2.1 Aircraft observations and ancillary data descriptions

Vertical distributions of aerosols and MBL cloud microphysical properties over the Azores were obtained during two intensive operational periods (IOPs) of the ACE-ENA, i.e., early summer 2017 (late June to July) and winter 2018 (January to February). Since the aerosol concentration and variability are much larger in the summertime of Azores, we mainly focus on July 2017 in this study. The Gulfstream-159 (G-1) of the ARM Aerial Facility (AAF) provides accurate measurements of aerosol size distribution, total aerosol number concentration, and chemical constituents below and above cloud layers during the summer IOP. The condensation particle counter (CPC) on board the G-1 can detect aerosol particles larger than 10 nm, and it can provide profiles of condensation nuclei number concentration (N_CN) when the aircraft ascends or descends. Note that N_CN measurements inside clouds can be contaminated and thus have large uncertainty. CCN number concentration (N_CCN) is obtained by the CCN-200 particle counter on board the G-1 aircraft. The N_CCN is a measurement under the controlled supersaturation of 0.35 % with a humidified particle size range from 0.75 to 10 µm (Rose et al., 2008). We analyze sulfate and organic-carbon (OC) mass concentrations measured by the Aerodyne High-Resolution Time-of-Flight Aerosol Mass Spectrometer (HR-ToF-AMS) and refractory black carbon (BC) from the Single Particle Soot Photometer (SP2). The accuracy of each individual instrument can be found in the instrument handbooks available on the ARM website.

We use cloud and drizzle microphysical property profiles retrieved from a combination of ground-based observations including a Ka-band ARM zenith radar, ceilometer, and microwave radiometer. The Fast Cloud Droplet Probe (Glienke and Mei, 2020) measured cloud droplet properties (diameter between 1.5 and 46 µm), and the Two-Dimensional Stereo Probe (2DS, Glienke and Mei, 2019) measured drizzle properties (diameter greater than 45 µm) that were used to evaluate the ground-based retrievals. Following Dong et al. (1998) and Frisch et al. (1998), cloud droplet size distribution was assumed as a lognormal distribution. Differently, drizzle size distribution was assumed as a normalized Gamma distribution, as suggested by O'Connor et al. (2005) and Ulbrich (1983). The retrieved cloud and drizzle properties were validated against the collocated aircraft in situ measurements during ACE-ENA (Wu et al., 2020). Both the time series and vertical profiles from the retrievals agree well with in situ observations. Treating the aircraft measurements as ground truth, the median retrieval uncertainties are estimated as ∼20 % for cloud droplet effective radius and ∼30 % for cloud droplet number concentration, liquid water content (LWC), and drizzle drop median radius.

To characterize long-range aerosol intrusions over a monthly timescale, we employ global aerosol reanalysis data, namely the Copernicus Atmosphere Monitoring Service (CAMS). It provides four-dimensional mass concentrations of aerosols and reactive gases with a horizontal spatial resolution of approximately 80 km and 60 vertical levels. The CAMS reanalysis was constructed by assimilating several satellite products of the atmospheric constituents into a global model and data assimilation system (Flemming et al., 2017). The assimilated satellite datasets include aerosol optical depth (AOD) from the Moderate Resolution Imaging Spectroradiometer (MODIS) and Advanced Along-Track Scanning Radiometer (AATSR), CO from Measurements Of Pollution In The Troposphere (MOPITT), and NO₂ and O₃ from the Ozone Monitoring Instrument (OMI), Global Ozone Monitering Experiment (GOME), etc.

2.2 Model description

The WRF model version 3.6 is employed in this study to simulate MBL clouds and their possible interactions with transported aerosols. Four nested domains are set up with horizontal resolutions of 19.2 km, 4.8 km, 1.2 km, and 300 m (Fig. 5). Even for the innermost domain, we try to cover as large of an area as possible, considering the highly heterogeneous meteorological conditions in the midlatitudes. The innermost domain is configured in a similar way with large-eddy simulations, and it uses the three-dimensional Smagorinsky first-order closure for eddy coefficient computation. Boundary layer parameterization is turned off for this domain. Note that 300 m horizontal resolution does not strictly meet the classic large-eddy simulation (LES) requirement, but recent simulations with similar resolutions successfully reproduced the structure and drizzle onset of MBL clouds (Wang and Feingold, 2009) and were used to study boundary layer cloud interactions with aerosols (Lin et al., 2016). A total of 65 stretched sigma levels are used with a 40 m vertical resolution within the MBL. The large-scale forcing is adopted from the ERA5 reanalysis data with 25 km horizontal resolution (European Centre for Medium-Range Weather Forecasts, 2017).

To accurately depict MBL cloud microphysical processes, a spectral bin microphysical (SBM) scheme is employed, which utilizes a pair of 33 bins to represent cloud or rain drops and aerosols separately without prescribed size distributions (Fan et al., 2012; Wang et al., 2013). Aerosol activation is explicitly calculated using the model-predicted water vapor supersaturation. The Kölher theory is used to calculate the critical radius. The hygroscopicity of sulfate is assumed for aerosols in each size bin. At each time step, aerosols with a radius greater than the critical radius are removed from the aerosol spectrum, and the mass of the activated droplets is added to the cloud spectrum. Aerosol regeneration from complete evaporation of droplets and/or raindrops is also considered in the SBM. Since the aerosol size distribution in the SBM ranges from a few nanometers to a few micrometers, the definition of aerosol in the model is closer to the condensation nuclei in the aircraft observation. Hence, observed vertical profiles of N_CN from selected cases are used for the initial and lateral boundary conditions of aerosols in the model. The model integrates from 12:00 UTC on the day before the selected case, and the first 12 h are considered as spin-up. Shortwave- and longwave-radiation-transfer calculations are accounted for by the Goddard and Rapid Radiation Transfer Model (RRTM) schemes, respectively. The radiative effect of aerosols above the cloud decks is not considered in the present model setup. We speculate that such an effect is small because of rather low aerosol optical depth over this remote region, even with the long-range-transported aerosols (aside from thick dust plumes from the Saharan Desert).

3.1 Characterization of aerosol vertical distribution using the CAMS reanalysis

Previous studies have shown that the CAMS aerosol product exhibits good agreement with ground-based observations such as AERONET and unassimilated satellite products such as the Multi-angle Imaging SpectroRadiometer (MISR) on a global scale (Christophe et al., 2019). The global spatial correlation of CAMS AOD with AERONET is about 0.83, and the bias in CAMS AOD seasonal variation is between −10 % and +20 %. Here we utilize this dataset to characterize the aerosol vertical distribution over the northeast Atlantic during the ACE-ENA field campaign. Vertical distributions and their temporal evolutions for five types of aerosols, including sulfate, organic carbon (OC), black carbon (BC), sea salt, and dust, over the whole month of July 2017 are displayed in Fig. 1 based on the CAMS aerosol reanalysis. Sulfate, OC, and BC are the predominant aerosol types possibly possessing an anthropogenic signature. BC and OC can also originate from biomass burning. Those aerosols share a similar spatiotemporal pattern in the free troposphere, indicating that they undergo similar long-range transport before arriving over the Azores islands. Marked and persistent low-altitude (1–2 km) pollution transport occurred between 1 and 13 July, as shown in the evolution of vertical profiles of sulfate, OC, and BC (Fig. 1a–c). High-altitude (3–6 km) pollution transport occurred between 6 and 20 July for those three aerosol types as well. Both modes of pollution transport occurred 50 % of the time during July 2017, indicating a high frequency of long-range transport over this area. The concentrations of OC, BC, and sulfate are generally low in the MBL, so aerosol penetration from the free troposphere into the lower MBL may not be significant during this month. One exception is sulfate during 18–21 July. Sulfate concentration experienced an increase in the MBL followed by a lag increase in the free troposphere. Since there is no significant transport signal before and during that time period, the elevated sulfate concentration within the boundary layer is likely due to some local sources such as oxidation of marine dimethyl sulfate (DMS). Typical high- and low-plume cases are represented by 18 and 12 July, respectively, so they will be investigated thoroughly in the later aircraft data analyses and model simulations.

Figure 1Temporal evolutions of vertical distributions for five types of aerosols as shown in (a) sulfate, (b) organic carbon, (c) black carbon, (d) sea salt, and (e) dust during July 2017 over the Azores based on the ECMWF-CAMS aerosol reanalysis product.

Download

The aerosols of natural sources, namely sea salt and dust, show different vertical distributions (Fig. 1d–e). Sea salt aerosols mainly reside near the surface and are rarely found above 1000 m. Dust particles are mainly found at high altitudes, typically above 3 km, during 5–14 July, indicating their long-range transport. However, the spatiotemporal pattern of dust in the free troposphere is quite distinctive from sulfate and smoke, implying different sources of long-range transport. Previous studies suggest the possible dust transport from the Saharan Desert to the northeast Atlantic region (Logan et al., 2014; Weinzierl et al., 2017). To address those issues, back-trajectory analyses were conducted, and the results are discussed later. During 15–19 July, dust particles are found within the boundary layer and even near the surface following the presence of a dust plume in the free troposphere earlier. Such a downward propagation does not occur for anthropogenic aerosols however, likely explained by the fact that dust particles are bigger in size with larger settling velocity.

3.2 Identification of source regions using back-trajectory analysis

The backward ensemble trajectories were computed using the NOAA Hybrid Single-Particle Lagrangian Integrated Trajectory (HYSPLIT; Stein et al., 2015) model based on the large-scale meteorological fields from the Global Data Assimilation System (GDAS) with a spatial resolution of 0.5^∘. We focus on three cases over 3 d to examine the sources of typical high- and low-altitude plumes of anthropogenic aerosols and mineral dust. The model uses an endpoint height of 1.5, 2.4, and 3 km for three selected cases to represent the air parcels in the anthropogenic low altitude, high altitude, and fine-mode dust plumes, respectively. To capture the different lengths of transport procedure, the model was backward-integrated for 7 d for the anthropogenic aerosol cases and 13 d for the mineral dust one. A total of 20 ensemble members are employed for each case. They agree with each other better on horizontal trajectory than vertical displacement. Larger differences are found among the ensemble members after 3 d for anthropogenic aerosols and after 2 d for dust.

The back-trajectory analyses confirm that the source region of sulfate, BC, and OC in the plumes is the North American continent (Fig. 2a, c), consistent with previous analyses of data from the earlier field campaign over the ENA site (Logan et al., 2014). The westerly jet carries the pollutants across the Atlantic Ocean, and it takes 3 to 4 d to arrive at the Azores. Temporal evolutions of trajectory vertical displacement reveal when aerosols are elevated from the planetary boundary layer (PBL) to the free troposphere, and such information can be used to pinpoint the aerosol source. Figure 2b and d suggest that aerosols are mainly from the central US in the high-plume case and from eastern US in the low-plume case. The curved trajectories in the low-plume case reflect the influence of the Bermuda or Azores High located to the south. The dust transports exhibit a much different pathway. Starting at 3 km altitude, the back-trajectory develops westward initially but sharply turns around and points to the North Africa (Fig. 2e, f). It suggests that the Sahara is the most likely source for the dust particles observed over the Azores.

Figure 2Back-trajectory analyses of air mass history starting from the ENA site for the three selected cases using the NOAA HYSPLIT trajectory model. Anthropogenic-aerosol-dominated plume with high altitude (Anthro_High_Alt) and low altitude (Anthro_Low_Alt), dust plume (Dust).

Note that back-trajectory analysis of air mass has its own limitations. For example, shipping emissions over the northern Atlantic Ocean are not considered in the present analysis. Also, the source attribution based on episodic events may be not representative of the climatological mean scenario. Therefore, the source attribution results here need to be further evaluated in future studies, which can utilize a more sophisticated approach such as source tagging in a global climate model (GCM), nudged by the reanalysis data (Wang et al., 2014).

3.3 Vertical distributions of different aerosols in aircraft observations

Aircraft observations during the ACE-ENA provide more accurate depictions of aerosol vertical distribution and aerosol layer heights relative to cloud layer heights, with differentiation of aerosol type and hygroscopicity. During the summer IOP, quite diverse aerosol vertical profiles are found. Here we focus on those with noticeable aerosol plumes in the free troposphere. Figure 3 shows two representative vertical distributions of aerosol mass concentrations averaged over the flights on 18 and 12 July, corresponding to the high- and low-altitude aerosol plumes, respectively. In the high-altitude-plume case, BC, OC, and sulfate concentrations all increase with height above clouds, indicating downward propagation of aerosol plumes and possible interaction with MBL clouds. BC and OC concentrations are even higher than that of sulfate in the free troposphere, suggesting the biomass-burning signature of the plume on that day. Conversely, within the MBL, there is a much higher concentration of sulfate in the MBL than those of BC and OC. This phenomenon is also captured by the CAMS aerosol reanalysis (Fig. 1a), lending support to the fidelity of the reanalysis dataset. For the low-altitude plume (Fig. 3b), the vertical gradients of aerosol concentrations are not clear above clouds. Meanwhile, the aerosol concentrations within 500 m right above clouds are higher than those near the cloud base (Fig. 3b), corroborating the physical contact between aerosol plumes and MBL clouds. Comparing Figs. 3 and 1, the CAMS reanalysis data generally agree with aircraft-observed aerosol profiles on the selected days, but the predicted aerosol mass mixing ratios are an order of magnitude higher in the reanalysis data. Those discrepancies point out that any quantitative usage of aerosol reanalysis product should be done with caution.

Figure 3Airborne-measured vertical profiles of sulfate (SO₄; red dots), organic carbon (OC; green dots), and refractory BC (rBC; black dots) mass mixing ratios averaged over multiple flights in two characteristic cases: (a) high-altitude aerosol plume on 18 July and (b) low-altitude aerosol plume on 12 July 2017. The highly uncertain and noisy aerosol observations due to cloud contamination are not shown (between two dashed lines), so the blank regions approximately denote cloud layer.

Download

Aerosol and CCN concentration vertical profiles are also available from the aircraft observations. For the high-altitude plume, N_CN reaches a peak of ∼600 cm⁻³ at 2.5 km and then decreases dramatically downwards to ∼180 cm⁻³ near the cloud top (∼1.1 km), which is even lower than N_CN values within the boundary layer, ranging from 200 to 300 cm⁻³ on that day (Fig. 4a). The measured 200 m average of N_CN above the cloud top is 185 cm⁻³, smaller than that below the cloud base, 290 cm⁻³ (Table 1). From the surface to a height of 2.5 km, the minimum N_CN occurs near the cloud top, reflecting the disconnection between MBL aerosols and those from long-range transport aloft. The characteristics of the N_CCN profile are similar to those of N_CN. In the low-altitude plume, both N_CN and N_CCN show a slower decline above the cloud layer (Fig. 4c, d). Also, the right-above-cloud-top N_CN and N_CCN at 1 km are higher than those below the cloud layer, indicating the physical contact of the aerosol plume with the cloud deck.

Figure 4Airborne-measured profiles of condensation nuclei (N_CN; blue) and cloud condensation nuclei (N_CCN; red) averaged over multiple flights in two cases with high- and low-altitude aerosol plumes. The highly uncertain and noisy aerosol observations within the cloud layer are not shown (between two dashed lines).

Download

Table 1Characteristics of condensation nuclei concentration and cloud vertical profiles for all eight cases during the summer phase of the DOE ACE-ENA field campaign.

¹ Average within 200 m above (below) cloud top (base). ² For continuous cloud layer. ³ Critical altitude is defined as the height at which above-cloud N_CN is equal to the below-cloud N_CN. N/A indicates not available.

Download Print Version | Download XLSX

During the summer IOP, the aircraft was deployed over 20 d to collect data. Among those days, only 8 of them have stable MBL clouds during the flight hours, according to the ground-based cloud radar. We summarize the aircraft-observed aerosol and cloud vertical distribution characterizations for those 8 d and eight cases in Table 1. Among those eight cases, 5 d show an increase in above-cloud N_CN along with height, and 1 d shows roughly constant N_CN above clouds, all of which indicate the existence of long-range transport of aerosols in the free troposphere and downward-propagating influence on the aerosol budget near the cloud top. Moreover, five out of the eight cases have above-cloud N_CN (within 200 m) significantly larger than below-cloud N_CN, implying the potential influence of free-troposphere aerosols on MBL clouds from another angle of view.