The Aerosol Atmosphere Boundary-Layer Cloud (A2BC) campaign took place in Hebei Province of the NCP, about 300 km south of Beijing, from May to December 2016. An intensive observation period (IOP) was from May to June when airborne observations were conducted using two airplanes to measure aerosol and cloud properties, respectively. A supersite was established in Xingtai (XT; 114.36∘ E, 37.18∘ N; 182 m above sea level, or a.s.l.). The National Science Foundation supported the project “Air chemistry Research In Asia” (ARIAs), with the goal of integrating in situ observations, satellite remote sensing, and chemical transport models to characterize and quantify tropospheric chemistry and composition over the NCP and to improve modeling tools used to eventually evaluate the effectiveness of air pollutant reduction policies. NASA's Korean US Air Quality (KORUS-AQ) experiment was conducted at roughly the same time in the downwind region. One of the airplanes (Y-12), which was based at Luancheng Airport (LC; 114.59∘ E, 37.91∘ N; 58 m a.s.l.), located in the southeast of Shijiazhuang, the capital of Hebei Province, measured aerosol properties. The airplane was flown to three locations in the area to conduct spirals from ∼ 300 to ∼ 3.5 km, as shown in Fig. 1. These locations include XT, Julu (JL; 115.02∘ E, 37.22∘ N; 30 m a.s.l.), and Quzhou (QZ; 114.96∘ E, 36.76∘ N; 40 m a.s.l.). All four sites are located to the east of the Taihang Mountains, with XT right at the foothill of the mountain range. A total of 11 flights were conducted during the A2BC and ARIAs IOP (Table 1).

Comprehensive measurements of aerosol optical properties were made during the field campaign using the instrumented turboprop Y-12 airplane operated by the Weather Modification Office of Hebei Meteorological Bureau. The typical speed of the aircraft is 60–70 m s-1, with ascent/descent rates of 2–5 m s-1. The aircraft was equipped with multiple aerosol and gas measurement instruments, and was rigorously tested and calibrated during a ground-based campaign to optimize instrument performance. Table 2 summarizes the instruments deployed on the plane and ground.

A full suite of instruments to measure aerosol and meteorological parameters was deployed at the Xingtai surface supersite. Those of concern to this study include aerosol optical remote sensing observations (CIMEL radiometer), in situ particle light extinction measurements (cavity attenuated phase shift spectroscopy (CAPS) particle light extinction monitor), and radiosondes for meteorological parameters.

Ground-based measurements collected at XT were matched in time and space with airborne measurements over the site. Figure 2 shows the vertical profiles of RH and T from the sounding done at 05:55 coordinated universal time (UTC, 13:55 local time) on 8 May 2016 and from aircraft measurements made over the site before the radiosonde launch (04:40–05:40 UTC, one descent spiral and two ascent spirals). Both RH and T derived from sounding data and measured from the aircraft follow the same general trends, with inversions near 800 and 2400 m.

For a final evaluation of the performance of the airborne instruments in measuring ambient aerosol characteristics, the σext measured by the nephelometer/PSAP on board the aircraft and by the ground-based CAPS PMex during research flight (RF) 6 (RF6; 21 May), RF7 (28 May), RF8 (28 May), and RF11 (11 June) is shown in Fig. 3. CAPS-measured σext collected during the spiralling part of the flight was averaged and all σext data were adjusted to a wavelength of 550 nm. In general, mean values of surface σext were consistent with (∼ 12 % smaller than) the corresponding aircraft-measured σext located at the bottom of the profiles.

There is an expected comparability of retrieval parameters (e.g., SSA) between ground-based observations and aircraft measurements. For example, the SSA values, derived from AERONET in Baltimore–Washington, DC, United States, were on average 0.011 lower than the values derived from aircraft profile measurements (Schafer et al., 2014). For aircraft–AERONET (based on the nephelometer combined with the PSAP, also used in this study) comparisons with ground-based AERONET at multiple sites over southern Africa, the SSA with a mean difference of 0.01 (rms: 0.03) in biomass burning aerosol measurements (Leahy et al., 2007). There is also a small difference between aircraft in situ (0.87) and CIMEL (0.85) measurements of SSA on a day with both an upper-level smoke aerosol layer and low-level dust layer in the Banizoumbou (Johnson et al., 2010). For this study, retrievals from a ground-based CIMEL automatic Sun-sky radiometer (CE-318) were compared with aircraft in situ measurements.

We calculated the columnar SSA from airborne measurements and its uncertainties following Leahy et al. (2007) and Schafer et al. (2014). The sampled SSA (ωsample) was calculated from σsca and σabs measured by a nephelometer and a PSAP, respectively. We assumed that in situ SSA measured between the minimum and maximum flight levels represents the entire column. In order to produce a column mean SSA (ωcolumn) and compare it with that retrieved from AERONET data, the SSA values were averaged for the duration of the profile after weighting the values according to aerosol loading. The columnar SSA was calculated with the same method for different types of aerosol layers in the vertical. Considering the vertical distributions of σsca and σabs, the SSA measured at higher altitudes (lower aerosol loading) is substantially lower than in the lower troposphere or the aerosol enrichment layer. Thus, the weighted mean is better than a simple average of ωsample, which would overestimate the absorption features of aerosol that has a negligible effect on radiation at the surface where the CIMEL radiometer is located.

Aerosol loading, denoted by σsca values, and the profiles were limited to samples collected at the lower level below 400 m and at the higher level greater than 2000 m to adequately represent the entire column. For every profile, weighting factors for each SSA measurement were generated that corresponded to the normalized magnitude of σsca. The measured SSA values were scaled proportionally to the aerosol loading at the altitude of the observation as in the following equation: ωcolumn=∑i=0Nσscaσsca(profile_mean)∗ωsampleN, where N equals the number of ωsample in the profile.

The SSA at 550 nm retrieved from the CIMEL (0.93) agreed well with airborne measurements (0.94) on 28 May (Table 3). The well-matched vertical profiles and ground-based data present a more complete profile of aerosol optical information in the lower atmosphere, demonstrating the high credibility and quality of airborne measurements. These SSA values obtained in the NCP are lower than those observed in Africa, and in the northeastern United States. The reason for the difference is probably due to different types of aerosols in these different locations: primarily mineral dust aerosols in Africa and secondary aerosols in the northeastern United States, and a mix of these two types of aerosols in the NCP.

A statistical summary of aerosol optical properties retrieved from aircraft measurements over the four target areas (LC, JL, QZ, and XT) is given in Table 4. Aerosol scattering coefficients (σsca) and backscattering coefficients (σbsca) are directly obtained from the nephelometer. The truncation correction was applied and the σsca values were corrected for the ambient RH. The mean σsca values at 550 nm measured at LC, JL, QZ, and XT were 57.10 ± 67.71, 95.57 ± 106.61, 87.00 ± 100.43, and 75.30 ± 84.58 Mm-1, respectively, and the mean σbsca value in each target area was 7.67 ± 7.6, 9.91 ± 9.19, 11.12 ± 10.51, and 9.21 ± 8.6 Mm-1, respectively. Aerosol absorption properties at LC and XT were stronger than those at JL and QZ. Note that each variable has a large standard deviation, and there were large variations associated with these aerosol optical properties.

Mean vertical distributions of σsca and σabs at 550 nm derived from aircraft measurements made over the four target areas are shown in rows 3 and 4 of Fig. 4. In general, the values of σsca and σabs decrease with altitude. Peaks in the σsca profiles at certain heights are seen at JL, QZ, and XT. The σsca profile at LC changes little below 3000 m. Compared with the other target areas, σabs values at JL are relatively lower near the ground and gradually decrease with height, which suggests that there were fewer light-absorbing and more scattering aerosols in this area.

Based on measurements of aerosol scattering and absorption coefficients, the aerosol extinction coefficient (σext, defined as the sum of σsca and σabs) and SSA (ω, defined as the ratio of σsca to σext) was calculated at 550 nm following Eq. (5): ωλ=σsca,λσext,λ=σsca,λσsca,λ+σabs,λ. The small mismatch in calculating ω was corrected by linearly extrapolating the σabs values to the wavelengths of the scattering measurements as defined. Both σext and ω are considered to be primary determinants for the direct radiative effect of aerosols. The vertical profile of σext closely follows that of aerosol mass concentration (Kim et al., 2015). SSA was described as an important factor in controlling whether an aerosol layer has a cooling or warming effect; it can also be used for studying the radiative forcing at the top of the atmosphere (Bergstrom and Russell, 1999). In this study, the ω550 in the range 0.68–0.99 with a mean value of 0.85 indicates the importance of absorbing aerosols.

The α was calculated using Eq. (6) where σsca,λ is the scattering coefficient at a given reference wavelength λ:

α(λ2/λ1)=-log⁡σsca,λ2/σsca,λ1log⁡λ2/λ1. In this study, we used two wavelengths, λ1=450 nm and λ2=700 nm.

An analogous expression can be written for the wavelength dependence of absorption. In general, α is a basic measure of the aerosol size distribution. It is related to the ratio of fine to coarse particles, with α less than 1 for coarse-mode aerosol such as dust particles and α ∼ 2.0 for very fine-mode aerosol such as smoke particles (Hamonou et al., 1999). Vertical profiles of the median values of ω550 and α are shown in rows 5 and 6 of Fig. 4. The profiles were generated by calculating the median value at each altitude layer from all the measured profiles. The variations seen may reflect changes in the origin of aerosol particles and transport routes (Léon et al., 2009), although the values fall well within the range of the standard deviations.

Angular-corrected data from the nephelometer is the scattered light intensity in the backward hemisphere of the particle (90–180∘). The backscattering fraction, βsca, is the ratio of the backscattering coefficient over the total scattering coefficient at a given wavelength (Garland et al., 2009). The ratio of forward scattering to backscattering varies with the particle size parameter and reflects the angular characteristics of particle scattering and the proportion of fine particles (diameters < 2 µm). During the field campaign, βsca remained at low values below 2000 m. Mean values of βsca measured in the four target areas were 0.13 ± 0.003, 0.11 ± 0.005, 0.12 ± 0.003, and 0.12 ± 0.006. Fluctuations above 2000 m suggest that particle sizes changed quickly, due presumably to different air masses transported from different directions.

During the experiment period, the majority of aerosol layers were well characterized by the sampled vertical profiles, and most aerosols resided below the maximum flight levels. Mie theory was applied to calculate the extinction profiles and the AOD and to estimate the impact of different aerosol vertical distributions on these optical properties. The AOD was calculated by integrating the extinction coefficient over height as AOD(z1∼z2)=∫Z1Z2σext(z)dz, where σsca(z) is the extinction coefficient at a height of z and Z2 is above most of the aerosol. Figure 5 shows the mean AOD at 550 nm for different altitude ranges. The regional mean AOD in each altitude range (< 1, 1–2, and 2–3 km) was 0.10 ± 0.08, 0.10 ± 0.07, and 0.03 ± 0.03, respectively. Standard deviations are greater than 50 % of the mean, suggesting that AODs varied greatly within an altitude range. The magnitude of AOD generally decreased with altitude. The effect of the transport of atmospheric pollutants is evident at QZ, where the largest AOD values were found in the 1–2 km layer instead of at the surface.

Generally driven by the weather conditions at flight time, a clean PBL is defined by a mean value of σsca less than 100 Mm-1 in every 100 m layer (except that adjacent to the surface layer). Figure 6a and c show the vertical profiles of σsca and ambient RH, retrieved from airborne measurements made under clean PBL conditions. Values of σsca near the surface ranged from 11.7 to 84.5 Mm-1, with an average value of 48.5 Mm-1. In general, σsca slowly decreased with height, which suggests that the atmosphere was relatively clean, with no distinct aerosol layer identified. The mean σsca profile decreases approximately exponentially with height, expressed as σsca,H=σsca,0⋅exp⁡(-(H-HRS)/Hp),(if H>HRS)σsca,0(if H≤HRS), where σsca,0 is σsca measured at the surface, H is the altitude above sea level, and Hp is the aerosol scale height (Hp represents the height where σsca is reduced to 1/e of its surface value); HRS represents a relative stable layer near the surface where the vertical variation of σsca was insignificant. In the cases of a clean PBL, σsca,0=124 Mm-1, Hp=1146 m, and HRS=837 m. A linear regression analysis shows the correlation coefficient, r2=0.96. The ambient RH under clean PBL conditions was divided into two groups: dry (RH: 27.4–36.8 %) and humid (RH: 53.1–83.6 %). Under dry conditions, σsca and ambient RH showed a good correlation (R2=0.62, Fig. 8a), while under humid conditions, the correlation was low (R2=0.23). To understand the sources and transport pathways of aerosols over the target areas during the field campaign, we calculated isentropic air mass back trajectories for 72 h using the NOAA Hybrid Single Particle Lagrangian Integrated Trajectory (HYSPLIT) model (Draxler and Hess, 1997; Stein et al., 2016) at 0.5, 1.5, and 2.5 km above mean sea level. The HYSPLIT model (http://ready.arl.noaa.gov/HYSPLIT.php, last access: 6 June 2018) was used along with the National Center for Environmental Prediction's Global Data Assimilation System 1∘ × 1∘ meteorological database to calculate backward trajectories terminated at the Xingtai supersite. Figure 9a shows 72 h air mass back trajectories under clean PBL conditions during the field campaign. Air masses most commonly originated from the northwestern region of the study area. Some of the trajectories can be traced back as far as Mongolia and Siberia, passing over the arid areas to the west of southern Hebei. Some clean and moist air masses originating from the less polluted southern and local areas were also seen.

The vertical distribution of air pollutants varied greatly from case to case due to a variety of influences. One of the crucial factors was PBL structure, which determines the vertical profile of σsca; see Fig. 6b. The corresponding ambient RH profiles (Fig. 6d), σsca as a function of RH (Eq. 9), and the back-trajectory analysis (Fig. 9b) are also shown. The PBL height is determined by the shapes of σsca vertical profiles. When pollution is confined primarily to the lower layer of the troposphere, the magnitude of σsca increased slightly with height up to an altitude where σsca decreased sharply. In this study, the mean rate of decrease was about 0.81 Mm-1 m-1. We defined the bottom of this layer as PBL height (HPBL). We normalized the curves by the shapes of σsca vertical profiles, and the PBL heights of different σsca and ambient RH profiles were adjusted to the same level to show the similarity of the shapes. The PBL heights, as determined by vertical profiles of σsca, ranged from 900 to 2000 m with an average value of ∼ 1400 m. The average σsca profile was determined as follows: σsca,H=σsca,PBL⋅exp⁡(-(H-HPBL)/Hp),(if H>HPBL)σsca,0+k⋅H,(if H≤HPBL), where HPBL is the normalized altitude of PBL height, Hp is the aerosol scale height in the free troposphere, and k is the changing rate of σsca in the PBL. In these cases, σsca,0=171 Mm-1, Hp=216 m, k = 0.03 Mm-1 m-1, and r2=0.9394. Figure 6d shows the ambient RH profiles under dry and humid conditions. The shapes of dry and humid RH profiles were similar in the PBL, but at the top of the PBL, the RH_dry profiles decreased dramatically, and the RH_humid profiles changed only slightly. Linear fits were made to determine the correlation between RH and σsca. Under dry conditions, there was a pronounced correlation (r2=0.95) between RH_dry and σsca profiles. But under humid conditions, the correlation coefficient was only 0.12, suggesting little impact of RH on σsca.

Most back trajectories under polluted conditions originated from the heavily populated and industrialized interior and coastal areas south of the study area (Fig. 9b). Some drier air masses (corresponding to RH_dry) were traced back to southern and local areas just as in the case of clean air masses. Moist air masses originated from the clean marine atmosphere to the southeast, then passed over the densely populated eastern/southeastern regions in the free troposphere before reaching the observation site. This analysis of both dry and moist air masses with aerosols in the lowest layer of the PBL shows that heavy local/regional pollution dominated during the field campaign and that the long-range transport of aerosols was less significant.

In addition to high aerosol concentrations in the lower PBL, upper-layer (referred to as the type A case, an example of which occurred on 2 June 2016) and multi-layer (referred to as the type B case, an example of which occurred on 6 June 2016) aerosol vertical distributions were also observed. In the type A case (Fig. 7a), the vertical profile of aerosol σsca retrieved from airborne measurements shows a mean σsca near the surface of about 45 Mm-1 and little variation below 1 km. A sharp increase in σsca was observed above with a peak value of 200 Mm-1 occurring around 2–3 km, and aloft the σsca profile decreased exponentially with height. For the type B case (Fig. 7b), mean σsca varied greatly with height, which suggests that multiple layers of aerosol particles were present in the PBL. The vertically inhomogeneous distribution of σsca suggests that aerosol particles in the PBL might be significantly affected by the long-range transport of air pollutants or local emissions in the study area. The profiles of σsca and ambient RH for type A and type B cases have similar shapes with a correlation coefficient of R2 = 0.91 and 0.59, respectively (Fig. 8c and d).

As shown in Fig. 9c, back trajectories for the type A case originated from less polluted regions in northeastern China which then moved toward the marine atmosphere over Bohai Bay and the densely populated region east of the study area. The aerosol enrichment in the upper layer of the PBL was probably due to regional transport and mixture of anthropogenic and sea-salt aerosols into the free troposphere. The back trajectories ending at 0.5, 1.5, and 2.5 km over the observation site for the type B case (Fig. 9d) show that air masses originated from the clean marine environment to the southeast, from the polluted environment to the southwest, and from the eastern coastal region. This could explain the strata, i.e., the multi-layered vertical distribution, of aerosols in the PBL.