ERA5 is used to detail the 3-D wind (u, v, ω), temperature, geopotential height (Z), and specific humidity (q) fields (Hersbach et al., 2020). Hourly and monthly data are available on a 0.25∘ longitude × 0.25∘ latitude grid, with 37 vertical levels ranging from 1000 to 1 hPa. The ERA5 wind structure compares favorably to that of other reanalysis data, such as the Modern-Era Retrospective Analysis for Research and Applications, version 2 (MERRA-2) and Japanese 55-year Reanalysis (JRA55), over the deployment region (Ryoo et al., 2021).

The boundary layer height (BLH) is estimated using, every 6 h, ERA5 specific humidity (q), based on a heuristic algorithm we have developed. The computed BLH is designed to include the decoupled cloud-topped planetary boundary layer (PBL) in which the cloud layer is above the well-mixed surface-based layer (see Fig. S1 in the Supplement). In this quasi-isentropic exchange, the air mass mixing occurs without changing the potential temperature rapidly. This BLH tends to be higher than the height of the well-mixed sub-cloud layer. The methodology for estimating the decoupled cloud-topped BLH is as follows: 1.

Calculate dq/dz (i.e., a vertical derivative of q, where q is specific humidity (g kg-1), and z is the vertical level in m) up to D, which is the maximum permitted BLH. D is set to 3 km over all the oceanic regions and islands south of 2∘ N and west of 10∘ E. D is 6 km over the land (except for the islands).

The first deployment of ORACLES was based in Walvis Bay (22.96∘ S, 14.51∘ E) on the Namibian coast of southwestern Africa. During September 2016, the AEJ-S was comparable to the climatological mean. The LLJ and the large-scale southern Atlantic anticyclone were stronger than the climatological mean. Sea surface temperatures (SSTs) over the SE Atlantic were warmer than the climatological means, but the monthly mean low CF was not noticeably lower than climatology in September 2016. Characteristics of the synoptic-scale condition and convective features during the September 2016 deployment are summarized in Table 1. Each column in Table 1 represents the dates of occurrence, both ER-2 (Earth Resources 2, NASA's high-altitude airborne science aircraft; NASA ER-2, 2014) and P-3 (NASA's aircraft designed for endurance and range, which are capable of long-duration flights; NASA P-3, 2022) flight days, and the longitude–latitude domain for which the synoptic features were captured.

Figure 2 shows the longitude–time cross section (i.e., Hovmöller diagrams) of every 6 h 600 hPa RH, BC, zonal wind, and heat low for September 2016. The AEJ-S is strongly associated with the heat low, consistent with climatology (Ryoo et al., 2021). The diurnal variations in the heat low are one of the major drivers of the AEJ-S (Fig. S2). Coherent variations in RH and the AEJ-S suggest the moisture transport from land is facilitated when the AEJ-S extends further west (Fig. 2a and b; see also Pistone et al., 2021, for examples). The similar westward transport feature is also shown in the water vapor field (Fig. S2). The AEJ-S strengthening from 8–15 September 2016 between 10∘ W and 20∘ E leads to the corresponding RH increases by 0–2 d (Fig. 2a, c, f).

When the AEJ-S is weak, the transport of RH and BC is also weak, as shown on 5–8 September 2016 (Fig. 2a, b, f). The relationship between BC–AEJ-S–RH strengthens as the month evolves, with higher BC mixing ratios occurring when the AEJ-S becomes stronger. Around 24 September, BC mass mixing ratios are high and the AEJ-S is strong, but RH and water vapor (not shown) decrease as they dissipate over the ocean.

A focus on three distinct flight days helps to elucidate the monthly evolution. The 8 and 14 September 2016 flight days were denoted as “routine” flight days (Fig. 3a and b), designed to build up representative statistics along a fixed track, whereas the targeted flight on 24 September 2016 (Fig. 3c) reached to the far northern edge of a coastal African smoke plume (Redemann et al., 2021). The routine flight on 8 September sampled dry conditions, with the moist plume located further north of 10∘ S (e.g., 5–10∘ S), while the moist plumes with RH ∼50 %–60 % in the range of ∼3–5.5 km were intercepted by the two other flights (Fig. 3d–f). Especially during the 24 September flight, the peak RH lines are aligned with the AEJ-S (black contour; Fig. 3f), where its maximum is observed around 5–10∘ S, 10–20∘ E, originating from the continent. Furthermore, on 24 September, high RH is observed throughout the vertical layers from the bottom, indicative of a slightly deeper marine boundary layer compared with the other 2 highlighted flight days (Fig. 3d–f).

Both water vapor and BC aerosol are transported by the strong zonal wind (e.g., AEJ-S), but these have different patterns on the different flight days (Fig.3g–l). On 8 September 2016, large moisture and BC gradients over the ocean around 5∘ W–10∘ E, 10–20∘ S are accompanied by a relatively weak zonal wind, with a mid-latitude cyclonic circulation expanding northward up to 20∘ S (Fig. 4a). On 14 September 2016, a stronger zonal wind transports moisture and BC over the ocean at 10∘ S, and the jet merges with the subtropical cyclonic circulation evident south of 20∘ S (Fig. 3h). By 24 September, as the heat low continues to strengthen, a large-scale mid-level cyclonic circulation becomes established, pushing the westward-moving mid-latitude jet stream further south (not shown). We separated the flight regions into region A and B because of (1) the different relationship of low cloud cover with the meteorological variables between the regions (see also Andersen et al., 2020) and (2) the latitudinal dependency of the LTS and low-cloud (e.g., stratocumulus) fraction (Wood, 2012). The detailed relationship is further discussed.

Figure 4 shows the time evolution of BC aerosol, wind, RH, and the cloud deck. For most of September 2016, the low CF is modulated by the synoptic-scale atmospheric circulation, including frontal passages. The width and the intensity of the developing cyclonic circulation shape the moisture and aerosol transport and low CF (Fig. 3j–l). The AEJ-S (wind at 600 hPa over region A) and the LLJ (wind at 925 hPa over region B) is weakly correlated in September 2016 (Pearson correlation (R) =0.32 with p value ∼0.11; see Fig. S3). On some days, the marine boundary layer also contains BC, such as from 21–24 September over region B. HYSPLIT (Hybrid Single-Particle Lagrangian Integrated Trajectory model) back trajectories and the aerosol forecasts used during the deployment indicate that the boundary layer BC is aerosol advected directly westward off of the continent that has not reached the altitude of the AEJ-S, residing near the boundary layer top (Dobracki et al., 2022), where it can be more easily entrained even when the subsidence is weak (not shown).

Aircraft observations do not show BC in the boundary layer for most of September (Shinozuka et al., 2020), except on 31 August 2016 (Diamond et al., 2018), indicating the CAMS reanalysis allows for too much entrainment into the boundary layer (reaching 0.8 ppb, parts per billion, at times). Simultaneously, the temperature increases, and the RH decreases (Fig. 4d). The temperature increase is large enough to suggest that the cause is advection from land, but this is not pursued further. The low CF is not much affected, although a slight decrease is found around 23 September. Similar features are also found in region A (not shown).

Whether convection is dry or moist can modulate how transported aerosols and moisture interact with clouds (Tao et al., 2012). To differentiate dry from moist convection, soundings and longitudinal cross sections of RH, winds, and θ over the ocean and the land are shown in Fig. 5. Over the ocean, the oceanic soundings show that the air is saturated when the temperature is equal to the dew point temperature (Alduchov and Eskridge, 1996) at around 900 hPa (Fig. 5a–c), forming stratocumulus. Moisture transported from the continent by the AEJ-S near 600 hPa can lead to occasional saturation, producing scattered mid-level clouds (Adebiyi et al., 2020). In the example of 24 September, the moist plume above stratocumulus resembles a dry-convection-type layer (above 700 hPa). This is indicated by the (1) dew point temperature following a nearly constant water vapor mixing ratio line and (2) air temperature roughly following the dry adiabatic lapse rate. This demonstrates that the dry convective layer propagates out from the land into the ocean (sounding plot of Fig. 5c and e).

Over the land, dry convection is more evident south of 10∘ S. The dew point nears saturation, ultimately favoring high RH, but with insufficient moisture for rainfall below 600 hPa (Fig. 5d–f). RH can exceed 60 %, and thin mid-level clouds can form at times near the top of the dry convection (Adebiyi et al., 2020). Dry convection over the Namibian–Kalahari dryland (16–17∘ S) can reach up to 500 hPa (∼ 5.5 km), especially on 24 September (Fig. 5f). This dry convection results in near-saturated conditions at the top of the deep continental boundary layer, explaining the moist plume that is transported westward by the AEJ-S at or north of the dry convection region. In contrast, moist convection is dominant north of 10∘ S (upper panels at 18∘ E, 5–6∘ S (CB) of Fig. 5d–f). All soundings over land around 5–6∘ S have significant convective available potential energy (CAPE). For all cases, the moisture plume around 600 hPa over the ocean is transported westward from land by the AEJ-S, although moisture can potentially be also modulated from a southern mid-latitude disturbance as well (e.g., 31 August 2016; Adebiyi and Zuidema, 2018; Kuete et al., 2020; Ryoo et al., 2021).

Low cloud structure and cover are influenced by the large-scale subsidence, LTS, and BLH (Klein and Hartman, 1993; Wood, 2015), but these relationships will be weaker at synoptic timescales (De Szoeke et al., 2016) compared to the monthly mean variations detailed in Ryoo et al. (2021) and in the complex SE Atlantic environment (Adebiyi and Zuidema, 2018). Here we examine the daily variability in the low CF over the course of 3 flight days in Fig. 6. Clearly shown is that LTS and low CF are spatially and temporally positively correlated (Fig. 6a–c, g). For example, on 14 and 24 September 2016, as the LTS is stronger, low CF increases over most of the regions. However, on 8 September 2016, a mid-latitude system has broken up the sheet of low clouds to the south of 20∘ S (Adebiyi and Zuidema, 2018), possibly reducing the regional low CF and lowering the LTS compared to other days (Fig. 4a). Deeper BLHs tend to have reduced low CF, consistent with entrainment drying. A depressed BLH (<1 km) promotes the occurrence of a high low CF, as seen on 24 September 2016 (Fig. 6b, c, f, h).

The passage of a developing frontal system, indicated by the mid-level (e.g., 600 hPa) trough located on the west of the sea level low pressure, alters the regional circulation, as shown for 8 and 24 September 2016 (Fig. 6f). This is also indicated by the largely different angle between the 600 hPa wind vectors and geopotential lines (Fig. 6d–f), which are more apparent when the low pressure develops south of 30∘ S. The divergence of the ageostrophic wind links the low-, mid-, and upper-level flow through the strong vertical motion (Holton, 2004), aiding anomalous ascent at the southern edge of the stratocumulus deck, cooling the lower free troposphere, reducing LTS, and ultimately decreasing the low CF (Adebiyi and Zuidema, 2018). The LTS–low CF relationship is stronger further south (region B; R∼0.73) than in region A. Figure S13 contains further information on the mid-level flow and SLP patterns on all September 2016 flight days. No sensitivity in the result was found when using either mid-level (e.g., 600 hPa) or upper-level (e.g., 500 hPa) geopotential height and wind to investigate the development and impact of the mid-latitude frontal system (not shown).

The second deployment occurred in São Tomé (0.34∘ N, 6.73∘ E). August marks the transition from austral winter (July) to austral spring (September) as the Southern Hemisphere warms, and the Kalahari heat low develops. During August 2017, the heat low development was delayed relative to the climatological mean (2000–2018). The August 2017 AEJ-S was weaker than in climatology due to an anomalous upper-level disturbance (wave) at around 10∘ S. The August 2017 lower free troposphere (∼700 hPa) was also drier over the SE Atlantic Ocean and the coast (around 0–15∘ E, 5–10∘ S) than in climatology, with a stronger LLJ at 925–950 hPa along the Namibian coast. Consistent with this, the southern Atlantic anticyclone was stronger and closer to the coast than the August climatological mean. The monthly mean low CF was noticeably lower than climatology in August 2017 and linked to a warm SST anomaly (Ryoo et al., 2021). The characteristics of synoptic-scale features during the flight days in August 2017 deployment are summarized in Table 2.

The strongest free tropospheric winds occur approximately 3∘ further north in August 2017 than in September 2016 or between 5 and 7∘ S (Sect. 3.1). The Hovmöller diagram in Fig. 7 indicates relatively dry conditions until 18 August 2017, with weak easterlies over the Atlantic Ocean (Fig. 7a and b). The BC aerosol layer coincides with the lower altitude of the AEJ-S and RH (Fig. 7b). The heat low strengthens considerably as the month progresses (Fig. 7c). Strongly easterly winds, or the AEJ-S, do not become apparent until 20 August, and their maximum height is lower (∼700 hPa, ∼3 km) in August than in September (Figs. 2, 5, 7d). In August 2017, there is another easterly jet above 500 hPa, and this is unusually strong compared to the climatology (Ryoo et al., 2021; Fig. 7d). Relatively dry free tropospheric conditions (RH is <∼60 %; q∼8 g kg-1) over the ocean continue until 18 August 2017 and, thereafter, shift to a moist regime after 20 August 2017 when moist plumes develop (Fig. 7e and f), coinciding with the development of the AEJ-S over the continent. A similar feature is evident in the water vapor mixing ratio (Fig. S4).

The flight tracks with their altitudes and RH superimposed for three August 2017 flights are shown in Fig. 8. The 18 August 2017 flight was coordinated with one from the CLARIFY (Haywood et al., 2021) campaign out of Ascension Island. The 21 August 2017 flight was a transit flight from Ascension Island back to São Tomé. The 30 August 2017 routine flight went along 5∘ E south to about 13∘ S and back. These flight days possess unique and interesting meteorological characteristics that help interpret the monthly evolution.

The remote atmosphere sampled on 18 August 2017 near Ascension Island was dry (RH <40 %) above ∼5 kft (∼ 1.5 km), with the flight back to São Tomé on 21 August encountering an abrupt change to a moister free troposphere at ∼7∘ W (Fig. 8e). This moisture tongue is tied to northeasterly wind advection from moist convection occurring mostly from the north of 5–10∘ S (Fig. 8b and e). The north–south flight along 5∘ E on 30 August 2017 sampled moist conditions more consistently (Fig. 8e and f). Contrary to the relatively uniform BLH on 18 August 2017 (mainly due to the long duration of the flight over the same location), the BLH on 21 and 30 August 2017 varies, with alternatively dry (RH <40 %) and moist (RH >60 %) boundary layers. The AEJ-S on 21 and 30 August 2017 is strong, penetrating further west (Fig. 8e and f). The BC aerosol is also modulated by the intensity and horizontal extent of zonal wind (e.g., AEJ-S) around 700 hPa around 5–7∘ S, with some recirculated back around the anticyclone to the continents (Fig. 8j–l).

Figure 9 illustrates that most of the BC is located between 2–3 km, increasing in altitude towards the end of August 2017. The BC altitude is generally lower than that in September (compare Figs. 9a and 4a), consistent with a lower AEJ-S core in August compared with September (Fig. 9a and b). BC and RH in the mid-troposphere (∼3–4 km) covary over region B after 20 August 2017. Interestingly, the moisture can extend higher in altitude than the BC within the ERA5/CAMS reanalysis. Boundary layer BC aerosols are slightly more abundant in August 2017 compared to September 2016 (Figs. 4d, 9d), and the smoky boundary layer is well observed near Ascension Island in August 2017 (Zhang and Zuidema, 2019; Pennypacker et al., 2020). The boundary layer BC mixing ratio is higher over region A than region B (Figs. 8h, 9a–b).

Clear relationships exist between low CF and the boundary BC–temperature–RH for a few cases (e.g., 7–9 and 17–19 August 2017), but in general, there is no clear statistically significant correlation. AEJ-S is also less associated with LLJ in August (Fig. S5). Mean subsidence in August 2017 is slightly weaker than in September 2016, and the correlation between subsidence and low CF is insignificant at the daily timescale (Fig. 9d), consistent with De Szoeke et al. (2016) and Adebiyi and Zuidema (2018).

Figure 10 communicates the vertical structure of the moisture during the three flights highlighted in Fig. 8. High humidity levels (RH >60 %) remain confined over the continents on 18 August 2017 (Fig. 10a). After 18 August 2017, moisture began to advect off of the continent at 700 hPa (Fig. 10b). Both moist and dry convection over land were more effective at moistening the mid-levels (700–600 hPa) on 21 and 30 August 2017, after the AEJ-S became active, than on 18 August 2017.

The cross sections show that the horizontal temperature gradient over land is weak, as expected for the tropics (Fig. 10a). Note that the moisture at all levels does not extend as far south as in September. Over the land, the temperature inversion layer sits above the unsaturated, dry layer over the land, and the dew point temperature follows a constant water vapor mixing ratio (21, 30 August in the bottom panels of Fig. 10e and f), confirming that dry convection occurs after 20 August 2017. Considering conditions were uniformly dry before 20 August, neither dry nor moist convection was likely to be initiated in August 2017 over the Kalahari dryland (NK at 18∘ E, 16–17∘ S). Here the surface temperature increases consistently from 18 to 30 August, and the dry convective layer deepens (Fig. 10d–f). This transports more moisture upward, south of 10∘ S, into the 850–600 hPa altitude range. Over the Congo basin (CB at 18∘ E, 5–6∘ S), the depth of the relatively moist near-surface layer is highest on 21 August and highly variable between the three soundings. Unlike September 2016, CAPE is generally low. This is consistent with the lower rainfall in that month as compared to September. Note that we see an increase in surface dew points from 18 to 21 August over land, which can help increase the RH in dry convection regions at 700 hPa and along the flight track in August 2017.

During August 2017, low CF is also positively correlated with LTS, the strongest in region B, with a similar correlation coefficient as in September 2016 (R∼0.57; top panel of Fig. 11g). The BLH tends to be higher over the open ocean (10–40∘ W, 10–40∘ S) and land than near the west coast of Africa (Fig. 11d–f). The mid-latitude jets are very strong but remain mostly confined south of 30∘ S. The mid-latitude frontal system nevertheless influences the local circulation, enhancing SE Atlantic anticyclones on the west and cyclones on the east of a mid-level trough, leading to cloud clearing on 18 August 2017 (Fig. 11a and d) and 21 August 2017 (Fig. 11b and e). The increased low CF is weakly associated with the low BLH (e.g., 22 August 2017 in Fig. 11g), and low CF is almost larger than 0.8, rendering poor correlations to meteorological variations. The distributions of LTS, BLH, and low CF differ over regions A and B (Fig. 11h), particularly with tight relation between LTS and low CF over region B.

The LTS is generally lower in August than in September over both regions, implying that the atmosphere is less stable in August, with LTS increasing as the month progresses. The decreased atmospheric stability may reflect the influence of the mid-latitude frontal system, especially in region B, where recurrent cloud clearing by local pressure and wind patterns are associated with the intensified mid-latitude frontal systems. The mid-level flow and SLP patterns for all August 2017 flight days are provided in Fig. S14.

The final deployment of ORACLES also occurred in São Tomé (0.34∘ N, 6.73∘ E) in October 2018. The equatorial deep convective band moves south in October (Adebiyi et al., 2015), and biomass burning is substantially less (Redemann et al., 2021; Zhang and Zuidema, 2021). Thus, despite a pronounced Kalahari heat low supporting strong free tropospheric easterlies to its north, less BC is ultimately advected westwards. Continental convective systems migrate southward from 5∘ S, and mid-latitude frontal systems also assist the transition to austral summer. The AEJ-S was slightly weaker compared to the climatological mean, as were the LLJ and the southern Atlantic anticyclone (Ryoo et al., 2021), especially as the month progressed; the strength of the AEJ-S and heat low over land exceeded the September values from early to mid-October but weakened by late October 2018 (Figs. 12, S11, and S12). The monthly mean low CF is not much lower than the climatological mean, despite a warmer-than-climatology SST. The low CF is the largest of all deployment months, with more mid-tropospheric moisture potentially reducing entrainment drying from above (Zhang and Zuidema, 2021). The characteristics of synoptic-scale and convective features during the flight days in October 2018 deployment are summarized in Table 3.

Figure 12 shows the 6 h time series as a function of longitude averaged over the 8–10∘ S latitude band (i.e., Hovmöller diagram) of RH, BC, and zonal wind at 600 hPa and the heat low during October 2018. BC is also well modulated by AEJ-S, but its magnitude decreases significantly by the end of October when the biomass burning season tapers off amidst increasing RH indicative of southward-moving convection (Fig. 12a and b). After the easterlies weaken, the BC aerosol outflow reduces (e.g., 7–14 October and after 26 October; Fig. 12a, b, f). The modulation of the AEJ-S by the heat low is clear, similar to the other deployments (Figs. 2 and 7). The AEJ-S leads to high RH by 1–2 d, indicating that AEJ-S can transport elevated moisture from land to the ocean (Figs. 12f and S6).

This flow changes dramatically as the month progresses, however. From 1–22 October 2018, high RH is propagated over the ocean by a strong AEJ-S. By 23–28 October 2018, RH remains high over the continent, but does not propagate westward, because the mid-tropospheric winds are weak (Fig. 12e and f). The low-level easterly winds are also weak (Fig. 12d).

Satellite imagery further indicates that substantial deep convection (brightness temperatures <230 K) develops over the continent north of 10∘ S, starting in mid-October 2018 and spreading southward to northern Namibia and Botswana by 23 October 2018 (Fig. 13a and c). Routine flights along 5∘ E occurred on 7 October 2018 to 15∘ S, and on 15 October 2018 to 14∘ S, with the 23 October 2018 survey flight going west along 5∘ S. Like the other deployment months, we chose three October flight cases based on their unique meteorological and aerosol features. RH exceeded 60 % for all three flights (Fig. 13f). Between 7 and 15 October 2018 (around 09:00–11:00 UTC; shown in Fig. 13d and e), a strong mid-level AEJ-S also advected moisture (RH >60 %), leading to moist (RH >60 %) conditions at ∼15 kft (∼4.5 km). Moist air is sampled along the flight track on 23 October 2018, along with an elevated marine BLH (Fig. 13c and f). The core latitude of the AEJ-S varies day by day.

The AEJ-S is strong on 7 and 15 October 2018 but weakens considerably around 10∘ S on 23 October 2018 (Fig. 13f). This could be tied to the collapse of the dome of warm, dry air over the southern African highlands and a breakdown of the 600 hPa anticyclone over southern Africa (Fig. 13i). This also led to weaker moisture transport to the ocean at the south of 10∘ S on 23 October 2018. Two possible causes are that (1) the mid-latitude waves associated with a strong upper-level disturbance influence the breakdown of the dome of hot dry air (Kuete et al., 2021) and that (2) convection reduces the temperature gradient by increasing high cloud cover over the region, limiting surface heating over the Kalahari–Namib dryland (Fig. 13c). The heat low reforms toward the end of the month (Fig. 13g–i). Different BC aerosol transport patterns are also observed in flight days in October 2018. The intensity of the horizontal transport of BC largely aligns with the strength of the AEJ-S and the recirculating flow (Fig. 13j–l). The extensive BC transport over the ocean is much more suppressed on 15 and 23 October than on 7 October 2018.

Figure 14 illustrates the time evolution of coupled BC aerosol–meteorological variables and how they relate to low CF in October 2018. The covariation between BC–RH in the mid-troposphere (3–4 km) is also observed in October 2018, with the RH increase extending higher in altitude than the BC. The aerosol loading decreases around mid-October 2018, while moisture variations continue; this was clearly shown when a frontal system passed by. An increased BC aerosol loading at 925 hPa correlates with an increased 925 hPa temperature up to 7 October (Fig. 14d). The subsidence–low CF relationship in October 2018 is somewhat different from that in September 2016 and August 2017 in that subsidence and low CF are positively correlated.

Latitudinal cross sections of moisture and winds, along with soundings from 3 representative flight days, indicate that the transport of moist plumes to the ocean is also a component of the AEJ-S in October (Fig. 15a–c). A noteworthy feature of the 7 October sounding over the ocean is its resemblance to the dry convection layer over land, indicating that this moist plume also likely originated from dry convection over the land and was then advected westward by the AEJ-S (Figs. 5c and 15a). On 7 and 15 October, dry convection was still dominant over the Namibia–Kalahari dryland (18∘ E, 16–17∘ S), while moist convection developed over the Congo basin (18∘ E, 5–6∘ S). Later, as dew point temperature no longer followed the constant water vapor mixing ratio line over the dryland and approached the temperature, this indicated that the dry convection transitioned to deep moist convection (top panel of soundings in Fig. 15d–f). On 23 October, the moist plumes with high RH (>60 %) reached 10–20∘ S (Fig. 15d–f), with moist convection over both Congo basins and Namibia–Kalahari dryland (see satellite image; Fig. 13c). Deep cloud layers, identified by locations where the dew point temperature matches the environmental temperature curves, are clearly shown on 23 October 2018 (soundings in both CB and NK in Fig. 15f). With such deep clouds, RH reaches ∼100 %, indicative of heavy rainfall.

The positive correlation between LTS and low CF is weaker in October 2018 (e.g., 10∘ W–0∘, 5–15∘ S) and statistically more insignificant than during the other 2 deployment months (see Fig. 16), partially because neither the LTS nor low CF vary much in October 2018, particularly over the north of 20∘ S. When the subsidence weakens compared to August and September, low CF and LTS remain mildly correlated, especially at intermediate LTS values (16–20 K; Fig. 16hs; Zhang et al., 2009). Similar to August and September, the LTS–low–CF correlation in October is more evident in region B than in region A (Figs. 6g and 11g). As LTS strengthens, the low–CF increases, as shown in Fig. 16a–c. The BLH is higher over the ocean (10–40∘ W, 10–40∘ S) and shallower near the west coast of Africa (Fig. 16d–f). Low CF variability is also relatively invariant with BC aerosol (this is also clear in region A at 0–10∘ E, 5–15∘ S), while a stronger LLJ tends to reduce low CF during October (Fig. S8c). The mid-tropospheric winds and LLJ, both strong during October 2018, are also the most correlated (R∼0.55; p value <0.01) of the 3 deployment months (see the AEJ-S–LLJ correlation analysis in Fig. S7). A sufficiently strong temperature inversion and moist and warm conditions associated with the mid-latitude frontal system may provide favorable conditions for both the AEJ-S and LLJ to develop together (Fig. 16a–f).

In particular, low CF slightly decreased on 23 October 2018 (Fig. 16c and g), but the high cloud cover increased (not shown), indicating not only day-to-day variability but also that a cloud regime transition to a deep convection (Fig. 15f) may occur, especially south of the flight region (<20∘ S). The low clouds also tend to respond to the change in the BLH, and large low CF is weakly associated with a lower BLH on some days (e.g., 6, 18, 25, and 30 October 2018; see the bottom plot in Fig. 16g), but the overall correlation between the BLH and low CF is low and statistically insignificant (Fig. 16g). This can be partially explained by large day-to-day variability from passing mid-latitude frontal systems. The developing mid-latitude frontal systems, indicated by (1) the reduced St. Helena high and LTS, (2) decreasing SLP slightly east of the mid-level trough, and (3) misalignment of mid-level wind vector and mid-level geopotential height, promotes a increased disturbance and vertical motion, eventually intensifying the frontal system (Holton, 2004). The mid-latitude upper-level disturbance also contributes to modulating the temperature gradient over SE Atlantic (Ryoo et al., 2021), ultimately affecting local circulation. Further examination of the mid-level flow and SLP patterns for all October flight days is shown in Fig. S15.