For this study, the regional-scale model framework COSMO-ART Consortium for Small-scale Modeling – Aerosols and Reactive Trace gases; is used. COSMO-ART is based on the operational weather forecast model COSMO of the German Weather Service (DWD). The ART extensions allow for an online treatment of aerosol dynamics and atmospheric chemistry. The model application of this study is accompanied by significant further developments of the emission parameterizations regarding mineral dust and BVOCs . Furthermore, a parameterization for trace gas emissions from gas flaring from the oil industry was developed to reproduce the specific pollution conditions of the research area . The model domain comprises Ivory Coast, Ghana, Togo, Benin and the Gulf of Guinea (red rectangles in Fig. ). The model setup is summarized in Appendix .

The simulations using the setup denoted in Table  are the result of a nesting into a 5 km COSMO-ART simulation (blue rectangle in Fig. a) using the ICON operational forecasts (approximately 13 km grid mesh size) as meteorological boundary conditions. These cover the time period 25 June to 3 July to allow for an aerosol–chemistry spin-up. The meteorological state is initialized every day at 00:00 UTC.

To assess the sensitivity of the ADE and AIE on the meteorological conditions, two factors FADE and FAIE were introduced in COSMO-ART, which allow for the scaling of the total aerosol mass and number densities, respectively, by simultaneously preserving the underlying aerosol distribution. All aerosol modes are changed uniformly by the factors. Note that the aerosol scaling only comes into consideration when deriving the aerosol optical properties (with respect to ADE) and the aerosol activation (with respect AIE). All aerosol dynamic processes remain unaffected by the scaling. Within a simulation the constraint FADE = FAIE is used to allow for physically consistent results. Table  summarizes the realizations used in this study. FADE=FAIE=1.0 is used as the reference case, whereas the factor variations 0.1, 0.25, 0.5, 2.0 and 4.0 are applied to assess the aerosol sensitivity.

The terms clean, reference and polluted should be seen in a relative sense as part of this experimental setup. They do not imply general evaluation of the SWA aerosol conditions. The period 2–3 July was selected due to the intense and persistent NLLS at the Savè supersite during that time . Furthermore, 3 July is the center of the monsoon post-onset phase and it is expected that the undisturbed monsoon conditions favor and support the process studies. Since the meteorological conditions show less variation from day to day, it is assumed that, even with a focus on a very short time period, insight can be achieved that can be generalized at least qualitatively to the length of the post-onset phase (22 June–20 July).

Within the DACCIWA project, an extensive field campaign took place in June–July 2016 in SWA (Fig. b) . The time period was selected to capture the onset of the WAM and a period characterized by increased cloudiness. The DACCIWA ground-based measurement campaign encompassed the time period from 13 June to 31 July 2016, including the three supersites Kumasi (Ghana), Savè (Benin) and Ife (Nigeria; red dots in Fig. b). A complete overview of the DACCIWA ground-based measurement campaign, their supersites, instrumentation and a first insight into the available data are presented in . The DACCIWA airborne measurement campaign captured the time period from 27 June to 17 July 2016 . The focus of this study is on Ivory Coast and therefore fewer observational data from the DACCIWA campaign are available for evaluation. However, a substantial evaluation with respect to meteorology and air pollution is realized with COSMO-ART over SWA with respect to other time periods and by focusing on the eastern part of the research area, which is not presented in this study but can be found in . For this study, observations of liquid cloud properties from the CDP-100 (cloud droplet probe, data revision 3) of the British Antarctic Survey (BAS) Twin Otter aircraft on 3 July 2016 are used for a comparison with COSMO-ART. The CDP-100 is a wing-mounted canister instrument including a forward-scatter optical system to measure the cloud droplet spectrum between 2 and 50 µm with a frequency of 1 Hz. The aerosol number density is evaluated using observations of the ATR42 SAFIRE (Service des Avions Français Instrumentés pour la Recherche en Environnement) for 3 July 2016. Additionally, the comparison of the modeled net downward shortwave and longwave radiation as well as the sensible and latent heat flux with Savè supersite observations is presented in Fig.  of Appendix . COSMO-ART reasonably reproduces the fluxes with lower fluxes with increasing aerosol as expected.

After describing the general AI properties in Sect. , this section presents the first insight into the aerosol influence on AI. To give an impression of the aerosol change related to the clean, reference and polluted case, Fig.  in Appendix  shows the mean total aerosol number and mass over Ivory coast in the lowest 2 km a.g.l. as a diurnal cycle from 2 July 15:00 UTC to 3 July 21:00 UTC. Figure  shows the horizontal wind speed difference at 250 m a.g.l. (hs) on 2 July 22:00 UTC between the clean and the reference case together with the corresponding AI front locations. The wind speed difference exhibits a filament structure in the zonal direction that covers nearly the entire SWA domain. Furthermore, it propagates inland with time (not shown).

Especially over Ivory Coast, a coherent pattern can be observed with a spatial shift between the two AI fronts with that of the clean case (black dashed line in Fig. ) ahead of the reference case front (black solid line in Fig. ). This anomaly pattern results from the fact that postfrontal wind speeds are generally higher than the prefrontal wind speed, as shown in Fig. a. To assess the aerosol impact on the vertical structure of the AI, Fig.  shows the meridional–vertical transect of wind speed and potential temperature for the clean (Fig. a) and polluted case (Fig. b) in the same way as presented for the reference case in Fig. a.

When comparing the results between the clean and polluted cases and by considering the reference case (Fig. a) as intermediate, aerosol-specific characteristics can be identified in addition to the general AI characteristics presented in Sect. . Whereas the temperature characteristics over the ocean are similar for the realizations, the inland temperature decreases with increasing aerosol amount. This is especially visible in the prefrontal area (Fig. ). In the polluted case the advective cooling is more effective, since the daytime inland near-surface air is a priori cooler due to a lower sensible heat flux from aerosol extinction. The reduced ocean–land temperature gradient in the polluted case leads to reduced temperature contrasts at the AI front (compare the 302 K isentrope for the clean case, bold line in Fig. a, and the polluted case, bold line in Fig. b). With the change in the ocean–land temperature gradient, the AI frontal position and the NLLJ strength and vertical extension are altered. The higher the aerosol amount, the more the AI front is lagging behind and the weaker the NLLJ. In the polluted case the vertical extension of the inland NLLJ and its wind speed in the jet axis are reduced by about 150 m and 2–3 m s-1, respectively. The AI frontal difference averaged over Ivory Coast at 21:00 UTC is 10 km between the clean and reference case and 20 km between the clean and the polluted case. With the decrease in temperature with increasing aerosol amount, the prefrontal wind speed generally increases (in contrast to the postfrontal area). This leads, with respect to the polluted case (Fig. b), to some areas of increased wind speed in the prefrontal area at a height that is typical of the NLLJ. Generally, the polluted case is characterized by a blurring of the prefrontal and postfrontal temperature and wind speed differences.

The temporal evolution of the median coastal distance of the AI front over Ivory Coast is presented in Fig.  for the six realizations. The polluted case (solid red) denotes a special case in Fig. . The other realizations with scaling factors between 0.1 and 2.0 show a systematic behavior. As expected, the front propagates inland with time. The higher the aerosol amount (dashed blue to solid green) the slower the inland propagation. This leads to a median spatial difference of about 27 km between factor 0.1 (dashed blue) and 2.0 (solid green) at 22:00 UTC.

Furthermore, Fig.  reveals two regimes, one before and one after 17:00 UTC. For the latter the frontal propagation diverges according to the aerosol amount as described above. Before 17:00 UTC an opposite behavior can be observed leading to the circumstances that with less aerosol the AI front is closer to the coastline. Therefore the starting point for the inland propagation at 15:00 UTC is not equal for all realizations but a reversed order can be observed compared to the situation at 22:00 UTC. The underlying mechanisms for the occurrence of these two regimes that switch around 17:00 UTC are assessed in the subsequent section. With a further increase in the aerosol amount, as realized in the polluted case (solid red line in Fig. ), the ocean–land temperature gradient is reduced as shown in Fig. b. The AI front evolution is therefore less pronounced than for the other realizations. In the eastern part of Ivory Coast the location of θs persists inland and does not form a coherent front near the coast. Averaging the frontal location over Ivory Coast therefore leads to a temporal evolution, which does not follow the behavior of the other realizations. By reducing the benchmark of θs from 302 to 301 K, the polluted case also follows the trend of a weaker frontal propagation with increasing aerosol (gray line in Fig. ). In this polluted case, with its cooler lower layers, the 301 K isentrope better represents the frontal location, as is also visible in Fig. b.

After diagnosing the characteristics of AI and the AI changes with changing aerosol amounts in Sect.  and , respectively, the question of the underlying feedback mechanism arises. The stationarity of the AI front near the coast in the early afternoon is related to the balance between the onshore-directed monsoon flow over the ocean and the turbulence over land e.g.. Therefore the change in turbulence can alter the balance and lead to differences in the AI front propagation (turbulence mechanism). With increased aerosol amounts, a near-surface cooling (relative to the realizations with less aerosol) is expected during daytime (as observed in Fig. b) either due to the ADE via scattering and absorption on aerosol particles or due to AIE via an increased albedo with reduced cloud droplet effective radii (Twomey effect). A reduced surface heating with increased aerosol amounts might lead to an increase in surface pressure. By considering the fact that the sea surface temperature (SST) is fixed in COSMO-ART and that the surface temperature over the ocean will therefore not be subject to substantial changes, a reduction in the land–sea pressure gradient can be expected, which could affect the AI front propagation (pressure gradient mechanism). In order to further elaborate this, Fig.  shows the spatial distribution of total cloud water (Fig. a) and precipitation (Fig. b) for the reference case on 2 July 15:00 UTC. The red line denotes the AI front. Clouds and precipitation occur primarily in the AI postfrontal area over Ivory Coast due to convergence and vertical lifting, upstream of mountain areas due to topographic lifting (especially at the Mampong Range and the Atakora Mountains) and via localized convection (primarily in the AI prefrontal area over Ivory Coast).

To shed light on potential effects from the turbulence mechanism and the pressure gradient mechanism, Fig.  shows the differences in surface net downward radiation (Fig. a), 2 m temperature (Fig. b) and sea level pressure (Fig. c) between the reference and the clean case. In addition, Fig. b, d and f present the same variables but for the areas that are cloud free in both realizations to exclude effects from displaced clouds and to highlight the ADE in a cloud-free environment.

The differences in surface meteorological quantities presented in Fig.  reveal a clear signal. The following values in brackets indicate the median and the 99th or 1st percentile of the surface quantities considering the cloud-free inland area. With increasing aerosol, more downward shortwave radiation is scattered and absorbed, leading to an average decrease in surface net downward shortwave radiation (-37 and -185 W m-2; Fig. b). The decrease in incoming shortwave radiation leads to a decrease in 2 m temperature (-0.5 and 2.5 K; Fig. d). The temperature decrease furthermore leads to a domain-wide inland surface pressure increase (+0.16 and +0.45 hPa; Fig. f). Omitting the negative pressure anomaly over Lake Volta due to the fixed SST, slightly higher domain-wide pressure anomalies are found (+0.17 and +0.45 hPa; Fig. f).

To prove the hypothesis that the surface pressure difference is caused by the temperature difference, the surface pressure over Ivory Coast at 15:00 UTC (over land and in cloud-free areas) is estimated from the pressure and temperature at 850 hPa using the barometric equation. This approach yields a spatially averaged value of +0.12 hPa that broadly agrees with the modeled value of +0.17 hPa (compare Fig. f). It can be concluded that the pressure changes are dominated by changes in low-level temperature. In the cloudy areas around the SWA mountains a higher pressure difference can be observed (Fig. e), indicating that cloud radiative effects are also contributing. To assess whether the reduction of incoming shortwave radiation due to clouds is related to a change in the cloud water content and therefore the optical thickness of the clouds or due to the Twomey effect with a change in cloud droplet number concentration (CDNC) and effective radius, Fig.  exhibits the empirical cumulative distribution function (ECDF) with respect to the COSMO-ART realizations of the CDNC (Fig. a), cloud droplet effective radius (Fig. b), cloud water (Fig. c) and precipitation (Fig. d). This figure corresponds to the cloud and precipitation patterns presented in Fig.  for 2 July 15:00 UTC.

A strong susceptibility of the CDNC and effective radii towards a change in the aerosol amount can be observed (Fig. ). The factor variation from 0.1 to 4.0 leads to an increase in the median CDNC by 1 order of magnitude from 100 to 1000 cm-3 (Fig. a) and a reduction in the median effective radius from 9 to about 3.5 µm. When considering the green and red curves in Fig. , which are related to an aerosol change symmetrically around the reference case (black), the effect on the CDNC and effective radius is nonlinear e.g.. An aerosol increase (solid green and red lines) has significantly stronger impacts than the aerosol decrease (dashed green and red lines).

In contrast to these remarkable changes, the effect on cloud water and precipitation (Fig. c, d, respectively) is insignificant. Except for the polluted case (solid red lines) all realizations show similar ECDFs, indicating that the aerosol increase neither leads to a cloud water increase due to precipitation suppression or due to enhanced water vapor condensation on the aerosol particles nor a cloud water decrease via enhanced evaporation. The polluted case shows a tendency of precipitation decrease (increase) for the weak (strong) precipitating areas related to an increase (a decrease) in cloud water. This effect of greater local rainfall amounts is in agreement with the findings of , likely via the convective-cloud invigoration mechanism. However, the deviations from the other realizations are small. Figure  reveals that the aerosol impact on radiation via the Twomey effect is very likely dominating the cloud–radiation interaction, whereas the cloud optical thickness impact (via a change in the amount of cloud water) is of minor importance. The weak precipitation response to the changing aerosol amount underlines the finding that radiation and its variation are the key players in the observed changes over SWA due to the ADE in and outside of clouds and the Twomey effect. There is ongoing work within DACCIWA with respect to large eddy simulations (LES) of aerosol–atmosphere interactions. It will be of interest to see whether the COSMO-ART results are consistent with the outcomes of these studies on smaller scales.

The turbulence and pressure gradient mechanisms are counteracting. With respect to the turbulence mechanism, a reduced heating weakens the turbulence in the PBL. Therefore the AI balance between the monsoon flow and the inland turbulence is shifted to the monsoon flow, favoring an inland propagation. Regarding the pressure gradient, a reduced heating decreases the land–sea pressure gradient, shifting the AI balance in the opposite direction and suppressing the inland propagation. When going back to the temporal evolution of the coastal distance of the AI front in Fig. , both mechanisms are evident and cause the two observed regimes before and after 17:00 UTC. The first regime includes the stationary phase of the AI front near the coast. With the decrease in incoming solar radiation with increasing aerosol, the turbulence decreases and therefore the stationary front location shifts inland. Unfortunately, the AI front detection via the θs method fails for the time period earlier than 15:00 UTC. Therefore the total difference in the stationary AI front location with changing aerosol cannot be assessed. Nevertheless, it is interesting that the location of the AI front during its stationary phase over Ivory Coast could be used as a proxy for the aerosol burden in that area (under otherwise identical conditions).

For the time period after 17:00 UTC when turbulence has decreased sufficiently, the pressure gradient mechanism dominates because the AI front in the clean case, although lagging behind at 15:00 UTC, is 11 km ahead of the reference case at 22:00 UTC (Fig. ).

Figure  summarizes the counteracting components of turbulence and pressure difference that govern the inland propagation of the AI front by comparing the temporal evolution of the differences between the reference and clean case (dashed lines) and the polluted and reference case (solid lines) in surface sensible heat flux (red, positive downward), surface latent heat flux (green, positive downward) and surface pressure (blue) spatially averaged for the AI prefrontal area over Ivory Coast.

The temporal evolution clearly shows that the sensible and latent heat flux differences (and the absolute value itself) decrease strongly with time in contrast to the pressure differences. After sunset (e.g. 18:24 UTC at Kumasi) the sensible and latent heat flux is negligible but the pressure differences continue. In fact, the altered land–sea pressure gradient is maintained until the AI front and the subsequent cool air mass have passed the area and compensates for the differences (not shown). It is expected that the high moisture in the monsoon layer prevents it from cooling significantly and reducing the differences that developed during daytime. The factor increase of 4 from 1.0 to 4.0 reduces (increases) the sensible heat flux (sea level pressure) more than the increase from 0.25 to 1.0, in agreement with the findings of the sensitivity of CDNC and effective radius in Fig. .

The monsoon flow over SWA is driven by the temperature gradient between the cool SSTs over the eastern equatorial Atlantic Ocean that are fixed in the model and the Saharan heat low that is not part of the modeling domain. With this location of the modeling domain, changes in the aerosol amount can serve as an amplifier for the monsoon flow that is able to increase or decrease the temperature gradients and thereby the AI front characteristics. In agreement with this, show that the sea breeze front over Cameroon weakens with enhanced aerosol number concentration. Longwave cooling is not significantly reduced, likely due to water vapor saturation in the monsoon layer (not shown). In contrast, the coherent differences in 2 m temperature and pressure, which were observed at 15:00 UTC (Fig. ), also persist during nighttime. The daytime heating of the land, stronger in the clean case and weaker in the polluted case, persists during night and exceeds potential effects from longwave cooling. The differences between the realizations are finally equalized by the passage of the AI front and postfrontal air mass.

In addition to the aerosol impact on AI, impacts on the SCT can also be observed. Figure  shows the vertical transect of modeled cloud water between Lomé and Savè (Fig. b) regarding the clean case (left) and the reference case (right) for 2 July 10:00 UTC (top) and 11:00 UTC (bottom). The red shaded area below the cloud layer denotes the development of the convective boundary layer (CBL) identified by dθ / dz<0. The black (red) solid line shows the top of this unstable layer regarding the reference (polluted) case to allow for comparison between the three realizations.

After sunrise the CBL starts to evolve. Via the same mechanism as described in Sect. , less shortwave radiation reaches the ground with increased amounts of aerosol and therefore the surface sensible and latent heat fluxes also decrease. This leads to a decelerated daytime CBL development and with that to a reduction of the cloud base height (Fig. , left). To underline the fact that this effect is visible not only in the Lomé–Savè transect but also for the entire SWA region, Fig.  shows the temporal evolution of the spatial average of total cloud cover (Fig. a), total cloud water (Fig. b) and the cloud base height (Fig. c) over SWA for the clean (blue dashed), reference (black solid) and polluted case (red solid).

Between 21:00 UTC and sunrise (05:30 UTC) the cloud cover increases (Fig. a) due to clouds that are advected onshore or develop inland. This is linked with a reduction in the mean cloud base (Fig. c). Between 01:00 and 07:00 UTC the clean case shows lower cloud base values than the reference and polluted cases. A detailed analysis reveals that this deviation is not related to NLLS but to mid-level clouds over the Lake Volta basin and in the northwestern part of the domain (not shown). After sunrise it is assumed that the NLLS intensifies via the vertical mixing of moisture in the developing convective PBL. With respect to the spatial average in Fig. c this leads to a reduction in mean cloud base height. The maximum cloud cover (Fig. a) is related to the minimum cloud base (Fig. c), underlining the dominance of NLLS. After reaching the cloud cover maximum, the SCT continues, which is related to a lifting of the cloud base and a decrease in cloud cover. For this SCT a clear temporal shift of about 1 h can be observed between the clean and the reference case and 2 h between the clean and the polluted case. The realizations with increased aerosol amounts react slower to the insolation after sunrise, reach the NLLS maximum coverage later and start later with the SCT as observed for the Lomé–Savè transect in Fig. . After 15:00 UTC this finally leads to a cloud cover that is increased compared to the clean case (Fig. a), implying an additional reduction in surface shortwave radiation that can be used for further cooling the surface and decelerating the AI front. The cloud water (Fig. b) shows a similar temporal shift with increasing aerosol amounts as for the cloud cover and cloud base. The weakening of the SCT with a higher aerosol burden leads to reduced amounts of cloud water after 13:00 UTC (Fig. b), likely due to reduced convective activity. However, during nighttime, the polluted case uniformly shows higher cloud water values than the clean and reference cases.

Figure  in Appendix  shows the cloud analysis restricted to clouds below 1500 m a.g.l. to assess the sensitivity of the spatial averaging towards the considered vertical column. The cloud cover (Fig. a) shows a similar temporal evolution as presented in Fig. a. The cloud water and cloud base temporal evolution in the lowest 1500 m a.g.l. (Fig. b, c) show less variation between the realizations compared to Fig. b and c. However, the temporal shift in the onset of the SCT is obvious in both figures. As expected, the initiation of the cloud base increase via the SCT occurs earlier when considering only the clouds below 1500 m a.g.l. in the averaging (compare Fig. c with Fig. c).

The aerosol feedback process study simulations presented in Sect.  and revealed several mechanisms relevant for SWA affecting the location and propagation of the AI front and the temporal evolution of the SCT. In the following section a proposal for a conceptual model will be presented.

This section aims to synthesize the findings that have been obtained with this aerosol feedback process study. We showed that AI affects the entire SWA domain through the course of the day via cold air advection, the NLLJ that can be found in the AI postfrontal area, and convergence-induced convection and precipitation. Two distinct meteorological responses to changes in the amount of aerosol via ADE and the Twomey effect were identified: a spatial shift of the Atlantic inflow (AI) and a temporal shift of the stratus-to-cumulus transition (SCT).

Figure  shows a conceptual scheme that combines both responses. The bigger loop is related to the first response (AI) and the smaller loop to the second (SCT).

Following the AI loop in Fig. , the increase in the amount of aerosol (number and mass) by a factor of 4 (0.25 to 1.0) is the initial perturbation of the system. The subsequent numbers in parenthesis are related to the median value over Ivory Coast (cloud-free inland areas) on 2 July 15:00 UTC to provide guiding values for the denoted changes.

Via ADE the aerosol increase leads to a decrease in surface net downward shortwave radiation (-37 W m-2) and surface temperature (-0.5 K). Previous studies showed that until the early afternoon, the AI front is stationary near the coast due to the balance between the monsoon flow from the sea and the sensible heat flux (turbulence) over land. With the afternoon decrease in sensible heat flux, the AI front propagates inland. This study showed that the decreased surface heating leads to a positive pressure anomaly over land (+0.16 hPa) and with that to a reduced land–sea pressure gradient. The latter is more persistent than the sensible heat flux that vanishes around sunset (compare Fig. ). The reduced pressure gradient leads to a reduced AI frontal velocity and therefore to a southward shift in the case of increased aerosol (11 km on 2 July 22:00 UTC). The postfrontal area is characterized by stronger wind speeds in the lowest 1000 m a.g.l. with the maximum around 250 m a.g.l. that is characteristic of the NLLJ. Therefore an AI frontal shift leads to a shift in the NLLJ inland propagation. Since the AI frontal propagation is linked to convergence-induced convection and convective precipitation, a meridional shift of the AI-related precipitation is also observed. These effects are primarily related to the afternoon but the AI frontal and NLLJ shift also leads to a shift in the inland propagation of coastal NLLS with a similar spatial magnitude as observed for the AI front (not shown).

The AI loop denoted in Fig.  includes a further mechanism related to the counteracting effects of the monsoon flow over the ocean and the sensible heat flux over land in the stationary phase of the AI front. With increasing aerosol the inland sensible heat flux decreases, which relocates the front farther from the coast. Therefore with increased aerosol the AI frontal inland propagation starts farther from the coast but is slower than in the low aerosol case due to the reduced land–sea pressure gradient as soon as the turbulence has declined after sunset.

The SCT loop is coupled to the AI loop via the decrease in surface shortwave radiation and temperature. This study pointed out that the deficit in surface heating due to ADE and cloud brightening via the Twomey effect lead to a decrease in sensible heat flux and therefore to a delayed development of the CBL. The lower CBL height leads to a lower cloud base and therefore to a later SCT and breakup of the closed cloud layer to scattered cumuli (compare Fig. a). Both loops are initialized after sunrise with the input in shortwave radiation. The SCT loop implies a positive cloud cover anomaly after 15:00 UTC with increasing aerosol. Sunset is around 18:30 UTC. Although the AI front already starts penetrating inland around 14:00–15:00 UTC, approximately a 3.5 h period is available for an additional surface cooling from the later cloud-layer breakup. This is a pathway for a further deficit in surface shortwave radiation and surface heating that could further weaken the AI loop as emphasized by the red arrow in the opposite direction in Fig. . However, the latter coupling between the two loops is only hypothesized. A future study needs to assess the significance of the contribution to inland surface pressure increase that comes from the deficit in shortwave heating via the later cloud-layer breakup.

The mechanisms described in Fig.  raise questions about the possibility to generalize these results. The AI feature is very likely a regular phenomenon under undisturbed monsoon conditions as confirmed by previous studies that focus on longer time periods. Within this process study the AI frontal shift was obvious for both days in the evening. However, the results presented above are related to Ivory Coast, which shows a more coherent AI frontal pattern than the eastern part of the domain, likely related to topographic features. This conceptual picture reveals radiation as a key player governing the feedbacks either via ADE or via a change in cloud albedo (Twomey effect). The AIE assessment within the process study reveals the known mechanisms, in particular the increase (decrease) in the CDNC (effective radius) with an increase in the aerosol number concentration. However, the AI-related clouds and precipitation reveal, aside from a meridional shift, no statistically significant difference. Although, the possibility for substantial effects from AIE cannot be excluded, a conceptual view as presented for the radiative effects has to be left for subsequent studies.