Increasing manmade air pollution likely to reduce rainfall in southern West Africa

Southern West Africa has one of the fastest growing populations worldwide. This has led to a higher water demand and lower air quality. Over the last three decades, most of the region has experienced decreasing rainfall during the little dry season (LDS, mid-July to end of August) and more recently also during the second rainy season (SRS, September–October), while trends during the first rainy season (FRS, mid-May to mid-July) are insignificant. Here we use spatio-temporal variations of precipitation, radiation, cloud and visibility observations from surface stations and from space to investigate whether the 5 increased pollution has contributed to suppressing rainfall by dimming incoming solar radiation. To isolate a potential aerosol influence, a multi-linear regression model based on sea-surface temperature (SST) indices is used. During both LDS and SRS weakly statistically significant but accelerating negative rainfall trends unrelated to known climatic factors are found. These are accompanied by a strong increase of pollution over the upstream tropical Atlantic caused by fire aerosol from Central Africa, particularly during the LDS. Over southern West Africa, where no long-term aerosol records are available, significant decreases 10 in horizontal visibility and incoming surface solar radiation are consistent with the hypothesized pollution impact. The latter trend is further enhanced by an increase in low-level cloudiness. The larger spread of potentially aerosol-related effects during the LDS is consistent with the stronger monsoon flow and less wet deposition. Negligible aerosol impacts during the FRS are likely due to the high degree of convective organization, which makes rainfall generation less sensitive to surface radiation. The overall coherent picture and the accelerating trends – some of which concealed by SST effects – should alarm policymakers in 15 West Africa to prevent a further increase in air pollution, as this could endanger water supply, and food and energy production for a large and growing population.

against changes in the station network (Chris Funk, personal communication) and a successful use in trend studies in SWA (e.g., Bichet & Diedhiou, 2018), large changes in surface station availability (e.g., https://data.chc.ucsb.edu/products/CHIRPS-2.0/ duce inhomogeneities in long-term trends. This has been noticed by Diem et al. (2019) for western Uganda, who stress the necessity to validate satellite-derived trends by ground-based measurements. Furthermore, the algorithm tends to smooth spa-130 tial inhomogeneities.
Therefore as an additional source of precipitation data in the region, the Karlsruhe African Surface Station Database (KASS-D), containing daily, quality-controlled rain gauge measurements from manned weather stations operated by National Weather Services (e.g., Vogel et al., 2018) has been used in this study. Only stations with at least 50% data coverage are considered to allow a meaningful trend analysis. As this criterion is hardly fulfilled before 1983 and after 2015, the main analysis is restricted 135 to this period. In addition, the trend analysis is repeated for the more recent, shorter timespan 2001-2017, for which surface observations of incoming solar radiation and more satellite data are available (see Section 2.4). Unfortunately, the availability of KASS-D data deteriorates during this period, mostly in Nigeria where our data base has little data after 2015. Despite this, we decided not to end this recent period in 2015 because trends become less meaningful for shorter timespans. Note that in the 1990s to 2010th, KASS-D contains daily data from many station ins SWA that have not been used in CHIRPS. trend calculations due to inconsistencies in the measurements. However, a shorter time series (2002)(2003)(2004)(2005)(2006)(2007)(2008)(2009)(2010)(2011)(2012)(2013)(2014)(2015) from Djougou, about 100 km northeast of Parakou, indicates that the observations in Parakou are consistent (not shown). No visibility data are available for Lamto, but human observer estimates of total cloud area fraction (TCAF) for January 2000-July 2016. From these (sub-)daily observations seasonal, respectively monthly, averages were calculated, again with a 50% data coverage criterion.
For a more complete look at radiation, monthly data of SDSR and effective cloud albedo (ECA) of the SARAH-2 data set 150 from the Satellite Application Facility on Climate Monitoring (CM SAF, Pfeifroth et al., 2017) are analyzed. The comparison of normalized trends (see Section 2.4) of SDSR and ECA allows us to estimate a residual potentially related to aerosol. This same technique of normalization is also applied to observations from the surface stations in Parakou and Lamto.
Monthly satellite measurements of aerosol optical depth (AOD) on a 1 • ×1 • horizontal grid from the Moderate Resolution Imaging Spectroradiometer (MODIS) (Platnick et al., 2017), i.e., the "combined dark target and deep blue AOD at 0.55 micron 155 for land and ocean: mean of daily mean", are used to calculate seasonal trends between July 2002 (beginning of data set) and October 2018. As these are monthly data, they are weighted when calculating season-averaged values. AOD for the LDS (15 July-31 August), for instance, is computed as AOD(LDS)=(0.5·AOD(July)+1.0·AOD(August))/1.5. Especially over land 5 https://doi.org/10.5194/acp-2020-463 Preprint. Discussion started: 29 May 2020 c Author(s) 2020. CC BY 4.0 License. clouds often inhibit AOD measurements from space, leading to many missing values in the MODIS monthly products. To cover the diurnal cycle, only months are used, when AOD data from both the Aqua and Terra platforms are available, which are then 160 averaged to obtain one single monthly value. For every year a seasonal mean value is computed if data are available for all months of the respective season. Again a sufficiently complete record with data for at least 50% of the seasons between 2002 and 2018 is required before calculating a trend.
To give climatological context to the trend analysis presented here, relative humidity, cloud cover, and meridional wind speed data from the fifth generation of European Centre for Medium-Range Weather Forecasts (ECMWF) atmospheric reanalyses The other three indices are computed from SST values from the Hadley Centre Sea Ice and Sea Surface Temperature (HadISST) data set. The same weighting as for the AOD data is used for the conversion from monthly climate indices to seasonal averages.

Methods for trend analysis
Long-term trends for seasonal rainfall totals are computed for every CHIRPS grid point and KASS-D station, applying Sen's 175 slope method (Sen, 1968;Hirsch et al., 1982;Hipel & McLeod, 1994). The resulting trends are tested on statistical significance using the Mann-Kendall test (Mann, 1945;Davison & Hinkley, 1997;Hipel & McLeod, 2005). Generally, trends are considered statistically different from zero if the two-sided p-value is smaller than the tested significance level α. All our tests are performed for α-values of 5% and 20%, the latter being a relatively weak criterion for statistical significance. As discussed in the context of "climate change attribution" by Lloyd & Oreskes (2018) and Knutson et al. (2019), the choice of significance 180 levels depends on how one intends to interpret the results. Small α-values reveal high confidence that a trend found in the data has some other reason than natural variability. However, a less strict significance level (α = 20% in our case) can still be useful for a study like ours that is dealing with the challenge of a relatively short and incomplete data record subject to multiple influence factors and possibly non-linearities. In such a situation choosing a large α implies reducing so-called "type II errors", i.e., retaining the null hypothesis of no trend beyond natural variability, although such a trend actually exists but is hard to detect 185 with the information at hand. Taking into account additional factors such as geographical distribution and seasonal behavior can help in the evaluation of a trend with weak statistical significance. Ultimately, it is the balance of all available evidence that does or does not suggest that an identified trend has others than natural causes (Knutson et al., 2019) and this is the philosophy we are following in this study. We feel that such an approach is particularly justified in the given situation, as we urgently need a risk assessment for a potential local human influence on rainfall in SWA.

190
Interannual variations of precipitation in SWA are, dependent on the exact region and season, influenced by (remote) climate indices. In order to distinguish between these climatic signals and other parameters such as (local) air pollution, we calculate at the far northeastern fringe of the study region show a lower correlation, indicating problems with data or stronger terrestrial forcings farther away from the ocean. The simple picture that emerges from this analysis is that a considerable part of the season-to-season variability is linearly controlled by the SST over the nearby Atlantic Ocean, with warmer waters leading to more rainfall. This is related to a weaker temperature and pressure gradient towards the Sahel, which leads to stronger convergence and higher moisture content in SWA (Losada et al., 2010).

220
The picture during the SRS is also dominated by the Atlantic 3 index but in the northern and especially eastern parts of the region, other climate indices yield better linear models (Fig. 2c). Given the weaker monsoon flow during this season, this may indicate a less close link to the nearby Atlantic Ocean and more room for teleconnections to influence rainfall. Correlation coefficients are generally lower than during the LDS but still reach 0.5 across considerable parts of the region, particularly in the coastal strip (marked by a pink line in Fig. 2d). It is striking that the regions that do not show the simplest model based 225 on the Atlantic 3 index also show lowest correlations, which again points towards data problems and/or more local influences such as topographic forcing by the Oshogbo Hills in Nigeria.
During the FRS, the Atlantic 3 index is also dominating but regions where other indices yield better linear models are larger than during the other two seasons ( Supplementary Fig. 1a). The correlation of the best models with rainfall is low during the FRS rarely exceeding values of 0.5 ( Supplementary Fig. 1b) showing a rather weak effect of ocean temperatures on rainfall 230 during this season. As before, the agreement between station and CHIRPS data in both the optimum index combination and the correlation is very good. The optimal linear models described above are now used to remove the dominant influences of SST fluctuations from the original time series. Trends in these residual time series will differ from those computed from the original time series and we will refer to them as "residual" trends and "full" trends, respectively. Trends and their statistical significance are also calculated 235 for all other observed parameters described in Section 2.3, namely visibility, clouds, radiation, and AOD, but no residual time series are calculated in these cases.

Indirect indicators for aerosol trends
Given the issues with aerosol observations in SWA described above we have to rely on indirect indicators to determine changes in air pollution. These are observations of cloud cover, radiation, and horizontal visibility. As there are not sufficient infor-240 mation for even the simplest radiative transfer estimates, we chose to estimate a potential aerosol effect on radiation through considering normalized trends. In this context, normalizing means subtracting the mean and dividing by the standard deviation of the respective time series. The dimensionless trends in such normalized time series can be directly compared to each other assuming no significant changes in variance during the considered time period (stationarity). If surface radiation was entirely determined by cloud cover, identical but opposite normalized trends would result. A non-zero sum of the normalized trends 245 points to an additional aerosol effect and / or changes in cloud optical thickness, which could at least in parts be related to aerosol. Positive sums (e.g., slight reduction in radiation but strong increase in cloud) would then indicate reduced aerosol, while negative sums (e.g., strong reduction in radiation but only slight increase in cloud) points to increased aerosol. This concept will be applied to the time series of CM SAF satellite data and surface observations from Parakou and Lamto in Section 3.3.

250
Another data set we use as indirect aerosol indicator are surface observations of horizontal visibility as estimated by human observers. Trends in this parameter are most likely an effect of changes in the aerosol burden but changes in low-level humidity and clouds may also have an effect, as they change incoming light and the split between direct and diffuse radiation. In order to examine a possible influence of clouds on visibility, the concept of normalized trends as explained above is also applied to horizontal visibility and low and medium cloud cover from station observations in section 3.3.

Results
In this section, trends in rainfall (section 3.2) and various indicators of aerosol burden (section 3.3.) are presented and related to each other. To put the results into context, the short section 3.1 summarizes the seasonal evolution of some key meteorological variables. Figure 3 shows the time mean seasonal evolution of key variables based on ERA5 and CHIRPS rainfall averaged over the main study region shown in Fig. 1. The main seasons as used in this paper are delineated by vertical black lines. During the main dry season during December-February rainfall drops to values well below 1 mm day −1 followed by a gradual ramp-up towards the FRS starting in mid-May. Peaks during the two rainy seasons FRS and SRS reach about 8 mm day −1 , while rainfall drops to below 4 mm day −1 during the LDS. The monsoonal retreat in October and November occurs much more rapidly than the 265 onset.

Climatological conditions
The arrows in Fig. 3 show average wind profiles for the three seasons of interest. Clearly the monsoon flow is strongest in the LDS reaching meridional wind speeds exceeding 4.5 m s −1 at 950 hPa with southerly winds up to 875 hPa, marking the depth of the monsoon layer. The FRS is also characterized by marked southerlies and an even deeper monsoon layer, while the SRS shows weaker flow reduced to a shallower layer. This is consistent with Guedje et al. (2019), who investigated upper air data 270 at the coastal station of Cotonou. As already discussed earlier, we anticipate that this should restrict the largest aerosol effects to the coastal strip, which contains the main pollution sources. The higher rainfall in SRS as compared to the LDS should also lead to more wet deposition, which would further support the concentration to the vicinity of sources. Wet deposition will of course also be enhanced during the FRS.
Another parameter of interest is the frequency of relative humidity of 95% and higher, as this determines the potential for wet 275 growth of aerosol, which strongly enhanced radiative effects (Deetz et al., 2018a;Haslett et al., 2019b). Here large differences are seen between the deep layer of very moist conditions in LDS as compared to the much shallower layers in the rainy seasons with maxima at 950hPa (FRS) and even 975hPa (SRS). Qualitatively, these results suggest that considerable aerosol effects during LDS could in fact be spread over a much larger region through the combination of faster transport, less wet removal, and wet aerosol growth. Finally the red curve in Fig. 3 shows the mean vertical profile of cloud cover restricted here to 06 UTC 280 data (corresponds to local time in the study region), the analysis time closest to the diurnal maximum (van der Linden et al., 2015). We anticipate this factor to influence the significance of aerosol changes to the surface energy balance, when incoming radiation is strongly reduced by cloud cover. Here differences between the three seasons are small in peak coverage, which typically occurs at 950hPa, while in the LDS the cloud layer appears to be somewhat thicker.
For the remainder of the paper, the analysis will concentrate on LDS and SRS, for which we found the largest indications 285 for aerosol effects. The corresponding results for the FRS are provided in the Supplementary Material and will be referred to along the way. Figure 3 offers relatively little in terms of potential reasons of fundamentally different behavior in the FRS, so we hypothesize that the much larger degree of convective organization during this season found by Maranan et al. (2018) is key in reducing sensitivity to local aerosol effects. In addition, spatial patterns of positive and negative rainfall trends during FRS change depending on the time period analyzed (cf. Supplementary Fig. S2a, b with S2d, e) and there are more pronounced 290 discrepancies between CHIRPS and station data. Regionally averaged trends are weak and statistically not significant on the 20% level ( Supplementary Fig. S2c, f).  Despite the relatively high correlations with SSTs over the Atlantic (see Fig. 2), the residual trends (Fig. 4b) do not differ fundamentally from the full trends. The main reason for this is the small warming of the tropical Atlantic during this season ( Fig. 4c) that dominates the statistical model at the majority of grid points and stations (see Section 2.4). Amongst the few 310 exceptions are Central Ghana and western Ivory Coast that both show a combination of all five climate indices to give the best statistical model (Fig. 2a). The Niño3.4 index, for example, is anti-correlated with rainfall and increases by 0.1 K per decade (not shown). This explains that residual trends are more positive than the full trends in regions where this index is part of the best linear model. Nevertheless, overall the highly negative residual trend points to a possible effect of increased dimming by aerosol. This would not stand in contrast with the positive trends at coastal stations as also seen in Fig. 4b, because the 315 additional aerosol radiative forcing needs time to take effect over land, while the immediate coast is dominated by advection from the ocean (and was most strongly changed by urbanization). The positive trends to the northeast of the Mampong Range in Central Ghana, the Atakora Mountains in Togo and Benin, and the Oshogbo Hills in northwest Nigeria could be an indication of delayed convective triggering over higher ground due to the reduced solar insolation, allows rainfall systems to travel further downstream. 320 Figure 4c shows the CHIRPS trends averaged over all grid points of the study region. Not surprisingly, both full and residual time series are quite similar with relatively small trends as compared to the large interannual variability. Subtraction of the SSTbased statistical model reduces the trend from −0.026 mm day −1 yr −1 to −0.022 mm day −1 yr −1 but, since the interannual variability is also reduced, this leads to statistical significance on the 20% level. The latter value translates to 23 mm per month over the 33-year period, which corresponds to nearly 20% of monthly rainfall during the LDS.

325
in Nigeria, while large parts of SWA show negative trends as low as −0.04 mm day −1 yr −1 (Fig. 5e). The change is most dramatic east of Lake Volta in Togo and Benin. Most of the large coastal cities, with the exception of eastern Ghana, now also show clearly negative residual trends.
Spatial averages over the coastal strip (Fig. 5f) show a corresponding shift to more negative trends. For the full time series this amounts to −0.050 mm day −1 yr −1 , while the residual trend is −0.087 mm day −1 yr −1 and thus 3.3 times larger than for 1983-2015. The residual trend is statistically significant on the 20% level, partly also due to a reduced year-to-year variability 365 when SST effects are removed. Over the 17-year period, this trend corresponds to 45 mm per month or 90 mm over the September-October period of the SRS. This is a substantial reduction relative to typical SRS totals of 350 mm and annual totals of 1400 mm (Sanogo et al., 2015).

Indirect indicators for aerosol trends
In the previous section, we have demonstrated that -once influences of SST changes are corrected for -large parts of SWA 370 have undergone an accelerating drying over recent decades. Seasonal and geographical patterns, together with the acceleration, are consistent with the hypothesis of a human influence through rapidly growing emissions of pollutants. What other evidence do we have to support this idea? As we will discuss in more detail in the next subsection, usable aerosol measurements are largely restricted to the ocean adjacent to SWA, impeding the establishment of a direct link to rainfall. Therefore we turn here to indirect indicators such as horizontal visibility (section 3.3.2) and SDSR (section 3.3.3), which need to be regarded in concert 375 with cloud cover that influences both quantities.   8b and 8c show normalized trends of SDSR and LLC for Parakou and Lamto, respectively. The normalized cloud trends correspond to those shown in Fig. 7d but are given here on a monthly basis as in Fig. 8a. For Parakou, this analysis results in a smoothing of the annual cycle of the negative SDSR trends but the minimum during the LDS and SRS remains (blue line in Fig.   8b). Cloud changes are near zero from January to May and then remain positive but small for the rest of the year (red line in Fig. 8b). Making the assumptions outlined in Section 5, this implies a residual trend that we assume to be due to aerosol during 465 the month from June to October with a peak in September (golden line in Fig. 8b). The corresponding analysis for Lamto (Fig.   8c) shows consistent behavior for SDSR, i.e., an overall smoother evolution after normalization with a minimum during the LDS and SRS, but a much larger positive trend in cloudiness. The residual -supposedly aerosol-related -trend is near zero from June to November and even negative in the rest of the year with a minimum in April. A caveat in this result is that Lamto does not provide the standard cloud cover reports available for Parakou but a more qualitative estimate called total cloud area 470 fraction (TCAF), which maybe less reliable, particularly when human observers change over time. If however we deem the results to be reliable, they would suggest that the wet season effect we see at Parakou and in most visibility estimates is weak in Lamto, for which we do not have any visibility data to back up the radiation analysis. Being a remote forest observatory and located relatively far west, so potentially not much affected by northeastward advection from the big coastal cities and biomass burning aerosol from Central Africa, a small aerosol signal is not implausible. Nevertheless, the signal during the rest of the 475 year remains surprising and may point to an increase in rainfall during the Sahelian dry season and accompanied washout of aerosol particles (Sanogo et al., 2015). As this part of the year is not in the focus of this study, we leave a more detailed analysis to future work.

Satellite-based aerosol estimates
Finally, Fig. 8d shows a corresponding analysis entirely based on satellite data and averaged over the study region (see  In this paper we have investigated the hypothesis whether the observed recent increase in manmade aerosol pollution over SWA could have impacted seasonal rainfall trends on decadal timescale. Given ongoing issues with climate models to realistically represent the West African monsoon, we decided to analyze this question based on available observational records alone. Given a large interannual variability, a strong influence from other climatic factors such as SSTs, and an overall relatively limited database, the investigation strategy was to look at this problem from as many different angles as possible to reach a 495 balance of evidence, even if statistical significance for individual factors may be low. Trend analyses are presented both for an extended period from 1983-2015 and a shorter more recent period from 2001-2017 (with some smaller deviations due to data availability). Most attention is given to the LDS, when the rainfall maximum is to the north of SWA, and to the SRS, while results for the FRS are less significant and thus only given in the Supplementary Material.
The main findings of this paper are: 500 -Interannual rainfall variability across most of SWA and during both LDS and SRS is positively correlated with SSTs over the tropical Atlantic, while the influence of other climatic factors is relatively weak. A multi-linear statistical model was produced to remove these effects in order to isolate residual trends that could be related to aerosol.
-During the LDS, SWA has dried significantly with the notable exception of the coastal areas between Accra and Lagos, where urbanization may have affected the local land-sea breeze circulation. Slightly warming SSTs in the Atlantic 505 during the last two decades have also supported rainfall increases in the coastal areas. Removing those effects leaves a statistically significant and considerable negative trend over most of SWA that has accelerated in recent years.
-During the SRS, SWA has become wetter in the far north and at the immediate coast with drying in between, particularly over Ivory Coast and Ghana. In the last two decades, the drying trend spread into the coastal zone and farther inland, restricting positive trends to central Ivory Coast, the northern Lake Volta area, and smaller parts of Nigeria. As in the 510 LDS, removing the influence of warming SSTs, creates a predominantly negative trend throughout most of SWA, but particularly in the coastal strip.
-Satellite estimates of AOD show large increases over the tropical Atlantic to the immediate south of SWA, but over land measurements are impeded by too frequent cloud cover. These increases are related to biomass burning aerosol from Central Africa being carried into SWA, particularly during the LDS, when this import adds significantly to local sources 515 (Haslett et al., 2019a).
-Station estimates of horizontal visibility suggest a substantial decrease over recent decades, particularly over coastal and low-lying inland locations. These cannot be explained by changes in cloudiness.
-Station and satellite estimates of SDSR suggest a marked reduction throughout most of the year but particularly during the LDS and SRS. This is partly explained by an increase in cloudiness and partly through an increase in aerosol, 520 consistent with the visibility analysis.
-These results cannot proof a local aerosol effect on rainfall but the balance of evidence strongly suggests such an association. Amongst the arguments to support such a claim are the geographical distribution (more pollution and rainfall suppression in the lowlands and even small improvements over higher ground), the seasonality (a larger affected area during the LDS due to faster transport, more pollution import from Central Africa and less wet deposition), and in particular the worrying acceleration of trends in the last two decades when pollution levels rose strongly (Liousse et al., 2014).
-With respect to possible mechanisms of an aerosol impact on rainfall, this work cannot provide any new evidence but can extrapolate findings of past studies based on shorter time periods or high-resolution model results, particularly those obtained in the framework of the DACCIWA project. These suggest little influence of aerosol on cloud properties making 530 indirekt effects rather unlikely to be the predominant pathway (Deetz et al., 2018b;Taylor et al., 2019). On the other hand, there is evidence that the high relative humidity at low levels, particularly during the LDS and SRS (Fig. 3), leads to wet growth of aerosol particles, which strongly enhances the direct effect (Deetz et al., 2018a;Haslett et al., 2019b). For the LDS Kniffka et al. (2019) have already demonstrated the high sensitivity of rainfall to incoming solar radiation, supporting the idea of a direct aerosol effect. Such work does not exist for the FRS and SRS, but the rainfall 535 classification by Maranan et al. (2018) suggests a predominance of long-lived, organized systems in the former and more locally triggered convection in the latter, possibly explaining the larger aerosol signal in SRS.
Despite the often indirect evidence presented here and admittedly in parts qualitative argumentation, we feel that the presented results are strong and convincing enough to justify more attention to this problem. While a negative influence of a long-term increase of manmade aerosol has been claimed for other regions such as southern Africa (Hodnebrog et al., 2016) 540 and eastern China (Huang et al., 2016), this study is the first to raise this issue for SWA. The consequences of this are twofold: First, scientists should increase efforts to better understand the mechanisms involved in the aerosol-rainfall connections using combinations of ground and satellite data in concert with models capable of representing the full complexity of the problem at hand. For the adjacent Sahel, Marvel et al. (2020) recently applied a multivariate fingerprinting technique to show that a greenhouse gas forcing signal is already detectable in the early 21st century while a forcing signal from global aerosol changes 545 can be expected to emerge only in the middle of this century. It would be interesting to conduct a comparable study for regional aerosol in SWA. Second, policymakers in SWA are advised to prevent a further increase in air pollution through suitable regulations and improved technology (Evans et al., 2018). This paper has shown that the aerosol-induced rainfall suppression is significant (order several 10s of mm per month) and has been accelerating in the last two decades, although some of the effect has been concealed by opposing effects from SST changes. Allowing the air pollution problem to further deteriorate in the 550 future could therefore cause significant socio-economic damage through impacts on human health and water supply, which in turn is closely linked to food security and energy production in SWA. September-31 October). Cloud cover and relative humidity range from 0 to 1 according to the scales at the bottom for each season. Symbols below the seasons' names mark the degree of convective organization according to Maranan et al. (2018).   ). b, c, Normalized trends of SRSR and cloud cover at Parakou (b) and Lamto (c). An aerosol signal (gold) is derived from SDSR and cloud cover (see Section 2.5 for details). While both stations measure SDSR, low and medium cloud cover (LLC) is only observed at Parakou. For Lamto the total cloud area fraction (TCAF, see Section 2.4) is used to calculate the aerosol signal. d, As b, c but for Satellite Application Facility on Climate Monitoring (CM SAF) data averaged over the entire study region (see Fig.   1). Instead of cloud cover the effective cloud albedo (ECA) is used to calculate the aerosol signal. In all panels a * (**) marks statistically significant trends on the α = 20% (5%) level.