The AErosol RObotic NETwork is a worldwide framework of CIMEL sun/sky photometers that retrieve aerosol optical data at discrete wavelengths ranging from 340 to 1,020 nm [19,23]. This study employs Level 2.0 data to ensure cloud screened, quality controlled retrievals of τ, τ_abs, α, α_abs, and ω_o in the wavelength range between 440 and 1,020 nm [24]. The retrievals are derived from inverted almucantar observations developed by [25] and [26]. We use the daily averaged data collected from three AERONET sites representing west-central, northern, and eastern China: 28 December 2007 and 20 March 2010 (SACOL) and 29 December 2007 and 21 March 2010 (Xianghe and Taihu). We use retrievals where τ₄₄₀ > 0.4 to ensure the least amount of uncertainty [19].

From Figure 1, SACOL (35.95°N, 104.14°E), the Semi-Arid Climate and Environment Observatory of Lanzhou University, is within a mineral dust source region in the Gansu province downwind from the Taklamakan and Gobi Deserts as well as within the confines of the Loess Plateau [7,17]. It is also nearby to Lanzhou City which has a population of over 3 million inhabitants along with strong urban and industrial influences [18]. Xianghe (39.75°N, 116.96°E) is in an aerosol mixture region affected by mineral dust, biomass aerosols, and pollution from nearby Beijing and other urban areas in northeast China [4,18]. Taihu (31.4°N, 120.21°E) lies within a heavily polluted region upwind of Shanghai but is itself surrounded by numerous industrial influences [9,18].

This study investigates the optical properties of mineral dust particles which are typically non-spherical in nature. In fact, the optical properties require special consideration since AERONET retrievals are primarily based on assumptions that atmospheric aerosols are homogeneous spherical particles [22,25,27]. As pointed out by [28], though there are shortcomings to using these assumptions in AERONET retrievals, it is still possible to gain useful information from the almucantar readings [25,28]. This study employs the use of a mineral dust model that is comprised of a spectral database calculated from mathematical and geometric theory [20,22]. The model contains theoretical calculations of the single scattering parameters of mineral dust aerosols by assuming a tri-axial ellipsoidal geometric shape which offers many advantages over calculations based on spherical (Mie Theory) and spheroidal assumptions which illustrates the most representative shape of mineral dust particles to date [20,22].

The [22] model uses a combination of Lorenz-Mie theory, the T-matrix method, discrete dipole approximation (DDA), and an improved geometric optics method (IGOM) to illustrate the optical properties of mineral dust-like aerosols. The results are compiled into a database that requires the knowledge of the particle aspect ratio (ratio of the semi-major to semi-minor axis assuming ellipsoidal geometry), size parameter (based on retrieved effective radius), and complex index of refraction. A more detailed discussion of the database as well as the methodology can be found in [22]. Though the database contains information about the phase matrix, extinction efficiency, extinction, scattering and absorption coefficients, and single scattering albedo, this study uses the latter to compare with AERONET retrievals.

We will consider the AERONET retrievals as truth and compare these retrievals with the theoretical calculations from the mineral dust model because the model will have inherent shortcomings when dealing with aerosol mixtures (e.g., external/internal mixtures and pollution coating). Since AERONET retrievals are based on spheroidal assumptions (rather than ellipsoidal) it would be informative to note the agreement between the two retrieval mechanisms in order to gain more insight on their limitations. This will be done by performing a comparison study using the AERONET retrieved effective radius and complex index of refraction (m_r + m_i) as inputs of the mineral dust model.

Figure 2(a,b) shows the 850 hPa geopotential heights for the 24–31 December 2007 (winter case) and 15–21 March 2010 (spring case) time spans. The mean synoptic patterns of both cases have similar trough (darker purple) locations over Northeastern China/Southern Russia. However, the ridge (red and orange) locations differ by ten degrees of latitude with the winter case ridge position (43°N) being at a higher latitude than the spring case (33°N). The position of the ridge in relation to the trough is also different between the two cases suggesting different wind flow patterns [8,29]. For example, there is more of a northwesterly flow over Xianghe than at Taihu and SACOL in the winter case but a west-northwesterly flow pattern is evident in the spring case at all three sites. From backward trajectory analysis (not shown) both storms had origins in the Loess Plateau (Western Gobi Desert) region but the wind patterns suggest dust from the eastern Gobi Desert may have impacted Xianghe more than SACOL and Taihu for the winter case. In contrast, the spring case wind pattern and subsequent track of the low pressure system advected dust from the Loess Plateau to all sites. The synoptic patterns for both cases illustrate how the absorptive properties of the mineral dust aerosols can change depending on the season which is discussed in the following section [7,8,29].

Figure 3(a,b) shows the size distributions of the winter and spring cases, respectively. For the winter case, SACOL and Xianghe have similar volume coarse mode magnitudes and primarily have unimodal distributions. There is a small fine mode contribution (r_eff-fine ~ 0.085 μm) at SACOL indicating pollution from the nearby city of Lanzhou [18]. There is a smaller, yet discernible fine mode contribution at Xianghe (r_eff-fine ~ 0.15) which indicates anthropogenic influences from Beijing and surrounding areas [19]. The coarse mode magnitude at Taihu is nearly half compared to the other sites but the fine mode contribution (r_eff-fine ~ 0.45) is likely due to industrial influences in the vicinity as well as urban influences from Shanghai [9]. The spring case exhibits a strong mineral dust signature at all three sites as there is a more pronounced unimodal coarse mode contribution. The lack of a discernible fine mode peak suggests that mineral dust dominates the size distribution in this case though pollution particles may still be present in the atmosphere.

The wavelength dependences of τ for the winter and spring cases are shown in Figure 4. SACOL has the weakest wavelength dependence of all three sites for both the winter and spring cases indicating coarse mode particles (α_440–870 of 0.29 and 0.11, respectively). This is due to a close proximity to various mineral dust source regions (Taklamakan and Gobi Desert and the Loess Plateau) [7,17]. However, the wavelength dependence for the winter case is stronger suggesting a noticeable more of a fine mode influence than in the spring case. At Xianghe, the stronger wavelength dependence and higher α_440–870 values (0.42–0.48) suggest more of a fine mode influence for both cases than at SACOL. Taihu has the strongest wavelength dependence in the winter case with α_440–870 of 0.62 indicating a mixture of coarse and fine modes due to an abundance of pollution particles resulting from stagnant winter season air patterns that allow for increased aerosol loading (τ₄₄₀ ~ 1.3) [32]. In the spring, there is much less wavelength dependence at Taihu suggesting coarse mode particles dominate the aerosol extinctive properties similar to the SACOL case. It is interesting to note that the α_440–870 value of 0.15 at Taihu illustrates the overwhelmingly large magnitude of dust particles in the spring as compared with the winter case. Since all sites have α_440–870 values less than 0.62, it can be concluded that coarse mode particles dominated the aerosol size properties along with some influences from the fine mode (e.g., pollution).

Figure 5 shows the wavelength dependence of τ_abs. The wavelength dependence is slightly stronger for the winter case than for the spring case. The winter α_abs440–870 value of 1.8 represents a typical value for dust due to the strong absorption in the visible and weak absorption in the near IR [32,33,34], while the spring α_abs440–870 value of 1.6 indicates a mineral dust influence with possible carbonaceous influences [18,32,34]. Taihu has a weaker wavelength dependence than Xianghe and nearly identical to SACOL in the winter case. The lower α_abs440–870 of 1.3 indicates a strong BC influence given by a weaker wavelength dependence across the visible and near IR [32]. It should be noted that OC tends to strengthen the absorption in the visible leading to an even higher α_abs440–870 value [33]. For the spring case, Taihu has the strongest wavelength dependence in both the visible and near IR indicating a robust mineral dust influence (α_abs440–870 ~ 2.5) [32,33,34]. The spring case at SACOL is more indicative of mineral dust influences with α_abs440–870 of 1.7. Table 1 summarizes the AERONET retrieved parameters used in this study. The following section will discuss how well the theoretically calculated ω_o(λ) at different wavelengths agree with AERONET retrievals at the three sites for the selected two cases.

Figure 6 and Figure 7 present the comparisons of ω_o(λ) values at wavelengths of 440–1,020 nm retrieved from three AERONET sites to those calculated by a mineral dust model for the winter and spring cases. In order to demonstrate the wavelength dependences of ω_o to the index of refraction and aerosol particle size, the index of refraction is held constant while the effective radius varies in Figure 6 and vice versa in Figure 7. Although the model results are summarized in Table 2, we focus on the following key discussion points at the SACOL, Xianghe and Taihu sites.

At the SACOL site, the model calculated ω_o(λ) agrees well with observation in the visible but overestimates in the longer wavelengths for the winter case (red) (Figure 6(a)). The spring case has slightly better agreement in the longer wavelengths (675 and 870 nm) but poor agreement otherwise. A possible explanation is due to the fact that two different real indices of refraction (m_r of 1.46 and 1.57 for winter and spring, respectively) are used which decreases the sensitivity of the modeled results because the optical properties of aerosol mixtures are poorly taken into account using these criteria. The Xianghe site has nearly the same indices of refraction for both the winter and spring cases with only one theoretical calculation (Figure 6(b)) where the model calculation and observation agree to within one standard deviation suggesting more sensitivity of a variable effective radius. This is likely due the aerosols having similar optical properties during both dust events (i.e., similar source regions and carbonaceous influences).

At Taihu, there is scarcely any agreement between theory and observation for the winter case suggesting that using a fixed m_i cannot account for the enhanced absorption from the large amount of carbonaceous aerosols present in this region (Figure 6(c)). In contrast, the spring case shows good agreement especially in the near IR due to the strong mineral dust influence that dominates the aerosol absorptive properties during this episode. It is interesting to note that the stronger absorption in the winter case has m_i of nearly five times higher than the spring case (more absorbing pollution aerosols during winter season). In addition to varying the effective radius and holding the index of refraction constant, we now vary the index of refraction and hold the effective radius constant.

When we vary the index of refraction in the model calculations, it is obvious that there is better agreement between the calculations and observations with more variability. This is especially evident at the SACOL and Taihu sites which exhibited different indices of refraction for the winter and spring cases (Figure 7). The calculated ω_o(λ) values at SACOL are still higher than the observed ones for the winter case, however, they are within one standard deviation (Figure 7(a)). There is also better agreement in the spring case as well. The Xianghe winter and spring cases do not drastically change because there are no big differences in particle size and index of refraction between the two cases (Figure 7(b)). At the Taihu site, it is evident that varying the index of refraction better accounts for the wide variety of absorbing aerosols in a mineral dust/carbonaceous mixture in both the winter and spring cases (Figure 7(c)).

To quantitatively estimate the impact of the dust events on the surface and Top-of-Atmosphere (TOA) radiation budgets, we use the AERONET calculated shortwave (SW) aerosol radiative effects (ARE). It is defined as the difference between the net bottom and top of the atmosphere (BOA and TOA) fluxes when dust aerosols are present, as well as the nearby clear-sky days (Fⁱ_BOA and F^h_TOA) and modeled SW fluxes without aerosol input (Fⁱ⁰_BOA and F^h0_TOA), and is given by ΔARE_BOA = Fⁱ_BOA – Fⁱ⁰_BOA (2a) and ΔARE_TOA = F^h_TOA – F^h0_TOA (2b) respectively [24,25]. This is a simulated product that uses a combination of sky radiance measurements and radiative transfer theory [24,25]. The criteria for selecting the clear days are: (1) τ₄₄₀ ≤ 0.3 and (2) clear sky occurrence before and after the dust events.

We note that there is a sharp increase in ARE during the dust events with ARE_BOA > ARE_TOA which suggests a stronger aerosol cooling effect at the surface than at the TOA (Table 3). The strongest ARE is observed at the Taihu site where the ARE values exceed −155 W·m⁻² near the surface and −60 W·m⁻² for the TOA spring case. This is due to dust mixing with carbonaceous aerosols (local pollution) leading to an overall enhanced absorption [14]. This is consistent with the [13], [14], and [15] studies that used Shouxian ARM Mobile Facility (AMF), Zhangye AMF, and AERONET data, respectively.