3 , Fig. 5a).Possible reasons for the decoupling of PM 10 and emissions along the Chinese coast are discussed below.
Emissions of anthropogenic PM 10 (Fig. 5c) were generally similar to those of BC (Fig. 5b), but the high PM 10 emission areas are widespread, in contrast to the distributions of strong BC emissions-mostly concentrated in relatively limited regions.Similar to the relationship between the BC emissions and derived PM 10 described above, the cities with high PM 10 loadings (>109 µg m −3 , Fig 5a) coincided with the areas with strong PM 10 emissions (>5000 Mg/year per grid, filled with orange-red-magenta with emissions from less to large, Fig. 5c).This was true for the Beijing-Tianjin municipalities, Hebei-Shanxi-Henan provinces, Hubei-Hunan provinces, and SCB-Guizhou Province.
Along the Chinese coast, including SDP, YRD and PRD, the relatively low PM 10 loadings (Fig. 5a) were in contrast to the strong PM 10 emissions there (Fig. 5c).For example, except for Tianjin, the cities along the coast including those around the Bohai Bay and SDP, in YRD and PRD all showed PM 10 concentrations less than 109 µg m −3 (Fig. 5a).Thus, the intensive PM 10 emissions in coastal regions bordering the Bohai Sea, the Yellow Sea, the East China Sea, and the South China Sea (Fig. 5c) evidently do not result in especially high PM 10 loadings.
The discrepancy between the concentrations and emissions of PM 10 along the coast can be explained by dilution of the pollutants by clean marine air and by wet scavenging of the PM by the more abundant precipitation there.With reference to the latter, an examination of the mean annual precipitaiton for fifty-one Chinese stations from 2000 to 2006 (Fig. 5d) shows that precipitation at the coastal sites is generally larger than those for the inland sites at the same latitude.Long-term averaged annual precipitations (1961 to 1990) for the areas along the Chinese coast were also larger than those for the inland regions at the same latitude (Fig. 5e).These findings support the idea that more PM is scavenged from the atmosphere by precipitation in coastal regions.
Moreover, as RH and liquid water content are related to the levels of SVM and nitrate in aerosols, they are believed to play a major role in the gas-particle partitioning of semivolatile species; volatilization loss of these SVM can result in underestimation of PM mass by the TEOM measurement heating at 50 • C (Sciare et al., 2007).Therefore, the PM 10 concentrations calculated from APIs may be more strongly affected by this artifact in the costal areas where RH is generally higher compared with inland sites.This may cause artificially lower PM 10 concentrations within our dataset along the coast.
In summary, except for the coastal areas, the distribution of PM 10 concentration was generally consistent with the anthropogenic emissions of BC and PM 10 over China.Additional information about anthropogenic emissions over China with respect to monthly and seasonal variations can be found in supplementary material 4-Monthly mean tropospheric NO 2 over Southeast Asia during October 2004 to February 2007 extracted from OMI (the Ozone Monitoring Instrument) version 1.0 (website, http://www.temis.nl/airpollution/no2col/no2regioomimonth col3.php)(Boersma, et al., 2007).

Latitudinal and longitudinal gradients of PM 10
The PM 10 concentrations for the northern zone generally were slightly higher than or comparable with those in the www.atmos-chem-phys.net/10/5641/2010/Atmos.Chem.Phys., 10, 5641-5662, 2010 Fig. 6.Comparison of PM 10 levels and variations for the three latitudinal zones in different seasons: the box-and-stem plots depict the minimum, the 1th, 5th, 25th, 50th (median), 75th, 95th, 99th percentile and the maximum for the PM 10 concentration, the red square in the box depicts the arithmetic mean PM 10 concentration, the median PM 10 concentrations are also presented.
middle zone during spring, summer and autumn (Table 4).
In the winter, however, PM 10 concentrations in the northern zone were much higher than in the other zones, presumably due to emissions from domestic heating in the colder northern zone coupled with the accumulation of pollutants in the stable lower boundary layer (especially during nighttime).Along these lines, the median PM 10 concentration at Urumchi during winter was 317 µg m −3 , which is about three times that in spring and autumn (92 µg m −3 and 116 µg m −3 , respectively), and more than five times that in summer (60 µg m −3 ).This high aerosol loading at Urumchi indicates very severe wintertime PM pollution in that city.Overall, the PM 10 concentrations in the southern zone were generally the lowest of the three zones (Table 4); they were only about half of those for the northern zone but more than half of those for the middle zone during all seasons.
Although the PM 10 concentrations showed a north-south difference in all the seasons, it was most pronounced in winter and less so in autumn (Table 4).
In addition to the distinct north to south decrease in PM 10 , further information on the patterns in PM 10 can be seen in the box and stem plots shown in Fig. 6; that is, both the concentration and the variability in PM 10 were the lowest in the southern zone.The PM 10 variances in the northern zone were greater than the middle zone during spring, comparable during summer, but lower during autumn and winter.Another general feature of the data is that the arithmetic mean PM 10 concentrations are typically larger than the medians, especially for the northern zone and the middle zone (Fig. 6); this indicates that the PM 10 concentrations in those parts of the country were periodically elevated by severe pollution episodes.
Possible reasons for the relationship between latitude and PM 10 loading include (1) greater emissions from domestic heating during winter in the north (see supplementary material 4), (2) stronger impacts from dust storms during spring for the cities in northern China, and (3) more efficient dilution of pollutants by the influx of clean maritime air and more efficient wet scavenging of PM by precipitation in southern China (Fig. 5d and e).In summary, the patterns in PM 10 reflect the combined impacts of emissions, transport, dilution and removal on aerosol loadings over a broad region.
In addition to the latitudinal gradient, a distinct longitudinal aerosol gradient is evident in the northern zone where PM 10 concentrations decreased from west to east (supplementary material 5).Weaker influences from desert dust as well as more efficient dilution with relatively clean marine air likely contribute to the low PM 10 levels in the eastern cities in northern China.A strong west-to-east gradient of BC between 30 • N and 35 • N also has been observed over the central Pacific Ocean (Kaneyasu and Murayama, 2000).On the other hand, for the cities in the middle and southern zones, PM 10 concentrations also generally exhibited a decreasing trend from west to east, but the longitudinal differences in those cases were smaller and insignificant.
The latitudinal and longitudinal gradients in PM 10 concentration as depicted here only show broad patterns in the urban aerosol over China; the spatial distribution of the atmospheric aerosol (as described in Sect.3.4) is undoubtedly more complicated.

Interannual variations
There were changes in the number of cities included in the dataset over the course of the study (Table 1), and to circumvent any biases from this, only the cities with full records (sixteen, eleven, and twelve cities in the northern, middle and southern zone, respectively) were considered in the following assessment of temporal trends in PM 10 (Fig. 7).A comparison between the entire dataset (all eighty-three cities in the three latitudinal zones) and the subset (thirty-nine cities with full records) shows similar patterns in PM 10 variations, but the arithmetic means (or medians) of the subset were generally higher.
We used the median values to assess interannual variations in the PM 10 concentrations for the three latitudinal zones for 2000 to 2006 (Fig. 7).This was done to reduce the influence of the rare exceptionally high values on the results.As noted above, whenever the reported API was 500, the PM 10 concentration was arbitrarily set to 600.This would have introduced some additional complications into the assessment if it had been based on the arithmetic or geometric mean PM 10 concentration, but it does not influence the assessments of trends based on the median PM 10 concentrations as presented here.
For the middle zone, the springtime PM 10 level varied in a manner similar to that seen in the northern zone (Fig. 7a), probably reflecting the widespread influence of dust events.There was an overall decreasing trend of summertime PM 10 concentration from 88 µg m −3 to 71 µg m −3 over the course of the study, but with higher levels (∼90 µg m −3 ) from 2002 to 2004 (Fig. 7b 7d).The annual PM 10 level varied around 100 µg m −3 , from 99 µg m −3 in 2000 to 97 µg m −3 in 2006 (Fig. 7e).Overall, the PM 10 level in the middle zone showed no significant change from the start to the end of the study (−1.7 µg m −3 , Fig. 7e).
As the springtime PM 10 level is strongly affected by dust events, the data for that season are considered separately to assess how anthropogenic emissions affect PM 10 .Our analysis of the data for summer, autumn and winter combined  7f).Comparing Fig. 7e with 7f, one can see that for the northern and middle zones, interannual variations in the annual median PM 10 levels differed from those in the SAW PM 10 levels, and this was probably due to influence of springtime dust on PM as discussed above.But for the southern zone, the SAW PM 10 level varied in a manner consistent with that in the annual PM 10 level (Fig. 7e and f), suggesting little influence from spring dust.
Taken together, in contrast to the small increase of urban PM 10 concentration in the southern zone (+ 12%), the PM 10 levels showed a decreasing trend for the northern cities (-19%) and a negligible decrease for the middle zone (-1.7%).This may be a result of the country's air pollution control efforts in the urban areas.These efforts include reducing stationary emissions (eliminating some small coal fired power plants, and managing heavy pollution industries), substituting natural gas for coal, controlling mobile source emission from vehicles, improving road conditions, and increasing urban vegetative cover (Duan et al., 2006;Chan and Yao, 2008;Fang et al., 2009).

Evaluation of urban PM 10 variation trend
Several studies have recently documented a decreasing trend of urban PM 10 level and a general improvement of urban air Atmos.Chem. Phys., 10, 5641-5662, 2010 www.atmos-chem-phys.net/10/5641/2010/quality in northern and middle China.For example, Chan and Yao (2008) indicated that the annual median PM 10 concentration in Beijing decreased from 180 µg m −3 in 1999 to 142 µg m −3 in 2005.A study in Lanzhou showed that the concentrations of magnetic minerals in the wintertime urban dustfall samples (collected during 1997-2005) have decreased, indicating an improvement of the winter air quality (Xia et al., 2008).Zhang et al. (2006) also reported that the air quality in Shenyang improved year-by-year from 1995 to 2004.In addition, PM 10 concentration in Nanchong decreased dramatically from 146 µg m −3 in January 2002-May 2003 (Wen et al., 2004) (Zhang et al., 2009).Note emission factor is a representative value that attempts to relate the quantity of a pollutant released to the atmosphere with an activity associated with the release of that pollutant, which is usually expressed as the weight of pollutant divided by a unit weight, volume, distance, or duration of the activity emitting the pollutant (e.g., kilograms of particulate emitted per megagram of coal burned, http://www.epa.gov/ttnchie1/ap42/).However, according to Streets et al. (2003) and Zhang et al. (2009), except for a decreasing trend in OC (organic carbon) (3.4 Tg→3.2 Tg), China's anthropogenic emissions, including SO 2 (20.4 Tg→31.0Tg), CO (116 Tg→166.9Tg) and NMVOC (nonmethane volatile organic compounds, 17.4 Tg→23.2Tg) increased by about 50% from 2000 to 2006, while those for NO χ (11.4 Tg→20.8Tg) and BC (1.05 Tg→1.8 Tg) increased by about 100%.Zhang et al. (2009Zhang et al. ( ) estimated 2006Zhang et al. ( and 2001 emissions for China using a same methodology, and they found that all species show an increasing trend during 2001 to 2006: 36% increase for SO 2 , 55% for NO χ , 18% for CO, 29% for VOC, 13% for PM 10 , and 14% for PM 2.5 , BC, and OC. How can the decreasing trend in PM 10 level from 2000 to 2006 be reconciled with the increases in fossil fuel usage (such as coal and petroleum) and anthropogenic emissions for the country?The most likely explanation is that increasingly stringent emission control standards and air quality management strategies for major cities have led to localized improvements of air quality.
Another possibility is that the sources for the pollution emissions have become more widely dispersed.Along these lines are the effects of growing urbanization.According to Chan and Yao (2008), from 1980 to 2005, the urban population in China increased from 19.6 to 40.5%; the number of cities increased to over 660, and more than 170 cities had over 1 million permanent residents in 2004 (National Bureau of Statistics China, 2005a, 2006a).Meanwhile, mega cities (conventionally defined as cities with populations over 10 000 000) emerged in the 1990s in China and city clusters developed in the proximity of the mega cities (National Bureau of Statistics China, 2005b, 2006b).The economic expansion, especially in rural and under-developed areas of China, has affected the geographical distributions of pollution emissions.This can be seen from the changes in the sources for major air pollutants, including SO 2 , NO x , NMVOC and BC from 2000 to 2006 (see supplementary material 6; Streets et al., 2003;Zhang et al., 2009).
More generally, aerosol loadings vary due to interactions among many processes including emissions (anthropogenic emission and natural dust production), transport (as well as convection influenced dispersion and dilution), photochemical transformation (new particle speciation and production of secondary aerosols), and deposition (dry and wet), with meteorology playing an overarching role.The causes for the variations in aerosol loadings in China are undoubtedly complicated, and results suggesting improvements in air quality in major cities in northern China, while encouraging, should be assessed cautiously; further study is obviously necessary on this issue.

Comparison of PM 10 with AOD/AOT
As described in Sect.3.4.1,areas found to have relatively high PM 10 concentrations such as central-eastern China, NCP, SCB, YRD and PRD (Fig. 5a) also were high AOD/AOT areas as identified by the radiation/optical measurements and satellite observations (Luo et al., 2000(Luo et al., , 2001;;Li et al., 2003;Wang et al., 2008) at 440 nm with surface PM 10 concentration (supplementary material 7).Our comparisons are limited to these two cities because of the limited overlap of the API PM 10 and AERONET AOT dataset.The AOTs at 440 nm generally showed good correlations with the surface PM 10 concentrations at Beijing and Guangzhou.Another important feature of the Beijing data is that the correlation between the AOT at 440 nm and the PM 10 concentration in summer (r = 0.63) is higher than in the other seasons (r = 0.34, 0.58 and 0.50 in spring, autumn and winter, respectively).Xia et al. (2006) presented a detailed comparison between AERONET AOT at 440 nm and PM 10 concentration in Beijing that covered a period of 33 months.Their study also found a higher correlation coefficient in summer (r = 0.77) compared with the other seasons (r = 0.37, 0.70, and 0.61 in spring, autumn and winter, respectively).They argued that the higher correlation between AOT and PM 10 in summer indicated that the aerosols are well mixed, and surface measurements then are a good indicator of columnintegrated values.These authors attributed the differences in seasonal and diurnal variations between AOT and surface PM 10 to the variation of atmospheric mixing layer height.Song et al. (2009) compared three years of MODIS AOD (at 550 nm) with spatial and seasonal variations of PM 10 concentration for forty-seven sites; they found variability in the relationships between AOD and surface PM 10 as well as regional differences in their seasonality over China.The correlation coefficients they calculated between PM 10 and AOD were +0.6 or higher in the southeastern coast but −0.6 or lower in the north-central region; and this regional discrepancy was largely attributed to the difference in the size distributions of aerosols.We discuss in more detail several factors that could cause differences between the surface PM 10 concentration and the AOD/AOT in supplementary material 8.
In summary, the results above indicate that AOD/AOT and the surface PM 10 measurements provide related but different information on atmospheric particulates.A myriad of factors cause both PM and AOD vary in complex ways.Conditions in the atmospheric boundary layer (convection and mixing layer height), aerosol size distributions and chemical composition, source regions and transport patterns, and wet scavenging all could result in differences between AOD/AOT and surface aerosol concentration.Further studies to assess the causes of long-term changes in surface PM and AOD as well as the relationships between them appear warranted.

Trends in PM 10 and AOD
We note that the decreasing trend of PM 10 level in northern China presented here appears to be different from the trend in visibility over a much longer period (1973( -2007( , Wang et al., 2009)).For the same period as this study (2000 to 2006), Wang et al. found an increase in the AOD in what the authors referred to as the "Asia (south)" region (which includes China, see Fig. 1 in Wang et al., 2009).This is in contrast to the decreasing trend we found for urban PM 10 in northern China.
What could be the cause of this inconsistency?In fact, the "Asia (south)" region in the Wang et al. study includes not only China but also several other countries in southern Asia (Fig. S7 in the supporting online material of Wang et al., 2009, http://www.sciencemag.org/cgi/content/full/323/5920/1468/DC1).It is well known that some of these countries, such as India (Venkataraman et al., 2002), Pakistan (Smith et al., 1996), Bangladesh (Salam et al., 2003) and Vietnam (Deng et al., 2008b), experience severe air pollution.Indeed, the AODs (0.55 µm) derived from visibility measurements at meteorological stations during 2000-2007 were much higher over the southern Asia countries (especially India and Bangladesh) than over China (Fig. S1 in the supporting online material of Wang et al., 2009, http://www.sciencemag.org/cgi/content/full/323/5920/1468/DC1).If the more heavily polluted areas of south Asia were to be excluded from the "Asia (south)" region in this analysis, the AOD tendency might well be different.Moreover, emissions (including intensive biomass burning emission) originating in southern Asia also are sources for air pollutants that can be transported across national borders into southwestern China and PRD (Deng et al., 2008b).
On the other hand, as mentioned in Sect.3.6.2, the PM 10 trends presented here reflect changes in aerosol loadings in major cities as they were based on APIs of eighty-six major Chinese cities, while the visibility data used in the Wang et al. (2009) study were available for about six hundred sites in mainland China, which were more representative of the variation in air pollution at regional scales.Therefore, the results of these two studies did not necessarily match each other.
More to the point, as discussed above, numerous factors can lead to differences between surface PM 10 and AOD/AOT.Along these lines, Li et al. (2005) reported that the seasonality of AOD over Urumchi did not precisely follow the monthly variations in common pollutants such as PM 10 , SO 2 and NO 2 .The difference in trends between AOD/AOT and surface PM 10 could be explained by changes in the emissions of anthropogenic fine particles that have high extinction efficiencies but contribute little to the PM 10 mass.For example, Zhang et al. (2004) found that the PM 2.5 /PM 10 ratio ranged between 0.5-0.7 with an average 0.6 for six cities (Guangzhou, Wuhan, Lanzhou, Chongqing, Qingdao, Beijing) and the PRD in China.High PM 2.5 /PM 10 ratios (e.g., larger than 0.6) are generally ascribed to secondary particulate formation; low ratios can be caused by fugitive dusts or dust delivered by long-distance transport (Chan and Yao, 2008).
Various studies also have shown conflicting patterns in AOD.For example, estimates of AOD based on emissions led Streets et al. (2006) to conclude that AOD over China decreased from 1995-1996 (∼0.305) to 2000 (0.29); and this was consistent with variation of surface shortwave irradiance measurements at fifty-two weather stations and with the rise trend in mean surface temperatures in China starting in the mid-1990s.However, Xia et al. (2007) reported that AOD at 550 nm over Beijing increased from about 0.28 in 1980 to about 0.68 in 2005.Using monthly AOD (500 nm) data from the Total Ozone Mapping Spectrometer (TOMS), Xie and Xia (2008) documented an increasing tendency of AOD in north China from 1997 to 2001, especially for the spring AOD in northeast China.These different results suggested that further investigation on AOD variation trend over China as well as its relationship with surface pollutants such as PM 10 is needed.

Conclusions
A comprehensive analysis of aerosol impacts on climate requires detailed information on chemical composition and optical properties of the aerosol as a function of particle size along with a full set of meteorological measurements.Unfortunately, data of this nature are only available from a few scattered sites and then for limited times over China.Further analysis of the spatial and temporal distribution of the aerosol levels and compositions is obviously necessary to assess its potential effects on climate as well as other effects such as adverse health and visibility impairment.Moreover, information on how the chemical components of the aerosol vary between different locations (e.g.coast and inland) also will be useful.
Most of EPA-China's air quality monitoring stations are located in densely populated urban areas in middle-eastern China, and this makes them suitable for addressing concerns over the health effects of air pollution.However, to study the broader impacts of air pollution, such as climate effects, as well as to evaluate transboundary transport, more monitoring needs to be conducted at rural or remote sites, especially in western China and along the country's borders.In addition, studies of specific types of air pollutants with the potential to influence climate, such as black carbon, sulfate and dust aerosol should be included in the monitoring programs.
Another broad objective for studies involving aerosol particles and their impacts is to link PM loadings to sources.Our studies have shown that spatial distribution of the urban PM 10 follows the anthropogenic emissions of BC and PM 10 over much of China but not along the Chinese coast; this suggests that pollutants are diluted by the influx of clean marine air and removed by scavenging of the particles due to more abundant precipitation there.
With further reference to air quality, our data indicate that the urban PM 10 level in the southern zone may have increased (+12%) from 2000 to 2006, suggesting that additional attention should be paid to the control of atmospheric particulate pollution in southern China.As the secondary aerosol is a major contributor to PM in the southern cities (Ye et al., 2003), controlling emissions of precursor gases may be helpful in improving the air quality in that region.
Decreasing trends in PM 10 level from 2000 to 2006 suggest improvements in the air quality of major cities in northern China.However, a comparison of the anthropogenic emissions over China between year 2000 and year 2006 shows that the sources for the pollution emissions have become more widely dispersed (see supplementary material 6; Streets et al., 2003;Zhang et al., 2009).The geographical distributions of pollution emissions have been affected by the economic expansion into rural and under-developed areas of China.In addition to stringent controls on large emission sources in urban areas, adequate attention also should be paid to the newer sources in rural and under-developed areas.
The frequencies of Asian dust events mirrored the API maxima for the northern cities, and this indicates that dust can contribute to the degradation of springtime air quality in northern China.Along the same lines, Li et al. (2007) observed important influences of dust events on spring air quality in Shijiazhuang city from 2002 to 2006.In that study, all of the springtime days with mid-level or more serious pollution were concurrent with or followed days of dust-events.Wang et al. (2004b) also reported a significant contribution of Asian dust to PM 10 pollution at Beijing during the springtime.This is significant because desert dust production will be difficult if not impossible to control, and some allowances for naturally-occurring PM may be appropriate for air quality standards.
Although the PM 10 dataset as used in this study is not sufficient for an in-depth analysis of aerosol impacts on climate, the results do provide insights into the spatial distributions and temporal variability of the atmospheric aerosol, especially those areas most strongly affected by particulate pollution.Undoubtedly, comparison of infromation on PM obtained using different methods can improve our understanding of the spatio-temporal patterns of the aerosol as well as the potential for climate and other effects.In addition, results of this study will contribute to a better understanding of air pollution in the major urban areas of China, information of this type should be useful for the development and implementation of effective air-pollution control strategies.

Fig. 4 .
Fig. 4. Variations of the daily air pollution indexes (APIs) for the (a) northern, (b) middle, and (c) southern latitudinal zones.Light gray dots linked with light gray straight solid lines denote the daily APIs for the individual cities, while cyan dots linked with black straight solid lines donate the daily geometric mean APIs for the three latitudinal zones.The numbers of occurrence days with high APIs for the cities in the three latitudinal zones are also presented at the top of (a), (b) and (c).

Figure
Figure 5a illustrates the spatial distribution of PM 10 for the eighty-six individual cities. High PM 10 (presented as the arithmetic mean in µg m −3 ) regions in northern China (127 to 192) include Urumchi (156) on the southern margin of the Junggar Basin; Lanzhou (192) and Xining (138) on the northeastern side of the Qinghai-Tibetan Plateau; Weinan (152) and Xi'an (144) in the Guanzhong Basin; Taiyuan (179), Datong (162), Yangquan (153) and Changzhi (132) in Shanxi Province in the basins to the west of the Taihang Mountains; Pingdingshan (156, a city with heavy coal industries) and Kaifeng (133) in Henan Province on the western margin of NCP; Beijing (159), Tianjin (140) and Shijiazhuang (168) in the northern part of NCP; Jinan (135) in the eastern part of NCP; and Shenyang (145), Anshan (132) and Fushun (127) in the southeastern part of the Northeast China Plain.In the middle zone, the high PM 10 (arithmetic mean in µg m −3 ) areas (119 to 147 µg m −3 ) were Chongqing (142), Chengdu (120) and Luzhou (132) in the Sichuan Basin (SCB), Changsha (147) and Wuhan (135) in central China, Nanjing (125) and Hangzhou (119) in YRD.The PM 10 levels in the cities in the southern zone were generally lower than those in the northern zone and the middle zone, and there were no areas with arithmetic mean PM 10 concentration >109 µg m −3 (Fig.5a).However, the arithmetic mean PM 10 concentrations (in µg m −3 ) exceeded 80 in four cities: Qujing (91), Guiyang (80.1),Guangzhou (85) and Shaoguan (87), and these were much higher than the total average PM 10 level of 58 µg m −3 for the southern zone.

Table 1 .
Locations, groups and air pollution index (API) data available period of the eighty-six cities in China.
• N, 91.03 • E A a Here API data available period was denoted by A: from June 2000 to February 2007, B: from June 2001 to February 2007, C: from June 2004 to February 2007, and D: from January 2006 to February 2007.

Table 2 .
Air quality classifications corresponding to the air pollution indexes (APIs), air quality management recommendations, and pollutant concentrations in China.Air pollution index (API) Air quality classification Air quality management recommendation Corresponding daily average pollutant concentration, µg m −3

Table 3 .
Proportions of days with PM 10 , SO 2 and NO 2 as the principal pollutant in the full set of air pollution index (API) records for the fourteen groups and three latitudinal zones.Group/Zone (City) Proportions of days as principal pollutant a

Table 4 .
Seasonal arithmetic mean PM 10 concentrations for the fourteen groups and three latitudinal zones.
a Concentrations in micrograms per cubic meter calculated from daily geometric mean PM 10 for the group/zone.b Here n stands for the number of daily geometric mean PM 10 concentrations for the group/zone.
(Yao et al., 2009) January 2006-February 2007.Shi et al. (2008)concluded that the air quality has improved in Shanghai, Nanjing, Hangzhou and Hefei from 2002 to 2005.Moreover, an analysis of data from more than 300 major cities in China indicated that the percentage of cities with PM 10 concentrations less than 100 µg m −3 increased from 36.8% in 2002 to 62.8% in 2006, and the cities with PM 10 concentrations greater than 150 µg m −3 decreased from 29.8% to 5.3%(Yao et al., 2009).It is worth noting that the variation trends in PM 10 concentration presented here for the three latitude zones in China reflect changes in aerosol loadings in major cities, but they are not necessarily representative of the variations in aerosol concentrations over less populated regions.We note that coal product usage in China increased from