Estimate of aerosol absorbing components of black carbon, brown carbon, and dust from ground-based remote sensing data of sun-sky radiometers

Authors

  • Ling Wang,

    1. State Environmental Protection Key Laboratory of Satellites Remote Sensing, Institute of Remote Sensing and Digital Earth of Chinese Academy of Sciences, Beijing 100101, China
    2. International Institute for Earth System Science, Nanjing University, Nanjing 210093, China
    Search for more papers by this author
  • Zhengqiang Li,

    Corresponding author
    1. State Environmental Protection Key Laboratory of Satellites Remote Sensing, Institute of Remote Sensing and Digital Earth of Chinese Academy of Sciences, Beijing 100101, China
    • Corresponding author: Z. Li, State Environmental Protection Key Laboratory of Satellites Remote Sensing, Institute of Remote Sensing and Digital Earth of Chinese Academy of Sciences, Beijing 100101, China. (lizq@irsa.ac.cn)

    Search for more papers by this author
  • Qingjiu Tian,

    1. International Institute for Earth System Science, Nanjing University, Nanjing 210093, China
    Search for more papers by this author
  • Yan Ma,

    1. State Environmental Protection Key Laboratory of Satellites Remote Sensing, Institute of Remote Sensing and Digital Earth of Chinese Academy of Sciences, Beijing 100101, China
    Search for more papers by this author
  • Fengxia Zhang,

    1. State Environmental Protection Key Laboratory of Satellites Remote Sensing, Institute of Remote Sensing and Digital Earth of Chinese Academy of Sciences, Beijing 100101, China
    Search for more papers by this author
  • Ying Zhang,

    1. State Environmental Protection Key Laboratory of Satellites Remote Sensing, Institute of Remote Sensing and Digital Earth of Chinese Academy of Sciences, Beijing 100101, China
    Search for more papers by this author
  • Donghui Li,

    1. State Environmental Protection Key Laboratory of Satellites Remote Sensing, Institute of Remote Sensing and Digital Earth of Chinese Academy of Sciences, Beijing 100101, China
    Search for more papers by this author
  • Kaitao Li,

    1. State Environmental Protection Key Laboratory of Satellites Remote Sensing, Institute of Remote Sensing and Digital Earth of Chinese Academy of Sciences, Beijing 100101, China
    Search for more papers by this author
  • Li Li

    1. State Environmental Protection Key Laboratory of Satellites Remote Sensing, Institute of Remote Sensing and Digital Earth of Chinese Academy of Sciences, Beijing 100101, China
    Search for more papers by this author

Abstract

[1] Black carbon (BC), brown carbon (BrC), and mineral dust (DU) are three major light absorbing aerosols, playing important roles in climate change. Better knowledge of their concentrations is necessary for more accurate estimates of their radiative forcing effects of climate. We present a method to retrieve columnar contents of BC, BrC, and DU simultaneously from spectral refractive indices and spectral single scattering albedo obtained from the sun-sky radiometer measurements. Then, this method is applied to investigate the columnar volume fractions and mass concentrations of BC, BrC, and DU in Beijing, China, based on measurements obtained from 2009 to 2010. Results show that among the three absorbing aerosols, DU dominates the largest volume fraction in the total aerosol volume (20–45%), followed by BrC (5–25%), and BC (< 5%). The retrieved monthly mean content of each absorbing component exhibits clear seasonal variation. BrC dominates in late fall and winter (40–92.5 mg/m2), whereas is extremely low in summer (< 10 mg/m2). DU dominates in spring, ranging from 270 to 405 mg/m2 (with volume fraction >30%), while during June–September, the DU fraction is generally lower than 30%. BC is characterized by low levels throughout the year. The monthly mean BC columnar mass concentration ([BC]) ranges from 2.7 to 7.3 mg/m2 with winter slightly higher than other seasons. As a preliminary validation, we compare our retrieved [BC] with in situ measurements. Similar day-to-day variation trends and good correlations are found between the retrieved [BC] and in situ measurements.

1 Introduction

[2] Recent modeling and field studies [e.g., Stier et al., 2007; Solomon et al., 2007; Bond et al., 2013] indicate that aerosol light absorption is an important factor of climate forcing. The direct radiative forcing of light-absorbing aerosols may be greater than that of methane and equal to about one third of that of carbon dioxide [Jacobson, 2001]. Hence, the net radiative effect of aerosols could change in sign from cooling to warming in regions with highly absorbing aerosols [Haywood and Shine, 1995].

[3] Absorbing aerosol has a warming effect through absorbing energy in the solar spectrum, and it includes three major types: black carbon (BC), mineral dust (DU), and some light absorbing organic carbon (often called as “brown carbon”, BrC) generated from biomass and coal combustion burning processes [Bahadur et al., 2012; Chung et al., 2012; Feng et al., 2013; Kirchstetter et al., 2004]. Several approaches can be used to evaluate their concentrations in particulate matter, such as in situ sampling of atmospheric aerosol particles with further laboratory analysis and simulating with chemistry transport model. In situ sampling can provide accurate measurement results but requires considerable effort, and it is also quite limited to spatial and temporal coverage. On the contrary, chemistry models can provide a large spatial and relatively fine temporal aerosol estimates, but the simulated results remain a large uncertainty [Bond and Bergstrom, 2006; Sato et al., 2003; Park et al., 2003]. These models sometimes would require a 200–400% increase in the BC emission inventories to match the measurements from aerosol robotics network (AERONET) [Sato et al., 2003]. Recently, some researches have started to give their attention to infer aerosol composition from AERONET retrievals, since it provides a wealth of information on aerosol optical and physical properties such as aerosol optical depth (AOD), single scattering albedo (SSA), and refractive indices [Dubovik et al., 2000b]. In addition, it can also provide a long-term view and extensive spatial coverage, as it comprises more than 300 surface sun-sky radiometers (or sun photometers) located throughout the world [Holben et al., 1998]. In 2005, Schuster et al. [2005] demonstrated an approach to retrieve BC concentration from AERONET imaginary refractive indices for the first time based on a three component assumption of BC and (NH4)2SO4 embedded in water host. Following his work, Dey et al. [2006], Arola et al. [2011], and Wang et al. [2012a, 2012b] included additionally absorbing component of BrC or dust into the three component model by considering the spectral dependent properties of the imaginary part of refractive indices. However, they did not consider the coexistence of BrC and dust in the retrieval, due to their similar spectral shape of imaginary refractive index given the spectral resolution (four bands) of the AERONET. Since dust and BrC/BC are different in particle size, it is possible to separate them based on size-related parameters. For example, Derimian et al. [2008] demonstrated a method to differentiate the absorption caused by dust and BC from SSA spectrum and developed equations relating spectral variability of SSA with the concentrations of the absorbing elements of iron in dust and BC. Bond et al. [2013] also used the information of the particle size difference between BC and dust to separate the BC contribution to aerosol absorption. In their retrieval, they assumed that BC controls the midvisible absorption of small (i.e., “fine-mode”) aerosol particles. Hence, they only considered aerosols less than 1 µm to avoid the influence of dust. However, these two researches did not consider the three absorbing aerosols existed simultaneously in an aerosol mixture either.

[4] In this study, we combine the spectral absorption and size information to assess whether these two kinds of information can be effective in deriving the concentrations of the three absorbing aerosols of BC, BrC, and DU simultaneously. The spectral absorption and size information is represented by imaginary refractive index (k) spectrum and SSA spectrum, respectively. We then implement this method to retrieve the column-integrated concentrations of BC, BrC, and DU over Beijing, China, using the AERONET data. In the last part, we validate the retrieved BC concentrations with the in situ measurements.

2 Measurements

2.1 Study Area

[5] Beijing (39.9°N, 116.3°E), the capital of China, is a unique area to study aerosol composition. The aerosol here can contain a mixture of BC, BrC, and DU considering the emission sources and influence factors. The high concentration of BC mainly comes from the large diesel fuel consumption, which is a result of the rapid industrialization and motorization of Beijing [He et al., 2001; Yang et al., 2000]. The BrC here mainly generates from less efficient and complete combustion processes of coal burning, which is the major energy source for power plants and home heating systems in northern China. In addition, straw burning in the rural areas around Beijing during the harvest time periods (April–May in spring and October–November in fall) can also contribute to the BrC in Beijing. At the same time, dust particles are common components in the atmosphere especially in spring season. During this time, dust storms frequently happen owing to Beijing's geographical location close to the dust source region of Mongolia.

2.2 Sun-Sky Radiometer Measurements

[6] A Cimel sun-sky scanning spectral radiometer CE-318 (Cimel Electronique, Paris, France) was installed on the roof of the Institute of Atmospheric Physics building (39°58′N, 116°23′E, 30 m high) in Beijing, China. The measurements with this instrument are part of the AERONET global network. The instrument makes direct sun measurements within a 1.2 full field of view every ~15 min at four standard channels of 440, 670, 870, and 1020 nm, with additional supplement channels like 940 nm [Holben et al., 1998]. However, the angular distribution of sky radiance is only measured at the four standard channels. Spectral AOD is obtained from direct sun measurements with an accuracy of ~0.01–0.02 [Holben et al., 1998]. Other aerosol optical properties, such as refractive indices, SSA, and the column integrated aerosol size distributions, are derived from sun and sky radiance measurements at 440, 675, 870, and 1020 nm by the inversion code proposed by Dubovik et al. [2000a, 2006]. The imaginary refractive index (k) is expected to have an uncertainty of 30–50%, the real refractive index (n) has an uncertainty of 0.04, and the SSA has an uncertainty of 0.03 when AOD (440 nm) > 0.4 [Dubovik et al., 2000b]. Hence, in our study, we only considered observations with AOD greater than 0.4 at 440 nm and utilize data from AERONET Version 2 inversion products.

3 Methods

3.1 Absorption Properties of Three Typical Absorbing Aerosols

3.1.1 k Spectra

[7] Figure 1 presents the k spectra of the three typical absorbing aerosols. BC has the strongest absorption and its imaginary refractive indices are relatively constant between 400 and 1100 nm [Bergstrom, 1972; d'Almeida et al., 1991]. Although dust and BrC are less absorbing compared to BC, their absorptions are wavelength dependent, exhibiting an enhanced absorption from the short visible (about 500 nm) toward the UV band. The enhanced absorption of dust is caused by iron oxide, primarily hematite and goethite [Koven and Fung, 2006; Sokolik and Toon, 1999; Wagner et al., 2012]. For BrC, it is caused by the presence of resonant ring structures [Yang et al., 2009]. Since BC shows different spectral dependent behavior with dust/BrC in the AERONET wavelength region of 440–1020 nm, the AERONET-retrieved k spectrum is helpful in separating absorption by BC from absorption by dust or BrC. However, it is difficult to distinguish between them merely from the k spectrum, as dust and BrC have similar k spectral-dependent behavior given the spectral resolution (four wavelength bands) of the AERONET instruments.

Figure 1.

Imaginary refractive index spectra of BC, BrC, and dust from UV to near-infrared. The vertical dot lines indicate the four standard wavelength bands of the AERONET sun-sky radiometers. In the legend, BrC_483 K and BrC_633 K mean the BrC generated under combustion temperature of 483 K and 633 K, respectively. DU_10% hematite and DU_1% hematite mean the dust particles contain 10% and 1% hematite, respectively.

3.1.2 SSA Spectra

[8] Different from the k spectrum, SSA spectrum is sensitive not only to changes in aerosol type, but also to the variability in aerosol size [Dubovik et al., 1998, 2002]. Dust and BrC are different concerning the particle size: dust aerosols are mainly coarse particles, while BrC aerosols generated from combustion processes are normally fine particles [Dubovik et al., 2002; Clarke et al., 2004; Arola et al., 2011]. Therefore, the difficulty in distinguishing them based on k spectra can be solved by examining the behavior of the SSA spectra.

[9] Figure 2 presents the SSA spectra for dust and BrC calculated using Mie theory for two cases. For each case, we assume their imaginary part is the same, namely they have the same k spectra. For case1, they have a lower imaginary part, with 0.0122 at 440 nm and 0.0013 at other three wavelengths, while for case2, a higher one is used, with 0.0525 at 440 nm and 0.005 at other three wavelengths (These two values were given by Chen and Bond [2010] for BrC generated under combustion temperature of 483 K and 633 K, respectively). The real part for dust is 1.57 as used by Schuster et al. [2005], 1.53 for BrC, at all wavelengths [Chen and Bond, 2010; Gillespie and Lindberg, 1992; Kirchstetter et al., 2004]. The volume particle size distributions of dust and BrC are adopted from the research results reported by Dubovik et al. [2002], who used bimodal lognormal functions to parameterize the particle size distributions. For dust, the median radius and the standard deviation are 0.14 µm and 0.43 µm for fine mode and are 2.11 µm and 0.64 µm for coarse mode. Considering that there is no specific size distribution for BrC, the size distribution of biomass burning aerosol is used instead. The median radius and the standard deviation are 0.14 µm and 0.43 µm for fine mode and are 3.24 µm and 0.78 µm for coarse mode.

Figure 2.

SSA spectra for dust and BrC simulated using Mie theory by assuming that they have the same k spectra, but different typical size distribution. Dust_case1 and BrC_case1 denote that they have an imaginary refractive index of 0.0122 at 440 nm and 0.0013 at other three wavelengths, while Dust_case2 and BrC_case2 denote that their imaginary refractive index is 0.0525 at 440 nm and 0.005 at other three wavelengths.

[10] The Mie simulated SSAs for dust and BrC shown in Figure 2 are consistent with the range of values reported in other published literatures. Dust SSA at 500 nm used in GOCART model is in the range of 0.6–0.9 [Chin et al., 2009], and BrC SSA at 440 nm given by Bahadur et al. [2012] based on sun photometer measurements is around 0.77. However, it should be noted the values given in Figure 2 are just for pure dust or BrC particles; hence, they may be different from the field measured ones. For example, Dubovik et al. [2002] found the dust SSA at 440 nm ranges from 0.9 to 0.95. This is because in the field measurements, pure dust or BrC aerosols are extremely rare. Dust aerosols may be contaminated from some nonabsorbing components, causing the aerosol less absorbing and having larger SSA value. Similarly, the BrC aerosols may be contaminated from BC which usually coexists with BrC but has stronger absorbing property, causing the aerosol more absorbing and having lower SSA value.

[11] In addition, from Figure 2, one can find that even though the dust and BrC have the same k spectra, their SSA spectra shapes between 670 and 870 nm are different. The shape of dust SSA from 670 to 1020 nm follows an increasing (or neutral) pattern, while for BrC, it follows a decreasing pattern. Thus, the major contributor to the enhanced absorption at the blue wavelength can be determined by examining the spectral behavior of SSA from 670 to 1020 nm. Accordingly, we propose here to use SSA(870 nm)–SSA(670 nm) (dSSA) to distinguish between dust and BrC further. dSSA is positive for dust-dominated aerosols, whereas negative for BrC-dominated aerosols.

[12] We used spherical Mie theory for dust scattering computations, rather than the spheroid code, but we estimate no significant error introduces to the conclusions of Figure 2. First, as we are inverting for SSA, it is less influenced by particle nonsphericity than angular measurements [Mishchenko et al., 1997]. Second, the increased SSA spectral behavior has been widely approved in the community, like Dubovik et al. [2002].

3.2 Forward Model for Aerosol Optical Property Simulation

3.2.1 Aerosol Composition Model

[13] Aerosol is a complicated mixture of liquid water and dry particles, such as BC, ammonium, and nitrate. Though there are numerous kinds of dry particles, they can be divided into two categories: one is the light-absorbing components like BC, dust, and BrC; the other is the scattering (nonabsorbing) components like sea salt, sulfate, and nitrate etc. We use ammonium sulfate (AS) here to surrogate for the nonabsorbing components to simplify the aerosol chemical composition model. This assumption is based on the following considerations. First, sulfate is the major contributor of the aerosol mass, and it usually associates with ammonium, contributing about 16–54% to the submicron aerosol mass fraction. Second, the refractive index of other nonabsorbing aerosols (including organic compounds, another important contributor to aerosol mass) mainly ranges from 1.51 to 1.57 [Larson et al., 1988], while the refractive index of AS just lies on a middle value of 1.53. Therefore, according to the optical properties of aerosol chemical components, we can treat aerosol as a mixture of five components: three light absorbing components (BC, BrC, and DU), one nonabsorbing component represented by AS and aerosol water (AW). This model is an extension of the four-component models of Arola et al. [2011] and Wang et al., 2012a, 2012b, contributing to that we add an extra information of spectral behavior of SSA to constrain the additional one component.

3.2.2 k Calculation Using Volume Mixing Rule

[14] We calculate an effective complex refractive index for aerosol mixture of BC, BrC, DU, AS, and AW using the volume averaged mixing rule [Heller, 1965]. The assumed complex refractive index values of these components are listed in Table 1. We consider a spectrally independent imaginary part (k), except for BrC and dust. For BrC, we fix k at 440 nm and 670 nm with the values used by Dey et al. [2006]. For dust, we use 0.010 and 0.004, respectively, which is the closest to the values reported by many research works related to exploring the refractive index of dust through laboratory and field measurements [Sokolik and Toon, 1999; Wagner et al., 2012]. We use the values of Soot A type for BC refractive index suggested by Bergstrom [1972]. The real parts at the four wavelengths n(λ) are assumed to be the same value for each aerosol component, considering that the little spectral dependence of n(λ) can be neglected in this study.

Table 1. Aerosol Components and Their Refractive Indices Used in the Composition Model
 Real Part (n)Imaginary Part (k)
 440–1020 nm440 nm670 nm870–1020 nm
DU1.570.0100.0040.001
BrC1.530.0630.0050.001
BC1.950.660.660.66
AS1.5310−710−710−7
AW1.33000

[15] Given the refractive indices and volume fractions of the five aerosol components, the complex refractive indices of the aerosol mixture are calculated from the following equations:

display math(1)
display math(2)

where i (=1,2,…,5) denotes the ith aerosol component. fi denotes the volume fraction of the component i. j (=1,2,…,4) is the summation index over the four retrieval wavelengths (440, 670, 870, and 1020 nm). In the k calculations, we only consider three absorbing aerosol components. This is because the imaginary refractive indices of AS and AW are extremely small; therefore, their presence has little influence on the imaginary refractive index of the mixture.

3.2.3 SSA Calculation Using Mie Theory

[16] In order to associate SSA with the content of absorbing component, we calculate dSSA through the flowing equation (refer to Appendix A for the detailed derivation of this formula):

display math(3)

[17] Where, τext670 and τext870 are the extinction optical depths at the wavelength of 670 and 870 nm, respectively; MAEi670 and MAEi870 are the mass absorption efficiencies of absorbing component i at the wavelength of 670 and 870 nm, respectively, often in units of m2/g; ρi and fi are the density and the volume fraction of absorbing component i, respectively; Vtotal is the total column-integrated volume concentration of the aerosol mixture. Among these parameters in equation ((3)), the values of SSA, Vtotal, and τext can be obtained from sun-photometer measurements; the values of density ρ of BC, BrC, and DU can be obtained from values presented in the literatures, which are listed in Table 2; the MAEs of BC, BrC, and DU can be simulated according to the published literatures and are also summarized in Table 2. Thus, all parameters in equation ((3)), except for the volume fractions f of the absorbing aerosols, are all the known characteristics.

Table 2. The Density and Mass Absorption Efficiencies of Three Absorbing Aerosol Components
 Density (g/cm3)MAE (m2/g)
440 nm670 nm870 nm
BC2.012.58.146.32
BrC1.80.9210.0670.050
DU2.60.1040.0450.035

[18] The MAE of each absorbing aerosol component listed in Table 2 is obtained as follows:

  1. [19] The MAE(λ) of BC

[20] A typical value of 10.0 m2/g at 550 nm is recommended by many researchers as the standard value [Bond and Bergstrom, 2006; Horvath, 1993; Liousse et al., 1993; Schuster et al., 2005]. Then, the BC MAE at other wavelengths can be inferred based on the fact that BC absorption follows an inverse-wavelength (λ−1) relationship from 370 to 950 nm.

  • 2.The MAE(λ) of BrC and DU

[22] They are both estimated using the Mie theory calculations in this study. The input parameters include refractive index, density, and volume particle size distribution. The values of refractive indices are adopted from Table 1, and the density is listed in Table 2. The parameters of volume particle size distribution for BrC and DU are consistent with the values used in SSA spectra simulation in section 'SSA Spectra'. The calculated MAEs of DU are 0.104, 0.045, and 0.011 at 440, 670, and 870 nm, respectively. The calculated MAEs of BrC are 0.921, 0.067, and 0.010 at 440, 670, and 870 nm, respectively. Those values are close to the values of BrC MAE given by Yang et al. [2009] (1.01, 0.32, and 0.02 at 470, 660, and 880 nm).

3.3 Applying Forward Model to Infer BC, BrC, and DU Content

[23] To derive the forward model, we iterate the volume fraction of the five components in the aerosol mixture until the calculated optical properties, i.e., averaged n at four bands, spectral k, and spectral SSA of the mixture, match best of the values retrieved from the sun photometer measurements. That means the minimized χ2 of equation ((4)):

display math(4)

where njmeas, klmeas, and dSSAmeas are the sun photometer measured real refractive indices, imaginary refractive indices, and dSSA, respectively; correspondingly njcal, kjcal, and dSSAcal are the model-calculated ones; j is the summation index over the four retrieval wavelengths (440, 670, 870, and 1020 nm).

4 Results and Discussions

4.1 Volume Fraction of Three Absorbing Aerosols in Beijing

[24] Figures 3a and 3b show the monthly averaged volume fractions of three typical absorbing aerosols over Beijing, based sun photometer measurements in 2009 (334 AERONET lev2.0 data) and 2010 (387 cases), respectively. One can find that the volume fractions for each absorbing aerosol in 2009 and 2010 have similar variation ranges and seasonality. Among the three absorbing aerosols, DU tends to occupy the largest volume fraction in the total aerosol volume ranging from 20 to 45%. BC occupies a volume fraction no more than 5% throughout the year. BrC varies in between BC and DU, ranging from about 5 to 25%.

Figure 3.

Monthly averaged volume fractions of absorbing aerosols in Beijing for the year of 2009 (a) and 2010 (b).

[25] There is an apparent seasonal variation for both BrC and DU. BrC fraction is higher at the beginning and end of the year than that from April to September. It possibly is caused by increased coal burning from winter heating during these time periods [Lin et al., 2009]. In Beijing, collective heating generally starts from November to the following mid-March, thus an increase in BrC amount is likely to happen during this period. On the other hand, the surface OC measurements in Beijing carried out by Cao et al. [2007] also showed similarly strong seasonality with the largest value in winter. DU tends to peak during spring seasons (March–May), then starts to decrease from June to September. However, it tends to increase again from October. High DU volume fraction in spring is related to the dust storm happened during this period [Eck et al., 2010]. The high values in winter are probably due to ash emissions related to coal burning and the bare surface with windy weather. On the other hand, in situ measurements in Beijing during winter showed that the dust usually accounts for ≥50% of the total aerosol mass concentrations at the surface [e.g., Duan et al., 2007; Yuan et al., 2008], indicating that the dust can also occupy a large fraction in winter.

4.2 Mass Concentration of Three Absorbing Aerosol Component in Beijing

[26] The component volume fractions shown in Figure 3 can be converted into column-integrated mass concentrations through multiplication by the component density and the size-integrated aerosol volume distribution. The details of the method can refer to Schuster et al. [2005]. The monthly averaged columnar mass concentrations of BC, BrC, and DU using two years’ data are shown in Figures 4a, 4b, and 4c, respectively.

Figure 4.

Monthly columnar mass concentration of (a) BC, (b) BrC, and (c) DU averaged for 2009–2010 period.

[27] The seasonal variations of the mass concentrations of BC, BrC, and DU in Beijing behave as expected. Compared to the other two absorbing components, [BC] ([] denotes the column-integrated mass concentration) is characterized by low levels throughout the year. The monthly mean [BC] ranges from 2.7 to 7.3 mg/m2 with winter slightly higher than other seasons. Assuming that BC is well mixed below a boundary layer height h, the surface concentration can convert into the column values by multiplying h. Using the typical boundary layer height of 1.0 km, the corresponding ground-level concentrations are 2.7–4.9 µg/m3 for summer and 4.0–7.3 µg/m3 for winter, which are comparable to those measured by Cao et al. (2007), who found the ground-level mass concentrations of BC in Beijing are 7.1 µg/m3 and 4.6 µg/m3 for winter (6–20 January) and summer (3 June to 31 July) periods.

[28] The monthly averaged columnar mass concentrations of BrC ([BrC]) are found to vary in a wide range throughout the year, with a maximum of 92.5 mg/m2 in November, and a minimum of 5.2 mg/m2 in August. This large discrepancy is mainly caused by the increasing coal consumption from November to the following March. It should be noted that the [BrC] predicted by Arola et al. [2011] is lower than ours, in the range 5–35 mg/m2. This difference can be explained by different data sets used in the retrieval. They only selected fine mode-dominated cases in the retrieval; hence, some large BrC fraction cases may be ignored. Additionally, the different imaginary refractive index and density of BrC used in the retrieval can also introduce some difference. Although the absolute value of [BrC] is higher than that predicted by Arola et al. [2011], the seasonality of [BrC] agrees well with their results and some other in situ measurements [Cao et al., 2007; Lin et al., 2009]. It can be noted that although BC and BrC are generated together during the combustion processes, the seasonal variation of [BC] is not notable compared to [BrC]. This can be explained by the following reasons: (i) the total amount of BC in the atmosphere is rather small, usually less than 10 mg.m−2 in urban areas [Schuster et al., 2005], hence the winter increase due to the increase in source emission from winter heating will not be very obvious; (ii) the emission sources of BC and BrC have differences. The BC emission sources are mainly from diesel fuel combustion processes, such as from motor vehicles, which are previous throughout the year, while the BrC emission sources in Beijing are mainly from coal burning, which have obvious seasonal characteristics.

[29] The dust mass concentration is also successfully obtained based on the method we presented. The seasonal variation of dust mass concentration is much more obvious than its volume fraction shown in Figure 3. High [DU] occurs in spring; the monthly average [DU] ranging from 270 to 405 mg/m2 during this period. Low [DU] mainly occurs in summer and early fall of September and the monthly average [DU] is around 50–100 mg/m2, which is about four to nine times lower than that in spring. During the heating period, there also exists a considerable [DU] (around 250 mg/m2). Some studies indicated that the magnitude of dust mass concentration in the surface level is in an order of 10 to 100 µg/m3 [Gong et al., 2003]. Correspondingly, the [DU] in the entire atmosphere column is estimated to be in an order of 10 to 100 mg/m2, assuming a typical boundary layer height of 1 km.

4.3 Significance on Simultaneously Considering BC, BrC, and DU

[30] Both BrC and DU have enhanced absorption at 440 nm; hence, more BC will be needed to account for the enhanced absorption without considering the BrC or DU as the absorbing components in the aerosols. This will in turn lead to overestimation of the positive radiative forcing effects, as the mass absorbing efficiency of BC is much larger than that of BrC and DU.

[31] Figure 5 exhibits the comparison between the retrieved [BC] when only considering BC as the absorbing component versus the retrievals using the method in this work, namely considering BrC and DU together with BC as the absorbing components. From Figure 4, we can find that [BC] is greatly overestimated (about by 80%) without considering BrC and DU in the aerosol mixture. This indicates that there exists a considerable amount of BrC and DU in the atmosphere over Beijing. In the previous work introduced by Wang et al. [2012b], the spectral behavior of imaginary refractive indices of aerosols in Beijing is analyzed. The results show that imaginary refractive indices of aerosols in Beijing have a strong wavelength-dependent characteristic, with k (440 nm) almost twice of k (670 nm). This also confirms the presence of spectral-dependent absorbing aerosols of BrC and DU in Beijing. Therefore, to include BrC and DU in the retrieval is significantly important for more accurate estimation of BC concentrations over cities like Beijing. In addition, this helps fully utilizing the spectral information of imaginary refractive indices in the aerosol composition retrieval. As shown in Figure 6, the calculated k (440 nm) is lower than the observed k (440 nm) when only considering BC as the absorbing component. Meanwhile, when including BrC and DU, the calculations become closer to the observations. Since the absorption by BrC and DU at 670 nm becomes very weak, there is no such influence for k (670 nm).

Figure 5.

Comparison of retrieved [BC] based on different assumption of the absorbing components in 2010 over Beijing.

Figure 6.

Scatter plot between simulated imaginary refractive indices and sun photometer measurements, left panel for k (440 nm) and right panel for k (670 nm).

4.4 Comparison With In Situ Measurements

[32] As a preliminary validation, we compare the retrieved [BC] with in situ measurements in Beijing. The ground-level BC mass concentration is obtained from Aethalometer (Magee Scientific AE51). The measurement campaign took place at the roof of Institute of Remote Sensing Applications, Chinese Academy of Sciences from 11 October to 31 October 2012. In order to perform rigorous comparisons, we use data collocated in time. For each retrieved BC from sun photometer, we chose the closest measured ones within a time interval of ±15 min. Figure 7 presents comparisons of daily mean BC mass concentration obtained from AE51 (in µg/m3) and retrieved from sun photometer over Beijing during October 2012. The retrieved BC and the measurements have consistent variation trends, with the square of correlation coefficient R2 of 0.64. The averaged BC mass concentrations are 6.84±2.48 and 4.33±2.86 for the retrieval and measurement, respectively. According to the simple equation of ground-level BC=(column BC)/(BC column height), the deduced mean BC column height is 1.57 km (=6.84/4.33), which is a reasonable boundary layer height in Beijing fall. It should be noted that, this value will approach to 3 km if it is calculated using the column [BC] retrieved without considering BrC and DU. The boundary layer height with a mean value of 3 km is obviously too high for Beijing fall, while it is confirmed to be around 1km by other studies, e.g. Zhang et al. [2006].

Figure 7.

Daily averaged BC mass concentration obtained from remote sensing retrievals and in situ measurements during October 2012 in Beijing.

5 Conclusions

[33] We employ the remotely sensed aerosol spectral refractive indices and spectral SSA to derive columnar contents of three major absorbing aerosol components of BC, BrC, and DU. The method developed in this work is a succession of several previous studies and has advantages in providing simultaneous determination of BC, BrC, and DU. We illustrate by investigating the columnar volume fractions and mass concentrations of absorbing aerosol components in Beijing, China from 2009 to 2010. Conclusions are as follows:

  1. [34] In order to constrain BC, BrC, and DU simultaneously in the retrieval, we employ extra information of SSA spectra, in addition to the spectral imaginary refractive index. We added DU into the four-component aerosol chemical model proposed by Arola et al. [2011] to form a five-component model (i.e., BC, BrC, DU AS, and water), which is more suitable to represent the aerosol chemical composition, for example in Beijing, in the retrieval.

  2. [35] In Beijing, among the three absorbing aerosols, DU dominates the largest volume fraction of the total aerosol volume amount (20–40%), while BC occupies the least (< 5%), and BrC varies from 5 to 20%.

  3. [36] The retrieved BC, BrC, and DU over Beijing show reasonable seasonal variation trends and are comparable with other published literatures. BrC dominates in late fall and winter (up to 92.5 mg/m2), whereas is extremely low in summer (< 10 mg/m2). This large discrepancy is attributed to the enhanced coal consumption from November to the following March. DU dominates in spring, in the range 270–405 mg/m2 due to the frequent occurrence of dust storms during this period. Compared to BrC and DU, BC is characterized by low levels throughout the year and the monthly mean [BC] ranges from 2.7 to 7.3 mg/m2 with winter slightly higher than other seasons.

  4. [37] It is necessary to include BrC and DU in the remote sensing retrieval model for more accurate estimation of BC concentrations. For example in Beijing, if BrC and DU are not considered in the retrieval, [BC] will overestimated by ~80%.

  5. [38] Consistent temporal variation trends and good correlations are found between retrieved [BC] and in situ measurements. The retrieved [BrC] and [DU] also shows good agreement with other literature results. More efforts will be engaged in validating them in Beijing. Further works are also needed to explore the optical (e.g., refractive indices, MAE) and physical (e.g., density) properties of the three absorbing aerosol components, to improve the estimation of their abundance, as the retrieved concentrations depend substantially on the assumed refractive indices and SSA (related to MAE, density, and size distribution etc.) of each component.

Appendix A

SSA Expressed in Terms of Contents of the Absorbing Components

[39] SSA is the ratio of light scattering to the extinction (sum of scattering and absorption):

display math(A1)

where τsca is the scattering optical depth, τext is the extinction optical depth, and τabs is the absorption optical depth, and can be treated as sums of the absorption optical depth of each absorbing component i:

display math(A2)

[40] In order to establish a link between SSA and the volume fractions of the absorbing components, we apply the following equation to calculate the absorption optical depth of each individual component τabs,i in equation ((A2)):

display math(A3)

[41] Where H is the height of the entire atmosphere, ci(h) is the mass concentration of absorbing component i in each layer h (e.g., µg/m3), kabs,i(h) represents mass absorption efficiency (MAE) of absorbing component i in each layer h, and is in units of m2/g. Assuming that MAE do not change vertically, i.e., kabs,i(h)=constant= MAEabs,i, equation ((A3)) can be transformed as follows:

display math(A4)

where Ci is the total mass of the absorbing component i in the entire atmospheric column, thus in unit of mass per unit area (e.g., µg/m2). It is the product of the density ρ and the column-integrated volume concentration V of the particles:

display math(A5)

[42] In equation ((A5)), ρi and Vi are the density and column-integrated volume concentration of the aerosol component i, respectively. Vi can be obtained using its volume fraction (fi) multiplied by the total columnar volume concentration of the aerosol mixture, Vtotal, which is in unit of volume per unit area (e.g., µm3/µm2). By substituting equations ((A5)), ((A4)), and ((A3)) into equation ((A1)), we SSA expressed in terms of the contents of the absorbing components:

display math(A6)

Acknowledgments

[43] This work was supported by National Basic Research Program of China (973 Program) under grant 2010CB950800 (2010CB950801), National Natural Science Foundation of China (Grant no. 41222007), and Chinesisch-Deutsches Forschungsprojekt (GZ659). The authors thank the principal investigator of Beijing AERONET site, Hongbin Chen, and staffs for providing the data. We are also grateful to the PHOTONS group of Laboratoire d'Optique Atmosphérique, University Lille 1, France, for helping in the maintenance of the Sun photometer network and calibration of the instruments.

Ancillary