Assessing Radiative Impacts of African Smoke Aerosols Over the Southeastern Atlantic Ocean

Abstract

Biomass burning smoke aerosols are efficient at attenuating incoming solar radiation. The Layered Atlantic Smoke Interactions with Clouds campaign was conducted from June 2016 to October 2017. The U. S. Department of Energy mobile Atmospheric Radiation Measurement site located on Ascension Island (AMF-ASI) identified several instances of smoke plume intrusions. Increases in surface and column measurements of aerosol loading were directly related to increases in fine mode fraction, number concentrations of aerosols (N_a), and cloud condensation nuclei (N_CCN). During periods of weak lower tropospheric stability, smoke particles were more likely to be advected downward either by boundary layer turbulence or cloud top entrainment under non-overcast sky conditions. Backward trajectory analysis illustrated that smoke aerosols reaching the AMF-ASI site were fine mode, less aged, strongly absorbing, and had shorter boundary layer trajectories while longer boundary layer trajectories denoted mixtures of weakly absorbing smoke and coarse mode marine aerosols. The most polluted smoke cases of August 2016 and 2017 revealed a notable contrast in radiative forcing per unit aerosol optical depth or radiative forcing efficiency (ΔF^eff) at the top of the atmosphere (TOA) and near-surface (BOA). The weakly (strongly) absorbing 2016 cases exhibited weaker (stronger) ΔF^eff at the TOA and BOA suggesting a warming (cooling) effect within the boundary layer. The 2017 cases featured the strongest ΔF^eff suggesting more of a cooling effect at the TOA and BOA due to mixing of fresh smoke with marine aerosols during transport.

Key Points

• African biomass burning smoke plumes feature highly variable top of the atmosphere and BOA radiative forcing efficiencies when mixed with marine aerosols

• Smoke aerosols with stronger radiative forcing efficiencies present a cooling effect which can stabilize the marine boundary layer

• Longer back trajectories suggest aging of smoke aerosols and mixing with marine aerosols which facilitate cloud development

Plain Language Summary

Fires are often observed in Central Africa throughout the year due to agricultural and natural processes. As a result of changes in pressure patterns along with wind direction and strength, the smoke aerosols can transport to regions thousands of kilometers away from the fires. The aerosols can aid in cloud development while at the same time, have variable impacts on the amount of solar radiation reaching the surface and returning to space which can cause either warming or cooling along their transport pathway. In the Southern Hemisphere tropics, there are few observations to document aerosol radiative forcing impacts on global climate. This study uses short-term and long-term measurements of aerosol physical properties in conjunction with meteorological analysis to diagnose the impact of smoke aerosols on incoming solar radiation and cloud development during transport.

1 Introduction

Aerosols have a direct impact on the radiative properties of the atmosphere such that fewer solar photons will reach the Earth's surface. This influences not only the Earth's surface temperature but also the amount of energy needed to drive numerous atmospheric dynamic processes (Pan et al., 2020; Tian et al., 2017). Carbonaceous aerosols derived from biomass burning via natural and anthropogenic activities have extensive global source regions (Che et al., 2022). Numerous studies have examined the relationship between biomass burning aerosols and their ability to readily activate as cloud condensation nuclei (CCN) with respect to aerosol transport and regional/seasonal variability (Gallo et al., 2023; Logan et al., 2020; McCoy et al., 2021; Redemann et al., 2021; Zheng et al., 2022). In addition, persistent high pressure systems facilitate large scale subsidence and drought. These conditions can lead to an increase in wildfire activity which is conducive to carbonaceous aerosol generation and will likely continue to increase in the near term (Hallar et al., 2017; Logan et al., 2020). It should be noted that cloud development is impacted because carbonaceous smoke CCN typically auto-convert into small cloud droplets which compete for available water vapor in strongly polluted environments (Adebiyi & Zuidema, 2018; Kacarab et al., 2020; Zheng et al., 2020).

In the Southern Hemisphere, extensive regions in Central Africa are the primary sources of nearly year-round biomass burning due to agricultural and wildfire activity, most notably within the sub-Sahel and Congo jungle terrains (Kacarab et al., 2020; Marquardt Collow et al., 2020; Redemann et al., 2021; Ryoo et al., 2021, 2022). Large scale mid-tropospheric (e.g., 600 hPa) wind patterns advect smoke aerosols westward over the southern Atlantic Ocean during the major burn season months of June-October (Adebiyi & Zuidema, 2016; Chaboureau et al., 2022; Zhang & Zuidema, 2019). Several multi-national and multi-agency efforts have investigated the Southern Hemisphere with respect to aerosols, clouds, precipitation, and meteorological patterns during the 2016–2017 period. A recent field campaign, Layered Atlantic Smoke Interactions with Clouds (LASIC), featured the deployment of a U. S. Department of Energy Atmospheric Radiation Measurement (ARM) mobile facility (AMF) along with surface, satellite, and aircraft measurements (Kacarab et al., 2020; Marquardt Collow et al., 2020; Ryoo et al., 2021, 2022; Zuidema et al., 2016).

Several studies have been published regarding LASIC and focused primarily on continental and marine aerosols, cloud development, and biomass burning smoke plume physico-chemical properties (Barrett et al., 2022; Che et al., 2022; Dedrick et al., 2022; Kuete et al., 2020; Marquardt Collow et al., 2020; Ryoo et al., 2021, 2022; Sedlacek et al., 2022; Zhang & Zuidema, 2019, 2021; Zuidema et al., 2016, 2018). This study seeks to relate the short-term variability of biomass burning smoke plume radiative properties to their impacts on near-surface and upper atmosphere radiative forcing and marine boundary layer cloud development. This addresses a continuing need to quantify how smoke aerosol radiative effects evolve during transport within the southeast Atlantic Ocean region (Marquardt Collow et al., 2020; Zhang & Zuidema, 2021). The following science questions will be addressed:

• SQ1: What roles do meteorology (e.g., synoptic scale wind patterns and aerosol transport), climatology, and extreme biomass burning events play in the observed aerosol-CCN relationships in the tropical Southern Hemisphere?

• SQ2: What observable direct radiative impacts do biomass burning aerosol layers exert on the atmosphere at the near-surface (BOA) and top of the atmosphere (TOA)? How do these radiative impacts change from year to year?

An examination of the short-term and long-term regional/seasonal influences of biomass burning aerosols on CCN is performed using ground-based data from the ARM mobile facility located on Ascension Island (AMF-ASI) in addition to ancillary meteorology and long-term average reanalysis data over the study domain which includes the Southern Atlantic Ocean and Central Africa regions. Both the 2016 and 2017 burn seasons are compared with one another and against a 21-year (2000–2020) regional long-term average (Kuete et al., 2020). In addition, the efficiency with which the biomass burning aerosols attenuate BOA and TOA shortwave radiation for both burn seasons is analyzed using a collocated Aerosol Robotic Network (AERONET) platform at the AMF-ASI site.

It is reasonable to hypothesize that synoptic scale pressure patterns help to drive the seasonal variations in aerosol transport via near-surface and upper-level winds. After the aerosols are generated, the biomass burning aerosols are assumed to age which suggests further alterations in their chemical structure thereby facilitating stronger hygroscopicity which then influences their absorptivity (Marquardt Collow et al., 2020; Sedlacek et al., 2022). Moreover, given the nearly 4,500 km distance between the AMF-ASI site and the major African mainland burn region as well as a predominant easterly wind component, there is an opportunity to investigate how biomass burning aerosols with different absorptive properties can activate as CCN when mixed with marine boundary layer aerosols having origins in the remote tropical Southern Atlantic Ocean (Dedrick et al., 2022; Zhang & Zuidema, 2019).

2 Materials and Methods

2.1 Ascension Island Surface Site

The AMF-ASI site houses an aerosol observing system (AOS) which provides continuous physical and chemical measurements of the marine boundary layer (Barrett et al., 2022; Zuidema et al., 2016). These measurements include the spectral scattering coefficient (σ_sp) and CCN number concentration (N_CCN), which are taken from a TSI 3563 Nephelometer and TSI 3776 CCN counter, respectively (Koontz, Flynn, et al., 2022; Koontz, Uin, et al., 2022; Kuang et al., 2022). Note that spectral σ_sp consists of three wavelengths representing the blue (450 nm), green (550 nm), and red (700 nm) channels (Barrett et al., 2022), and the green channel is used in this study. The total aerosol number concentration (N_a) is a derived product calculated from scanning mobility particle sizer bin measurements whose particle sizes range from 40 to 700 nm (Barrett et al., 2022). Daily averaged AOS measurements which include N_CCN values taken at a supersaturation of 0.2% are used in this study (Logan et al., 2020). This study focuses on cases where σ_sp and N_CCN values exceed 20 Mm⁻¹ and 200 cm⁻³, respectively, to demarcate the transition between clean and polluted conditions (Logan et al., 2018, 2020). The fine mode fraction (FMF) is calculated as the daily averaged ratio of the 1–10 μm size cuts of σ_sp at the green wavelength and differentiates influences from biomass burning (FMF > 0.4) and marine aerosols (FMF << 0.4) in this study. The AOS aerosol data set (σ_sp and FMF) contains nearly every date from 2 June 2016 to 31 October 2017, but the N_a and N_CCN data sets start after 26 June 2016 and end on 31 October 2017. Therefore, the July and August data sets for 2016 and 2017 are used in this study.

2.2 The Aerosol Robotic Network (AERONET) Measurements

The AOS hosts a dedicated AERONET site which uses a sun/sky photometer to retrieve column information on the radiative properties of aerosols and water vapor (Holben et al., 1998; Sinyuk et al., 2020). Data products regarding aerosol optical properties such as aerosol optical depth (AOD), Ångström exponent (Angstrom exponent (AE) or spectral dependence of AOD with wavelength), single scattering co-albedo (ratio of absorption to extinction AOD at 440 nm), and radiative forcing efficiency per unit AOD (ΔF^eff) which is given by the following expression:

$$${\increment}{F}^{\text{eff}}550\,\text{nm}=\frac{{\increment}{F}_{550}}{{\text{AOD}}_{550}}$$$

where AOD₅₅₀ is interpolated using the spectral Angstrom equation given by:

$$${\text{AOD}}_{550}={\text{AOD}}_{440}\,\ast \,{(550/440)}^{\left({-\text{AE}}_{440-870}\right)}.$$$

García et al. (2012) and Logan, Xi, and Dong (2013) analyzed the radiative forcing efficiency of mineral dust and carbonaceous aerosols. García et al. (2012) concluded that African biomass burning aerosols could exert strong influences on the warming and cooling potential of the atmosphere at the near surface (BOA) and top of the atmosphere (TOA) while Logan, Xi, and Dong (2013) study added that the mixtures of dust and carbonaceous aerosols could enhance the radiative forcing via absorption of shortwave energy and lead to more cooling at the surface. Daily averaged, Level 2.0 version 3 data (see Sinyuk et al., 2020) are used in this study. Note that the AMF-ASI AERONET data set used in this study did not explicitly include AOD at 550 nm. Clean conditions are denoted by AOD₄₄₀ less than 0.2 while polluted conditions are denoted by AOD₄₄₀ greater than 0.3 in this study. FMF will be compared with AE to introduce higher confidence in the representativeness of column versus surface measurements when smoke plumes pass directly over the AMF-ASI site.

2.3 Reanalysis and Satellite Products

The 600 hPa wind product from the National Centers for Environmental Prediction (Kalnay et al., 1996) is used as a proxy for the magnitude and direction of the mid-troposphere smoke aerosol transport to the AMF-ASI site (Chaboureau et al., 2022; Hsieh & Cook, 2007; Kuete et al., 2020; Ryoo et al., 2022). The NASA Modern Era Retrospective Analysis for Research and Applications Reanalysis version 2 (MERRA-2) platform incorporates a sophisticated data assimilation system from the Goddard Earth Observing System version 5 (GEOS-5) model (Buchard et al., 2017; Gelaro et al., 2017). The model considers data from a combination of satellite retrievals, aircraft measurements from various field campaigns, and AERONET measurements (Buchard et al., 2017; Marquardt Collow et al., 2020). MERRA-2 provides the monthly area averaged surface organic carbon species concentration (proxy for wildfire smoke activity and plume transport) with a spatial resolution of 0.5° × 0.625° and the study domain is bounded by 0°–25°S and 40°W–40°E. The long-term average of surface organic carbon concentration from 2000 to 2020 is compared with the 2016 and 2017 burn seasons over the primary burn areas of Central Africa. The MERRA-2 daily low-level cloud fraction product is used to quantify cloud presence over the AMF-ASI site during episodes of smoke plume intrusions (Marquardt Collow et al., 2020). Reanalysis sounding data derived from NCEP is employed to calculate the lower tropospheric stability (LTS) parameter in the study region based on Wood and Bretherton (2006) and Ryoo et al. (2021) where the potential temperature difference (Δθ) between 1,000 and 700 hPa is used. The 2016 and 2017 fire source points from the Moderate Imaging Spectroradiometer (MODIS) and Visible and Infrared Imaging Radiometer Suite (VIIRS) sensors are used in conjunction with the MERRA-2 data to show the source areas of the biomass burning aerosols with respect to the long-term average.

3 Results

3.1 Climatology and Source Regions of Southeastern Atlantic Smoke

Figure 1 shows the monthly mean column aerosol loading over Ascension Island. As documented by previous studies, the generation of biomass burning smoke approaches a maximum during September. The southern branch of the African Easterly Jet (AEJ-S) strengthens (i.e., increase in mid-level (600 hPa) wind speed) in July and achieves a maximum in October (Adebiyi & Zuidema, 2016). This is a large contributing factor to the long-range, westward transport of central African biomass burning smoke plumes over the remote South Atlantic Ocean (Adebiyi & Zuidema, 2016; Chaboureau et al., 2022; Kuete et al., 2020; Ryoo et al., 2021). Note that non-refractory aerosols (e.g., sea salt) can also contribute to the total aerosol profile, especially during clean episodes (Zuidema et al., 2018). However, it is assumed that during the major burn season months (June-October), biomass burning aerosols are the dominant contributors to aerosol radiative forcing during polluted episodes.

Figure 1

1998–2020 Atmospheric Radiation Measurement site located on Ascension Island median, 10th, 25th, 75th, and 90th quartiles, and monthly mean (star symbols) aerosol optical depth (AOD₄₄₀) retrieved from Aerosol Robotic Network. Mean monthly 2016 (gold asterisk) and 2017 (green asterisk) values for the burn season months are added for comparison in this study.

Figure 2 shows that the months of June, July, and August have the highest surface aerosol loading of the 2016 and 2017 yearly averages, consistent with previous studies (Adebiyi & Zuidema, 2016; Ryoo et al., 2022). The extreme aerosol loading values (σ_sp exceeding 100 Mm⁻¹) in the monthly mean surface aerosol data can be attributed to sporadic episodes of dense smoke plumes advecting from the continent during June, July, and August. In Figure 2a, the mean σ_sp values approach an absolute maximum in August which denotes a higher likelihood of aerosols from smoke plumes advected to the surface. Note that this contrasts with the maximum AERONET column measurement monthly mean occurring in September due to weather patterns being more conducive to the smoke plumes remaining primarily aloft during transport over the remote Southern Atlantic Ocean. The FMF values near or above 0.4 illustrate the influences of biomass burning smoke episodes during every calendar month (Figure 2b). August has the highest mean FMF with skewness in the distribution toward more extreme values. Though not too surprising, the biomass burning smoke aerosols contributed to the increases in N_CCN with the largest increase occurring during August (Figure 2c).

Figure 2

2016–2017 monthly median, 10th, 25th, 75th, and 90th quartiles, and mean of (a) σ_sp, (b) fine mode fraction, and (c) N_CCN at the Atmospheric Radiation Measurement-Ascension Island site. Mean values for each month are denoted by an asterisk.

The 2016 and 2017 years are compared/contrasted with one another to give context in discerning how biomass burning smoke aerosol behavior evolved during LASIC while also considering how each year's burn season months of July and August compared with the long-term climate average. In Figure 3, a total of 13 cases (red pluses) were identified by AERONET in 2016. Four cases in July (3, 16, 23, and 24 July) were classified as moderately polluted (0.2 < AOD₄₄₀<0.3) while four cases in August (13, 14, 15, and 30 August) were strongly polluted (AOD₄₄₀ > 0.3). The remaining five cases occurred during relatively clean marine episodes (0.1 < AOD₄₄₀ < 0.2) beginning on 27 July and ending by 4 August. This pattern was corroborated by AMF-ASI surface measurements which showed corresponding increases (decreases) in the σ_sp, N_a, N_CCN, and FMF parameters during polluted (clean) episodes. Note that there were periods when AERONET retrievals indicated less presence of smoke aerosols than surface measurements. For example, strongly polluted σ_sp, N_a, and N_CCN values were evident when AOD was considered moderately polluted. This was likely due to a combination of increases/decreases in cloud fraction (see lower panel of Figure 3) and boundary layer instability which caused more smoke aerosols to be observed at the surface. However, the AMF-ASI surface site likely missed the smoke plumes that passed overhead during clear-sky conditions when the boundary layer was less turbulent (e.g., stable conditions).

Figure 3

Aerosol physical properties (upper panel) and lower tropospheric stability (LTS, lower panel) during the 2016 and 2017 burn seasons at the Atmospheric Radiation Measurement site located on Ascension Island site. The daily mean of the aerosol observing system observations of σ_sp, N_a, N_CCN, and fine mode fraction (FMF) (primary y-axis) are shown along with clean (0.1 < AOD₄₄₀ < 0.2), moderately polluted (0.2 > AOD₄₄₀ > 0.3) and strongly polluted (AOD₄₄₀ > 0.3) cases (red pluses). Note that the FMF and AOD₄₄₀ are plotted on the secondary y-axis. The σ_sp values (green line) are multiplied by 10 as a scaled fit. Clean N_CCN conditions are near or below the blue line (100 cm⁻³) while moderately polluted conditions are denoted between the purple and red lines (200–300 cm⁻³). Strongly polluted conditions are above the red line (300 cm⁻³). Daily mean cloud fraction (red line) and LTS (blue line) are shown in the lower panel. LTS values are normalized to the highest value of the 2-month period for scale.

The LTS parameter exhibited notable decreases during the polluted cases, and cleaner cases featured elevated LTS values suggesting boundary layer turbulence and stability, respectively. A scatterplot of the August 2016 cases is provided in Figure A1 (see Appendix A) to further demonstrate the negative relationship (R ∼ −0.33) between LTS and aerosol loading for the most polluted and cleanest cases during August. Note that the relationship for the July cases was nonexistent (R ∼ 0). Furthermore, the increase in N_a and N_CCN values also denote that cloud development can be enhanced by the presence of smoke aerosols during periods of decreased LTS due to (a) rising air parcels containing hygroscopic CCN and (b) subsequent cloud top entrainment of free troposphere smoky air parcels assuming the presence of marine stratocumulus clouds (Marquardt Collow et al., 2020; Zhang & Zuidema, 2019). There are some slight increases in cloud fraction around the times of the smoke plumes, corresponding to the episodes of decreased LTS. However, it should be noted that the low-level cloud fraction does not explicitly differentiate between marine boundary layer stratocumulus clouds and stratus clouds. From MODIS visible imagery, it can be shown that both cloud types are observed in advance of smoke plumes, during the smoke plume episodes, and during episodes of clean conditions. In addition, Marquardt Collow et al. (2020) illustrated the relationship between cloud fraction (using a 0.25 threshold) and smoke plume intrusions. From Figure 2 of their study, it can be shown that cloud fractions exceeding 0.25 not only occurred during smoke plume episodes (i.e., inferring possible marine stratocumulus cloud top entrainment) but also occurred during the strongly polluted cases (and declining LTS values) shown in this study (Figure 3, lower panel).

Figure 4 gives insight into the physico-chemical and radiative properties of the AERONET smoke aerosol cases for the 2016 burn season. The selected cases are grouped into four regions categorized by their AE and single scattering co-albedo values (Logan, Xi, Dong, et al., 2013, 2014; Tian et al., 2017). Region II is the primary area for biomass burning aerosols while Regions III and IV denote clean marine and/or mineral dust aerosols. The four most polluted cases all occurred during August and had distinct absorptive and size characteristics. For example, the 13 and 30 August cases had the highest AOD values (∼0.8) but were less absorbing than the 14 August and 15 August cases. The difference in absorptive properties can be partially explained by investigating the mode of generation of the aerosols. Smoke aerosols from fresh fires may contain a dense coating of refractory black carbon (rBC) within the earliest periods of combustion (Sedlacek et al., 2022; Zuidema et al., 2018). As these aerosols are advected away from the source region, this coating can be oxidized or impacted by marine aerosols thereby altering their radiative properties as well as enhancing their ability to undergo net condensational growth. Zuidema et al. (2018) pointed out that the 14 August and 15 August cases were likely exposed to higher moisture conditions during transport. This helps to explain the smaller AE values for these two cases. The 13 August and 30 August cases likely underwent aging possibly due to clear sky photolytic reactions due to lower cloud extent (e.g., Marquardt Collow et al., 2020; Figure 2). It should also be noted that the smoke aerosol composition (e.g., organic/inorganic chemical species) could have differed between the case days. Therefore, a trajectory analysis for each case is used to add support to how the smoke aerosols evolved during transport.

Figure 4

Classification of the selected 2016 cases using Aerosol Robotic Network retrievals of Angstrom exponent (AE) and single scattering co-albedo (after Logan, Xi, Dong, Li, & Cribb, 2013; Tian et al., 2017) with mean and standard deviation (1σ). The most polluted cases (AOD₄₄₀ > 0.3) are denoted in red. Following Zheng et al. (2020), AE > 1 denotes fine mode dominance while AE < 1 denotes coarse mode dominance.

The smoke plume heights were reported by several studies to be located between 600 and 700 hPa (e.g., Kuete et al., 2020; Marquardt Collow et al., 2020; Ryoo et al., 2021). Thus, the 600 hPa winds illustrate the westward transport of biomass burning smoke plumes from the primary continental burn region to the AMF-ASI site via the remote ocean (Figure 5a). The developing AEJ-S and Kalahari wind patterns are evident north and south of the burn region, respectively (Kuete et al., 2020). The backward trajectories of the four cases identified in Figure 4 are illustrated in Figure 5. The primary burn region is generally located between 5°S–15°S (e.g., Marquardt Collow et al., 2020), and a total of 4,968 individual fires were recorded for August 2016 (Figure 5a). Additionally, an analysis of the burn season months of 2016 with respect to the long-term average (2000–2020) revealed a positive anomaly of surface organic carbon concentration (Figures 5b and 5c). The 14 August and 15 August cases (Figure 5b) exhibited relatively short transport pathways at the near-surface (500 m AGL) and free troposphere (3,000 m) altitudes that were mostly over the remote South Atlantic. Zuidema et al. (2018) pointed out that the rBC values for these cases were the highest in the 2-year period. Note that the sub-cloud (1,500 m AGL) trajectory for 14 August passed directly over the continental fire region. This can explain co-albedo values between 0.14 and 0.15. Though the rBC was elevated during the 13 August (Figure 5c) and 30 August (Figure 5d) cases, the trajectories were longer at the sub-cloud and free troposphere altitudes which can explain the lower co-albedo values (∼0.1) and supports a higher likelihood mixing with non-refractory aerosols.

Figure 5

(a) 600 hPa wind vectors and fire location (gold pluses). Backward trajectories of air parcels to the Atmospheric Radiation Measurement site located on Ascension Island site (red box) plus MERRA-2 August monthly mean organic carbon surface concentration minus the long-term average for the (b) 14–15 August (c) 13 August, and (d) 30 August 2016 cases.

In Figure 6, a total of 13 cases (red pluses) were identified by AERONET in 2017. There were seven clean cases in July and August (4, 5, 6, 7, 8, 9, and 12 July; 21 August), two moderately polluted July cases (22 and 23 July), and three strongly polluted cases in August (10, 11, and 14 August). Similar to 2016, the pattern of increases and decreases in the σ_sp, N_a, N_CCN, and FMF parameters followed the AERONET retrievals. The LTS parameter showed similar decreases during the moderately polluted July cases with less evident decreases during the strongly polluted August cases. Similar to August 2016, a scatterplot of the August 2017 cases is shown in Figure A1, demonstrating a weaker negative correlation (R ∼ −0.23). In addition, the overall cloud fraction was lower in 2017 than 2016 but seemed to have a slight positive correlation (R₂₀₁₇ ∼ 0.26 vs. R₂₀₁₆ ∼ 0) with the increase/decrease in aerosol and N_CCN quantities. That is, the cloud fraction noticeably increased (decreased) when a smoke plume advanced into (passed) the region in 2017. Figure 7 showed that only the 23 July and 14 August cases were in Region II while the remaining cases were contained within Region III with the 10 August (11 August) case being the most (least) absorptive in terms of co-albedo values. Though the 2017 cases appear to be more absorbing and the overall size of the aerosols noticeably larger given the lower AE, the overall aerosol loading is less than the 2016 cases.

Figure 6

Same as Figure 4 but for the 2017 cases.

Figure 7

Same as Figure 4 but for the 2017 cases.

The 600 hPa winds in Figure 8a showed a stronger wind maximum around 10°N and stronger winds south of 20°S during 2017 compared to 2016. However, the wind speed and direction closest to the burn region during 2017 were comparable to 2016. There were fewer fires in 2017 (4,260 vs. 4,968 in 2016) with a corresponding negative surface organic carbon concentration anomaly over the northern burn region. Backward trajectory analysis of the three most polluted August cases suggested longer transport pathways passing over the remote Southern Atlantic. The 10 August case had a short sub-cloud trajectory originating over the remote ocean in a zone of anomalously strong organic carbon concentration. In addition, the near-surface trajectory originated in the vicinity of an anomalously strong southern burn region. The free troposphere trajectory originated over the continent far removed from appreciable fire activity. Thus, the locations of the trajectories likely explained why this case featured the largest co-albedo (0.17). The 11 August and 14 August cases featured a combination of trajectories that originated off of the African coast away from primary burn regions. Moreover, the trajectories seemed to favor a collection of smoky air parcels that drifted off of the mainland and likely mixed with remote marine aerosols. In turn, the co-albedo value decreased to 0.1 for these cases. Overall, the aerosol cases during 2017 had a larger coarse mode aerosol influence in comparison to the 2016 cases.

Figure 8

Same as Figure 5 but for (a) August 2017 600 hPa winds and fire locations and the (b) 10 August, (c) 11 August, and (d) 14 August cases.

3.2 Comparison of Biomass Burning Smoke Aerosol Radiative Properties

Figure 9 gives a closer look at the efficacy of how the smoke aerosols activated as CCN for the 2016 and 2017 moderately and strongly polluted cases (σ_sp > 20 Mm⁻¹). The 2016 cases featured a strong correlation (R ∼ 0.837) between σ_sp and N_CCN. However, the 2017 burn season exhibited a noticeable increase in correlation (R ∼ 0.878). That is, it took fewer particles to activate as CCN in 2017 than in 2016. Recall that the 2016 aerosol cases had a higher mean AE than the 2017 cases while the co-albedo values (absorption) were on average higher for the 2017 cases (Figures 4 and 7). The difference in the 2016 and 2017 correlations may be closely related to the radiative and chemical properties of the smoke aerosols and how they evolve during transport. That is, as the smoky air parcels advected off of the continent over the remote ocean, physical mixing with marine aerosols likely acted to alter the physical properties of the smoke aerosols to a larger degree in 2017 than in 2016. Therefore, a deeper discussion of the contrasting direct and semi-direct radiative properties of the 2016 and 2017 smoke aerosol cases is presented.

Figure 9

Scatterplot of σ_sp and N_CCN for the polluted cases denoted by σ_sp greater than 20 Mm⁻¹.

3.3 Discussion of Radiative Impacts of Smoke Aerosols Within the Boundary Layer and TOA

Table 1 presents a summary of the AERONET derived spectral AOD, AE, co-albedo, and radiative forcing efficiencies (ΔF^eff) of the selected 2016 and 2017 burn season cases. Radiative forcing efficiency is a good measure of the extinction potential of aerosols per unit AOD which loosens the dependence on aerosol loading (e.g., García et al., 2012). In this way, a stronger comparison of radiative forcing due to smoke aerosols between the two burn seasons can be analyzed with respect to how strongly the aerosols attenuate incoming and outgoing solar radiation. When considering the absolute magnitude of ΔF^eff during smoke plume cases, strongly negative near-surface (BOA) ΔF^eff values denote weaker transmission of incoming shortwave radiation to the surface and thus more absorption of energy at the height of the aerosol layer. In addition, strongly negative upper atmosphere (TOA) ΔF^eff values are a direct result of diminished outgoing shortwave radiation being scattered within the upper atmosphere, possibly by clouds (Marquardt Collow et al., 2020). García et al. (2012) pointed out that AERONET retrievals during clean conditions (e.g., AOD₄₄₀ < 0.1) can create known artifacts by inflating ΔF^eff to higher values at the TOA and BOA. The artifacts are primarily due to an atmosphere dominated by multiple scattering effects from marine aerosols rather than smoke aerosol extinction (e.g., Region III aerosols in Figures 4 and 7). Furthermore, the difference between TOA and BOA ΔF^eff becomes smaller as AOD decreases (Marquardt Collow et al., 2020).

Table 1. Aerosol Robotic Network Retrieved Aerosol Radiative Properties for the 2016 and 2017 Cases

Aerosol loading	Day_of_Year	AOD 440	AOD 550	Radiative forcing efficiency (BOA)	Radiative forcing efficiency (TOA)	AE	Co-albedo
2016 Cases
Weak loading	27 July 2016	0.13	0.11	−1,686.81	−962.50	0.74	0.07
	29 July 2016	0.07	0.07	−4,524.30	−1,090.20	0.30	0.08
	31 July 2016	0.10	0.08	−2,844.19	−722.40	0.65	0.16
	1 August 2016	0.10	0.08	−3,782.53	−644.26	0.72	0.13
	3 August 2016	0.15	0.12	−1,365.82	−698.36	0.79	0.06
Moderate loading	3 July 2016	0.24	0.19	−819.50	−531.15	0.89	0.04
	16 July 2016	0.18	0.16	−997.43	−657.61	0.52	0.03
	23 July 2016	0.20	0.17	−1,089.35	−593.25	0.65	0.07
	24 July 2016	0.22	0.19	−807.13	−482.94	0.65	0.07
Strong loading	13 August 2016	0.83	0.63	−285.42	−104.86	1.30	0.10
	14 August 2016	0.59	0.45	−478.83	−131.92	1.24	0.15
	15 August 2016	0.41	0.32	−724.24	−182.53	1.25	0.14
	30 August 2016	0.78	0.58	−203.06	−87.97	1.37	0.09
2017 Cases
Weak loading	21 August 2017	0.12	0.11	−2,264.09	−707.91	0.68	0.12
	4 July 2017	0.07	0.06	−3,495.83	−1,206.49	0.40	0.15
	5 July 2017	0.05	0.05	−5,815.02	−1,078.17	0.45	0.21
	6 July 2017	0.05	0.04	−6,076.91	−901.18	0.33	0.22
	7 July 2017	0.04	0.04	−7,901.90	−1,434.11	0.40	0.22
	8 July 2017	0.04	0.04	−7,784.09	−1,390.95	0.35	0.16
	9 July 2017	0.06	0.05	−5,422.18	−1,144.58	0.61	0.18
	12 July 2017	0.05	0.05	−6,174.96	−1,182.11	0.42	0.21
Moderate loading	22 July 2017	0.26	0.21	−1,174.69	−265.44	0.94	0.17
	23 July 2017	0.29	0.23	−1,043.38	−266.63	1.01	0.15
Strong loading	10 August 2017	0.34	0.28	−641.74	−159.42	0.85	0.17
	11 August 2017	0.41	0.34	−427.38	−208.23	0.90	0.10
	14 August 2017	0.41	0.33	−536.10	−221.47	1.06	0.11

• Note. The most polluted cases (AOD₄₄₀ > 0.3) are denoted in red.

Seven polluted cases (AOD₄₄₀ > 0.3) were previously identified in Figures 4 and 7. Of the four polluted cases in 2016, the 13 August and 30 August cases featured the highest AOD₄₄₀ values of 0.83 and 0.78 with corresponding BOA ΔF^eff values of −285 and −203 W m⁻² and TOA values of −104 and −88 W m⁻², respectively. However, the 14 August and 15 August cases had lower AOD₄₄₀ values of 0.59 and 0.41 along with stronger BOA ΔF^eff values of −478 and −724 W m⁻² and TOA ΔF^eff values of −132 and −182 W m⁻², respectively. The 13 August and 30 August 2016 cases exhibited weaker attenuation at the surface and outgoing radiation at the TOA than the strongly absorbing 14 August and 15 August 2016. García et al. (2012) suggested that smoke aerosols that are most efficient at attenuating incoming solar radiation at the BOA should also have a pronounced attenuation at the TOA. This is consistent with the 14 August and 15 August cases which had larger co-albedo values although the 13 August and 30 August cases had higher AOD₄₄₀ values and thus demonstrates the competing influences of smoke radiative effects versus smoke aerosol loading which is mitigated by using ΔF^eff (García et al., 2012). Given the elevated rBC and weaker ΔF^eff values for the 13 August and 30 August cases, it is possible that these aerosols allowed for more absorption of incoming solar radiation within the marine boundary layer than the 14 August and 15 August cases. This suggests that the former two cases had more of a warming effect at the top of the marine boundary layer (Marquardt Collow et al., 2020; Zhang & Zuidema, 2021; Zuidema et al., 2018). The decrease in LTS values for the 13 August, 14 August, 30 August, and to a lesser extent, 15 August cases suggest sufficient boundary layer turbulence for cloud formation. The overall relationship between LTS and cloud fraction for the four cases was not direct, but the marine boundary layer was demonstrated to be dominated by stratocumulus clouds as per satellite (not shown) and discussed in Zhang and Zuidema (2021). Atmospheric soundings taken from near the AMF-ASI site for the 2016 and 2017 cases are provided in Appendix B for perspective.

Of the three polluted cases in 2017 (Figure 6), the 10 August case had a lower AOD₄₄₀ value (0.34) than the 11 August and 14 August cases (0.41). The 10 August 2017 case had the lowest AOD₄₄₀ value of all seven cases (0.34) and BOA (TOA) ΔF^eff values of −642 (−159) W m⁻². The 11 August (14 August) 2017 case had a BOA ΔF^eff value of −427 (−536) W m⁻² and TOA ΔF^eff value of −208 (−221) W m⁻². The 10 August case had a higher co-albedo and stronger BOA ΔF^eff than the 11 August and 14 August cases. The 10 August 2017 case was comparable to the 14 August and 15 August 2016 cases in terms of ΔF^eff. However, the weakly absorbing 11 August and 14 August 2017 cases had the strongest negative ΔF^eff values at the BOA and TOA which contrasted with the weakly absorbing 13 August and 30 August 2016 cases. That is, more incoming radiation was attenuated, leaving less radiation to be realized at the TOA for the 2017 cases likely cooling the marine boundary layer.

Zhang and Zuidema (2019) previously investigated the impacts of the 2016 and 2017 smoke aerosols on the marine boundary layer heating profile during LASIC. They showed that the 2016 cases had an overall higher rBC mass concentration than the 2017 cases during smoke plume intrusions over the AMF-ASI site (see Figure 1 of Zhang & Zuidema, 2019). Moreover, their study found that low-cloud development was perturbed by the smoke aerosols differently in both years as a result of the changes in boundary layer dynamics due to the warming and cooling tendencies of the smoke aerosols. Marquardt Collow et al. (2020) investigated the warming and cooling effects of smoke aerosols of the cases used in this study. They found that the biomass burning aerosols had an overall cooling effect on the BOA while having a warming effect aloft that was independent of cloud cover. When considering ΔF^eff, it is evident that under polluted, non-overcast conditions, smoke aerosols can have varying influences on the heating/cooling of the marine boundary layer which needs to be considered. Thus, it is important to continue investigating the magnitude of the direct and semi-direct radiative impact of smoke aerosols in modulating marine boundary layer cloud development by carefully accounting for short- and long-term variations in aerosol radiative properties with respect to large-scale meteorological patterns such as humidity, hygroscopicity, turbulent kinetic energy, and low-level/mid-tropospheric winds.

4 Summary and Conclusion

This study leverages selected polluted cases from the LASIC field campaign during the 2016 and 2017 burn season months of July and August to investigate and quantify biomass burning smoke aerosol radiative effects due to aerosol transport and boundary layer dynamics. Biomass burning smoke plumes were evident over the AMF-ASI site given by corresponding increases in both surface (σ_sp) and column aerosol loading (AOD), Ångström exponent (AE), and FMF. In line with previous studies, the biomass burning smoke aerosols readily activated as CCN, given the observed increases in surface N_a and N_CCN values during both seasons. The LTS parameter decreased during periods of smoke plume intrusions suggesting a combination of marine boundary layer turbulence (and possible cumuliform cloud top entrainment) which mixed the smoke aerosols down to the surface. Back trajectory analysis revealed that smoky air parcels having longer transport pathways, especially at sub-cloud and near surface altitudes, contained mixtures of smoke and marine aerosols. In contrast, shorter back trajectories featured strongly absorbing, fine mode smoke aerosols that were less aged.

The radiative properties of the biomass burning smoke aerosols varied from 2016 to 2017. In general, smoke plumes that advected over the AMF-ASI site having co-albedo values exceeding 0.14 were most efficient (per unit AOD) at attenuating incoming shortwave radiation at the near-surface (BOA) and allowing less radiation back to space (TOA). The weakly absorbing 2016 cases were less efficient at attenuating incoming radiation at the BOA and TOA than the 2017 cases suggesting contrasting impacts to the atmospheric heating profiles due to longer transport and stronger mixing of smoke and marine aerosols during the 2017 cases. This also had direct implications on cloud development in that the 2017 smoke aerosols exhibited a stronger ability to activate as CCN.

Future work will consist of conducting a radiative transfer analysis to examine how smoke aerosol radiative forcing efficiency can positively or negatively impact cloud radiative forcing. Moreover, further analysis of the relationship between strongly and weakly absorbing smoke aerosols and boundary layer CCN is needed to quantify radiative effects on cloud fraction and precipitation processes, especially over a climatological scale.

Acknowledgments

T.L, L.W., A. A., & A. G. are supported by the National Science Foundation (AGS-2031750) at Texas A&M University and X. D., B. X., & X. Z. are supported by the National Science Foundation (AGS-1700728) at the University of Arizona. The authors wish to thank the two anonymous reviewers whose constructive suggestions and comments strengthened the manuscript. The authors are grateful for the software package SounderPy used to visualize the atmospheric soundings, for boundary layer environmental data analysis, and to calculate the lower tropospheric stability parameter for the 2016 and 2017 cases.

Appendix A

The negative correlation between lower tropospheric stability and surface aerosol loading (e.g., scattering coefficient) is apparent in the August 2016 and 2017 cases suggesting boundary layer instability which acts to advect aerosols in the free troposphere down toward the surface.

Figure A1

Scatterplot of lower tropospheric stability and σ_sp for the August 2016 and 2017 cases.

Appendix B

Boundary layer instability is evident in the seven cases investigated in this study. It is interesting to note that the convective available potential energy was highest in the cases with the highest column and surface aerosol loading (e.g., 13 August 2016 and 11 August 2017 cases). This suggests more overturning within the boundary layer not only advected more aerosols over the Atmospheric Radiation Measurement-Ascension Island site but also mixed down more aerosols to the surface. Hence, modeling of these cases can help disentangle atmospheric physical processes from aerosol radiative influences (Figures B1-B4).

Figure B1

Sounding profile at Ascension Island for the 14 August and 15 August 2016 cases.

Figure B2

Sounding profile at Ascension Island for the 13 August and 30 August 2016 cases.

Figure B3

Sounding profile at Ascension Island for the 10 August and 11 August 2017 cases.

Figure B4

Sounding profile at Ascension Island for the 14 August 2017 case.

Open Research

Data Availability Statement

Ground-based measurements provided by the U.S. Department of Energy as part of the Atmospheric Radiation Measurement (ARM) user facility during the 2016–2017 Layered Atlantic Smoke Interactions with Clouds (LASIC) campaign: cloud condensation nuclei (Koontz, Uin, et al., 2022, https://doi.org/10.5439/1323894); scattering coefficient (Koontz, Flynn, et al., 2022, https://doi.org/10.5439/1228051); aerosol number concentration (Kuang et al., 2022, https://doi.org/10.5439/1476898). NASA MERRA-2 organic carbon surface concentration data are provided by the Global Modeling and Assimilation Office (GMAO) (2015), https://doi.org/10.5067/FH9A0MLJPC7N. Vector wind data at the 600 hPa level is provided by the NOAA Physical Sciences Laboratory, Boulder, Colorado, USA, from their website at https://psl.noaa.gov/data/composites/day/. We thank the PIs, Rick Wagener, Lynn Ma, and their staff for establishing and maintaining the ARM-ASI AERONET site used in this investigation (https://aeronet.gsfc.nasa.gov/cgi-bin/webtool_opera_v2_inv?stage=3&region=Atlantic&state=Ascension_Island&site=ARM_Ascension_Is&place_code=10&if_polarized=0). Atmospheric soundings and potential temperature data used in calculating the lower tropospheric stability parameter are provided using SounderPy software (Gillett, 2024).