Total Ozone Mapping Spectrometer (TOMS) observations of increases in Asian aerosol in winter from 1979 to 2000



[1] Emission inventories indicate that the largest increases in SO2 emissions have occurred in Asia during the last 20 years. By inference, the largest increases in aerosol, produced primarily by the conversion of SO2 to sulfate, should have occurred in Asia during the same time period. Decadal changes in regional aerosol optical depths are calculated by analyzing Total Ozone Mapping Spectrometer (TOMS) vertical aerosol optical depths (converted to 550 nm) from 1979 to 2000 on a 1° by 1° global grid. The anthropogenic component of the TOMS aerosol record is maximized by examining the seasonal cycles of desert dust and boreal fire smoke and identifying the months of the year for which the desert dust and boreal fire smoke are least conspicuous. Gobi and Taklimakan desert dust in Asia is prevalent in the TOMS record during spring, and eastern Siberian smoke from boreal forest fires is prevalent during summer. Aerosol trends are calculated on a regional basis during winter (November–February) to maximize the anthropogenic component of the aerosol record. Large increases in aerosol optical depths between 1979 and 2000 are present over the China coastal plain and the Ganges River basin in India. Aerosol increased by 17% per decade during winter over the China coastal plain, while SO2 emissions over the same geographical region increased by 35% per decade.

1. Introduction

[2] Tropospheric aerosol influences the Earth's climate in several ways [Intergovernmental Panel on Climate Change (IPCC), 2001]. Particles directly cool the atmosphere by scattering a fraction of the incoming solar radiation back to space. Particles such as desert dust and carbonaceous soot absorb shortwave and longwave radiation and contribute to atmospheric heating. Particles also have indirect effects since they can serve as condensation nuclei. For a given amount of condensable water vapor, aerosols can increase the number of smaller cloud droplets, and thereby increase cloud reflectivity and contribute to atmospheric cooling [Twomey et al., 1984]. Smaller droplets fall slower than larger droplets, and thus rainfall rates can be influenced by particles. Decadal changes in regional aerosol concentrations therefore are of interest, due to the effects of aerosol upon the earth's radiative and hydrological processes.

[3] Tropospheric aerosol particles, of several compositional types, are produced by a number of natural and human causes. Desert dust is injected into the atmosphere if surface winds are strong enough to loft dust into the air. Desert dust is observed in northern latitudes over Africa, the Middle East, western and central Asia [Prospero et al., 2002]. Soil dust is also generated in settled areas due to ongoing human activities, including agriculture, transportation, construction, and industry. Smoke from forest fires, due to natural (i.e., lightning) and human causes, is observed over central Africa, the Amazon, Southeast Asia, and eastern Siberia. Sulfate is observed over many regions of the Earth and is derived from the conversion of dimethyl sulfide (DMS) from the ocean and by the chemical transformation of SO2 emissions from natural (e.g., volcanic activity) and human (i.e., generation of electricity by fossil fuel combustion, residential heating, transportation, and other industrial) activities.

[4] Our discussion of satellite observations of decadal changes in aerosol naturally gravitates toward Asia, since double digit increases in population, and emissions that produce aerosols (such as that due to the release of SO2) have taken place in Asia. A third of the world's population of 6.3 billion people ( currently resides in India and China. The population of India increased by 23% between 1991 and 2001 to a present value near 1.03 billion ( The population of China increased 26% per decade from 0.55 billion to 1.2 billion between 1950 and 1995, and is presently estimated to be near 1.3 billion (

[5] A number of studies have quantified decadal changes in regional emissions of SO2. Smith et al. [2001] tabulate regional SO2 emissions from 1980 to 2000 in 5 year time intervals for 12 geographical regions. Total global SO2 emissions [see Smith et al., 2001, Table 3] vary between 67 and 72 Tg S per year, peaking in 1990. Emissions decrease in eastern Europe, the United States (from 12 in 1980 to 8.1 Tg S per year in 2000), and western Europe, while emissions increase over China from 7.8 in 1980 to 18 Tg S per year in 2000. Streets et al. [2000] estimate that Asian SO2 emissions grew by 16% between 1990 and 1997, from 16.8 to 19.6 Tg S per year, respectively. Smith et al. [2003] tabulates annual global SO2 emissions from 1850 to the present on a 1° by 1° longitude-latitude grid. These studies all point out that largest increases in SO2 emissions have occurred in Asia. By inference, largest increases in sulfate should also be associated with Asia.

[6] Though sulfate is an important aerosol type, it is not the only anthropogenic aerosol of interest. Black carbon [Streets et al., 2001a] and organic carbon (produced by gaseous organic precursors [Jacobson et al., 2000]) are also of importance. Total Asian emissions for the year 2000 are estimated to be 17.1 Tg S, 10.4 Tg organic carbon, and 2.54 Tg black carbon [Streets et al., 2003]. Global tracer-transport aerosol modeling predicts annual average aerosol loadings of sulfate, organic carbon, and black carbon with a similar relative ordering: 30, 20, and 2 mg m−2 over China [Qian et al., 2003].

[7] The diversity of aerosol types is also readily apparent in chemical analysis of Asian aerosol. Beijing PM2.5 (aerosol with diameters less than 2.5 μm) particles in 1999–2000 were comprised of 30% organic material, 7% elemental carbon, 12% crustal (e.g., desert dust), 15% sulfate, and 10% nitrate [He et al., 2001]. XiAn, located south of the Loess Plateau in China, has aerosol whose elemental composition is a combination of crustal elements (e.g., Si and Ca) and sulfur [Zhang et al., 2002]. Once transported over long distances, Asian PM2.5 aerosol at the Crater Lake, Oregon and Mount Lassen IMPROVE (Interagency Monitoring of Protected visual Environments) sites from 1989–1999 had fractions ∼30% mineral, 28% organic compounds, 10% sulfate, and 4% elemental carbon [VanCuren, 2003]. VanCuren [2003] noted a strong statistical association between soil dust and combustion products, and concludes that soil elements are not exclusively due to desert dust events.

[8] Biomass is a very important energy source in developing countries [Hall, 1991]. Biomass (e.g., fuel wood, crop residues, and dung) is burned in a variety of cooking stoves in rural regions, and accounts for 45% of the total primary energy consumption in India [Gadi et al., 2003]. Of the 2898 Tg of global biomass that is burned annually, it is estimated that 481 and 584 Tg are consumed in India and China, respectively [Yevich and Logan, 2003].

[9] With the development of satellite technology that can observe changes in atmospheric aerosol on a global basis, it is of interest to examine long-term satellite records for aerosol trends. The series of Total Ozone Mapping Spectrometer (TOMS) experiments, in particular, allows one to study changes in column ozone and aerosol optical depths from 1979 to the present, with a data gap from 1994 to mid 1996. The TOMS aerosol record is unique in its capability to detect all aerosol types over land. Other experiments, such as the advanced very high resolution radiometer (AVHRR) and the Moderate Resolution Imaging Spectroradiometer (MODIS), have difficulty retrieving desert dust optical depths over land due to the high reflectivity of land surfaces. The dark surface albedo of land surfaces free of ice and snow at the TOMS near ultraviolet wavelengths allows for the detection of all aerosol types over land.

[10] In the following sections of this paper we examine TOMS aerosol optical depth data on a 1° by 1° global grid from 1979 to 2000 [Torres et al., 2002] to determine regional patterns of aerosol and their temporal trends, and relate these trends to regional SO2 emissions over the same time period. Comparisons of TOMS aerosol optical depths and the Smith et al. [2003] inventory will focus upon the winter months of November–February, since (as discussed below) Asian deserts and boreal fires produce copious amounts of Asian aerosol during spring and summer, respectively. The TOMS aerosol data record is discussed in section 2. This section displays global maps of annually averaged and winter optical depths, and also presents time series of monthly averages of the optical depths for several selected desert, boreal forest, and populated regions in Asia. SO2 emission inventories are discussed and presented in contour maps in section 3. Linear trends of the TOMS and SO2 data between 1979 and 2000 for selected regions are calculated in section 4 for winter (November–February) in order to maximize the anthropogenic component, and to lessen the desert and boreal forest fire components, of the aerosol record. These linear trends are placed in context to other factors, such as decadal changes in Asian atmospheric variables, in section 5. The results of the paper are summarized in section 6.

2. TOMS Aerosol Optical Depth Data

[11] Torres et al. [2002] present TOMS aerosol optical depths from 1979 through 2000, tabulated on a 1° by 1° global grid between 89.5°S and 89.5°N. Each data point is a monthly averaged value. Measurements of the backscattered radiance (Iλ) at two wavelengths λ1 and λ2 are used in the aerosol retrieval, which is based upon the theoretical framework presented by Torres et al. [1998]. Particles are characterized by comparing measured and precomputed tabulated values of spectral contrast (Iλ1/Iλ2) and the TOMS radiance at the longer wavelength (Iλ2). Wavelength pairs at 340 and 380 nm are used for the Nimbus 7 and Meteor3 data, and pairs at 331 and 360 nm for the Earth Probe and Advanced Earth Observing Satellite (Adeos) TOMS data. The Nimbus 7 TOMS instrument operated from October 1978 to April 1993. Earth Probe TOMS data dates from mid-1996 to the present.

[12] Sulfate, dust, and carbonaceous aerosols are the three main aerosol types in the TOMS aerosol retrieval look up table. The algorithm specifies a land type associated with an observation based on the CERES/SARB surface map (available and presented at Dust and sulfate particles are assumed over arid or semiarid environments, while carbonaceous and sulfate particles are assumed over other land types. When the TOMS aerosol index [Herman et al., 1997] does not indicate the presence of absorbing aerosol (i.e., desert dust or smoke), then sulfate is the assumed particle type. The retrieval generates the total vertical optical depth of the aerosol at 380 nm, which is then converted to the optical depth at 550 nm by applying a conversion factor based upon the theoretical wavelength-dependent variations of the different aerosol types. The aerosol optical depth detection limit is 0.1 at 380 nm over both oceans and continents, which becomes a detection limit of 0.07 at 550 nm.

[13] TOMS optical depths of UV-absorbing aerosols are within 30% of ground-based values determined by the Aerosol Robotic Network (AERONET), while nonabsorbing optical depths agree to within 20% [Torres et al., 2002]. Table 3 of Torres et al. [2002] indicates that AERONET and TOMS optical depths have a correlation coefficient of 0.89, based upon 496 comparisons, and the bias of the TOMS data is 0.058 (i.e., TOMS optical depths are larger than AERONET values). For error analysis purposes, we use a random error of 30% and a systematic bias of 0.06 for an individual 1° × 1° monthly averaged TOMS optical depth.

[14] Before we focus on Asian aerosol, it is instructive to note that the magnitudes of the aerosol optical depths that are observed by TOMS over Asia (presented in following text) are similar in magnitude to those observed elsewhere. Torres et al. [2002] display time series for seven regions outside of Asia: Saharan Desert, eastern Atlantic Ocean, western Atlantic Ocean, the Sahel, southern Africa, South America, and the eastern United States. TOMS aerosol optical depths usually vary between 0.2 and 2, and time series display considerable temporal variations. Saharan desert dust peaks in April–July and is minimal in winter, with monthly averages of aerosol optical depth near 0.8 and 0.25, respectively. Southern Brazil and southern Africa biomass fires produce maxima in August–September. Southern Brazilian fires spike aerosol optical depths to values between 0.5–1.2, while baseline values are near 0.3.

[15] Figures 1 and 2display annually averaged TOMS aerosol optical depths for 1979–1981 and 1997–2000, and the ratios of these averages for the two time periods. These years are before the El Chichon eruption of 1982 and after the Mt. Pinatubo eruption of June 1991, and were selected to avoid volcanic influences. Large optical depths in Figures 1 and 2 are apparent over the Sahel in Africa, the Amazon, Siberia and Canada and also over Southeast Asia and Indonesia. Biomass fires are likely responsible for many of the large optical depths observed over these regions. Ratios of the optical depths, presented in the bottom panels of Figures 1 and 2 are largest over the Amazon, eastern Siberia, off-shore of Mexico, over Indonesia, and the China coastal plain. Beijing and a large portion of the population of China are located in this coastal plain.

Figure 1.

TOMS aerosol total optical depths at 550 nm for 1979–1981 and 1997–2000 for all seasons and the ratios of the optical depths for the two time periods.

Figure 2.

Same as Figure 1, except that the geographical focus is upon Asia.

[16] Figure 3 presents a time series of TOMS aerosol optical depths averaged over the globe (dashed line) and over the south Pacific (solid line). These times series are displayed to indicate the volcanic influence in the TOMS data, and to quantify baseline (lowest) optical depths. Optical depths over the South Pacific have minima near 0.2, while global averaged minima are near 0.4. The absence of data from May 1993 through mid-1996 is due to a data gap in the TOMS record. In the graphs presented below for other geographical regions, optical depths vary between 0.3 and values greater than unity, with the larger optical depths being due to desert dust and boreal fire episodes.

Figure 3.

Monthly averages of TOMS aerosol total optical depths for the south Pacific (20°–30°S, 100°–120°W) and the globe (70°S–70°N, all longitudes). There is a data gap in the TOMS record from 1993 to 1996. EC and MP mark the eruptions of El Chichon and Mt. Pinatubo in April 1982 and June 1991, respectively.

[17] The Mt. Pinatubo influence in 1991 is clearly present in the south Pacific time series in Figure 3 since optical depths increase from a low background value near 0.25 to 0.6 after the eruption. The Pinatubo volcanic influence is also present in the global time series. Trend analysis of the TOMS aerosol data (discussed below) excludes years (1991–1992) that are influenced by the Mt. Pinatubo volcanic cloud.

[18] Desert regions upwind of the China coastal plain are another important influence upon the TOMS time series for eastern China. Figure 4 presents monthly averaged optical depths from 1979–2000 for the China coastal plain and the Gobi desert, which is 1200 km west of Beijing. Both time series exhibit yearly variations in optical depth on the order of 0.5. Monthly averages for the two regions are displayed in Figure 5, while monthly averages of Taklimakan desert optical depths are presented in Figure 6. Though the Taklimakan desert is 3300 km westward of Beijing, this desert is a prolific source region of dust during the spring. Note that the Taklimakan optical depths are twice as large as those over the Gobi desert. Both desert regions have maximum optical depths in spring, and minimum values during winter.

Figure 4.

Monthly averages of TOMS aerosol total optical depths for 1979–2000 for the China coastal plain (35°–45°N, 115°–125°E) and the Gobi desert (40°–50°N, 100°–110°E).

Figure 5.

Monthly averages of TOMS aerosol total optical depths for 1979–2000 for the China coastal plain (35°–45°N, 115°–125°E) and the Gobi desert (40°–50°N, 100°–110°E).

Figure 6.

Monthly averages of TOMS aerosol total optical depths for 1979–2000 for the Taklimakan desert (35°–45°N, 80°–90°E).

[19] Xuan and Sokolik [2002] note that the Taklimakan and central Gobi deserts are the two main dust sources in northern China, with annual PM10 (particulate matter with diameters less than 10 μm) emission rates of 0.38 ton ha−1 yr−1 and 0.24 ton ha−1 yr−1, respectively. It is well known that the Gobi desert can produce copious amounts of desert dust, and that this dust can travel over large distances. Gobi desert dust in April 1998 was carried across the Pacific and increased aerosol loading on the west coast of the United States from a normal value between 10–25 μg m−3 to 65 μg m−3 [Husar et al., 2001]. An analysis of IMPROVE sampling network data by VanCuren and Cahil [2002] demonstrated that fine Asian dust is transported across the North Pacific nearly continuously, except in winter.

[20] The springtime maximum in desert dust, displayed in Figures 4 and 5, is a well established temporal characteristic [Littmann, 1991; Parungo et al., 1994; Qian et al., 2002]. As discussed by Natsagdorj et al. [2003], 61% of dust storms in Mongolia occur in the spring, 22% of the time in October and November, and 10% in the winter.

[21] Siberia is another source region of aerosol that potentially can enhance the aerosol load over China, since boreal fires are numerous, intense, and persist over several months time. There are approximately 30,000 boreal forest fires in Siberia each year [Kasischke et al., 1999]. Smoke from particularly intense fires can dominant yearly averages, e.g., smoke from the boreal fires in 1998 is prominent in the middle panel of Figure 2 in eastern Siberia (50°–70°N, 120°–140°E). The monthly averages presented in Figure 7 for this region indicate that these fires persisted during summer over four months, and that the optical depths during July–October were greater than 2.

Figure 7.

Monthly averages of TOMS aerosol total optical depths for 1998 over eastern Siberia (50°–70°N, 120°–140°E). The fires in Siberia in 1998 were some of the most intense during the last 20 years.

[22] It is well established that desert dust in spring is transported over eastern China. Figure 4 of Chung and Yoon [1996] displays trajectories of Gobi and Taklamakan desert dust for April 1993 that passes over eastern China. The chief layers of dust usually traveled between 1 and 3 km, moved by a steering flow at 850–500 hPa. Two layers of Gobi desert dust, near the surface and between 1.5 and 3.5 km, decreased visibility over eastern China during April 1988 [Wang et al., 2000]. Figures 4 and 5 of Xiao et al. [1997] present contours of dust concentrations of Gobi desert dust that pass over eastern China at 1 and 5 km during a two week period in March 1994.

[23] Trajectories for 1994–2000 were calculated to illustrate that desert dust can be transported into the China coastal plain region also during winter. Several thousand five day trajectories on the 290 K potential temperature surface, corresponding to trajectory pressures near 780 hPa, were calculated from the Gobi and Taklimakan deserts to estimate the fraction of the time that desert dust can encounter the China coastal plain during November–February. European Center for Medium range Weather Forecasts (ECMWF) winds for 1994–2000 were used in the calculations. The China coastal plain was encountered 83 and 9% of the time for trajectories starting in the Gobi and Taklimakan, respectively, during November–February. As noted above, however, dust storms in Mongolia only occur 10% of the time in winter.

[24] Insight into the seasonal characteristics of the composition of Asian aerosol is also available from the chemical analysis of lower tropospheric aerosol. Prospero et al. [2003] discuss chemical analyses of aerosols at Midway Island (28 N, 177 W), 5400 km eastward of the China coast, from 1981–2000. Their chemical analyses [see Prospero et al., 2003, Figure 5] isolated non-sea-salt (nss) sulfate, anthropogenic sulfate, natural nss-sulfate, total nitrate, anthropogenic nitrate, and mineral dust components. During April–May most of the aerosol observed at Midway is mineral dust. The ratio of anthropogenic sulfate (μg m−3) to the sum of total nss sulfate plus nitrate plus mineral dust (μg m−3) at Midway, based upon Figure 5 of Prospero et al. [2003], is presented in Figure 8 for each month of the year. Anthropogenic sulfate is most prevalent during November–January and accounts for approximately 24% of the total non sea salt aerosol.

Figure 8.

The ratio of anthropogenic sulfate (μg m−3) to the sum of non-sea-salt sulfate plus nitrate plus mineral dust (μg m−3) for each month at Midway, based upon the data displayed in Figure 5 of Prospero et al. [2003].

[25] This chemical analysis is in concert with other observations. SO2 and NOx daily concentrations in XiAn China during 1996–1997 were largest during winter [Zhang et al., 2002], and consistent with the fact that the home residential ‘heating season’ peaks between November and March. Monthly averaged non-sea-salt and black carbon aerosol concentrations at the Amami (29°N, 128°E) and Miyako (25°N, 125°E) islands in the East China sea during 1991–1994 were relatively high in winter to spring and low in summer [Kaneyasu and Takada, 2004].

[26] On the basis of the TOMS monthly variations of desert dust (Figures 5 and 6) and boreal smoke (Figure 7), and the chemical analyses of aerosol at Midway Island (Figure 8) and at other ground sites, the anthropogenic component of the TOMS data record is maximized if the winter months of the data record are examined. We therefore will examine the time trend of aerosol over the China coastal plain during winter (November–February), and exclude months for which desert dust and boreal fire smoke dominate the time series.

[27] TOMS aerosol optical depths for the months of November–February for 1979–1981 and 1997–2000 are presented in the top and middle panels of Figure 9. The focus of Figure 9 is upon Asia, since cloudy conditions in winter limit the number of observations over the eastern United States, Europe, and Russia [Torres et al., 2002]. Large optical depths in 1997–2000 are located over the coastal plain of China, the Sichuan basin (near 30°N and 110°E), the Ganges river basin of India, and northern Pakistan. Ratios of aerosol optical depths (1997–2000 to 1979–1981) are presented in the bottom panel of Figure 9. Largest ratios are located over the China coastal plain, northern India and Pakistan, and the Taklimakan and Gobi deserts.

Figure 9.

TOMS aerosol total optical depths at 550 nm for 1979–1981 and 1997–2000 for winter months (November–January) and the ratios of the optical depths for the two time periods. C, G, T, and I in the top panel indicate the China coastal plain, Gobi, Taklimakan, and India regions used in our analysis.

[28] Large ratios in Figure 9 are also apparent over the Bay of Bengal and the Arabian Sea, and are interpreted as due to the outflow of aerosol from India and Bangladesh. As reviewed by Ramanathan et al. [2001], anthropogenic haze spreads over most of the North Indian Ocean each year from December to April. Aerosol offshore of India is visible in several satellite data sets. Leon et al. [2001] and Chu et al. [2003] present Meteosat-5 and MODIS views of aerosol offshore of India. Model data assimilations and a discussion of the direct radiative forcing by aerosols offshore of India during the Spring 1999 Indian Ocean Experiment (INDOEX) is presented by Collins et al. [2002].

[29] The desert regions in Figure 9 have winter aerosol optical depths less than 0.4, while the optical depths over the China coastal plain are above 0.5. Large ratios in the bottom panel of Figure 9 coincide with regions of Asia (i.e., the China coastal plain and the geographical arc from northern Pakistan through the Ganges river basin and Bangladesh) in which the population density is greater than 250 people per km2 over an extended geographical region (e.g., see the world population density map available at

3. SO2 Surface Emissions

[30] The Smith et al. [2003] emission inventory (contact: is tabulated for all four seasons in one year increments on a 1° × 1° longitude-latitude grid from 1850 to 2000. Data files are included for low and high cases, i.e., emissions near ground level, and emissions lofted to about 100 m or greater through a combination of tall stacks and/or thermal buoyancy, respectively. Figures 10 and 11 display the sum of the winter low- and high-case SO2 emissions of Smith et al. [2003] for 1979–1981 and 1997–2000. Largest SO2 emissions in 1997–2000 are over the China coastal plain, and selected spots in the eastern United States and Europe. Ratios of emissions for the two time periods are presented in the bottom panels of Figures 10 and 11. The ratio at a given grid point was set to zero if the SO2 emission value was less than 0.25 × 103 mt S. SO2 emissions increased by factors greater than 1.6 from 1997–2000 to 1979–1981 over the China coastal plain, India, Pakistan, Bangladesh, Iran, and Indonesia. Emissions decrease over the eastern United States, western Europe, and Russia, with ratios of emissions near 0.7, 0.3, and 0.4, respectively, for the two time periods.

Figure 10.

Smith et al. [2003] SO2 emissions for 1979–1981 and 1997–2000 and the ratios of the emissions for the two time periods. The sum of surface “low emissions” and “high emissions,” i.e., emissions that are lofted to altitudes above 100 m, are graphed. The ratio at a given grid point was set to zero if the SO2 emission value was less than 0.25 × 103 mt S.

Figure 11.

Same as Figure 10, except that the geographical focus is upon Asia.

[31] Figure 11 SO2 emissions are similar to those presented in Figure 9 of Streets et al. [2003]. China is the dominant region of SO2 emissions in both inventories, followed by India and Pakistan. Total SO2 emissions in the year 2000 for China and India are cited by Streets et al. [2003] to be 20,385 and 5536 Gg. Ratios of Chinese to Indian SO2 emissions for the year 2000, based on the Streets et al. [2003] and Smith et al. [2003] inventories, are 3.7 and 4.8, respectively. The Smith et al. [2003] emissions for China and India in 2000 are 38% and 5% larger than those of Streets et al. [2003] due to two factors: different assumed coal sulfur content, and the assumption that the amount of coal consumed in China is underestimated by official reports.

[32] Our focus is on the winter months, the time of year in which combustion related emissions are maximal. Streets et al. [2003] point out that the five primary combustion related emissions in China (CO, SO2, NOx, black carbon, and nonmethane volatile organic carbon), excluding open biomass burning, are largest in the winter months. On the basis of Streets et al.'s [2003] data, black carbon and SO2 emissions in January typically account for 0.17 and 0.1 of the yearly total, i.e., fractions above a uniform monthly average of 0.83.

[33] It should be noted that SO2 emissions did not increase in a monotonic manner in China during the last 50 years. Streets et al.'s [2000] and Smith et al.'s [2003] data indicate that SO2 emissions from China decreased after 1995. The main reason for the decline is attributed to a reduction in industrial coal use [Carmichael et al., 2002]. Smith et al.'s [2003] emissions decrease by 10% from 1996 to 2000 over the China coastal plain. Prospero et al. [2003] point out that anthropogenic sulfate aerosol observed at Midway Island during 1998–2000 may indicate a decrease in anthropogenic emissions, compared to years prior to 1995 [see Prospero et al., 2003, Figure 8].

4. Linear Trend Analysis of the TOMS and SO2 Data

[34] Regional averages of monthly TOMS aerosol optical depths for individual months of November–February are presented in Figure 12 for the China coastal plain and the Gobi desert. The Gobi and China coastal plain data points are displayed together in Figure 12 since if Gobi desert dust optical depths increased as fast, or faster, than the China coastal plain optical depths, then transport of desert dust to the coastal plain could account for the majority of the observed increase over the coastal plain. Each data point is an average of individual 1° × 1° optical depths for the latitude and longitude ranges indicated in Table 1. A representative error bar is also presented in Figure 12. Since many (e.g., on the order of 50 or greater) individual 1° × 1° optical depths are used to define a regional average, the error bar is essentially the 0.06 optical depth bias cited in section 2 (i.e., the 30% random error of an individual 1° × 1° optical depth is reduced by a factor near 7 or greater when the root mean square of bias and random error is calculated).

Figure 12.

TOMS monthly averages of total aerosol optical depths for the China coastal plain (30°–42°N, 115°–125°E) and the Gobi desert (40°–50°N, 100°–110°E) for November, December, January, and February.

Table 1. Trend Analysis of Aerosol Optical Depth and SO2 Emissions for Several Geographical Regionsa
RegionLatitudeLongitudeAerosol, % decade−1Aerosol Correlation Coefficient rSO2, % decade−1
  • a

    The uncertainties in the aerosol (% decade−1) rates of change are 2σ values.

China coastal plain30°–42°N115°–125°E17.3 ± 6.70.8235
India10°–30°N70°–90°E10.6 ± 4.90.7447
Gobi desert40°–50°N100°–110°E6.6 ± 8.90.39Low SO2
Taklimakan desert35°–45°N80°–90°E−1.2 ±11.3−0.06Low SO2
South Pacific20°–30°S120°–100°W−7.9 ± 8.3−0.44Low SO2
Globe70°S–70°N180°W–180°E−3.1 ± 3.8−0.38−10

[35] A number of data points from Figure 12 are excluded from the calculation of the winter averages presented in Figure 13. Data points in 1991 and 1992 are excluded due to the possible influence by the Mt. Pinatubo eruption (see Figure 3). Other data points, with optical depths greater than 0.7, were excluded in the China coastal plain for November and December of 1997. Contour maps of TOMS optical depths for these months (not shown) display large optical depths over the Beijing region, with few other data points throughout the coastal plain. Less than 20 1° × 1° grid boxes contribute to each of these monthly averages. Though the optical depths over Beijing are reasonable (they are not larger than desert dust storm or boreal optical depths displayed in Figures 5 and 6), they produce a locally biased average for the China coastal plain, and are treated as outliers. Finally, the optical depth near 0.8 in February 1984 is excluded since inspection of the optical depth contour map indicates a strong maximum near Langzhou (35 N, 105 E), a region which is west of the coastal plain and which has a 5–10% annual average dust storm frequency [Littman, 1991]. Desert dust storm activity likely is responsible for the large optical depth in February 1984.

Figure 13.

Winter averages of TOMS total aerosol optical depths, averaged for consecutive months from November through February for the China coastal plain (30°–42°N, 115°–125°E) and the Gobi desert (40°–50°N, 100°–110°E). Linear fits and decadal trends are indicated in the graph. The 17.3 ± 6.7% increase per decade for the China coastal plain has a correlation coefficient (r) of 0.82 and is significant, while the 6.6 ± 8.9% increase per decade for the Gobi desert aerosol data (r = 0.37) is not significant.

[36] Winter averages (November–February) for 1979–2000 are presented in Figure 13. Linear trends of the winter data for the Gobi and China coastal plain data were calculated and are presented in Figure 13. Product moment correlation r coefficients [see Chatfield, 1998, equation (8.5)] were calculated to indicate the degree of correlation of each individual time series. Decadal changes in aerosol optical depths and correlation coefficients are presented in Table 1 for the Gobi desert, China coastal plain, and other geographical regions.

[37] Aerosol increased by 17.3 ± 6.7% per decade over the China coastal plain for 1979–2000. There are 15 elements in the sample set for the China coastal plain, and the ±6.7% per decade uncertainty is a 2σ uncertainty. Since the 95% critical point for the absolute value of the correlation coefficient for a sample set of 15 is 0.50 [see Chatfield, 1998, Table 14], and the correlation coefficient for the seasonal averages over the China coastal plain is high (0.82), the 17.3 ± 6.7% per decade change is significant. The 6.6 ± 8.9% per decade increase over the Gobi desert sample of 16 points has a correlation coefficient of 0.39 and is not statistically significant. We feel that the data in Figure 13 indicate that aerosol optical depths over the China coastal plain increased significantly between 1979 and 2000, and that the 17.3 ± 6.7% increase over the China coastal plain is not likely due to desert dust emissions, although we cannot rule out the possibility that some portion of the increase is due to dust emissions.

[38] A 10.6 ± 4.9% per decade increase in aerosol is calculated for the India region (see Figure 14). The India winter averages have a correlation coefficient of 0.74. There are 17 samples for this fit, the correlation coefficient is greater than the critical point of 0.50, and the decadal increase is statistically significant. Taklimakan desert, south pacific, and global aerosol decadal changes are single digit values (−1.3 ± 11.3, −7.9 ± 8.3, and −3.1 ± 3.8% per decade, respectively), and are not statistically significant (the correlation coefficients are −0.06, −0.44, and −0.39, respectively).

Figure 14.

Winter averages of TOMS total aerosol optical depths, averaged for consecutive months from November through February for the India region (10°–30°N, 70°–90°E). The 10.6 ± 4.9% increase per decade has a correlation coefficient (r) of 0.74 and is significant.

[39] The sensitivity of the linear trend to the large optical depths excluded in the regions of the China coastal plain was assessed by calculating the trend with all of the large optical depths (except those for the volcanic years of 1991 and 1992). The least squares calculation produced a linear trend of 15.7 ± 5.9% per decade, similar to the 17.3 ± 6.7% per decade trend cited above.

[40] The sensitivity of the trends to the choice of the selected geographical regions was also assessed. If the geographical box focuses upon the Beijing region (35°–45°N, 115°–125°E), then the linear trend is 18.6 ± 6.9% per decade. For the geographical arc from northern Pakistan through the Ganges river valley and Bangladesh (20°–32°N, 70°–95°E) the linear trend is 12.3 ± 4.7% per decade. These trends differ little from the trends cited above.

[41] Table 1 also presents decadal trends for the Smith et al. [2003] SO2 (sum of low and high case) emissions for the same geographical regions. Regions for which the surface emissions are below 500 t of S per year are labeled as “Low SO2”. Of the six regions presented in Table 1, largest decadal changes in TOMS aerosol optical depths, largest correlation coefficients, and largest increases in SO2 emissions are associated with the China coastal plain and India.

[42] The increases in the SO2 emissions for China (35% per decade) and India (47% per decade), specified in Table 1, are similar to rates of change reported in other research. Garg et al. [2001] and Streets et al. [2001b] estimate that India's SO2 emissions grew at rates of 5.5% and 4.9% per year during the 1990s. Streets et al. [2001b] estimate that China SO2 emissions grew 3.6% per year from 1985 through 1996. (SO2 rates are calculated by dividing the time derivatives by the time averaged mean SO2 values). Total SO2 emissions in the year 2000 for China and India are cited by Streets et al. [2003] as 20,385 and 5536 Gg.

[43] While the trend analysis in Table 1 focuses upon the years for which TOMS data is available, it should be mentioned that desert dust emissions did not increase monotonically throughout the 20th century. China dust storm frequency of occurrence was appreciably larger during the 1950s and 1960s [Parungo et al., 1994; Qian et al., 2002], in comparison to the TOMS record period starting in 1979. Reforestation efforts and changes in meteorology may be responsible for these decadal changes in dust storm frequency. Three hundred million young trees were planted in northern China, beginning in the 1950s [Parungo et al., 1994] in an effort to decrease dust storm severity. Decadal variations in cyclonic activity and precipitation [Qian et al., 2002] are also likely causes for the observed variations in Asian dust storms for the time periods centered on 1950–1970 and 1980 to present. Qian et al. [2002] postulate that increased warming in Mongolia reduced the meridional temperature gradient and the frequency of strong cyclonic winds in northern China during the last fifty years, which has produced less intense dust storms.

5. Other Considerations

[44] It is useful to place the decadal changes in aerosol opacity and the tabulated SO2 emissions in context with changes in other meteorological variables. For example, decadal changes in precipitation will modulate the amount of aerosol that undergoes wet deposition, and therefore the aerosol opacity that is observed by satellites. There are several studies of decadal changes of relevant Asian atmospheric variables.

[45] Kaiser [2000] calculated trends in cloud amount, surface air pressure and humidity for China for 1954–1994. Cloudiness changed by −3% per decade in the China coastal plain, and annual mean station pressure increased by 0.4 hPa per decade for many stations in central China.

[46] Zhai and Eskridge [1997] quantified changes in precipitable water (PW) over China from 1970–1990 from an analysis of 44 radiosonde stations. In general, PW and precipitation decreases from south to north. PW is lowest in January, and is quite low northward of 35 N [see Zhai and Eskridge, 1997, Figure 3], e.g., PW over the Beijing region in January is 5 mm, while that over the Tibetian plateau is 2.5 mm. China experienced an annual warming trend from 1970–1990 of 0.26 C/decade, and increased PW of 1.2 mm decade−1, though trends for winter are indeterminate. Hu et al. [2003] analyzed monthly precipitation and temperature from 160 meteorological stations in China for 1951–2000, and detected an annually averaged warming trend, and a drying trend in north and northeast China.

[47] Increases in SO2 should be reflected in increases in acid deposition. As discussed by Streets et al. [1999], the acidity measurement record at Ryori on the Pacific coast of Japan has indicated an increase in acidity of precipitation since the late 1970s. Monitoring data for Japan, however, suggest that strong acids are partly neutralized by windblown dust (kosa, or yellow sand) rich in calcium. Chung et al. [2001] note that a high value of pH in rainwater has been generally observed at a rural site in Korea from 1990–1999 when yellow sand was prevalent in East Asia, while airflows from south and central China often resulted in precipitation with low pH values.

[48] Contour maps of Asian SO2 emissions and sulfur deposition, based upon model simulations, indicate that high sulfur deposition regions follow closely the spatial deposition and magnitude of SO2 emissions [Arndt et al., 1997]. Model simulations indicate that total sulfur deposition over the China coastal plain likely increased between 10 and 20% during 1975–2000 [Carmichael et al., 2002].

[49] We conclude that the magnitudes of the percent decadal changes in cloudiness, PW, and precipitation are not as influential as the double digit changes in population, and its associated human activities, that have increased aerosol optical depths over China during the last two decades.

6. Discussion

[50] The populations of China and India have increased 26% and 23% per decade during the last 20 years, with accompanying increases in coal, oil, biomass consumption, and land-altering activities. Inventories of SO2 emissions for different regions of the world indicate that largest increases during the last 20 years are associated with China, India, and Pakistan. By inference, largest increases in anthropogenic aerosol are expected to be associated with these regions. Analysis of the 20 year TOMS aerosol optical depth data set from 1979 to 2000 for winter months (November–February), months in which desert dust and boreal fire smoke influences are less than in other months, demonstrates that large increases in aerosol optical depth are associated with the coastal plain of China and the geographical arc from northern Pakistan through the Ganges river basin.

[51] The percent increases per decade in population, SO2 emissions, and aerosol optical depths associated with these regions are double digit values. Aerosol increased by 17 and 11% per decade during winter over the China coastal plain and India, while SO2 emissions increased by 35% and 47% per decade, respectively. Winter aerosol optical depths increased by factors greater than 1.2 over the Bay of Bengal and the Arabian Sea during the last two decades (see Figure 9).

[52] All other factors being the same, the increase in aerosol burden is expected to be approximately proportional to the increase in SO2. Barth and Church [1999] used the NCAR Community Climate Model version 3 (CCM3) to determine the contributions of southeast China and Mexico City to the global aerosol burden. Southeast China emitted 11.6% of the global anthropogenic sulfur emissions and contributed 9% to the global sulfate burden in their simulations. When emissions were doubled the aerosol burden contributed by China increased by a factor of 2.2.

[53] Our observed optical depth decadal change over China of 17% is half as large as the 35% per decade change in SO2 emissions. There are several factors that may contribute to this discrepancy. A primary factor is that sulfate aerosol composes only a portion of the total Asian aerosol budget. Anthropogenic aerosol at Midway Island during the last 20 years composed just 24% of the total non sea salt aerosol (see Figure 8). Gobi desert dust optical depths (see Figures 5 and 12) are more than half as large as those over the China coastal plain in winter. The differences in the aerosol and SO2 decadal rates (i.e., 17% versus 35%) therefore are likely due to the presence of nonsulfate aerosol over the coastal plain.

[54] The differences in the aerosol and SO2 decadal rates (i.e., 11% versus 47%) over India, however, are not accounted for by this argument, since aerosol measured during the February 1999 INDOEX field campaign was 10% mineral dust and 85% anthropogenic aerosol [Lelieveld et al., 2001], half of which is generated by biofuels. Time trends of SO2, black carbon, organic carbon, and soil dust generated in settled regions, however, need not follow each other in the same manner, and the transfer rates of these aerosol types from the planetary boundary layer into the free troposphere need not be the same. Further studies are required to elucidate these trends at a level of temporal and spatial detail comparable to that of the SO2 emission inventories.

[55] The differences in the aerosol and SO2 decadal rates, cited above for both China and India, require further study, with consideration of a number of topics beyond the scope of the present paper. For example, high emissions of SO2 (see section 3) have increased faster than low emissions in China and India. If the conversion rate from SO2 to sulfate is different for high and low emissions, then optical depth trends may not follow the total emissions trend. If there have been changes in boundary layer meteorology during the last several decades, then these changes could influence aerosol trends in the free troposphere.

[56] Finally, the large amplitudes of the desert dust and boreal forest smoke optical depths that are presented in Figures 57 point to the need to accurately identify specific aerosol types (i.e., desert dust, boreal forest fire smoke, sulfate aerosol, and carbonaceous aerosol) from satellite and ground-based remote-sensing data. In addition to the further analysis of the TOMS data record, other multiwavelength aerosol optical depth data, such as that of the Moderate Resolution Imaging Spectroradiometer (MODIS) and Ozone Monitoring Instrument (OMI) experiments, can be used to identify aerosol composition and trends on a regional basis during the 21st century.


[57] This research was supported by divisional funds of the Atmospheric Chemistry Division at the National Center for Atmospheric Research (NCAR). NCAR is sponsored by the National Science Foundation. Appreciation is expressed to Gabrielle Petron, James Smith, Bill Randel, Aaron Goldman, Mary Barth, William Collins, and Daniel Marsh for helpful comments. Appreciation is also expressed to Terry Galloway, whose mentoring of Steven Massie in 1970–1974 provided the inspiration for this paper.